Analytical approach to piezoelectric model synthesis with the use of Cauer’s method for system design

Jacek Harazin; Andrzej Wróbel

doi:10.12688/f1000research.140943.3

Home Browse Analytical approach to piezoelectric model synthesis with the use...

ALL Metrics

Views

Downloads

Get PDF

Get XML

Export

▬

✚

Research Article

Revised

Analytical approach to piezoelectric model synthesis with the use of Cauer’s method for system design

[version 3; peer review: 2 approved, 2 approved with reservations]

Jacek Harazin ¹, Andrzej Wróbel¹

PUBLISHED 06 Sep 2024

Author details Author details

¹ Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, Silesian Voivodeship, 44-100, Poland

Jacek Harazin
Roles: Data Curation, Formal Analysis, Funding Acquisition, Investigation, Methodology, Software, Visualization, Writing – Original Draft Preparation

Andrzej Wróbel
Roles: Conceptualization, Project Administration, Resources, Supervision, Validation, Writing – Review & Editing

OPEN PEER REVIEW

REVIEWER STATUS

Abstract

Background

Piezoceramic materials have unique property which enables direct and bilateral conversion between mechanical and electrical energy. This ability facilitates significant miniaturisation of technology and opens many opportunities in design of new actuators and energy harvesters. Mathematical modelling of piezoelectric modules is notoriously hard due to complex constitutive equations defining mechanical and electrical energy conversion.

Methods

The article presents research on a new synthesis method based on the Cauer’s method. Mechanical damping is introduced with the use of Rayleigh’s approximation. A discrete electromechanical model is formed based on the Mason’s piezoelectric model. The proposed approach allows modelling of piezoelectric systems based on a set of characteristic frequencies. The method allows a more general approach to the problem of modelling new systems, as opposed to application-oriented methods seen in literature. A non-standard model analysis method using edge graphs and structural numbers is also verified as a potential alternative for matrix-based method. The authors compare their precision and computation requirements.

Results

The structural method of mechanical model analysis gave identical results as the reference matrix method. However, the non-classical algorithm took much longer to calculate and was using more memory. The electromechanical model analysis has shown an error of 5% in comparison to resonance frequencies taken from a reference plate specification. The calculated magnitude of displacement was well above the capability of a 3.5mm thick piezoelectric plate.

Conclusions

The synthesis method presented in this paper allows synthesizing piezoelectric cascade models based on limited information in form of characteristic frequencies. Currently this method allows a coarse approximation of the real piezoelectric parameters with limited number of inputs. The additional method of analysis based on structural numbers offers a promising alternative to matrix calculations but requires a more thorough investigation of the computational power required to determine whether it can compete with existing algorithms.

Keywords

piezoelectricity, mathematical model, synthesis

Corresponding author: Jacek Harazin

Competing interests: No competing interests were disclosed.

Grant information: The article was financed from the subvention for maintenance and development of research potential of young scientists of the Silesian University of Technology.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Copyright: © 2024 Harazin J and Wróbel A. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite: Harazin J and Wróbel A. Analytical approach to piezoelectric model synthesis with the use of Cauer’s method for system design [version 3; peer review: 2 approved, 2 approved with reservations]. F1000Research 2024, 12:1273 (https://doi.org/10.12688/f1000research.140943.3) First published: 05 Oct 2023, 12:1273 (https://doi.org/10.12688/f1000research.140943.1) Latest published: 06 Sep 2024, 12:1273 (https://doi.org/10.12688/f1000research.140943.3)

Revised Amendments from Version 2

The introduction section was rebuilt using more up to date references including numerical analysis of piezoelectric models. References to the research works using the Eshelby tensor and Mori-Tanaka methods were also added, along with references mentioning electro-mechanical homogenization methods. The research gap statement was rewritten to make it more conscise and clear. Additional figure was added to better explain the edge graphs and structural method used in the article.

Conclusions were also condensed and formed into a series of bullet points to make them clearer and more approachable by the readers, as per reviewer's requests.

See the authors' detailed response to the review by Pedro M. Ferreira
See the authors' detailed response to the review by Saurabh K. Yadav
See the authors' detailed response to the review by Guangtao Lu

Introduction

Piezoelectric materials have found extensive applications in mechanical engineering due to their unique properties. The mechanics of the piezoelectric phenomenon allow a significant reduction in the size of systems due to direct conversion between mechanical and electrical energy within the structure of the material. The effect is reversible which means that piezoelectric materials can be used as both actuators and sensors in a broad spectrum of applications. Due to many stringent requirements imposed by the highly specialised applications of piezoelectric systems, technological processes are constantly being improved to create piezoelectric materials with ever better technical properties. Piezoelectric crystals with increased coupling coefficients are created by extending the monocrystalline structure throughout their volume while multilayer piezoelectric composites of polymers and ceramics are developed to improve their mechanical properties. The demands on precision and performance of piezoelectric transducers and new material structures are a challenge for engineers and researchers who use numerical methods to model the behaviour of piezoelectric systems.¹^–¹⁷ A set of constitutive equations for piezoelectricity¹⁸ has been established to form the basis for the electromechanical coupling effect modelling. Because of the complex dimensional relations between electric fields and mechanical stresses, mathematical modelling of piezoelectric components has been approached with many different methods. Soluch used matrix tensor equations in his work¹⁹ to model wave propagation inside different piezoelectric volumes. He pointed out however, that modelling the wave propagation in all three dimensions is too complex of a task due to the number of reflections affecting the wave. Lumped models have been used to model the behaviour of piezoelectric transducers using a combination of basic elements used in theory of control such as capacitors, coils and resistors.⁹ Most notable is the Maxwell model used by researchers like Goldfarb and Celanovic⁸ to model the effect of hysteresis. Adriaens et al. in their work⁷ have also tried to combine electrical and mechanical elements from both domains to better model the damping related to energy dissipation inside the piezoelectric material. There have also been works trying to approach the problem of piezoelectric material coefficient determination using FEM (finite element method) analysis, like the research work of Lahmer et al.¹⁰ Complex piezoelectric structures utilizing multi-phase composites with inclusions made of piezoelectric materials with varying mechanical properties created a necessity for models capable of representing the mechanics between different material and grain boundaries. Numerous micromechanical models have been created based on the Eshelby tensor and Mori-Tanaka method such as the work of Sangryun Lee et al.¹ where the researchers tried to predict effective properties of the composite material made from different piezoelectric materials. Nada Tassi et al. in their work² have also tried using micromechanical modelling to model the behaviour of piezoelectric components under large deformation and high electric field. Fateh E et al. used electro-mechanical homogenization-based model to model the behaviour of piezoelectric materials at α → β phase transition, while also considering damage accumulating at interfacial boundaries.³ There are also analytical methods that can be labelled as “non-classical” which rely on machine learning algorithms such as the work of Jeffrey Hu and Yuqi Song,⁴ where the researchers try to predict the piezoelectric modulus using machine learning and graph neural networks. Wang and Tsai²⁰ used block diagrams to model the behaviour of a thickness mode piezoelectric transducer and Wróbel²¹ employed a method relying on edge graphs to model piezoelectric relations in the system based on the matrix of constitutive equations.

There is a research gap however when it comes to the subject of a full piezoelectric system synthesis where majority of the initial parameters are unknown. Many of the developed models are used on a case-by-case basis where system variables defining piezoelectric behaviour are fully or partially known or they are being empirically determined.¹⁰^–¹⁶^,²⁰ Most of presented methods focus on the analysis of different model configurations and providing data on the model behaviour depending on the input information. Those methods can be used to reliably prototype new transducers but need a large prior in-depth knowledge about desired configuration of materials. This article is focused on the development of a new method for piezoelectric system synthesis where the only initial information comes in the form of established resonant and anti-resonant frequencies.

Wilhelm Cauer was among the pioneers of electrical model synthesis, creating algorithms based on transfer function polynomials, defining large range of electrical systems by using mathematical decomposition. Cauer synthesis²²^,²³ relies on decomposing transfer function polynomials into continued fractions. The set of coefficients of a continued fraction represents values defining each system component properties (like impedance or capacitance). This procedure allows any transfer function of a complex system to be broken down into a series of components represented by a lumped model. Moreover, because of existing analogies between electrical and mechanical elements, this method can be applied to both electrical and mechanical models.²⁴^–²⁷

Presented method allows a more general approach to the problem of modelling new systems, as opposed to application-oriented methods focused on modelling systems for very specific use cases, proposed by cited researchers. The algorithm based on Cauer synthesis method is used to construct an initial discrete mechanical model which serves as the base for piezoelectric model synthesis. Mechanical damping is introduced to the model with the help of Rayleigh’s method.²⁸ The transition between purely mechanical model and a discrete electromechanical model of the piezoelectric plate is done with the combination of Mason’s model and analogies between mechanical and electrical components. Because the initial information is restricted only to a desired set of characteristic frequencies, there are many unknown piezoelectric parameters which have to be determined in this process. To solve this problem, authors have chosen an approach where most of the parameters are being approximated by using the obtained mechanical properties of an already synthesized system and a set of equations derived from the constitutive equations of piezoelectricity.

Part of this article has been devoted to a study on the potential alternative for matrix-based analysis of mechanical models with damping, by employing edge graphs and structural number algebra.²⁴^,²⁵^,²⁷^,²⁹^–³¹ The topic of analysis of mechanical systems is indirectly related to the topic of synthesis of piezoelectric systems. The studied analysis methods are used to verify synthesized piezoelectric systems. The article presents a comparison of results obtained with both methods on a variety of mathematical models. A study was also carried out to test the performance of computer algorithms developed on the basis of the non-classical method against standard matrix-based computer algorithms. Other methods based on edge graphs bare close naming resemblance like the ones used in¹¹^,²⁰ which may cause confusion. As a matter of clarification, this article is focusing on the method using edge graphs in combination with structural number algebra. Structural numbers visually resemble matrix calculation as each structural number construct consist of rows and columns of numbers. The number of rows and columns, however, is tightly connected to the shape of an edge graph which a given structural number relates to.

An example of a structural number created based on a graph in Figure 1 is as follows:

(1)

D = [ab] [acd] [de] = [\begin{matrix} a & a & a & b & b & b & b & b \\ c & c & d & a & a & c & c & d \\ d & e & e & d & e & d & e & e \end{matrix}],

Figure 1. An example of a graph taken as a basis for a structural number formulation.

where:

$a, b, c, d, e$ – graph edges,

$D$ – graph structural number.

Arithmetic and logic operations based on structural number algebra were described in depth by professor Stanisław Bellert in his book²⁹ and in articles.³⁰^,³¹ More details can also be found in previous article published by authors.³² The article also directly relates to adaptations of this method used by other researchers for analysis and synthesis of various systems.²^,²¹^,²⁷

Methods

Mechanical model synthesis

The synthesis process begins with a selection of predetermined resonant and anti-resonant frequencies. The set of frequencies may be freely defined according to future applications or may result from a vibration spectral analysis of the object of interest. In this method, the identified set of frequencies is the only input information. It is used to generate a transfer function polynomial, which can later be transformed into a chain fraction equation. Cauer’s algorithm mentioned in Refs. 25, 27, 33, 34 was used to create four different types of mechanical models depending on the allocation of coefficients in the chosen transfer function polynomial. Depending on the polynomial, the cascade model resulting from the chain fractioning process can be constrained on one end or not constrained at all. Additionally, the model can be excited by either a kinetic or dynamic force. A combination of those factors creates four different types of models with their own transfer functions (detailed examples of each type and associated transfer function structures are given in Refs. 25, 27). The model chosen for the purposes of this article is a cascade model with one of its ends constrained and which is being excited by the dynamic force. The choice is motivated by the close resemblance of such a model to a fixed piezoelectric stack, where individual plates can be described as successive stages of the model. According to the models proposed in Refs. 25 and 27, a hypothetical transfer function for an infinite degree cascade model constrained on one end and excited with a dynamic force takes the factorial form:

(2)

T (s) = H \frac{(s^{2} + ω_{1}^{2}) (s^{2} + ω_{3}^{2}) \dots (s^{2} + ω_{n - 2}^{2}) (s^{2} + ω_{n}^{2})}{s (s^{2} + ω_{2}^{2}) (s^{2} + ω_{4}^{2}) \dots (s^{2} + ω_{n - 1}^{2})},

where:

T(s) – response of a system,

H – response amplification factor,

ω – consecutive frequencies (even are resonant and odd are anti-resonant),

s – Laplacian variable.

The factorial form is transformed into a standard polynomial form to define the coefficients for successive powers of the equation:

(3)

T (s) = H \frac{a_{i} s^{n} + a_{i - 1} s^{n - 2} + \dots + a_{1} s^{2} + a_{0}}{b_{i} s^{n - 1} + b_{i - 1} s^{n - 3} + \dots + b_{1} s},

where:

a_i – numerator coefficients,

b_i – denominator coefficients.

After the initial calculation of the polynomial coefficients, the transfer function is converted into a continued fraction by using the Cauer’s method in its first form.²² The method involves breaking down a polynomial into a chain fraction using simple mathematical operations. The first form of Cauer synthesis is implemented by dividing the coefficient of the greatest power of a numerator by the coefficient of the greatest power of a denominator. The resulting fraction is being multiplied by the denominator and subtracted from the polynomial function, leaving the rest, which is then inverted. This process is repeated until all powers of the transfer function are reduced:

(4)

I_{p}^{1} = \frac{a_{i} s^{n}}{b_{i} s^{n - 1}} = \frac{a_{i}}{b_{i}} s,

(5)

R_{p}^{1} = \frac{a_{i} s^{n} + a_{i - 1} s^{n - 2} + \dots + a_{1} s^{2} + a_{0}}{b_{i} s^{n - 1} + b_{i - 1} s^{n - 3} + \dots + b_{1} s} - I_{p}^{1},

(6)

R_{p}^{j + 1} = \frac{1}{R_{p}^{j}} - I_{p}^{j + 1},

where:

$I_{p}^{i}$ – fraction extracted by division of parameters next to the highest power “s”,

$R_{p}^{i}$ – rest of the original polynomial after subtraction.

A continued fraction resulting from the process can be used to determine the values of discrete component parameters in the synthesized system:

(7)

\frac{T (s)}{H} = I_{p}^{1} ∙ s + \frac{1}{I_{p}^{2} ∙ s + \dots + \frac{1}{I_{p}^{n - 1} ∙ s + I_{p}^{n} ∙ s}} .

Obtained values can be used in either mechanical or electrical cascade systems which gives this method a high degree of flexibility. An example of an infinite stage mechanical system resulting from the calculations have been shown in Figure 2.

Figure 2. A mechanical cascade system composed of an infinite number of stages composed of mass “m” and spring “c” pairs, synthesized from the transfer function polynomial with the Cauer’s method.

The newly obtained mechanical model consists of springs and inertial components which can be used to determine the stiffness and mass/density of the piezoelectric material. Damping elements can be added to the model to account for energy dissipation which is an inherent property of piezoelectric materials. The mechanical quality factor is often used to indicate piezoelectric mechanical damping ability. The factor is inversely proportional to the amount of damping generated by the material and can be derived from Rayleigh’s method.²⁸^,³⁵ However, damping introduces complex derivatives into the equations of motion for any mechanical or electrical system. Every system can be considered under the case of subcritical, critical, and over-critical damping. Depending on the strength of damping elements, a system can be in a state of damped oscillations (subcritical damping) or the oscillations may stop immediately after the harmonic excitation disappears (critical damping), they do not occur (over-critical damping). Because of that, there are three separate integral solutions that describe each case respectively. A resonance (additive interference of harmonic oscillations) occurs only under the case of subcritical damping. In order to limit the range of possible solutions, it was assumed that all models would only be considered under subcritical damping, thus eliminating cases where vibrations cannot resonate. The decision was made because the project was focused on piezoelectric applications where vibration generation was their main purpose. Another assumption limited the range of solutions to those with a Rayleigh damping coefficient related to stiffness. According to literature,¹³ most cases of piezoelectric damping are restricted to damping with respect to element stiffness only. The two assumptions simplify the term for the damping coefficient. Based on the obtained equations, it is then possible to determine the damping parameters for each stage in relation to the spring stiffness:

(8)

β < \frac{2}{ω_{1} + ω_{n}},

(9)

b_{i} = β c_{i},

where:

$β$ – stiffness damping coefficient,

$b_{i}$ – damping value,

$c_{i}$ – spring stiffness.

To verify the model in its current state, an intermediate electromotive force was also added at each stage. Its role was to simulate the actuation of piezoelectric plates by an electrical stimulus, similar to.¹⁴^,²⁰ The electromotive force acts as a simplified force generated by the electromechanical coupling used to describe the conversion of energy conversion between the mechanical and electrical part of a piezoelectric plate. The mechanical model with added damping and electromotive forces is shown in Figure 3.

Figure 3. A discrete cascade mechanical model with added damping elements.

The model masses are marked with “m” symbols, springs with “c” symbols, dampers with “b” symbols and electromotive forces with “G” symbols. The displacement of each mass was marked with “x”.

Mechanical model verification

The synthesized model had to be verified after the addition of new damping elements and additional forces. An analysis of the model response was carried out in terms of its magnitude along the frequency axis. Two analysis methods were used to verify the existing mechanical model. The classical matrix method served as a reference for an additional method using structural numbers.²¹^,²⁷^,²⁹^–³¹ The use of a second method is considered as part of the ongoing research³² aimed at evaluating the accuracy and computational load of structural numbers in comparison to the matrix calculation. Current state of the algorithm based on structural numbers should still not be considered as final as there are more improvements to be made in terms of its compatibility with computation software. Starting with the classical method, a set of equations of motion was created to describe relations between parts of the model. To avoid overcomplicating the example, only two degrees of freedom were considered. The equations of motion for a two-stage model derived from the Lagrangian equations were as follows:

(10)

\{\begin{cases} G_{1} = m_{1} {\ddot{x}}_{1} + b_{1} ({\dot{x}}_{1} - {\dot{x}}_{2}) + c_{1} (x_{1} - x_{2}) \\ G_{2} = m_{2} {\ddot{x}}_{2} - b_{1} ({\dot{x}}_{1} - {\dot{x}}_{2}) + b_{2} {\dot{x}}_{2} - c_{1} (x_{1} - x_{2}) + c_{2} x_{2} \end{cases} .

Under the assumption into account, that the system is only considered in the case of subcritical damping, an integral of a second-order differential equation³⁶ for the variable displacement x can be described by the equation:

(11)

\{\begin{cases} x_{1} = A_{11} sinωt + A_{12} cosωt \\ x_{2} = A_{21} sinωt + A_{22} cosωt \end{cases},

where:

$A_{i 1}$ – real part of the i^th element amplitude response,

$A_{i 2}$ – imaginary part of the i^th element amplitude response.

Combining (10) with (11) and separating the components with sine and cosine coefficients, then reordering the equations by combining all the coefficients standing next to consecutive amplitudes, yields a coefficient matrix:

(12)

[\begin{matrix} k_{1} - m_{1} ω^{2} & - b_{1} ω & - k_{1} & b_{1} ω \\ b_{1} ω & k_{1} - m_{1} ω^{2} & - b_{1} ω & - k_{1} \\ - k_{1} & b_{1} ω & - m_{2} ω^{2} + k_{1} + k_{2} & - b_{1} ω - b_{2} ω \\ - b_{1} ω & - k_{1} & b_{1} ω + b_{2} ω & - m_{2} ω^{2} + k_{1} + k_{2} \end{matrix}] .

Based on the obtained coefficient matrix (12) and a vector of actuating forces, a vector of amplitudes can be calculated. To obtain absolute displacement, the sine and cosine coefficient must be combined using Pythagoreans’ theorem:

(13)

\begin{array}{l} A_{1} = \sqrt{{(A_{11})}^{2} + {(A_{12})}^{2}} \\ A_{2} = \sqrt{{(A_{21})}^{2} + {(A_{22})}^{2}} \end{array}

To calculate phase-shift of each stage along the frequency spectrum, a tangent between the real and imaginary component of displacement can be calculated:

(14)

\begin{array}{l} φ_{1} = atan \frac{A_{11}}{A_{12}} \\ φ_{2} = atan \frac{A_{21}}{A_{22}} \end{array} .

The set of derived equations (13, 14) allows the amplitude response and phase shifts to be determined for any cascade system built using standard discrete elements representing model inertia, stiffness, and damping. Calculations are relatively quick and easy when considering systems with a small number of degrees of freedom. However, calculations become much more complex and computationally intensive for systems with a large number of degrees of freedom. This is an effect of the number of elements being multiplied in the determinant of a matrix. The numbers get big really fast because of the accumulation of the powers when all the matrix components are being combined. This problem prompted the search for alternative solutions for rapid analysis of more complex systems.

One of the alternatives to matrix calculations is being developed by the researchers at the Silesian University of Technology.²¹^,²⁴^,²⁵^,²⁷ The non-classical method of system analysis involves the use graphs and structural numbers to illustrate relations between forces in a system. The structural number method offers an alternative solution for deriving equations of motion by using a graphical representation of relations in the system. Each edge of the graph corresponds to a single discrete relation between two elements of a system, represented by nodes. All the edges and nodes in the graph are assigned numbers called “structural numbers”. These numbers form structures similar to matrices based on edge connections with nearby nodes. Using the algebra of structural numbers, it is then possible to derive a system of equations similar to equations of motion. A detailed description of this method in the context of solving mechanical cascade systems without damping was presented in Ref. 32. The structural number algebra is an entirely separate method of performing mathematical operations and therefore requires extensive explanation. It is strongly recommended to consult the book of the original author of this method²⁹ or related articles³⁰^,³¹ for the details on its usage.

In the context of this article, several necessary changes had to be introduced to account for damping elements in the model. The basic method does not allow a simple decomposition of the sine and cosine coefficients when damping is present in the system. A standard approach includes resolving the equations with motion derivatives replaced with Laplacian transform. The issue arises when the final equation obtained with structural number algebra is turned back into its derivative form. When motion derivatives are replaced with the integral solution for sub critically damped motion as in (11), the equation becomes extremely tangled. To overcome this problem, an intermediate method has been developed, which realises the mentioned decomposition of sine and cosine components (11) prior to the application of the structural number algebra, Effectively, that means that the entire graph of relations was duplicated, separating the edges corresponding to sine and cosine coefficients. Each case is then being calculated separately until the last step. Both resulting graphs are shown in Figure 4.

Figure 4. Graph of the cascade model where: a) shows the sine coefficients b) shows the cosine coefficients.

The nodes of each graph represent the degrees of freedom of the model they are describing. Each edge corresponds to forces acting between each degree and is being described by the formula of elements creating each force. “ω” is the frequency, “m” is the mass of each element, “c” corresponds to spring loads and “b” to damping.

The structural numbers of both graphs are synonymous, the only difference being the opposite sign in the first derivatives of the sine and cosine function, which relate to the damping coefficients. As with the matrix method, the sine graph corresponds to the real part of the calculation and the cosine part to the imaginary part. This time, in order to obtain absolute values of the displacements, the merging (similar to (13)) has to be performed for the whole structural number and all the simultaneity functions.²⁷^,²⁹^,³² The amplitude response of each stage can be calculated by solving the equations:

(15)

A_{11} = |\frac{(\binom{Sim}{z} (\frac{\partial D (ω)}{\partial [1]}) G_{11})}{\partial D (ω)} + \frac{(\binom{Sim}{z} (\frac{\partial D (ω)}{\partial [2]}; \frac{\partial D (ω)}{\partial [1]}) G_{21})}{\partial D (ω)}|,

(16)

A_{1} = \sqrt{{(A_{11})}^{2} + {(A_{12})}^{2}},

(17)

D (ω) = \sqrt{D_{s} {(ω)}^{2} + D_{c} {(ω)}^{2}},

where:

$D_{s} (ω)$ – structural number for the sine graph,

$D_{c} (ω)$ – structural number for the cosine graph,

$A_{ij}$ – amplitude calculated with simultaneity function.

The process of obtaining the necessary structural numbers to derive equations (15–17) and the meaning of a simultaneity function has been explained in more detail in.²⁹^,³² The computer algorithm which was later used for this purpose is described in detail under the section dedicated to the algorithm testing.

Electromechanical model synthesis

The synthesized mechanical model serves as a basis for the construction of a combined electromechanical model of a piezoelectric system. Constitutive equations and analogy between mechanical and electrical components were used to derive parameters of the piezoelectric model.⁸^,⁹^,¹³^,¹⁴ The electromechanical coupling coefficient for piezoelectric coupling perpendicular to the plate surface was determined based on resonant, anti-resonant frequency pairs:

(18)

k_{33}^{2} = \frac{\frac{π}{2}}{1 + \frac{∆ f}{f_{r}}} tan \frac{\frac{π}{2} + \frac{∆ f}{f_{r}}}{1 + \frac{∆ f}{f_{r}}}, ∆ f = f_{a} - f_{r},

where:

$k_{33}$ – electromechanical coupling factor in the 3-3 direction,

$f_{a}$ – anti-resonant frequency of a given stage,

$f_{r}$ – resonant frequency of a given stage.

A broader term defining electromechanical coupling for a general case, called an effective coupling coefficient was used to approximate the capacity of piezoelectric plates³⁷:

(19)

\frac{k_{eff}^{2}}{1 - k_{eff}^{2}} = \frac{∆ f^{2}}{f_{r}^{2}},

(20)

C_{p} = \frac{k_{eff} C_{m}}{1 - k_{eff}},

(21)

C_{m} = \frac{1}{c_{m}},

where:

$k_{eff}$ – effective electromechanical coupling factor,

$C_{p}$ – plate capacitance,

$C_{m}$ – equivalent capacitance from mechanical stiffness,

$c_{m}$ – mechanical stiffness of a lumped piezoelectric model.

Another property of a piezoelectric material frequently mentioned in specifications is the mechanical quality factor. This property describes the amount of energy that is being lost during dynamic operation of a piezoelectric plate. It corresponds to mechanical losses in a system. An approximate value of a mechanical quality factor can be calculated using the equation derived from the established Rayleigh’s method:

(22)

Q_{m} = \frac{1}{βω} .

Electrical permittivity can be also approximated if piezoelectric capacitance is known by using an equation:

(23)

ε_{33}^{T} = \frac{C_{p} h}{A},

where:

$ε_{33}^{T}$ – electrical permittivity of a piezoelectric material in 3-3 direction under constant stress,

$h$ – plate thickness,

$A$ – plate cross section field.

Lastly, based on the lumped stiffness of the mechanical model, a piezoelectric material stiffness can be obtained using Hooke’s law:

(24)

c_{33}^{E} = c_{m} \frac{h}{A},

where:

$c_{33}^{E}$ – piezoelectric stiffness in 3-3 direction under constant electrical field.

Piezoelectric discrete model verification

To analyse piezoelectric response with calculated parameters, a discrete electromechanical model was created based on the Butterworth-Van Dyke model and other lumped piezoelectric models available in the literature.⁷^,⁸^,²⁰^,³⁸ The basic form of constitutive equations for piezoelectric effect in a 33 mode was used:

(25)

\{\begin{matrix} D_{3} = ε_{33}^{T} E_{3} + d_{33} T_{3} \\ S_{3} = d_{33} E_{3} + s_{33}^{E} T_{3} \end{matrix},

where:

$D_{3}$ – electric displacement field,

$S_{3}$ – mechanical strain,

$E_{3}$ – electric field,

$T_{3}$ – mechanical stress,

$d_{33}$ – piezoelectric charge constant,

$s_{33}^{E}$ – mechanical compliance constant (inverse of mechanical stiffness constant).

The equation (25) was modified using Hooke’s law and by converting equations for underlying variables:

(26)

E_{3} = \frac{U_{p}}{h},

where:

U_{p}

– voltage across the piezoelectric plate thickness,

(27)

D_{3} = \frac{Q_{p}}{A},

where:

Q_{p}

– charge across the piezoelectric plate surface.

By combining the constitutive equations with (26,27) and plugging the equations of motion for the initial mechanical model (10) into the mechanical input of the electromechanical system, a lumped piezoelectric model was created. Relations inside a single piezoelectric plate are described by a system of equations:

(28)

\{\begin{matrix} m \ddot{x} + b \dot{x} + c_{m} x = - c_{m} d_{33} U_{p} + F \\ U_{0} C_{p} = x c_{m} d_{33} + U_{p} ε_{33}^{T} \frac{A}{h} (1 - k_{33}^{2}) \end{matrix},

where:

$F$ – external force applied to a piezoelectric plate,

$U_{0}$ – external voltage applied to a piezoelectric plate,

$m$ – lumped piezoelectric mass,

$b$ – lumped piezoelectric damping,

$c_{m}$ – lumped piezoelectric stiffness,

$U_{p}$ – lumped piezoelectric capacitance,

The discrete piezoelectric model has been shown in Figure 5. Inputs and outputs were extracted from (28) by grouping parameters and converting the equation into a Laplace form. The resulting matrix was used to create an algorithm for piezoelectric model analysis:

(29)

[\begin{matrix} X (s) \\ U_{p} (s) \end{matrix}] ∙ [\begin{matrix} m s^{2} + bs + c_{m} & c_{m} d_{33} \\ c_{m} d_{33} & ε_{33}^{T} \frac{A}{h} (1 - k_{33}^{2}) \end{matrix}] = [\begin{matrix} F \\ U_{0} C_{p} \end{matrix}] .

Figure 5. Discrete piezoelectric model composed of mechanical lumped model combined by a transformer with an electrical network.

“F” represents an external force, “c” is the mechanical stiffness, “b” corresponds to damping, “N” is the electromechanical coupling, “C” is the electrical capacitance of the piezoelement and “U” represents the voltage exciting the circuit. “a,b,h” correspond to geometric dimensions of the piezoelectric element, while “x” represents the displacement.

The top left coefficient of the coefficient matrix (29) refers to purely mechanical transformations within the piezoelectric plate. Top right and bottom left coefficients refer to the electromechanical coupling. The bottom right coefficient refers to purely electrical transformations. The matrix can be easily expanded by adding more stages to the model as well as it can be calculated using the matrix method mentioned in (11).

Algorithm testing

A series of computer programs were prepared to test the accuracy and computational load of used algorithms. All the programs were written using mathematical software called MATLAB in version R2019b (RRID:SCR_001622). The code was prepared using basic functions offered by the standard MATLAB suite.

The polynomial equation was created using “for” loops which added consecutive elements to its product form, depending on the number of frequencies given as an input. The MATLAB software operates interchangeably between the product form and exponential form of polynomials, so the conversion was done automatically. The Cauer synthesis was realised with basic program loops which continuously calculated the remainders of the polynomial function, as was explained with equations (4,5,6). Based on the order of elements in the resulting chain fraction, consecutive parameters in form of spring loads and masses were extracted from the equation in each iteration of the loop.

The analysis method using matrix calculations was realised by first creating the equations of motion (10) depending on the number of masses given as an input. The displacement was described using a substitute parameter “X” which then was replaced in bulk with derivatives of the sine cosine integral (11) by using MATLAB function “subs”. Next, the equations were transformed into a matrix form and solved with the inbuilt MATLAB solver. The input vector of external forces was used to manipulate the output amplitudes of the matrix formula.

The structural number method was constructed from ground up, using vector transformation operations inside MATLAB. The structural numbers were represented by matrices consisting of simple numbers corresponding to numbered graph edges. Each number was bound to a set of equations representing forces acting inside the system. The method of assigning numbers to equations was explained in.³² The structural number was formed by adding vectors (each vector consists of numbers corresponding to edges connected to a single graph vertex). The process was done using inbuilt MATLAB function called “combvec”. After that, several sorting operations were done in a loop to eliminate any repeats in the matrix representing the graph structural number. Further steps done to determine all simultaneity functions were also done by appropriately sorting and pruning the structural number matrix to create each structural number operation. To obtain final equations from the graph algebra, an intermediate step was introduced, which replaced each number with the equation that was bound to it. The step was also done using the “subs” function.

The eletromechanical system analysis was an extended matrix algorithm that was based on the equation (28). No new functions were introduced to make the algorithms and every step was made in analogy to the matrix method used to analyse purely mechanical systems.

The code has been saved in a proprietary MATLAB format “.mlx” and placed in repositories of Mendeley Data site.⁴²^,⁴³ The code placed in the repository⁴² contains algorithms used for the synthesis of mechanical systems and their comparison through the matrix and structural number methods. The repository⁴³ contains the algorithm used to synthesise and analyse a full eletromechanical model, simulating the real behaviour of a piezoelectric module. MATLAB is using an open standard for their file formats, so it is possible to open files using standard Windows tools. For reproduction purposes a user has to convert the extension of the file format into “.zip” and open its contents through the “.xml” file stored inside the package. It must be noted however, that the precision and calculation times may vary, depending on the type of software used to reproduce the results and even on the hardware used to process the computation tasks. It is also necessary to note that some of the code may have to be rewritten because of the use of MATLAB specific syntax that may be different in other software. For the sake of consistency and repeatability, all calculations were done on a single computer with a Ryzen 7 2700X processor and a 4 × 8 GB kit of DDR4 CL16 RAM clocked at 3000 Mhz.

Algorithms used to compare the precision and computation time of both matrix and structural number methods were published in an on-line repository.⁴² To compare the precision of both algorithms the chosen input data consisted of a collection of resonant (41 kHz and 55 kHz) and anti-resonant (47 kHz) frequencies. The frequencies were chosen arbitrarily based around piezoelectric product specifications regarding different piezoelements in the shape of cubes and cylinders, found on the market. Both algorithms were analysing a model synthesized on the basis of chosen frequencies which do not relate to any specific real piezoelectric element. The necessary system component parameters were calculated using the synthesis algorithm. The damping ratio was left at 100% and the scaling factor H was set to a value of 0.01. The dynamic forces were set to 1N of harmonic force with a cosine wave applied to both degrees of freedom. The program named “comparison”⁴² was used to compare the precision of both algorithms and produce the resulting graphs and calculate deviations for the extrema. The results obtained with a matrix method were chosen as the reference point for the results obtained with the non-classical method, using structural numbers.

A simple test was also carried out to check how computationally demanding are both methods in comparison to each other. For that purpose, two programs called “Matrix_method” and “Structural_number_method” were written in MATLAB R2019b. Both algorithms were written with a fixed set of model parameters and a fixed set of equations. To avoid long computation times, parameters were chosen for a two-degree system resonating at frequencies of 10 rad/s and 22 rad/s with an antiresonant frequency of 17 rad/s. The computation time was measured for both methods using the inbuilt MATLAB timer function “tic/toc”. The timer measured seconds that the program took to form the equation and calculate the amplitudes in both cases. Each program was run 10 times with the exact same settings and all calculation time measurements were noted down with an average time derived additionally. Before each run, all the variables generated in the previous runs were cleared to avoid any potential changes in computation speed because of reusing the stored results.

To further approximate the computational load of each algorithm, both were expanded to cover a bigger variety of available systems. Necessary automations were introduced to enable program scalability for systems with larger number of degrees of freedom. Another test was conducted with an aim to determine the calculation time for each algorithm depending on the increasing number of degrees of freedom of the analysed model. The time criterion is insufficient to conclusively determine the computational load of algorithms, however. An additional test was made using the inbuilt Matlab profiler module to verify the program RAM usage during calculations and check how much memory is used by each algorithm during execution. The test was done to roughly estimate the “space requirement” of each algorithm along with the “time requirement”, which are used to estimate algorithm complexity.³⁹ Because of increased algorithm size and because of tests being made on models with higher number of degrees of freedom, a decision was made to take five time and memory measurements for each test run. Both algorithms were rewritten and placed in a new online repository.⁴⁴

For the analysis of the electromechanical system, model with a single degree of freedom was chosen. The resonant frequency was set to 590kHz, based on a thickness vibration mode of a piezoelectric plate listed on the market (https://www.steminc.com/PZT/EN/piezoelectric-plate-45x45x35mm-55-khz). Much higher resonant frequency was chosen this time, to test model behaviour with higher frequencies. The scaling factor H was set to 0.0025. The damping factor β was calculated from an existing mechanical quality factor Q_m of 83 using equation (8) which was inversed. The remaining input parameters such as plate dimensions, plate stiffness and capacitance were either taken from an existing piezoelectric material specification or derived using the known physical properties (18-27) (http://www.steminc.com/piezo/PZ_property.asp). The aim of this study was to verify the precision in which a model is capable to simulate the real resonant frequency based on provided material and geometry data. The amplitude response was not a subject of this test, because of insufficient information provided in the piezoelement specification and the inability to conduct real tests at that time. All parameters used in the case of this article are also contained in the program called “analysis_piezoelectric_system” in the on-line repository.⁴³

Results

A comparison of the efficiency and accuracy of the two methods was done with the comparing program. A two-stage mechanical system with two resonant frequencies of 41kHz and 55kHz (257611rad/s and 345575rad/s respectively) was used for the comparison. The graphs in Figure 6 show the displacement amplitudes for both degrees of freedom of the considered case, obtained by matrix calculation and structural numbers.

Figure 6. Model response calculated for a) first vibrating mass, and b) second vibrating mass.

Blue line represents a response calculated with matrix method; red line represents the response calculated with structural number algebra.

Differences were spotted between the results obtained by the two methods. To gain a better understanding, derivatives were taken from the model response equations to calculate the local extremes obtained by both functions. Table 1 gives a direct and relative comparison between the frequencies obtained with both functions. Table 2 gives the calculated values of the displacement amplitude for each characteristic frequency.

Table 1. Comparison of frequencies between model responses obtained with both methods.

	Matrix method	Structural numbers	Indirect difference
1^st stage	257611 rad/s	257611 rad/s	0%
1^st stage	345575 rad/s	345575 rad/s	0%
2^nd stage	257611 rad/s	257611 rad/s	0%
2^nd stage	345575 rad/s	345575 rad/s	0%

Table 2. Comparison of amplitude responses obtained with both methods on the nodes.

	Matrix method	Structural numbers	Direct difference	Indirect difference
1^st stage	0.0101 mm	0.0101 mm	0 mm	0%
1^st stage	0.0017 mm	0.0027 mm	0.001 mm	58.82%
2^nd stage	0.0033 mm	0.0033 mm	0 mm	0%
2^nd stage	0.0004 mm	0.0006 mm	0.0002 mm	50%

The nodal frequencies calculated with the structural method match perfectly with the reference matrix method. There were significant differences in the obtained amplitudes, especially around the second resonant frequency of the system in both stages. The alternative method using structural numbers proved sufficient to check the nodal frequencies. However, when it comes to accuracy in terms of amplitude response, this method was found to be subject to considerable error. Calculations were performed for models with more than two degrees of freedom and the discrepancy between amplitudes of each stage determined with both methods was found to be random. It was expected that the error may result from a mistake performed during the formulation of the structural number algorithm.

Another problem with the structural number method was that the authors were unable to establish the equations necessary to calculate phase shifts in the system’s response.

Table 3 shows the computation time and an average time from all measurements done on both algorithms based on matrix and structural number methods.⁴² The test showed that matrix method took an average of 13.972 s to calculate the amplitude response of a two-stage system. The alternative method took only 8.637 s on average, which is 38% faster than the classical method.

Table 3. Computation time for both methods used to analyse a simple mechanical system.

Trial	Matrix method	Structural numbers
1	12.815 s	7.531 s
2	12.824 s	7.649 s
3	12.761 s	7.577 s
4	12.759 s	7.546 s
5	13.099 s	7.563 s
6	12.905 s	7.578 s
7	12.839 s	7.626 s
8	13.162 s	7.574 s
9	13.080 s	7.570 s
10	13.088 s	7.627 s
Average	12.933 s	7.584 s

The first study was done on algorithms in which the full parameterisation of mathematical relationships has not yet been applied. Both algorithms were rewritten and extended to increase versatility of the program. A second study was carried out, in which the computation time and required memory were measured for algorithms used to calculate systems with different number of degrees of freedom. Each test was run five times, and an average result was taken. Systems with up to six degrees of freedom were tested with each method. Results of the study were placed in Table 4.

Table 4. Computation time and memory usage for parametrized algorithms.

Deg. of Fr. trials	Matrix method		Structural number method
Deg. of Fr. trials	Time [s]	Memory [kb]	Time [s]	Memory [kb]
2.1	2.549	48	3.109	16
2.2	2.517	32	3.054	32
2.3	2.539	32	3.071	32
2.4	2.567	32	3.06	32
2.5	2.553	32	3.072	32
2.AVG	2.545	35.2	3.0732	28.8
3.1	6.252	1024	9.617	20
3.2	6.209	1024	9.527	28
3.3	6.185	28	9.412	28
3.4	6.158	28	9.429	28
3.5	6.196	68	9.534	28
3.AVG	6.2	434.4	9.5038	26.4
4.1	13.434	512	34.863	372
4.2	13.21	7844	34.793	7868
4.3	13.22	28	34.79	7872
4.4	13.144	1024	34.732	7900
4.5	13.083	28	34.617	7860
4.AVG	13.2182	1887.2	34.759	6374.4
5.1	25.737	1760	165.642	31996
5.2	25.262	384	171.528	7896
5.3	25.138	880	171.156	32480
5.4	25.283	1324	165.678	31136
5.5	25.149	10792	164.773	26908
5.AVG	25.3138	3028	167.7554	26083.2
6.1	49.889	3428	1369.374	782492
6.2	46.77	3240	1318.753	667988
6.3	47.255	5172	1237.701	804440
6.4	47.379	7840	1265.202	750964
6.5	47.792	4056	1280.015	733920
6.AVG	47.817	4747.2	1294.209	747960.8

The results of the second study indicated a significant increase in the time and required memory for mathematical operations performed by the structural number method relative to the classical matrix method. The difference in calculation time and memory usage of the modified algorithms increased significantly with the number of degrees of freedom of the analyzed systems. The data on memory usage was collected using an inbuilt profiler module inside Matlab which has been shown in Figure 7.

Figure 7. An inbuilt Matlab profiler module used to collect the data on the algorithm memory usage.

An error made in the first iteration of the program was also largely removed during modification of the algorithms and the results obtained with both methods in the second study were matching to the 0.001 μm. An example of results obtained for a system with three degrees of freedom was shown in Figure 8 and placed in Table 5. The comparisons between two algorithms for other studied models are also available in the data repository.⁴⁴

Figure 8. A graph showing model response obtained with analysis done using both methods.

Green line represents model response calculated with the matrix algorithm. Red dots show datapoints calculated with structural numbers. The obtained results show a perfect match.

Table 5. Results obtained for both methods in the resonance frequency data points.

Frequency [Hz]	Amplitude [μm]
Frequency [Hz]	Matrix method	Structural method
45000	22.0713	22.0717
75000	6.9048	6.9048
93000	1.6714	1.6714

The discrete electromechanical model⁴³ was verified on a system simulating the behaviour of a single piezoelectric plate with specifications very close to the real ones taken from (http://www.steminc.com/piezo/PZ_property.asp). The plate had to vibrate at a resonance frequency of 590kHz (3707079rad/s). Mechanical parameters of the plate were determined with the Cauer synthesis method. Piezoelectric plate dimensions were taken from (https://www.steminc.com/PZT/EN/piezoelectric-plate-45x45x35mm-55-khz). Resulting graph of an amplitude response for piezoelectric plate displacement and voltage were shown in Figure 9.

Figure 9. Model response calculated for a) displacement curve, and b) voltage curve along the frequency axis.

The model had a resonance near the frequency of 560 kHz (3516624 rad/s) which gives an error of 5% in relation to the real plate specification. The calculated magnitude of displacement at the resonant frequency is 2.6 mm which is well above the capability of a 3.5 mm thick piezoelectric plate. The voltage shows a significant increase in piezoelectric voltage in the resonant region and a complete reduction near the frequency of 590 kHz which was supposed to be the original resonant frequency.

Conclusions

The aim of this work was to demonstrate a method that allows a piezoelectric model to be synthesized by using only a set of resonant/anti-resonant frequencies as the input data. By combining Cauer’s method with Rayleigh’s damping approximation and Van Dyke-Butterworth model, an electromechanical model can be created to simulate the behaviour of a real piezoelectric component. The purpose of this method is to streamline the process of prototyping advanced piezoelectric systems capable of generating and sensing vibrations in various resonant frequencies by establishing a rough set of approximate material properties or geometrical dimensions a given system needs based on a limited information. Based on the results presented in this article, several conclusions can be made about the current state of this concept.

• The synthesized electromechanical model has shown a small deviation from the target resonant frequency. An initial hypothesis is that this error could be corrected by applying a correction factor to the input frequency.
• The calculated amplitude response for the resonant frequency was several orders of magnitude greater than the expected value of displacement for a plate of given thickness. It is expected that the error originates from the differences between piezoelectric material properties at its resonance and material specification which is measured for a frequency of 1 kHz.
• Currently, it is only possible to synthesize models composed of elements placed in series. This imposes a heavy restriction on modelling capabilities. The method could be extended to include elements connected in parallel.
• Authors in Refs. 40, 41 mention significant differences in piezoelectric material properties measured at their resonance in comparison to their specification which usually contains properties measured at much lower frequencies. This discrepancy has to be verified with more testing.
• The reference plate specification (more information here) didn’t contain enough data to reliably compare the results obtained from the model analysis. As a result, only a rough comparison of material properties could be conducted.

In light of those conclusions, it is clear that presented mathematical model needs further work and empirical validation before it can be considered widely viable.

The method presented in this article relies heavily on mechanical model synthesis in its initial stage. In order to proceed with confidence, it was necessary to verify the mechanical model before using it in later stages of the work. However, for more complex systems the classical matrix calculation creates a significant computational load, which extends the calculation time. An additional goal was to develop an alternative method for analysis using graphs and structural numbers and compare its precision and computing time. Conclusions were made based on results obtained from two rounds of tests.

• The developed analysis method gave promising initial results in terms of computation time. The accuracy of calculated characteristic frequencies between the matrix method and the proposed alternative was on par in terms of calculated peaks near resonant frequencies.
• Based on the results of the second study the structural number method was found much less efficient in terms of both computation time and memory required. The results were progressively worse with the increasing number of calculated degrees of freedom.

The observed discrepancy may be attributed to the structure of the code written using Matlab software. Purely algebraic operations were used to determine the structural numbers and mathematical formulas. To determine specific numerical values, the algorithm sequentially substituted numerical values of input variables in a closed loop. The use of extensive closed loops in computer software can lead to intensive RAM usage. The formulas determined for systems with several degrees of freedom are also extensive and have coefficients with high powers, which in turn can lead to intensive CPU resource usage. Due to the fact that the code was prepared using a dedicated language that is part of the Matlab software, it is unclear whether the observed differences result from poor optimization of the prepared code or whether the code on which the matrix method is based is better optimized, as it is an integral part of the software and exists in the form of a ready-made module.

The structural number method definitely needs further fine-tuning and refactoring to improve its performance, but it was shown that it can be an alternative to matrix calculations. Checking the performance of the algorithm using different calculation software may also be a good way forward, as the underlying problems may as well result from some program limitations.

Data availability

Source data

Reference data for the specific piezoelectric module has been sourced from a piezoelectric materials vendor: https://www.steminc.com/PZT/EN/piezoelectric-plate-45x45x35mm-55-khz.

A list of detailed material properties used for algorithms and the electromechanical model analysis was taken from: http://www.steminc.com/piezo/PZ_property.asp.

Underlying data

Mendeley Data: Model algorithms. https://doi.org/10.17632/gdt8yr9m3d.4 ⁴²

This project contains the following data:

- Comp_time_test.xlsx
- Comparison_value_set.mat

Mendeley Data: Piezoelectric electromechanical model (first iteration), https://doi.org/10.17632/pydnfmkfdk.3 ⁴³

This project contains the following underlying data:

- electromech_model.mat

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).

Software availability

Model algorithms

Source code available from:

Archived source code available from: https://doi.org/10.17632/gdt8yr9m3d.4 ⁴²

License: CC0 1.0

Piezoelectric electromechanical model

Source code available from:

Archived source code available from: https://doi.org/10.17632/pydnfmkfdk.3 ⁴³

License: CC0 1.0

Improved analysis algorithms for structural and matrix methods

Source code available from:

Archive source code available from: https://doi.org/10.17632/xk443p7m7d.1 ⁴⁴

License: CC0 1.0

References

1. Sangryun L, Jiyoung J, Seunghwa R: Micromechanics-based prediction of the effective properties of piezoelectric composite having interfacial imperfections. Compos. Struct. 2020; 240:112076. Publisher Full Text
2. Nada T, Lahcen A, Nadia F, et al.: Nonlinear micromechanical modeling of fully coupled piezo-elastic composite under large deformation and high electric field. Compos. Struct. 2023; 315:116991. Publisher Full Text
3. Fateh EM, Amar M, Fahmi Z, et al.: A Coupled Electro-Mechanical Homogenization-Based Model for PVDF-Based Piezo-Composites Considering α → β Phase Transition and Interfacial Damage. Polymers. 2023; 15:2994. PubMed Abstract | Publisher Full Text | Free Full Text
4. Jeffrey H, Yuqi S: Piezoelectric modulus prediction using machine learning and graph neural networks. Chem. Phys. Lett. 2022; 791:139359. Publisher Full Text
5. Nguyen B, Zhuang X, Rabczuk T: Numerical model for the characterization of Maxwell-Wagner relaxation in piezoelectric and flexoelectric composite material. Comput. Struct. 2018; 208:75–91. Publisher Full Text
6. Larbi W: Numerical modeling of sound and vibration reduction using viscoelastic materials and shunted piezoelectric patches. Comput. Struct. 2020; 232: 105822. Publisher Full Text
7. Adriaens H, Koning W, Banning R: Modelling piezoelectric actuators. IEEE ASME Trans. Mechatron. 2000; 5: 331–341. Publisher Full Text
8. Goldfarb M, Celanovic N: Modelling piezoelectric stack actuators for control of micromanipulation. IEEE Control. Syst. 1997; 17: 69–79. Publisher Full Text
9. Richter B, Twiefel J, Wallaschek J: Piezoelectric Equivalent Circuit Models.Priya S, Inman DJ, editors. Energy Harvesting Technologies. Boston, MA: Springer; 2009. Publisher Full Text
10. Lahmer T, Kaltenbacher M, Kaltenbacher B, et al.: FEM-based determination of real and complex elastic, dielectric, and piezoelectric moduli in piezoceramic materials. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 2008; 55(2): 465–475. PubMed Abstract | Publisher Full Text
11. Goldstein A: Self-Tuning multimodal piezoelectric shunt damping. J. Braz. Soc. Mech. Sci. Eng. 2011; 33: 428–436. Publisher Full Text
12. Batra R, Dell’isola F, Vidoli S, et al.: Multimode vibration suppression with passive two-terminal distributed network incorporating piezoceramic transducers. Int. J. Solids Struct. 2005; 42: 3115–3132. Publisher Full Text
13. San Miguel J, Buiochi F, Adamowski J: Numerical modelling of a circular piezoelectric ultrasonic transducer radiating in water. ABCM Symposium Series in Mechatronics. 2000; 2: 458–464.
14. Yamaoka S, Ohnishi K: Reaction force estimation of piezoelectric actuator by charge observation. IEEE International Conference on Mechatronics (ICM). Vicenza, Italy IEEE Xplore. 2013. Publisher Full Text
15. Delibas B, Arockiarajan A, Seemann W: Rate dependent properties of perovskite type tetragonal piezoelectric materials using micromechanical model. Int. J. Solids Struct. 2006; 43: 697–712. Publisher Full Text
16. Gan J, Zhang X: A review of nonlinear hysteresis modelling and control of piezoelectric actuators. AIP Adv. 2019; 9(4): 040702. Publisher Full Text
17. Zhe L, Xing-Rong H, Le F: Numerical modeling of fluid–structure–piezoelectric interaction for energy harvesting. Comput. Methods Appl. Mech. Eng. 2023; 414: 116164. Publisher Full Text
18. IEEE: Standard on Piezoelectricity. ANSI/IEEE; 1988.
19. Soluch W, et al.: Wstęp do piezoelektryki. Warszawa: Wydawnictwo Komunikacji i Łączności; 1980.
20. Wang S-H, Tsai M-C: Dynamic modelling of thickness-mode piezoelectric transducer using the block diagram approach. Ultrasonics. 2011; 51: 617–624. PubMed Abstract | Publisher Full Text
21. Wróbel A: Classical and non-classical research methods and experimental studies of piezoelectric plates. Gliwice: Wydawnictwo Politechniki Śląskiej; 2018.
22. Cauer W: Ideale Transformatoren und lineare Transformationen. Elektrische Nachrichtentechnik (ENT). 1932; 9: 157–174.
23. Cauer W: Theorie der linearen Wechselstromschaltungen. Berlin: Akademie-Verlag; 1960; vol. 2. .
24. Buchacz A, Dymarek A, Dzitkowski T: Synthesis of discrete, continuous and discrete continuous vibrating systems represent-ed by graphs. Sixth International Scientific and Engineering Conference - Machine-Building and Technosphere on the Border of the XXI Century. 1999; 3: 243–245.
25. Buchacz A, Żurek K: Modelowanie i badanie wpływu zjawiska piezoelektrycznego na charakterystyki układu mechanicznego. Wydawnictwo Politechniki Śląskiej. 2010.
26. Smith M: Synthesis of mechanical networks: the inerter. IEEE Trans. Automat. Contr. 2002; 47: 1648–1662. Publisher Full Text
27. Białas K, Buchacz A, Dzitkowski T: Synteza drgających aktywnych układów mechanicznych z tłumieniem w ujęciu grafów biegunowych i liczb strukturalnych. Wydawnictwo Politechniki Śląskiej. 2010.
28. Nakamura N: Extended Rayleigh Damping Model. Front. Built Environ. 2016; 2. Publisher Full Text
29. Bellert S, Woźniacki H: Analiza I synteza układów elektrycznych metodą liczb strukturalnych. Warszawa: WNT; 1968.
30. Pikoń A, Wojnarowski J: Application of structural numbers to generating the characteristics of mechanical systems. Mech. Mach. Theory. 1995; 30:1027–1037. Publisher Full Text
31. Bellert S: Topological analysis and synthesis of linear systems. J. Frankl. Inst. 1962; 274:425–443. Publisher Full Text
32. Harazin J, Wróbel A: Research on a cascade model synthesis with the use of classical and non-classical methods in the context of new piezoelectric stack applications. Int. J. Mech. Sci. 2021; 12: 959–969. Publisher Full Text
33. Białas K, Buchacz A, Gałęziowski D: Application of Electric Subsystem as Implementation of Reduction of Mechanical Vibrations. Solid State Phenom. 2015; 220-221:9–14. Publisher Full Text
34. Białas K, Buchacz A: The Influence of Changing the Parameters of Electrical Components Implementing the Active Reduction of Vibration. Appl. Mech. Mater. 2014; 657:614–618. Publisher Full Text
35. Adamowski JC, Nader G, Silva ECN: Effective Damping Value of Piezoelectric Transducer Determined by Experimental Techniques and Numerical Analysis. ABCM Symposium Series in Mechatronics. 2004; 1: 271–279.
36. Giergiel J: Tłumienie drgań mechanicznych. Kraków: Wydawnictwo Akademii Górniczo-Hutniczej; 1984.
37. Trindade M, Benjeddou A: Effective Electromechanical Coupling Coefficients of Piezoelectric Adaptive Structures: Critical Evaluation and Optimization. Mech. Adv. Mater. Struct. 2009; 16: 210–223. Publisher Full Text
38. Buchacz A, Płaczek M: Modelowanie i badanie jednowymiarowych drgających układów mechatronicznych. Gliwice: Wydawnictwo Politechniki Śląskiej; 2011.
39. Arora S, Barak B: Computational complexity: A modern approach. Cambridge University Press; 2007.
40. Mirzoyan V, Sahakyan A, Mirzoyan P: Behavior of dielectric permittivity of piezomonocrystall near the resonant frequency. Phys. Math. Sci. 2014; 3:62–65.
41. Park Y, Zhang Y, Majzoubi M, et al.: Improvement of the standard characterization method on k₃₃ mode piezoelectric specimens. Sens. Actuators A: Phys. 2020; 312:112124–112124. Publisher Full Text
42. Harazin J: Model algorithms. Mendeley Data. 2022. Publisher Full Text
43. Harazin J: Piezoelectric electromechanical model (first iteration). Mendeley Data. 2023. Publisher Full Text
44. Harazin J: Improved analysis algorithms for structural and matrix methods. Mendeley Data. 2024. Publisher Full Text

Comments on this article Comments (0)

Version 3

VERSION 3 PUBLISHED 05 Oct 2023

Author details Author details

¹ Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, Silesian Voivodeship, 44-100, Poland

Jacek Harazin
Roles: Data Curation, Formal Analysis, Funding Acquisition, Investigation, Methodology, Software, Visualization, Writing – Original Draft Preparation

Andrzej Wróbel
Roles: Conceptualization, Project Administration, Resources, Supervision, Validation, Writing – Review & Editing

Competing interests

No competing interests were disclosed.

Grant information

The article was financed from the subvention for maintenance and development of research potential of young scientists of the Silesian University of Technology.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Article Versions (3)

version 3

Revised

Published: 06 Sep 2024, 12:1273

https://doi.org/10.12688/f1000research.140943.3

version 2

Revised

Published: 02 Apr 2024, 12:1273

https://doi.org/10.12688/f1000research.140943.2

version 1

Published: 05 Oct 2023, 12:1273

https://doi.org/10.12688/f1000research.140943.1

© 2024 Harazin J and Wróbel A. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Download

Export To

metrics

	Views	Downloads
F1000Research	-	-
PubMed Central Data from PMC are received and updated monthly.	-	-

Citations

SEE MORE DETAILS

CITE

how to cite this article

Harazin J and Wróbel A. Analytical approach to piezoelectric model synthesis with the use of Cauer’s method for system design [version 3; peer review: 2 approved, 2 approved with reservations]. F1000Research 2024, 12:1273 (https://doi.org/10.12688/f1000research.140943.3)

NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.

track

receive updates on this article

Track an article to receive email alerts on any updates to this article.

Open Peer Review

Current Reviewer Status: ?

Key to Reviewer Statuses VIEW HIDE

ApprovedThe paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approvedFundamental flaws in the paper seriously undermine the findings and conclusions

Version 3

VERSION 3

PUBLISHED 06 Sep 2024

Revised

Views

Reviewer Report 09 Sep 2024

Pedro M. Ferreira, Universidade Nova de Lisboa, Lisbon, Lisbon, Portugal

Approved

https://doi.org/10.5256/f1000research.170982.r321408

All the comments have been taken into account, ... Continue reading

CITE

Report a concern

Respond or Comment

Views

Reviewer Report 09 Sep 2024

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

Approved

https://doi.org/10.5256/f1000research.170982.r321409

The manuscript has been greatly improved, and the authors have responsed to all the main concerns. As the authors replied to comments, they are still trying their best to further investigate and solve some main concerns the two reviewers point ... Continue reading

CITE

Report a concern

Respond or Comment

Version 2

VERSION 2

PUBLISHED 02 Apr 2024

Revised

Views

Reviewer Report 13 Sep 2024

Zine Ghemari, University Mohamed Boudiaf,, M’sila,, Algeria

Approved with Reservations

https://doi.org/10.5256/f1000research.164274.r292326

The synthesis method described in this paper enables the creation of piezoelectric cascade models using limited information in the form of characteristic frequencies. At present, this method provides a rough approximation of the actual piezoelectric parameters with a limited number of inputs. An alternative analysis method based on structural numbers shows promise compared to matrix calculations, but it needs further investigation to assess the computational power required to determine its competitiveness with current algorithms.
The authors should revise this article according to the following comments:
1- do you show the economic profit of this work?
2- Add recent work on piezoelectricity such as:
- Improvement of the piezoelectric sensor by the progress of the measurement accuracy." IEEE International Conference on Smart Materials and Spectroscopy (SMS’2018), Yassmine Hammamet (Tunisia).
- Improvement of the vibration analysis technique by optimizing the parameters of the piezoelectric accelerometer
- Extraction of the relative sensitivity model and improvement of the piezoelectric accelerometer performances
- Mechanical resonator sensor characteristics development for precise vibratory analysis
3-Correct any spelling mistakes

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Maintenance, Piezoelectricity, Sensors

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

CITE

Report a concern

Respond or Comment

Views

Reviewer Report 28 Aug 2024

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

Approved with Reservations

https://doi.org/10.5256/f1000research.164274.r261934

Thank the authors for their response and detailed revisions. The current version has been greatly improved, and however there are still some concerns that need be addressed. Moreover, impressed by the innovative effort.

If the name

If the name of the proposed method is included in the title, the novelty of this manuscript will be more easily got by the readers.
Figure 3: what do these symbols, ₁X_0s, ₁X_1s and so on, stand for?
The material type and some key parameters of PZT should be clearly addressed instead of just putting a website link.
The resolution of the figures is low and their quality should be further improved.
Table 2 and 5: how to confirm the accuracy of the data?
Based on the results, the proposed method can be used to analyze the characteristics of piezoelectric systems accurately. However, the computation cost is much larger than the traditional method. Therefore, what is the merit or advantage of this method?

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Structural health monitoring by uisng lamb waves

CITE

Report a concern

Author Response 10 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

10 Sep 2024

Author Response
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will ... Continue reading
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will have to be placed in another version of the article if necessary. I will try to address the issues point by point, as previously.

The aim of this article’s title was to focus the attention of the reader on the aspect of piezoelectric model synthesis, with the main “tool” being the Cauer’s method. The presented analysis method using structural numbers was placed within the article as it was necessary to provide means to validate the synthesis method. The research on the alternative using structural numbers was done in parallel and there were plans initially, to also use structural numbers for the synthesis itself, hence this method ended up being part of the article. I realise the outcome of that idea is not ideal and the title may be somewhat ambiguous but at the same time the title is not too concise already and I’m afraid I would only make it worse by trying to fix this issue.

₁X₀and ₁X₁are nodes of the graph of relations. Each node represents the displacement of any object being part of a physical model. So, if we had an example of a model with two moving masses, a graph of such model would contain three nodes. The ₁X₀node is special in that it represents any relations of a model with anything we consider being “outside” of the model’s scope. In the case of this article, the ₁X₀node represents one of the fixed ends of the model, where the last spring and damper are attached. The first number symbolizes an entire isolated system with its own relations. So, if we had two models working in parallel, the first one would have “1” at the front, and the second one would have “2”. The second number is used to number nodes within each system.

I agree that putting just the link to the used material catalogue might have introduced unnecessary ambiguity into the article. I did a mistake of not properly disclosing that I chose the SM411 material to test the model response which would have made it slightly better. If You consider this a major issue with the article, then I will try to prepare another version with clearly disclosed information about the used material.

I will try to improve the quality of figures in the next version if necessary.

The entire code used to perform the study and obtain the data in tables 2 and 5 was placed in an online repository (Mendeley data) under the reference 42. The algorithm is written in such a way that it should automatically collect all the data and print the amplitude responses with both studied methods. The only caveat is that the algorithm was written using the proprietary Matlab code format, so it is recommended that the algorithm is used with Matlab r2019b to ensure the same results.

The method of system analysis presented in the article has a markedly different approach to the way in which the relationships occurring within the analysed system are represented.
The response equations of the system are obtained based on transformations on structural numbers, which are in principle less complex than matrix equations. This led the authors to believe that the algorithmisation of this method would allow the response to be calculated with less computing power than in the case of the matrix method. Unfortunately, this could not be proved in the article. However, it is not clear whether the inferior performance of the method obtained from calculations performed in Matlab is directly due to the shortcomings of the method itself, or whether it is the result of poor optimisation of the code itself, or even of the environment, which is optimised for the well-known matrix method. Therefore authors assumed that it is still worth sharing the results and the method may have potential, however more work is necessary to determine the cause of worse performance.

Thank you again for your time spent on preparing your review. I hope my answers have cleared at least some of the issues raised. I will be eagerly waiting for any further responses.

Best regards,
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will have to be placed in another version of the article if necessary. I will try to address the issues point by point, as previously.

The aim of this article’s title was to focus the attention of the reader on the aspect of piezoelectric model synthesis, with the main “tool” being the Cauer’s method. The presented analysis method using structural numbers was placed within the article as it was necessary to provide means to validate the synthesis method. The research on the alternative using structural numbers was done in parallel and there were plans initially, to also use structural numbers for the synthesis itself, hence this method ended up being part of the article. I realise the outcome of that idea is not ideal and the title may be somewhat ambiguous but at the same time the title is not too concise already and I’m afraid I would only make it worse by trying to fix this issue.

₁X₀and ₁X₁are nodes of the graph of relations. Each node represents the displacement of any object being part of a physical model. So, if we had an example of a model with two moving masses, a graph of such model would contain three nodes. The ₁X₀node is special in that it represents any relations of a model with anything we consider being “outside” of the model’s scope. In the case of this article, the ₁X₀node represents one of the fixed ends of the model, where the last spring and damper are attached. The first number symbolizes an entire isolated system with its own relations. So, if we had two models working in parallel, the first one would have “1” at the front, and the second one would have “2”. The second number is used to number nodes within each system.

I agree that putting just the link to the used material catalogue might have introduced unnecessary ambiguity into the article. I did a mistake of not properly disclosing that I chose the SM411 material to test the model response which would have made it slightly better. If You consider this a major issue with the article, then I will try to prepare another version with clearly disclosed information about the used material.

I will try to improve the quality of figures in the next version if necessary.

The entire code used to perform the study and obtain the data in tables 2 and 5 was placed in an online repository (Mendeley data) under the reference 42. The algorithm is written in such a way that it should automatically collect all the data and print the amplitude responses with both studied methods. The only caveat is that the algorithm was written using the proprietary Matlab code format, so it is recommended that the algorithm is used with Matlab r2019b to ensure the same results.

The method of system analysis presented in the article has a markedly different approach to the way in which the relationships occurring within the analysed system are represented.
The response equations of the system are obtained based on transformations on structural numbers, which are in principle less complex than matrix equations. This led the authors to believe that the algorithmisation of this method would allow the response to be calculated with less computing power than in the case of the matrix method. Unfortunately, this could not be proved in the article. However, it is not clear whether the inferior performance of the method obtained from calculations performed in Matlab is directly due to the shortcomings of the method itself, or whether it is the result of poor optimisation of the code itself, or even of the environment, which is optimised for the well-known matrix method. Therefore authors assumed that it is still worth sharing the results and the method may have potential, however more work is necessary to determine the cause of worse performance.

Thank you again for your time spent on preparing your review. I hope my answers have cleared at least some of the issues raised. I will be eagerly waiting for any further responses.

Best regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 10 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

10 Sep 2024

Author Response
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will ... Continue reading
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will have to be placed in another version of the article if necessary. I will try to address the issues point by point, as previously.

The aim of this article’s title was to focus the attention of the reader on the aspect of piezoelectric model synthesis, with the main “tool” being the Cauer’s method. The presented analysis method using structural numbers was placed within the article as it was necessary to provide means to validate the synthesis method. The research on the alternative using structural numbers was done in parallel and there were plans initially, to also use structural numbers for the synthesis itself, hence this method ended up being part of the article. I realise the outcome of that idea is not ideal and the title may be somewhat ambiguous but at the same time the title is not too concise already and I’m afraid I would only make it worse by trying to fix this issue.

₁X₀and ₁X₁are nodes of the graph of relations. Each node represents the displacement of any object being part of a physical model. So, if we had an example of a model with two moving masses, a graph of such model would contain three nodes. The ₁X₀node is special in that it represents any relations of a model with anything we consider being “outside” of the model’s scope. In the case of this article, the ₁X₀node represents one of the fixed ends of the model, where the last spring and damper are attached. The first number symbolizes an entire isolated system with its own relations. So, if we had two models working in parallel, the first one would have “1” at the front, and the second one would have “2”. The second number is used to number nodes within each system.

I agree that putting just the link to the used material catalogue might have introduced unnecessary ambiguity into the article. I did a mistake of not properly disclosing that I chose the SM411 material to test the model response which would have made it slightly better. If You consider this a major issue with the article, then I will try to prepare another version with clearly disclosed information about the used material.

I will try to improve the quality of figures in the next version if necessary.

The entire code used to perform the study and obtain the data in tables 2 and 5 was placed in an online repository (Mendeley data) under the reference 42. The algorithm is written in such a way that it should automatically collect all the data and print the amplitude responses with both studied methods. The only caveat is that the algorithm was written using the proprietary Matlab code format, so it is recommended that the algorithm is used with Matlab r2019b to ensure the same results.

The method of system analysis presented in the article has a markedly different approach to the way in which the relationships occurring within the analysed system are represented.
The response equations of the system are obtained based on transformations on structural numbers, which are in principle less complex than matrix equations. This led the authors to believe that the algorithmisation of this method would allow the response to be calculated with less computing power than in the case of the matrix method. Unfortunately, this could not be proved in the article. However, it is not clear whether the inferior performance of the method obtained from calculations performed in Matlab is directly due to the shortcomings of the method itself, or whether it is the result of poor optimisation of the code itself, or even of the environment, which is optimised for the well-known matrix method. Therefore authors assumed that it is still worth sharing the results and the method may have potential, however more work is necessary to determine the cause of worse performance.

Thank you again for your time spent on preparing your review. I hope my answers have cleared at least some of the issues raised. I will be eagerly waiting for any further responses.

Best regards,
Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will have to be placed in another version of the article if necessary. I will try to address the issues point by point, as previously.

The aim of this article’s title was to focus the attention of the reader on the aspect of piezoelectric model synthesis, with the main “tool” being the Cauer’s method. The presented analysis method using structural numbers was placed within the article as it was necessary to provide means to validate the synthesis method. The research on the alternative using structural numbers was done in parallel and there were plans initially, to also use structural numbers for the synthesis itself, hence this method ended up being part of the article. I realise the outcome of that idea is not ideal and the title may be somewhat ambiguous but at the same time the title is not too concise already and I’m afraid I would only make it worse by trying to fix this issue.

₁X₀and ₁X₁are nodes of the graph of relations. Each node represents the displacement of any object being part of a physical model. So, if we had an example of a model with two moving masses, a graph of such model would contain three nodes. The ₁X₀node is special in that it represents any relations of a model with anything we consider being “outside” of the model’s scope. In the case of this article, the ₁X₀node represents one of the fixed ends of the model, where the last spring and damper are attached. The first number symbolizes an entire isolated system with its own relations. So, if we had two models working in parallel, the first one would have “1” at the front, and the second one would have “2”. The second number is used to number nodes within each system.

I agree that putting just the link to the used material catalogue might have introduced unnecessary ambiguity into the article. I did a mistake of not properly disclosing that I chose the SM411 material to test the model response which would have made it slightly better. If You consider this a major issue with the article, then I will try to prepare another version with clearly disclosed information about the used material.

I will try to improve the quality of figures in the next version if necessary.

The entire code used to perform the study and obtain the data in tables 2 and 5 was placed in an online repository (Mendeley data) under the reference 42. The algorithm is written in such a way that it should automatically collect all the data and print the amplitude responses with both studied methods. The only caveat is that the algorithm was written using the proprietary Matlab code format, so it is recommended that the algorithm is used with Matlab r2019b to ensure the same results.

The method of system analysis presented in the article has a markedly different approach to the way in which the relationships occurring within the analysed system are represented.
The response equations of the system are obtained based on transformations on structural numbers, which are in principle less complex than matrix equations. This led the authors to believe that the algorithmisation of this method would allow the response to be calculated with less computing power than in the case of the matrix method. Unfortunately, this could not be proved in the article. However, it is not clear whether the inferior performance of the method obtained from calculations performed in Matlab is directly due to the shortcomings of the method itself, or whether it is the result of poor optimisation of the code itself, or even of the environment, which is optimised for the well-known matrix method. Therefore authors assumed that it is still worth sharing the results and the method may have potential, however more work is necessary to determine the cause of worse performance.

Thank you again for your time spent on preparing your review. I hope my answers have cleared at least some of the issues raised. I will be eagerly waiting for any further responses.

Best regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern

Views

Reviewer Report 09 Aug 2024

Pedro M. Ferreira, Universidade Nova de Lisboa, Lisbon, Lisbon, Portugal

Approved with Reservations

https://doi.org/10.5256/f1000research.164274.r307986

The manuscript presents a novel approach to synthesizing piezoelectric systems, starting with only known resonant and anti-resonant frequencies. This method offers a broader perspective on system modelling utilizing an algorithm based on the Cauer synthesis method, an initial discrete mechanical model is constructed, forming the foundation for piezoelectric model synthesis. The manuscript is well-composed and systematically organized. However, there are a few concerns and suggestions that should be addressed before it can be accepted for indexing. In my view, these issues require more detailed clarification before the manuscript can be considered for indexing.

The introduction to the present literature review is poor, particularly in the following excerpt: "Researchers largely focus on modelling simple piezoelectric components with lumped models and matrix tensor equations or try to approach the problem using FEM (finite element method) analysis." There are several models that could be referenced in the introduction, such as Eshelby-Type Models, Micromechanics-based models, or Asymptotic Homogenization-Based Models, which are widely used to model the behaviour of piezoelectric composites. Additionally, this excerpt needs to be supported with more current references. Keywords that could help in finding up-to-date references include: "Piezoelectric composites, Electro-mechanical properties, Analytical models, Numerical models."
In the last paragraph of the introduction, the excerpt “The number of rows and columns, however, is tightly connected to the shape of an edge graph which a given structural number relates to,” is very unclear. It should be rewritten for better clarity and explained in more detail.
The author compares the computation time and memory usage of parameterized algorithms, highlighting a significant increase in both when using the structural number method compared to the classical matrix method (as shown in Table 4). The authors should clarify why the calculation time and memory usage of the modified algorithms increase so dramatically as the number of degrees of freedom in the analysed systems rises.
The conclusion is overly long and needs to be more concise, with an emphasis on the limitations of the authors' analytical approach. For better clarity, consider presenting the main conclusions and limitations as bullet points.

Is the work clearly and accurately presented and does it cite the current literature?

Partly
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Piezoelectric Composites, Piezoelectric Behaviour, Advanced Materials, Smart Materials

CITE

Report a concern

Author Response 06 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

06 Sep 2024

Author Response

Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links ... Continue reading Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links to the numerical methods mentioned in the review. I also rewrote the research gap statement to make it more readable. The conclusion section has also been revised. Below I have included corresponding point by point responses to the various comments made in the review.

1. After rereading the introduction section, I must agree that it was too vague and contained too much information that could be considered irrelevant in the scope of the article. To amend that, I have completely rewritten most of this section, while also adding mentions and new references to current works (2020-2023) related with Eshelby-Type/micromechanical models. I must admit that I was so focused on the mathematical synthesis aspect in the context of piezoelectric models that I didn’t encounter works on models mentioned in your review. Thank You for bringing this issue up because that also helped me in finding new potential paths to expand my research in the future.

2. I have decided to extend the explanation with a graphical clue in form of a new figure 1, which shows an example of a graph, from which a new structural number is being created. The example of a structural number has also been placed under new equation 1. I think that a simple figure might provide much more insight than my extensive explanations, while also conserving more space. In hindsight, I should have provided a descriptive figure in the first place.

3. I agree that this issue was left with a vague explanation. It was already pointed out that the conclusions were overly long and realising that I was trying to make the explanation concise. Unfortunately, I made it too vague instead. I have tried to amend this by adding more detailed explanation in the conclusions. To summarize the explanation, the uncovered issue may be due to the structure of the written code for the method based on structural number algebra. It was made using closed loops which were combined with quite complicated substitution operations and complex algebraic representations of each graph edge. Combined, these three elements could contribute to the perceived great increase in both computation time and RAM requirement. The matrix method was largely an unmodified module readily available as an integral part of the Matlab software which may have been the reason for its much better performance. The new method had to be written from ground up which probably left a lot of the performance on the table due to my lack of programming skill and also the inability to better access program’s memory and CPU handling operations.

4. Most of the conclusions section was reworked according to raised issues. I have tried to form several bullet points to streamline the thought process and make the text more accessible to readers. I wasn’t able to put the entire section in a set of bullet points, but I made an attempt to better separate the conclusions from both experiments described in the article to avoid confusion. I hope that will satisfy the requirements.
Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links to the numerical methods mentioned in the review. I also rewrote the research gap statement to make it more readable. The conclusion section has also been revised. Below I have included corresponding point by point responses to the various comments made in the review.

1. After rereading the introduction section, I must agree that it was too vague and contained too much information that could be considered irrelevant in the scope of the article. To amend that, I have completely rewritten most of this section, while also adding mentions and new references to current works (2020-2023) related with Eshelby-Type/micromechanical models. I must admit that I was so focused on the mathematical synthesis aspect in the context of piezoelectric models that I didn’t encounter works on models mentioned in your review. Thank You for bringing this issue up because that also helped me in finding new potential paths to expand my research in the future.

2. I have decided to extend the explanation with a graphical clue in form of a new figure 1, which shows an example of a graph, from which a new structural number is being created. The example of a structural number has also been placed under new equation 1. I think that a simple figure might provide much more insight than my extensive explanations, while also conserving more space. In hindsight, I should have provided a descriptive figure in the first place.

3. I agree that this issue was left with a vague explanation. It was already pointed out that the conclusions were overly long and realising that I was trying to make the explanation concise. Unfortunately, I made it too vague instead. I have tried to amend this by adding more detailed explanation in the conclusions. To summarize the explanation, the uncovered issue may be due to the structure of the written code for the method based on structural number algebra. It was made using closed loops which were combined with quite complicated substitution operations and complex algebraic representations of each graph edge. Combined, these three elements could contribute to the perceived great increase in both computation time and RAM requirement. The matrix method was largely an unmodified module readily available as an integral part of the Matlab software which may have been the reason for its much better performance. The new method had to be written from ground up which probably left a lot of the performance on the table due to my lack of programming skill and also the inability to better access program’s memory and CPU handling operations.

4. Most of the conclusions section was reworked according to raised issues. I have tried to form several bullet points to streamline the thought process and make the text more accessible to readers. I wasn’t able to put the entire section in a set of bullet points, but I made an attempt to better separate the conclusions from both experiments described in the article to avoid confusion. I hope that will satisfy the requirements.
Competing Interests: No competing interests were disclosed. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 06 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

06 Sep 2024

Author Response

Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links ... Continue reading Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links to the numerical methods mentioned in the review. I also rewrote the research gap statement to make it more readable. The conclusion section has also been revised. Below I have included corresponding point by point responses to the various comments made in the review.

1. After rereading the introduction section, I must agree that it was too vague and contained too much information that could be considered irrelevant in the scope of the article. To amend that, I have completely rewritten most of this section, while also adding mentions and new references to current works (2020-2023) related with Eshelby-Type/micromechanical models. I must admit that I was so focused on the mathematical synthesis aspect in the context of piezoelectric models that I didn’t encounter works on models mentioned in your review. Thank You for bringing this issue up because that also helped me in finding new potential paths to expand my research in the future.

2. I have decided to extend the explanation with a graphical clue in form of a new figure 1, which shows an example of a graph, from which a new structural number is being created. The example of a structural number has also been placed under new equation 1. I think that a simple figure might provide much more insight than my extensive explanations, while also conserving more space. In hindsight, I should have provided a descriptive figure in the first place.

3. I agree that this issue was left with a vague explanation. It was already pointed out that the conclusions were overly long and realising that I was trying to make the explanation concise. Unfortunately, I made it too vague instead. I have tried to amend this by adding more detailed explanation in the conclusions. To summarize the explanation, the uncovered issue may be due to the structure of the written code for the method based on structural number algebra. It was made using closed loops which were combined with quite complicated substitution operations and complex algebraic representations of each graph edge. Combined, these three elements could contribute to the perceived great increase in both computation time and RAM requirement. The matrix method was largely an unmodified module readily available as an integral part of the Matlab software which may have been the reason for its much better performance. The new method had to be written from ground up which probably left a lot of the performance on the table due to my lack of programming skill and also the inability to better access program’s memory and CPU handling operations.

4. Most of the conclusions section was reworked according to raised issues. I have tried to form several bullet points to streamline the thought process and make the text more accessible to readers. I wasn’t able to put the entire section in a set of bullet points, but I made an attempt to better separate the conclusions from both experiments described in the article to avoid confusion. I hope that will satisfy the requirements.
Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links to the numerical methods mentioned in the review. I also rewrote the research gap statement to make it more readable. The conclusion section has also been revised. Below I have included corresponding point by point responses to the various comments made in the review.

1. After rereading the introduction section, I must agree that it was too vague and contained too much information that could be considered irrelevant in the scope of the article. To amend that, I have completely rewritten most of this section, while also adding mentions and new references to current works (2020-2023) related with Eshelby-Type/micromechanical models. I must admit that I was so focused on the mathematical synthesis aspect in the context of piezoelectric models that I didn’t encounter works on models mentioned in your review. Thank You for bringing this issue up because that also helped me in finding new potential paths to expand my research in the future.

2. I have decided to extend the explanation with a graphical clue in form of a new figure 1, which shows an example of a graph, from which a new structural number is being created. The example of a structural number has also been placed under new equation 1. I think that a simple figure might provide much more insight than my extensive explanations, while also conserving more space. In hindsight, I should have provided a descriptive figure in the first place.

3. I agree that this issue was left with a vague explanation. It was already pointed out that the conclusions were overly long and realising that I was trying to make the explanation concise. Unfortunately, I made it too vague instead. I have tried to amend this by adding more detailed explanation in the conclusions. To summarize the explanation, the uncovered issue may be due to the structure of the written code for the method based on structural number algebra. It was made using closed loops which were combined with quite complicated substitution operations and complex algebraic representations of each graph edge. Combined, these three elements could contribute to the perceived great increase in both computation time and RAM requirement. The matrix method was largely an unmodified module readily available as an integral part of the Matlab software which may have been the reason for its much better performance. The new method had to be written from ground up which probably left a lot of the performance on the table due to my lack of programming skill and also the inability to better access program’s memory and CPU handling operations.

4. Most of the conclusions section was reworked according to raised issues. I have tried to form several bullet points to streamline the thought process and make the text more accessible to readers. I wasn’t able to put the entire section in a set of bullet points, but I made an attempt to better separate the conclusions from both experiments described in the article to avoid confusion. I hope that will satisfy the requirements.
Competing Interests: No competing interests were disclosed. Close
Report a concern

Version 1

VERSION 1

PUBLISHED 05 Oct 2023

Views

Reviewer Report 29 Nov 2023

Saurabh K. Yadav, Institute of Infrastructure Technology Research and Management, Ahmedabad, Gujarat, India

Approved with Reservations

https://doi.org/10.5256/f1000research.154351.r222666

The manuscript proposes a new approach for parameter estimation of piezoelectric structures based on the Cauer’s method, and compares its computational performance with the traditional matrix method. The authors claim that their method can synthesize a piezoelectric model using only a set of resonant and anti-resonant frequencies as the input data, and approximate the piezoelectric parameters using the synthesized mechanical model and some physical equations. The authors also use a non-classical method based on structural numbers and edge graphs to analyze the mechanical model and reduce the computational load. The manuscript is well-written and organized, and the mathematical derivations are clear and rigorous. However, I have some concerns and suggestions that need to be addressed before the manuscript can be accepted for publication.

The main merit and novelty of the manuscript are not clearly stated in the abstract, introduction and conclusion. The authors should highlight how their method differs from the existing methods in the literature, and what are the advantages and limitations of their method.
The authors use a single piezoelectric plate as an example to verify their electromechanical model. However, the results show that the model has a significant error in the resonant frequency and the displacement amplitude. The authors should explain the sources of this error and how it can be reduced or corrected. Moreover, the authors should test their model on more realistic and complex piezoelectric structures, such as stacks, rings, beams, etc., and compare their results with experimental data or other numerical methods.
The authors compare the computational performance of their method with the matrix method using a two-degree system. However, this is not a fair comparison, as the matrix method can handle systems with any number of degrees of freedom, while the structural number method is limited by the complexity of the edge graph. The authors should compare the computational performance of both methods on systems with different numbers of degrees of freedom, and use more appropriate metrics, such as the number of floating point operations, the memory usage, the accuracy, etc.
The authors use some references that are published in non-English languages, such as Refs. 8, 11 and 29. The authors should provide English translations of these references, or use more accessible sources that are published in English. The authors should also update their literature review with some recent works on piezoelectric modelling and parameter estimation.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Yes
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Mathematical modelling

CITE

Report a concern

Author Response 13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

13 Apr 2024

Author Response

Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review ... Continue reading Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your concerns directly:

1. Let me repeat my answer to the second reviewer here, as it addresses the same concerns. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.
The limitation of this method is that it can only be used to synthesize models composed of components connected in series. So far, it can only be used to synthesise specific piezoelectric plate configuraions, where each component in a stack is responsible exclusively for a single resonant frequency set as the input. I have modified the article to better reflect those statements.

2. Following up from my response to the second reviever, I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. It has to be noted as well, that this method relies on a lumped representation of an otherwise continous system. The main issue of this approach is that the lumped parameters are approximations of the real parameters and an electromechanical model currently can't perfectly match the behavior of a piezoelectric plate, where the interactions and their magnitude changes between the layers of a single piezoelement. There are also boundaries between piezoelectric elements, made with adhesives and isolating layers, which also introduce many hidden variables in the real model behaviour. The presented method is still in its infancy and requires more testing and refinement. I would like to provide a consice report on the empirical results in a follow up article though, as including all the information about the procedure and test stand would make this article very bloated.

3. I was made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. New suite of tests has uncovered more information about the structural number algorithm optimisation and shed some light on its shortcomings. I have modified the manusript to showcase the second study. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation and i will try to consult a specialist on the topic with an intent of providing some collaboration with more indepth information about algorithm complexity of the proposed method.

4. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

I hope that my response provides an exhaustive explanation to the issues raised.

Best Regards,
Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your concerns directly:

1. Let me repeat my answer to the second reviewer here, as it addresses the same concerns. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.
The limitation of this method is that it can only be used to synthesize models composed of components connected in series. So far, it can only be used to synthesise specific piezoelectric plate configuraions, where each component in a stack is responsible exclusively for a single resonant frequency set as the input. I have modified the article to better reflect those statements.

2. Following up from my response to the second reviever, I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. It has to be noted as well, that this method relies on a lumped representation of an otherwise continous system. The main issue of this approach is that the lumped parameters are approximations of the real parameters and an electromechanical model currently can't perfectly match the behavior of a piezoelectric plate, where the interactions and their magnitude changes between the layers of a single piezoelement. There are also boundaries between piezoelectric elements, made with adhesives and isolating layers, which also introduce many hidden variables in the real model behaviour. The presented method is still in its infancy and requires more testing and refinement. I would like to provide a consice report on the empirical results in a follow up article though, as including all the information about the procedure and test stand would make this article very bloated.

3. I was made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. New suite of tests has uncovered more information about the structural number algorithm optimisation and shed some light on its shortcomings. I have modified the manusript to showcase the second study. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation and i will try to consult a specialist on the topic with an intent of providing some collaboration with more indepth information about algorithm complexity of the proposed method.

4. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

I hope that my response provides an exhaustive explanation to the issues raised.

Best Regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

13 Apr 2024

Author Response

Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review ... Continue reading Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your concerns directly:

1. Let me repeat my answer to the second reviewer here, as it addresses the same concerns. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.
The limitation of this method is that it can only be used to synthesize models composed of components connected in series. So far, it can only be used to synthesise specific piezoelectric plate configuraions, where each component in a stack is responsible exclusively for a single resonant frequency set as the input. I have modified the article to better reflect those statements.

2. Following up from my response to the second reviever, I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. It has to be noted as well, that this method relies on a lumped representation of an otherwise continous system. The main issue of this approach is that the lumped parameters are approximations of the real parameters and an electromechanical model currently can't perfectly match the behavior of a piezoelectric plate, where the interactions and their magnitude changes between the layers of a single piezoelement. There are also boundaries between piezoelectric elements, made with adhesives and isolating layers, which also introduce many hidden variables in the real model behaviour. The presented method is still in its infancy and requires more testing and refinement. I would like to provide a consice report on the empirical results in a follow up article though, as including all the information about the procedure and test stand would make this article very bloated.

3. I was made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. New suite of tests has uncovered more information about the structural number algorithm optimisation and shed some light on its shortcomings. I have modified the manusript to showcase the second study. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation and i will try to consult a specialist on the topic with an intent of providing some collaboration with more indepth information about algorithm complexity of the proposed method.

4. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

I hope that my response provides an exhaustive explanation to the issues raised.

Best Regards,
Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your concerns directly:

1. Let me repeat my answer to the second reviewer here, as it addresses the same concerns. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.
The limitation of this method is that it can only be used to synthesize models composed of components connected in series. So far, it can only be used to synthesise specific piezoelectric plate configuraions, where each component in a stack is responsible exclusively for a single resonant frequency set as the input. I have modified the article to better reflect those statements.

2. Following up from my response to the second reviever, I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. It has to be noted as well, that this method relies on a lumped representation of an otherwise continous system. The main issue of this approach is that the lumped parameters are approximations of the real parameters and an electromechanical model currently can't perfectly match the behavior of a piezoelectric plate, where the interactions and their magnitude changes between the layers of a single piezoelement. There are also boundaries between piezoelectric elements, made with adhesives and isolating layers, which also introduce many hidden variables in the real model behaviour. The presented method is still in its infancy and requires more testing and refinement. I would like to provide a consice report on the empirical results in a follow up article though, as including all the information about the procedure and test stand would make this article very bloated.

3. I was made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. New suite of tests has uncovered more information about the structural number algorithm optimisation and shed some light on its shortcomings. I have modified the manusript to showcase the second study. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation and i will try to consult a specialist on the topic with an intent of providing some collaboration with more indepth information about algorithm complexity of the proposed method.

4. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

I hope that my response provides an exhaustive explanation to the issues raised.

Best Regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern

Views

Reviewer Report 29 Nov 2023

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

Approved with Reservations

https://doi.org/10.5256/f1000research.154351.r222661

This manuscript proposed a new approach for parameter estimation of piezoelectric structures based on the Cauer’s method, and its computational performance is compared with the traditional method. However, in my opinion, some concerns need to be addressed more clearly before ... Continue reading

What’s the main merit of the manuscript? It should be clearly addressed in the abstract, introduction and conclusion.
To compare the computational performance of the new model, the model response and computation time are compared with the matrix method. What are the errors of these two models? The two frequencies of the new model are compared with the ones of matrix method. Are the results of the matrix method accurate? What are the actual frequencies?

It is the same during the amplitude comparison. Moreover, to compare the computational performance of the algorithm, there are many metrics, such as time cost, the number of floating points (FLOPs) multiple adds (MAdds), and so on. In addition, what is the meaning of only decreasing the computation time?
Page 4: “A combination of those factors creates four different types of models…” It refers to Refs. 8 and 11. However, Refs. 8 and 11 are published in non-English. Therefore, the details of the four models are suggested to be introduced in the manuscript.
Experimental results: the results should be further analyzed.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Partly
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Structural health monitoring by uisng lamb waves

CITE

Report a concern

Author Response 13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

13 Apr 2024

Author Response

Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have ... Continue reading Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your querries to the best of my ability:

1. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.

2. The predetermined frequencies and amplitudes were chosen arbitrarily and do not reflect any real piezoelectric system. A comparison between both analysis methods was done on the same model synthesized with the Cauer method. The matrix method was considered as a reference method because its a known and established way to calculate model response. The structural number method was compared against it and it was a pure comparison between two analytical methods.
I was also made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation. I will need to consult a specialist on that matter and conduct more tests but that would probably go beyond the scope of this article and should be contained in a separate paper focused only on the comparison between both methods.

3. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

4. I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. However, i would like to envelope the empirical analysis, along with all the necessary test procedures and the test stand description in a second article, serving as the continuation for this study.

I hope that i have provide an exhaustive explanation to the issues raised.

Best Regards,
Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your querries to the best of my ability:

1. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.

2. The predetermined frequencies and amplitudes were chosen arbitrarily and do not reflect any real piezoelectric system. A comparison between both analysis methods was done on the same model synthesized with the Cauer method. The matrix method was considered as a reference method because its a known and established way to calculate model response. The structural number method was compared against it and it was a pure comparison between two analytical methods.
I was also made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation. I will need to consult a specialist on that matter and conduct more tests but that would probably go beyond the scope of this article and should be contained in a separate paper focused only on the comparison between both methods.

3. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

4. I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. However, i would like to envelope the empirical analysis, along with all the necessary test procedures and the test stand description in a second article, serving as the continuation for this study.

I hope that i have provide an exhaustive explanation to the issues raised.

Best Regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

13 Apr 2024

Author Response

Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have ... Continue reading Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your querries to the best of my ability:

1. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.

2. The predetermined frequencies and amplitudes were chosen arbitrarily and do not reflect any real piezoelectric system. A comparison between both analysis methods was done on the same model synthesized with the Cauer method. The matrix method was considered as a reference method because its a known and established way to calculate model response. The structural number method was compared against it and it was a pure comparison between two analytical methods.
I was also made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation. I will need to consult a specialist on that matter and conduct more tests but that would probably go beyond the scope of this article and should be contained in a separate paper focused only on the comparison between both methods.

3. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

4. I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. However, i would like to envelope the empirical analysis, along with all the necessary test procedures and the test stand description in a second article, serving as the continuation for this study.

I hope that i have provide an exhaustive explanation to the issues raised.

Best Regards,
Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your querries to the best of my ability:

1. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.

2. The predetermined frequencies and amplitudes were chosen arbitrarily and do not reflect any real piezoelectric system. A comparison between both analysis methods was done on the same model synthesized with the Cauer method. The matrix method was considered as a reference method because its a known and established way to calculate model response. The structural number method was compared against it and it was a pure comparison between two analytical methods.
I was also made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation. I will need to consult a specialist on that matter and conduct more tests but that would probably go beyond the scope of this article and should be contained in a separate paper focused only on the comparison between both methods.

3. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

4. I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. However, i would like to envelope the empirical analysis, along with all the necessary test procedures and the test stand description in a second article, serving as the continuation for this study.

I hope that i have provide an exhaustive explanation to the issues raised.

Best Regards,
Competing Interests: No competing interests were disclosed. Close
Report a concern

Comments on this article Comments (0)

Version 3

VERSION 3 PUBLISHED 05 Oct 2023

Open Peer Review

Reviewer Status

Reviewer Reports

	Invited Reviewers
	1	2	3	4
Version 3 (revision) 06 Sep 24	read		read
Version 2 (revision) 02 Apr 24	read		read	read
Version 1 05 Oct 23	read	read

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, China
Saurabh K. Yadav, Institute of Infrastructure Technology Research and Management, Ahmedabad, India
Pedro M. Ferreira, Universidade Nova de Lisboa, Lisbon, Portugal
Zine Ghemari, University Mohamed Boudiaf,, M’sila,, Algeria

Comments on this article

All Comments(0)

Add a comment

Browse by related subjects

Back to all reports

Reviewer Report

5 Views

09 Sep 2024 | for Version 3

Pedro M. Ferreira, Universidade Nova de Lisboa, Lisbon, Lisbon, Portugal

5 Views Cite this report Responses(0)

Approved

All the comments have been taken into account, the manuscript is in a position to be approved.

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Piezoelectric Composites, Piezoelectric Behaviour, Advanced Materials, Smart Materials

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

8 Views

09 Sep 2024 | for Version 3

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

8 Views Cite this report Responses(0)

Approved

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Structural health monitoring by uisng lamb waves

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

3 Views

13 Sep 2024 | for Version 2

Zine Ghemari, University Mohamed Boudiaf,, M’sila,, Algeria

3 Views Cite this report Responses(0)

Approved With Reservations

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Maintenance, Piezoelectricity, Sensors

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

16 Views

28 Aug 2024 | for Version 2

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

16 Views Cite this report Responses(1)

Approved With Reservations

If the name of the proposed method is included in the title, the novelty of this manuscript will be more easily got by the readers.
Figure 3: what do these symbols, ₁X_0s, ₁X_1s and so on, stand for?
The material type and some key parameters of PZT should be clearly addressed instead of just putting a website link.
The resolution of the figures is low and their quality should be further improved.
Table 2 and 5: how to confirm the accuracy of the data?
Based on the results, the proposed method can be used to analyze the characteristics of piezoelectric systems accurately. However, the computation cost is much larger than the traditional method. Therefore, what is the merit or advantage of this method?

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Structural health monitoring by uisng lamb waves

Respond to this report

Responses (1)

Author Response

10 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

Thank You for another insightful review. Unfortunately I have already sent a new version of the article prior to receiving your comments on version 2 so any possible amendments will have to be placed in another version of the article if necessary. I will try to address the issues point by point, as previously.

The aim of this article’s title was to focus the attention of the reader on the aspect of piezoelectric model synthesis, with the main “tool” being the Cauer’s method. The presented analysis method using structural numbers was placed within the article as it was necessary to provide means to validate the synthesis method. The research on the alternative using structural numbers was done in parallel and there were plans initially, to also use structural numbers for the synthesis itself, hence this method ended up being part of the article. I realise the outcome of that idea is not ideal and the title may be somewhat ambiguous but at the same time the title is not too concise already and I’m afraid I would only make it worse by trying to fix this issue.
₁X₀and ₁X₁are nodes of the graph of relations. Each node represents the displacement of any object being part of a physical model. So, if we had an example of a model with two moving masses, a graph of such model would contain three nodes. The ₁X₀node is special in that it represents any relations of a model with anything we consider being “outside” of the model’s scope. In the case of this article, the ₁X₀node represents one of the fixed ends of the model, where the last spring and damper are attached. The first number symbolizes an entire isolated system with its own relations. So, if we had two models working in parallel, the first one would have “1” at the front, and the second one would have “2”. The second number is used to number nodes within each system.
I agree that putting just the link to the used material catalogue might have introduced unnecessary ambiguity into the article. I did a mistake of not properly disclosing that I chose the SM411 material to test the model response which would have made it slightly better. If You consider this a major issue with the article, then I will try to prepare another version with clearly disclosed information about the used material.
I will try to improve the quality of figures in the next version if necessary.
The entire code used to perform the study and obtain the data in tables 2 and 5 was placed in an online repository (Mendeley data) under the reference 42. The algorithm is written in such a way that it should automatically collect all the data and print the amplitude responses with both studied methods. The only caveat is that the algorithm was written using the proprietary Matlab code format, so it is recommended that the algorithm is used with Matlab r2019b to ensure the same results.
The method of system analysis presented in the article has a markedly different approach to the way in which the relationships occurring within the analysed system are represented.
The response equations of the system are obtained based on transformations on structural numbers, which are in principle less complex than matrix equations. This led the authors to believe that the algorithmisation of this method would allow the response to be calculated with less computing power than in the case of the matrix method. Unfortunately, this could not be proved in the article. However, it is not clear whether the inferior performance of the method obtained from calculations performed in Matlab is directly due to the shortcomings of the method itself, or whether it is the result of poor optimisation of the code itself, or even of the environment, which is optimised for the well-known matrix method. Therefore authors assumed that it is still worth sharing the results and the method may have potential, however more work is necessary to determine the cause of worse performance.

Thank you again for your time spent on preparing your review. I hope my answers have cleared at least some of the issues raised. I will be eagerly waiting for any further responses.

Best regards,

View more View less

Competing Interests

No competing interests were disclosed.

Back to all reports

Reviewer Report

17 Views

09 Aug 2024 | for Version 2

Pedro M. Ferreira, Universidade Nova de Lisboa, Lisbon, Lisbon, Portugal

17 Views Cite this report Responses(1)

Approved With Reservations

The introduction to the present literature review is poor, particularly in the following excerpt: "Researchers largely focus on modelling simple piezoelectric components with lumped models and matrix tensor equations or try to approach the problem using FEM (finite element method) analysis." There are several models that could be referenced in the introduction, such as Eshelby-Type Models, Micromechanics-based models, or Asymptotic Homogenization-Based Models, which are widely used to model the behaviour of piezoelectric composites. Additionally, this excerpt needs to be supported with more current references. Keywords that could help in finding up-to-date references include: "Piezoelectric composites, Electro-mechanical properties, Analytical models, Numerical models."
In the last paragraph of the introduction, the excerpt “The number of rows and columns, however, is tightly connected to the shape of an edge graph which a given structural number relates to,” is very unclear. It should be rewritten for better clarity and explained in more detail.
The author compares the computation time and memory usage of parameterized algorithms, highlighting a significant increase in both when using the structural number method compared to the classical matrix method (as shown in Table 4). The authors should clarify why the calculation time and memory usage of the modified algorithms increase so dramatically as the number of degrees of freedom in the analysed systems rises.
The conclusion is overly long and needs to be more concise, with an emphasis on the limitations of the authors' analytical approach. For better clarity, consider presenting the main conclusions and limitations as bullet points.

Is the work clearly and accurately presented and does it cite the current literature?

Partly
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Piezoelectric Composites, Piezoelectric Behaviour, Advanced Materials, Smart Materials

Respond to this report

Responses (1)

Author Response

06 Sep 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

Thank You for your insightful review. In line with Your remarks, I have rewritten most of the introductory section of the article, adding more up-to-date literature references and including links to the numerical methods mentioned in the review. I also rewrote the research gap statement to make it more readable. The conclusion section has also been revised. Below I have included corresponding point by point responses to the various comments made in the review.

1. After rereading the introduction section, I must agree that it was too vague and contained too much information that could be considered irrelevant in the scope of the article. To amend that, I have completely rewritten most of this section, while also adding mentions and new references to current works (2020-2023) related with Eshelby-Type/micromechanical models. I must admit that I was so focused on the mathematical synthesis aspect in the context of piezoelectric models that I didn’t encounter works on models mentioned in your review. Thank You for bringing this issue up because that also helped me in finding new potential paths to expand my research in the future.

2. I have decided to extend the explanation with a graphical clue in form of a new figure 1, which shows an example of a graph, from which a new structural number is being created. The example of a structural number has also been placed under new equation 1. I think that a simple figure might provide much more insight than my extensive explanations, while also conserving more space. In hindsight, I should have provided a descriptive figure in the first place.

3. I agree that this issue was left with a vague explanation. It was already pointed out that the conclusions were overly long and realising that I was trying to make the explanation concise. Unfortunately, I made it too vague instead. I have tried to amend this by adding more detailed explanation in the conclusions. To summarize the explanation, the uncovered issue may be due to the structure of the written code for the method based on structural number algebra. It was made using closed loops which were combined with quite complicated substitution operations and complex algebraic representations of each graph edge. Combined, these three elements could contribute to the perceived great increase in both computation time and RAM requirement. The matrix method was largely an unmodified module readily available as an integral part of the Matlab software which may have been the reason for its much better performance. The new method had to be written from ground up which probably left a lot of the performance on the table due to my lack of programming skill and also the inability to better access program’s memory and CPU handling operations.

4. Most of the conclusions section was reworked according to raised issues. I have tried to form several bullet points to streamline the thought process and make the text more accessible to readers. I wasn’t able to put the entire section in a set of bullet points, but I made an attempt to better separate the conclusions from both experiments described in the article to avoid confusion. I hope that will satisfy the requirements.

View more View less

Competing Interests

No competing interests were disclosed.

Back to all reports

Reviewer Report

21 Views

29 Nov 2023 | for Version 1

Saurabh K. Yadav, Institute of Infrastructure Technology Research and Management, Ahmedabad, Gujarat, India

21 Views Cite this report Responses(1)

Approved With Reservations

The main merit and novelty of the manuscript are not clearly stated in the abstract, introduction and conclusion. The authors should highlight how their method differs from the existing methods in the literature, and what are the advantages and limitations of their method.
The authors use a single piezoelectric plate as an example to verify their electromechanical model. However, the results show that the model has a significant error in the resonant frequency and the displacement amplitude. The authors should explain the sources of this error and how it can be reduced or corrected. Moreover, the authors should test their model on more realistic and complex piezoelectric structures, such as stacks, rings, beams, etc., and compare their results with experimental data or other numerical methods.
The authors compare the computational performance of their method with the matrix method using a two-degree system. However, this is not a fair comparison, as the matrix method can handle systems with any number of degrees of freedom, while the structural number method is limited by the complexity of the edge graph. The authors should compare the computational performance of both methods on systems with different numbers of degrees of freedom, and use more appropriate metrics, such as the number of floating point operations, the memory usage, the accuracy, etc.
The authors use some references that are published in non-English languages, such as Refs. 8, 11 and 29. The authors should provide English translations of these references, or use more accessible sources that are published in English. The authors should also update their literature review with some recent works on piezoelectric modelling and parameter estimation.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Yes
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Mathematical modelling

Respond to this report

Responses (1)

Author Response

13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

Thank you for the critical review. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your concerns directly:

1. Let me repeat my answer to the second reviewer here, as it addresses the same concerns. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.
The limitation of this method is that it can only be used to synthesize models composed of components connected in series. So far, it can only be used to synthesise specific piezoelectric plate configuraions, where each component in a stack is responsible exclusively for a single resonant frequency set as the input. I have modified the article to better reflect those statements.

2. Following up from my response to the second reviever, I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. It has to be noted as well, that this method relies on a lumped representation of an otherwise continous system. The main issue of this approach is that the lumped parameters are approximations of the real parameters and an electromechanical model currently can't perfectly match the behavior of a piezoelectric plate, where the interactions and their magnitude changes between the layers of a single piezoelement. There are also boundaries between piezoelectric elements, made with adhesives and isolating layers, which also introduce many hidden variables in the real model behaviour. The presented method is still in its infancy and requires more testing and refinement. I would like to provide a consice report on the empirical results in a follow up article though, as including all the information about the procedure and test stand would make this article very bloated.

3. I was made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. New suite of tests has uncovered more information about the structural number algorithm optimisation and shed some light on its shortcomings. I have modified the manusript to showcase the second study. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation and i will try to consult a specialist on the topic with an intent of providing some collaboration with more indepth information about algorithm complexity of the proposed method.

4. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

I hope that my response provides an exhaustive explanation to the issues raised.

Best Regards,

View more View less

Competing Interests

No competing interests were disclosed.

Back to all reports

Reviewer Report

31 Views

29 Nov 2023 | for Version 1

Guangtao Lu, Wuhan University of Science and Technology, Wuhan, Hubei, China

31 Views Cite this report Responses(1)

Approved With Reservations

What’s the main merit of the manuscript? It should be clearly addressed in the abstract, introduction and conclusion.
To compare the computational performance of the new model, the model response and computation time are compared with the matrix method. What are the errors of these two models? The two frequencies of the new model are compared with the ones of matrix method. Are the results of the matrix method accurate? What are the actual frequencies?

It is the same during the amplitude comparison. Moreover, to compare the computational performance of the algorithm, there are many metrics, such as time cost, the number of floating points (FLOPs) multiple adds (MAdds), and so on. In addition, what is the meaning of only decreasing the computation time?
Page 4: “A combination of those factors creates four different types of models…” It refers to Refs. 8 and 11. However, Refs. 8 and 11 are published in non-English. Therefore, the details of the four models are suggested to be introduced in the manuscript.
Experimental results: the results should be further analyzed.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Partly
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Structural health monitoring by uisng lamb waves

Respond to this report

Responses (1)

Author Response

13 Apr 2024

Jacek Harazin, Faculty of Mechanical Engineering, Politechnika Slaska, Gliwice, 44-100, Poland

Thank you very much for the critical review and Your suggestions regarding the article clarity. I would also like to apologise for a long break before my reply. I have prepared several corrections to the article regarding your review and they will be published with the new version. I've also tried to address your querries to the best of my ability:

1. The article showcases a new method to synthesize cascading electromechanical models using the Cauer method coupled with Van Dyke-Butterworth models to represent piezoelectric stacks with clumped mechanical and electrical components. Each of those methods have been already used by researchers but i have found no literature on the particular approach presented in this article. The novelty revolves around the general approach to synthesis with very limited input information. Most researchers focus on specific applications where piezoelectric parameters and the analysed structure is largely known or predetermined experimentally. Here, only the frequencies are known and there is an infinite number of possible models that could realise those predetermined resonant frequencies. Coupled with further steps presented in the article, this method offers a possibility to design piezoelectric stacks with different material properties or geometric dimensions, capable of resonating in predetermined frequencies.

2. The predetermined frequencies and amplitudes were chosen arbitrarily and do not reflect any real piezoelectric system. A comparison between both analysis methods was done on the same model synthesized with the Cauer method. The matrix method was considered as a reference method because its a known and established way to calculate model response. The structural number method was compared against it and it was a pure comparison between two analytical methods.
I was also made aware that a comparison of calculation time used exclusively is a bad metric for algorithm complexity. I have tried looking for more literature on the matter and added more references to the article. Given my abilities and available resources, i have prepared a second study which focused more on the memory criterion. I've used the available Matlab evaluation modules to estimate the average RAM usage between both algorithms on a variety of models, along with a comparison between the average time it took both algorithms to complete. I am aware that this method may also not be the best way to evaluate algorithm complexity and only presents a rough estimation. I will need to consult a specialist on that matter and conduct more tests but that would probably go beyond the scope of this article and should be contained in a separate paper focused only on the comparison between both methods.

3. I have provided additional literature references to publications of the authors involved in the development of the structural number method where they descibe mentioned models in english. I hope that will provide the necessary background which was lacking previously.

4. I agree and i can already attest that the method is currently being verified empirically. The synthesis method has been verified already with three real piezoelectric stacks with different geometrical dimensions which has shown some discrepancy between model results and real plate behaviour. The source of those differences was traced back to how some of the piezoelectric parameters are determined in the factory (under 1kHz wave stimulation) versus plate properties in the resonance proximity. However, i would like to envelope the empirical analysis, along with all the necessary test procedures and the test stand description in a second article, serving as the continuation for this study.

I hope that i have provide an exhaustive explanation to the issues raised.

Best Regards,

View more View less

Competing Interests

No competing interests were disclosed.

Alongside their report, reviewers assign a status to the article:

Approved - the paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations - A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approved - fundamental flaws in the paper seriously undermine the findings and conclusions

[1] 1. Sangryun L, Jiyoung J, Seunghwa R: Micromechanics-based prediction of the effective properties of piezoelectric composite having interfacial imperfections. Compos. Struct. 2020; 240:112076. Publisher Full Text

[2] 2. Nada T, Lahcen A, Nadia F, et al.: Nonlinear micromechanical modeling of fully coupled piezo-elastic composite under large deformation and high electric field. Compos. Struct. 2023; 315:116991. Publisher Full Text

[3] 3. Fateh EM, Amar M, Fahmi Z, et al.: A Coupled Electro-Mechanical Homogenization-Based Model for PVDF-Based Piezo-Composites Considering α → β Phase Transition and Interfacial Damage. Polymers. 2023; 15:2994. PubMed Abstract | Publisher Full Text | Free Full Text

[4] 4. Jeffrey H, Yuqi S: Piezoelectric modulus prediction using machine learning and graph neural networks. Chem. Phys. Lett. 2022; 791:139359. Publisher Full Text

[5] 5. Nguyen B, Zhuang X, Rabczuk T: Numerical model for the characterization of Maxwell-Wagner relaxation in piezoelectric and flexoelectric composite material. Comput. Struct. 2018; 208:75–91. Publisher Full Text

[6] 6. Larbi W: Numerical modeling of sound and vibration reduction using viscoelastic materials and shunted piezoelectric patches. Comput. Struct. 2020; 232: 105822. Publisher Full Text

[7] 7. Adriaens H, Koning W, Banning R: Modelling piezoelectric actuators. IEEE ASME Trans. Mechatron. 2000; 5: 331–341. Publisher Full Text

[8] 8. Goldfarb M, Celanovic N: Modelling piezoelectric stack actuators for control of micromanipulation. IEEE Control. Syst. 1997; 17: 69–79. Publisher Full Text

[9] 9. Richter B, Twiefel J, Wallaschek J: Piezoelectric Equivalent Circuit Models.Priya S, Inman DJ, editors. Energy Harvesting Technologies. Boston, MA: Springer; 2009. Publisher Full Text

[10] 10. Lahmer T, Kaltenbacher M, Kaltenbacher B, et al.: FEM-based determination of real and complex elastic, dielectric, and piezoelectric moduli in piezoceramic materials. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 2008; 55(2): 465–475. PubMed Abstract | Publisher Full Text

[11] 11. Goldstein A: Self-Tuning multimodal piezoelectric shunt damping. J. Braz. Soc. Mech. Sci. Eng. 2011; 33: 428–436. Publisher Full Text

[12] 12. Batra R, Dell’isola F, Vidoli S, et al.: Multimode vibration suppression with passive two-terminal distributed network incorporating piezoceramic transducers. Int. J. Solids Struct. 2005; 42: 3115–3132. Publisher Full Text

[13] 13. San Miguel J, Buiochi F, Adamowski J: Numerical modelling of a circular piezoelectric ultrasonic transducer radiating in water. ABCM Symposium Series in Mechatronics. 2000; 2: 458–464.

[14] 14. Yamaoka S, Ohnishi K: Reaction force estimation of piezoelectric actuator by charge observation. IEEE International Conference on Mechatronics (ICM). Vicenza, Italy IEEE Xplore. 2013. Publisher Full Text

[15] 15. Delibas B, Arockiarajan A, Seemann W: Rate dependent properties of perovskite type tetragonal piezoelectric materials using micromechanical model. Int. J. Solids Struct. 2006; 43: 697–712. Publisher Full Text

[16] 16. Gan J, Zhang X: A review of nonlinear hysteresis modelling and control of piezoelectric actuators. AIP Adv. 2019; 9(4): 040702. Publisher Full Text

[17] 17. Zhe L, Xing-Rong H, Le F: Numerical modeling of fluid–structure–piezoelectric interaction for energy harvesting. Comput. Methods Appl. Mech. Eng. 2023; 414: 116164. Publisher Full Text

[18] 18. IEEE: Standard on Piezoelectricity. ANSI/IEEE; 1988.

[19] 19. Soluch W, et al.: Wstęp do piezoelektryki. Warszawa: Wydawnictwo Komunikacji i Łączności; 1980.

[20] 20. Wang S-H, Tsai M-C: Dynamic modelling of thickness-mode piezoelectric transducer using the block diagram approach. Ultrasonics. 2011; 51: 617–624. PubMed Abstract | Publisher Full Text

[21] 21. Wróbel A: Classical and non-classical research methods and experimental studies of piezoelectric plates. Gliwice: Wydawnictwo Politechniki Śląskiej; 2018.

[22] 22. Cauer W: Ideale Transformatoren und lineare Transformationen. Elektrische Nachrichtentechnik (ENT). 1932; 9: 157–174.

[23] 23. Cauer W: Theorie der linearen Wechselstromschaltungen. Berlin: Akademie-Verlag; 1960; vol. 2. .

[24] 24. Buchacz A, Dymarek A, Dzitkowski T: Synthesis of discrete, continuous and discrete continuous vibrating systems represent-ed by graphs. Sixth International Scientific and Engineering Conference - Machine-Building and Technosphere on the Border of the XXI Century. 1999; 3: 243–245.

[25] 25. Buchacz A, Żurek K: Modelowanie i badanie wpływu zjawiska piezoelektrycznego na charakterystyki układu mechanicznego. Wydawnictwo Politechniki Śląskiej. 2010.

[26] 26. Smith M: Synthesis of mechanical networks: the inerter. IEEE Trans. Automat. Contr. 2002; 47: 1648–1662. Publisher Full Text

[27] 27. Białas K, Buchacz A, Dzitkowski T: Synteza drgających aktywnych układów mechanicznych z tłumieniem w ujęciu grafów biegunowych i liczb strukturalnych. Wydawnictwo Politechniki Śląskiej. 2010.

[28] 28. Nakamura N: Extended Rayleigh Damping Model. Front. Built Environ. 2016; 2. Publisher Full Text

[29] 29. Bellert S, Woźniacki H: Analiza I synteza układów elektrycznych metodą liczb strukturalnych. Warszawa: WNT; 1968.

[30] 30. Pikoń A, Wojnarowski J: Application of structural numbers to generating the characteristics of mechanical systems. Mech. Mach. Theory. 1995; 30:1027–1037. Publisher Full Text

[31] 31. Bellert S: Topological analysis and synthesis of linear systems. J. Frankl. Inst. 1962; 274:425–443. Publisher Full Text

[32] 32. Harazin J, Wróbel A: Research on a cascade model synthesis with the use of classical and non-classical methods in the context of new piezoelectric stack applications. Int. J. Mech. Sci. 2021; 12: 959–969. Publisher Full Text

[33] 33. Białas K, Buchacz A, Gałęziowski D: Application of Electric Subsystem as Implementation of Reduction of Mechanical Vibrations. Solid State Phenom. 2015; 220-221:9–14. Publisher Full Text

[34] 34. Białas K, Buchacz A: The Influence of Changing the Parameters of Electrical Components Implementing the Active Reduction of Vibration. Appl. Mech. Mater. 2014; 657:614–618. Publisher Full Text

[35] 35. Adamowski JC, Nader G, Silva ECN: Effective Damping Value of Piezoelectric Transducer Determined by Experimental Techniques and Numerical Analysis. ABCM Symposium Series in Mechatronics. 2004; 1: 271–279.

[36] 36. Giergiel J: Tłumienie drgań mechanicznych. Kraków: Wydawnictwo Akademii Górniczo-Hutniczej; 1984.

[37] 37. Trindade M, Benjeddou A: Effective Electromechanical Coupling Coefficients of Piezoelectric Adaptive Structures: Critical Evaluation and Optimization. Mech. Adv. Mater. Struct. 2009; 16: 210–223. Publisher Full Text

[38] 38. Buchacz A, Płaczek M: Modelowanie i badanie jednowymiarowych drgających układów mechatronicznych. Gliwice: Wydawnictwo Politechniki Śląskiej; 2011.

[39] 39. Arora S, Barak B: Computational complexity: A modern approach. Cambridge University Press; 2007.

[40] 40. Mirzoyan V, Sahakyan A, Mirzoyan P: Behavior of dielectric permittivity of piezomonocrystall near the resonant frequency. Phys. Math. Sci. 2014; 3:62–65.

[41] 41. Park Y, Zhang Y, Majzoubi M, et al.: Improvement of the standard characterization method on k₃₃ mode piezoelectric specimens. Sens. Actuators A: Phys. 2020; 312:112124–112124. Publisher Full Text

[42] 42. Harazin J: Model algorithms. Mendeley Data. 2022. Publisher Full Text

[43] 43. Harazin J: Piezoelectric electromechanical model (first iteration). Mendeley Data. 2023. Publisher Full Text

[44] 44. Harazin J: Improved analysis algorithms for structural and matrix methods. Mendeley Data. 2024. Publisher Full Text

Analytical approach to piezoelectric model synthesis with the use of Cauer’s method for system design

Abstract

Background

Methods

Results

Conclusions

Keywords

Revised Amendments from Version 2

Introduction

(1)

Figure 1. An example of a graph taken as a basis for a structural number formulation.

Methods

Mechanical model synthesis

(2)

(3)

(4)

(5)

(6)

(7)

Figure 2. A mechanical cascade system composed of an infinite number of stages composed of mass “m” and spring “c” pairs, synthesized from the transfer function polynomial with the Cauer’s method.

(8)

(9)

Figure 3. A discrete cascade mechanical model with added damping elements.

Mechanical model verification

(10)

(11)

(12)

(13)

(14)

Figure 4. Graph of the cascade model where: a) shows the sine coefficients b) shows the cosine coefficients.

(15)

(16)

(17)

Electromechanical model synthesis

(18)

(19)

(20)

(21)

(22)

(23)

(24)

Piezoelectric discrete model verification

(25)

(26)

(27)

(28)

(29)

Figure 5. Discrete piezoelectric model composed of mechanical lumped model combined by a transformer with an electrical network.

Algorithm testing

Results

Figure 6. Model response calculated for a) first vibrating mass, and b) second vibrating mass.

Table 1. Comparison of frequencies between model responses obtained with both methods.

Table 2. Comparison of amplitude responses obtained with both methods on the nodes.

Table 3. Computation time for both methods used to analyse a simple mechanical system.

Table 4. Computation time and memory usage for parametrized algorithms.

Figure 7. An inbuilt Matlab profiler module used to collect the data on the algorithm memory usage.

Figure 8. A graph showing model response obtained with analysis done using both methods.

Table 5. Results obtained for both methods in the resonance frequency data points.

Figure 9. Model response calculated for a) displacement curve, and b) voltage curve along the frequency axis.

Conclusions

Data availability

Source data

Underlying data

Software availability

Model algorithms

References

Comments on this article Comments (0)

Open Peer Review

Comments on this article Comments (0)

Open Peer Review

Reviewer Status

Reviewer Reports

Comments on this article

Browse by related subjects

Competing Interests Policy

Stay Updated