Multi-physics coupling mechanical behavior prediction method of piezoelectric semiconductor device

The multi-physical coupling nonlinear problem of piezoelectric semiconductor devices is solved through the gridded physical information neural network model, and the problems of large computing resources and inaccurate prediction of traditional methods are solved, efficient and accurate mechanical behavior prediction is achieved, and design is enhanced. value of application.

CN118940715AActive Publication Date: 2024-11-12ZHEJIANG UNIV

Patent Information

Application Number
CN202410928757.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12
Estimated Expiration
2044-07-11

AI Technical Summary

Technical Problem

The prior art is difficult to effectively deal with the multi-physical coupling nonlinear problem in piezoelectric semiconductor devices. The traditional method consumes a lot of computing resources and is inaccurate in prediction, making it difficult to obtain high-precision results in complex systems.

Method used

Using a grid-based physical information neural network model, a parallel physical information neural network with the same structure is constructed, and a loss function is constructed in combination with physical laws. The model is trained to predict the displacement, potential and electron concentration of piezoelectric semiconductor devices, and use adaptive Parameter adjustments are made to improve model accuracy.

Benefits of technology

It realizes efficient multi-physics coupled mechanical behavior prediction, improves computational efficiency and prediction accuracy, and can provide accurate design guidance in the absence of data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940715A_ABST
    Figure CN118940715A_ABST
Patent Text Reader

Abstract

The invention discloses a method for predicting a multi-physics coupling mechanical behavior of a piezoelectric semiconductor device, which predicts the multi-physics coupling mechanical behavior through a gridding inversion full-connection neural network model, and has remarkable advantages in solving a piezoelectric semiconductor multi-field coupling nonlinear problem. Flexible modeling of a complex coupling relation can be realized, a nonlinear effect can be coped, and efficient optimization is realized; meanwhile, through parallel training of the physical information neural network, under the condition that the calculation problem is complex and the calculation amount is large, the loss function in the model training process is reduced, the trained model can predict the approximate solution of the problem more rapidly, and therefore the calculation efficiency and the prediction accuracy are improved. Besides, according to the prediction method, the physical law is fused in the model training process, so that the designed model has high generalization ability, the precision of a model prediction result can be improved under the condition of lack of data, and the practical application value of the model designed through the method is enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-physical field mechanical behavior prediction of piezoelectric semiconductors, and specifically relates to a multi-physical field coupled mechanical behavior prediction method for piezoelectric semiconductor devices. Background Art

[0002] Piezoelectric semiconductor devices are a type of semiconductor device that uses the piezoelectric effect to convert mechanical energy into electrical energy or vice versa. These devices can be used to make optoelectronics, electronics, microelectronics, integrated circuits, chips, memory, and microwave devices, and have broad application potential in energy, communications, aerospace, and medical health. As the demand for energy sustainability and intelligence increases, piezoelectric semiconductor technology will continue to be driven by research and application to meet the evolving market needs.

[0003] Piezoelectric semiconductor devices often involve the coupling of multiple physical fields, including mechanical fields, electric fields, and thermal fields. Understanding the interactions between these physical fields is crucial to establishing a multi-field coupling computational model. At present, there are the following difficulties in solving the multi-field coupling nonlinear problems of piezoelectric semiconductors: (1) Piezoelectric semiconductor devices are simultaneously affected by electric fields, strain fields, and temperature fields. These fields are coupled with each other, resulting in complex physical phenomena and nonlinear behaviors; (2) The nonlinear effects of piezoelectric semiconductor devices become more significant under multi-field coupling, such as geometric nonlinearity and nonlinear carrier transport. These nonlinear effects require complex numerical methods to process; (3) Due to the complexity of multi-field coupling nonlinear problems, numerical simulations often require high-precision numerical methods and large-scale computing resources, which places high demands on computing power and algorithms. Traditional multi-field coupling computational methods usually use finite element analysis (FEA) to establish physical models. This method can describe simple geometric shapes and physical field distributions, but it is more difficult to deal with multi-physical field coupling, nonlinear, and multi-scale problems of piezoelectric semiconductor devices that belong to piezoelectric dielectric devices. In addition, based on existing solvers, a large amount of computing resources and time are required, resulting in situations where the solution cannot be obtained. In addition, the finite element model needs to be verified by actual test data, which may be difficult to obtain or costly in complex systems. When the physical model and data are more complex, traditional physical information neural networks (PINNs) have problems such as low computational efficiency and accuracy, and inaccurate predictions in processing high-dimensional multi-field nonlinear calculations. Therefore, there is an urgent need for an efficient computational method to solve the problem of multi-physics field coupled nonlinear mechanical analysis of piezoelectric semiconductors. Summary of the invention

[0004] In view of the defects of existing solution methods, the present invention provides a piezoelectric semiconductor multi-field coupling nonlinear mechanical analysis method based on a gridded physical information neural network, which solves the problem that traditional methods cannot solve or process the complex multi-physical field nonlinearity of piezoelectric semiconductors. This further provides effective guidance for the design, optimization and application of actual devices.

[0005] In a first aspect, the present invention provides a method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device, the prediction method comprising the following steps:

[0006] Step 1: Acquire experimental or simulation data of a piezoelectric semiconductor device as a data set, wherein the data in the data set include displacement, potential, and electron concentration of the piezoelectric semiconductor.

[0007] Step 2: construct a gridded inversion fully connected neural network model; the gridded inversion fully connected neural network model includes multiple parallel physical information neural networks with the same structure. The physical information neural network includes an input layer, a hidden layer and an output layer.

[0008] Step 3: Construct the loss function of the gridded inversion fully connected neural network model, which includes the physical information error term and the data error term of each physical information neural network. Use the data set to train the gridded inversion fully connected neural network model constructed in step 2.

[0009] Step 4: Collect the local displacement, potential and electron concentration of the piezoelectric semiconductor device to be tested, and input them into the trained gridded inversion fully connected neural network model to predict the overall displacement, potential and electron concentration of the piezoelectric semiconductor device.

[0010] Preferably, in step 3, each physical information neural network loss function Loss k The expression is:

[0011]

[0012] in, is the physical information error term; is the data error term; k = 1, 2, ..., M; M is the number of parallel physical information neural networks.

[0013] The expression of the loss function Loss of the entire model is:

[0014]

[0015] Preferably, the method for obtaining the physical information error term is as follows:

[0016] According to the piezoelectric semiconductor control equation, constitutive equation and nonlinear strain-displacement equation, the three physical fields of force-electricity-carrier are coupled to construct a group of partial differential equations:

[0017] N a U a (x,y,z)-q=0

[0018] Among them, N a is the parameter matrix; U a (x, y, z) is the physical quantity matrix; q is the external load; a=1,2,...,A; A is the number of categories of the physical quantity matrix.

[0019] Obtain physical information error terms based on partial differential equations The expression is:

[0020]

[0021] Where n is the size of the data set; is the predicted physical quantity matrix.

[0022] Preferably, the data error term The method to obtain is:

[0023]

[0024] in, is the actual physical quantity matrix.

[0025] As a preference, the parameter matrix N a There is an adaptive parameter λ in the training process, which can adaptively adjust the parameter λ to minimize the physical information error term.

[0026] Preferably, in step 2, the number of hidden layers of each physical information neural network is 5, and the number of neurons in each layer is 50.

[0027] In a second aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory stores the computer program; and the processor executes the aforementioned method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device.

[0028] In a third aspect, the present invention provides a readable storage medium storing a computer program; when the computer program is executed by a processor, it is used to implement the aforementioned method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device.

[0029] The present invention has the following beneficial effects:

[0030] 1. The present invention predicts the multi-physical field coupled mechanical behavior through a gridded inversion fully connected neural network model, and has significant advantages in solving the multi-field coupled nonlinear problem of piezoelectric semiconductors. The prediction method can realize flexible modeling of complex coupling relationships through a data-driven method, cope with nonlinear effects, and achieve efficient optimization; at the same time, by adding parallel trained physical information neural networks, when the calculation problem is more complex and the amount of calculation is large, the loss function in the model training process is reduced, so that the trained model can predict the approximate solution of the problem more quickly, thereby improving the calculation efficiency and the accuracy of the prediction.

[0031] 2. The present invention predicts the mechanical behavior of multi-physical field coupling through a gridded physical information neural network, provides new ideas and tools for solving the nonlinear problems of multi-field coupling of piezoelectric semiconductors, and mines hidden patterns and laws from large amounts of data; at the same time, the present invention integrates physical laws in the process of training the model, so that the designed model has a strong generalization ability, and can improve the accuracy of the model prediction results in the absence of data, thereby enhancing the practical application value of the model designed by the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is the flow chart of the gridded physical information neural network in the present invention.

[0033] Figure 2 This is a schematic diagram of the gridded physical information neural network model in the present invention.

[0034] Figure 3 Schematic diagram of the displacement of electrons in the x direction in Example 1.

[0035] Figure 4 Schematic diagram of the displacement of electrons in the z direction in Example 1.

[0036] Figure 5 Schematic diagram of potential distribution along the x-axis in Example 1.

[0037] Figure 6 Schematic diagram of electron concentration distribution along the x-axis in Example 1. DETAILED DESCRIPTION

[0038] In order to make the purpose, method and implementation process of the present invention clearer, piezoelectric semiconductors are taken as the research object, and a three-dimensional multi-field coupled nonlinear equation group is established by combining the equilibrium equation, the electrostatic Gaussian law and the semiconductor physics equation. The multi-field coupling of piezoelectric semiconductors is studied by using meshed physics information neural networks (MPINNs). The present invention is further explained below with reference to the accompanying drawings.

[0039] Example 1

[0040] like Figure 1 As shown, a method for predicting the multi-physics field coupled mechanical behavior of a piezoelectric semiconductor device includes the following steps:

[0041] Step 1: Obtain a data set of a three-dimensional piezoelectric semiconductor circular plate through the COMSOL three-dimensional finite element model. The size of the data set is 2000; the data set includes the displacement, potential and electron concentration of the piezoelectric semiconductor structure. In order to make the prediction results more accurate, the data is normalized. The normalization formula is:

[0042]

[0043] Among them, a norm is the normalized data; a is the original data; a max and a min are the maximum and minimum values ​​in the original data respectively.

[0044] Step 2: Figure 2 As shown, a gridded inversion fully connected neural network model is constructed; the gridded inversion fully connected neural network model includes two parallel physical information neural networks with the same structure. Each physical information neural network includes an input layer, a hidden layer and an output layer, and uses the tanh function as an activation function; in each physical information neural network, the number of hidden layers is 5, and the number of neurons in each layer is 50.

[0045] Step 3: Construct the loss function of the gridded inversion fully connected neural network model. The loss function of this model includes physical information error term and data error term. Physical information error term and data error term The method to obtain is as follows:

[0046] According to the piezoelectric semiconductor control equation, constitutive equation, nonlinear strain-displacement relationship, and the relationship between electric field and electron concentration gradient, the three physical fields of force-electricity-carrier are coupled to construct a set of partial differential equations.

[0047] The governing equation for piezoelectric semiconductors is:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] Among them, T xx , Tyy , T zz , T xy , T xz , T yz is the stress component; D x ,D y , D z is the electric displacement component; J x ,J y , J z is the electron current density; ρ is the mass density; is the second-order differential of the displacement component; n is the electron concentration; is the doping concentration; q is the external load.

[0054] The constitutive equation is:

[0055] T ij =c ijkl S kl -e kij E k ,

[0056] D k =e kiji ε ij +ε kl E k ,

[0057] J k =qnμ ij E k +qD ij n ,k ,

[0058] Among them, c ijkl is the elastic stiffness coefficient; S kl is strain; E k is the electric field; is the electric potential; e ij is the piezoelectric constant; ε ij is the dielectric constant; μ ij is the electron mobility; D ij is the electron diffusion coefficient.

[0059] The nonlinear strain-displacement relationship is:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] The relationship between the electric field and the electron concentration gradient is:

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] The constructed partial differential equations are:

[0074]

[0075]

[0076]

[0077]

[0078]

[0079] Convert the constructed partial differential equations into matrix form:

[0080]

[0081] Among them, U is the physical quantity matrix; K is the parameter matrix. The physical quantity matrix U and the parameter matrix K are specifically:

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105] Introducing the adaptive parameter λ into the parameter matrix K, the original matrix form of partial differential equations is converted into:

[0106]

[0107] Where N is the parameter matrix, which is expressed as:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] By using the parameter matrix N a The adaptive parameter λ is set in the training process, which can adaptively adjust the parameter λ to minimize the physical information error term. According to the transformed partial differential equation, the physical information error term is obtained. The expression is:

[0122]

[0123] Where n is the size of the data set; is the predicted physical quantity matrix; k = 1, 2, ..., M; M is the number of parallel physical information neural networks; a = 1, 2, ..., A; A is the number of physical quantity matrix categories.

[0124] Data error term The expression is:

[0125]

[0126] in, is the actual physical quantity matrix.

[0127] Each grid loss function Loss k The expression is:

[0128]

[0129] The expression of the loss function Loss of the entire model is:

[0130]

[0131] The gridded inversion fully connected neural network model constructed in step 2 of the data set training was trained 50,000 times, and the loss function value was continuously updated using the gradient optimization algorithm. If the loss function Loss < ε, it means that the model training is completed; where ε is the preset threshold.

[0132] Step 4: Input the local displacement, potential and electron concentration of the piezoelectric semiconductor device to be tested into the trained gridded inversion fully connected neural network model to predict the displacement, potential and electron concentration of the piezoelectric semiconductor device as a whole. The distribution results of each physical quantity along the x-axis diameter are as follows: Figures 3 to 6 shown.

[0133] Step 5. Use PINNs and MPINNs to predict the overall displacement, potential and electron concentration of the piezoelectric semiconductor device respectively, and use the mean square error and prediction time to evaluate the prediction results of the two methods. The evaluation results are shown in Table 1. It can be seen from Table 1 that the method provided by the present invention not only has a higher efficiency, but also has a higher accuracy in the prediction results.

[0134] Table 1

[0135]

[0136] Example 2

[0137] A method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric dielectric device. The prediction object of this embodiment is a piezoelectric dielectric device other than a piezoelectric semiconductor device. In this embodiment, a gridded inversion fully connected neural network model with the same structure as that of embodiment 1 is used to predict the displacement and potential of the entire piezoelectric dielectric device based on the local displacement and potential of the piezoelectric dielectric device. During the prediction process, the physical information error term used in the loss function of the gridded inversion fully connected neural network model is constructed based on the displacement and potential of the piezoelectric dielectric device.

Claims

1. A method for predicting the multi-physics field coupled mechanical behavior of piezoelectric semiconductor devices. Features: The prediction method includes the following steps: Step 1, obtaining experimental or simulation data of a piezoelectric semiconductor device as a data set, wherein the data in the data set are key physical quantities of the piezoelectric semiconductor device; Step 2: construct a gridded inversion fully connected neural network model; the gridded inversion fully connected neural network model includes multiple parallel physical information neural networks with the same structure; the physical information neural network includes an input layer, a hidden layer and an output layer; Step 3: construct a loss function of the gridded inversion fully connected neural network model, the loss function includes the physical information error term and the data error term of each physical information neural network; use the data set to train the gridded inversion fully connected neural network model constructed in step 2; Step 4: Collect the local displacement, potential and electron concentration of the piezoelectric semiconductor device to be tested, and input them into the trained gridded inversion fully connected neural network model to predict the overall displacement, potential and electron concentration of the piezoelectric semiconductor device.

2. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 1, Features: In the step 1, the key physical quantities include: displacement, potential and electron concentration of the piezoelectric semiconductor device.

3. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 2, Features: In step 3, each physical information neural network loss function Loss k The expression is: in, is the physical information error term; is the data error term; k=1,2,...,M; M is the number of parallel physical information neural networks; The expression of the loss function Loss of the entire model is:

4. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 3, Features: The method for obtaining the physical information error term is as follows: According to the piezoelectric semiconductor control equation, constitutive equation and nonlinear strain-displacement equation, the three physical fields of force-electricity-carrier are coupled to construct a group of partial differential equations: NaU(x,y,z)-q=0 Where Na is the parameter matrix; Ua(x, y, z) is the physical quantity matrix; q is the external load; a=1,2,...,A; A is the number of physical quantity matrix categories; Obtain physical information error terms based on partial differential equations The expression is: Where n is the size of the data set; is the predicted physical quantity matrix.

5. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 3, Features: The data error term The method to obtain is: in, is the actual physical quantity matrix.

6. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 4, Features: The parameter matrix N a There is an adaptive parameter λ in the training process, which can adaptively adjust the parameter λ to minimize the physical information error term.

7. The method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device according to claim 2, Features: In the step 2, the number of hidden layers of each physical information neural network is 5, and the number of neurons in each layer is 50.

8. A computer device, Features: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory stores the computer program; and the processor executes the method for predicting the multi-physical field coupled mechanical behavior of a piezoelectric semiconductor device as described in claim 1.

9. A readable storage medium, Features: A computer program is stored; when the computer program is executed by a processor, it is used to implement the multi-physical field coupled mechanical behavior prediction method of a piezoelectric semiconductor device as described in claim 1.

Citation Information

Patent Citations

  • Deep learning-based multi-physics field model coupling type solving method

    CN115146527A

Cited By

  • Multi-physics field TCAD proxy modeling method

    CN121525613A