Method for predicting nonlinear behavior of composite material under thermal load based on neural network
Through the prediction method based on neural network, the composite material constitutive model is constructed using finite element model analysis and data set training, which solves the problem of complex and time-consuming traditional methods and low prediction accuracy, and realizes the nonlinear behavior prediction under the thermal load of composite materials with high precision.
Patent Information
- Application Number
- CN202510038699.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-10
- Publication Date
- 2025-05-06
AI Technical Summary
The prediction method of nonlinear behavior under thermal load of traditional composite materials is more complex and time-consuming, and the prediction accuracy is difficult to guarantee.
The prediction method based on neural network is adopted, and the temperature-stress-strain curve obtained by tensile heat load test is analyzed for finite element model, the data set is generated, and the target parameters are obtained by using neural network training to construct a composite material constitutive model based on neural network.
This method can accurately simulate the constitutive behavior of complex nonlinear composite materials, avoid the prior assumptions in traditional methods and simplify the model, improve prediction accuracy, and reduce computational costs.
Smart Images
Figure CN119940016A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of artificial intelligence and composite materials, and in particular to a prediction method for nonlinear behavior of composite materials under thermal load based on a neural network. Background Art
[0002] With the rapid development of materials science and engineering technology, composite materials have been widely used in aerospace, automobile manufacturing, construction engineering and other fields due to their excellent performance. Among them, due to the difference in thermal expansion coefficient between fiber and matrix material, thermal load will cause non-uniform stress and strain in composite materials. This non-uniformity makes the mechanical response of composite materials show nonlinear characteristics. Therefore, accurately predicting the nonlinear behavior of composite materials under thermal load is of great significance to improve their design efficiency and performance prediction accuracy.
[0003] In most cases, the behavior of materials at high temperatures and strain rates is highly nonlinear, and the effects of many factors on flow stress are also nonlinear, which reduces the accuracy of the commonly used regression method prediction and limits the application field. In addition, the selection, development and numerical implementation of constitutive equations are time-consuming. The prediction of nonlinear behavior of traditional composite materials under thermal loads is mostly mathematical modeling methods, which usually require prior assumptions on the analytical form of the behavior law and the establishment of a conceptualized mathematical calculation model. This method is relatively complex and time-consuming, and the prediction accuracy is difficult to guarantee.
[0004] Therefore, there is an urgent need to provide a neural network-based prediction method for the nonlinear behavior of composite materials under thermal loads. Summary of the invention
[0005] In order to solve the problem that traditional modeling methods are complex and time-consuming to generate, and the prediction accuracy is difficult to ensure, an embodiment of the present invention provides a prediction method for nonlinear behavior of composite materials under thermal load based on a neural network.
[0006] In one aspect, a method for predicting nonlinear behavior of composite materials under thermal load based on a neural network is provided, the method comprising:
[0007] Based on the temperature-stress-strain curve obtained from the tensile thermal load test, the finite element model analysis of the composite specimen is carried out to extract the true temperature, stress, strain and stress increment, strain increment and temperature increment of the specimen at each analysis step to generate a data set;
[0008] Using the data set to train a pre-built neural network, so as to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model;
[0009] Based on the target parameters, a material constitutive subroutine is written to construct a composite material constitutive model based on a neural network.
[0010] On the other hand, a device for predicting nonlinear behavior of composite materials under thermal load based on a neural network is provided, which is used to implement the steps described in any method embodiment of the specification, and the device comprises:
[0011] An analysis unit is used to perform finite element model analysis on the composite specimen based on the temperature-stress-strain curve obtained from the tensile thermal load test, extract the true temperature, stress, strain and stress increment, strain increment and temperature increment of the specimen at each analysis step, and generate a data set;
[0012] A training unit, used to train a pre-built neural network using the data set, so as to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model;
[0013] The construction unit is used to write a material constitutive subroutine based on the target parameters and construct a composite material constitutive model based on a neural network.
[0014] On the other hand, a computer device is provided, which includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of the above-mentioned method.
[0015] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned method are implemented.
[0016] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program implements the steps of the above method when executed by a processor.
[0017] The technical solution provided by the present invention can at least bring the following beneficial effects:
[0018] The use of artificial neural network technology can infinitely approach the mapping of any continuous function. This feature can accurately simulate the complex nonlinear constitutive behavior of composite materials. Unlike traditional modeling methods, the artificial neural network constitutive model is completely driven by experimental data and abandons any prior assumptions or simplified models, thus ensuring that the model output is not restricted by the specific temperature-stress-strain curve shape. It not only saves time and effort, but also makes the constitutive model more objective and accurate in reflecting the response characteristics of composite materials. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0020] Figure 1 It is a flow chart of a method for predicting nonlinear behavior of composite materials under thermal load based on a neural network provided by one embodiment of the present invention;
[0021] Figure 2 A temperature-stress-strain curve obtained from a composite material tensile thermal load test provided by an embodiment of the present invention;
[0022] Figure 3 is a structural diagram of a neural network provided by an embodiment of the present invention;
[0023] Figure 4 This is a graph of prediction results of a composite material constitutive model based on a neural network provided by an embodiment of the present invention;
[0024] Figure 5 It is an error rate between a training set and a test set in a neural network learning process provided by an embodiment of the present invention;
[0025] Figure 6 It is a result comparison diagram of a predicted value and an experimental value provided by an embodiment of the present invention;
[0026] Figure 7 It is a structural diagram of a prediction device for nonlinear behavior of composite materials under thermal load based on a neural network provided by one embodiment of the present invention;
[0027] Figure 8 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0028] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0029] The specific implementation of the above concept is described below.
[0030] Please refer to Figure 1, an embodiment of the present invention provides a method for predicting nonlinear behavior of composite materials under thermal load based on a neural network, the method comprising:
[0031] Step 100: Based on the temperature-stress-strain curve obtained from the tensile thermal load test, a finite element model analysis is performed on the composite material specimen, and the true temperature, stress, strain, stress increment, strain increment and temperature increment of the specimen at each analysis step are extracted to generate a data set;
[0032] Step 102: using the data set to train a pre-built neural network, so as to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model;
[0033] Step 104: Based on the target parameters, write a material constitutive subroutine to construct a composite material constitutive model based on a neural network.
[0034] In the embodiment of the present invention, the artificial neural network technology is used to infinitely approach the mapping of any continuous function. This characteristic can accurately simulate the complex nonlinear constitutive behavior of composite materials. Unlike traditional modeling methods, the artificial neural network constitutive model is completely driven by experimental data and abandons any prior assumptions or simplified models, thereby ensuring that the model output is not restricted by the specific temperature-stress-strain curve shape, which not only saves time and effort, but also makes the constitutive model more objective and accurate in reflecting the response characteristics of composite materials.
[0035] Described below Figure 1 How the various steps are performed.
[0036] For step 100:
[0037] In some implementations, step 100 may include:
[0038] Establish a uniaxial tensile test of composite materials under thermal load, record the true stress and strain of the material at each temperature, and obtain the temperature-stress-strain curve;
[0039] A finite element model is established based on the composite material specimen, and the temperature-stress-strain curve is input for finite element calculation, so as to apply temperature-stress load combinations to the finite element model in sequence; wherein, a uniformly spaced stress load is applied to one end of the finite element model;
[0040] For each temperature-stress load combination, perform:
[0041] According to the finite element calculation results, the reaction force and displacement of the finite element model are extracted, and the stress and strain at each analysis step under the current temperature condition are calculated;
[0042] Calculate the stress increment, strain increment and temperature increment of the finite element model at each analysis step under the current temperature-stress load combination, and generate a training sample;
[0043] The data set is generated using the stress increment, strain increment and temperature increment of the finite element model at each analysis step under each temperature-stress load combination.
[0044] In this embodiment, reference can be made to Figure 2 The schematic diagram of the temperature-stress-strain curve is shown in Figure 1. The finite element model of the composite specimen is shown in Figure 2. Figure 4 shown.
[0045] The stress increment Δσ, strain increment Δε and temperature increment ΔT of the finite element model at each analysis step are calculated using the following formula:
[0046] Δε=ε t -ε t-1
[0047] Δσ=σ t -σ t-1
[0048] ΔT=T t -T t-1
[0049] Where: t , ε t and T t is the stress, strain and temperature of the finite element model in the current analysis step; σ t-1 , ε t-1 and T t-1 are the stress, strain, and temperature of the finite element model in the previous analysis step.
[0050] In this embodiment, a data set is established through tensile thermal load testing and finite element analysis, and a neural network is trained using the data set. The neural network is directly driven by temperature-stress-strain experimental data to output target parameters for constructing a constitutive model.
[0051] Regarding step 102:
[0052] In some embodiments, the neural network model is trained as follows:
[0053] Get the training set obtained by splitting the data set;
[0054] Based on the strain increment of each training sample in each analysis step in the training set, the plastic strain rate in each analysis step is generated, and the natural logarithm of the ratio of the plastic strain rate to the reference strain rate is used to preprocess the plastic strain rate in each analysis step;
[0055] The stress increment of each training sample in each analysis step in the training set is used as a label, and each training sample is normalized to generate input data;
[0056] Initialize the neural network, the output layer and the hidden layer are initialized separately, and the bias items are initialized using random distribution;
[0057] The initialized neural network is trained and optimized using the input data and labels of each training sample in the training set until a neural network model that meets expectations is obtained.
[0058] In this embodiment, since the plastic deformation rate in the constitutive equation usually enhances the logarithm of the plastic strain rate, the plastic strain rate is preprocessed by calculating the natural logarithm of the ratio of the plastic strain rate to the reference strain rate to obtain the stress. The plastic strain rate is preprocessed by the following formula:
[0059]
[0060] In the formula, is the plastic strain rate, is the reference strain rate, σ1 is the stress obtained after preprocessing in the current analysis step;
[0061] The input data is obtained as follows:
[0062]
[0063] In the formula, ε p is the plastic strain of the current analysis step, T is the temperature of the current analysis step, is the input data of the current analysis step in the training sample, which consists of three parts: x1, x2, and x3. x1, x2, and x3 are normalized values.
[0064] In some embodiments, reference may be made to Figure 3 The neural network model includes an input layer, a plurality of hidden layers and an output layer; wherein the number of hidden layers is equal to the number of incremental steps in the finite element model analysis, the number of incremental steps is the number of analysis steps minus one, the hidden layers and the incremental steps are in a one-to-one correspondence, and the first hidden layer corresponds to the first analysis step;
[0065] The hidden layer is used to perform chain derivation of the output data based on the input data, weight terms and bias terms of the current analysis step, and superimpose with the previous hidden layer to obtain the final output data. Finally, the flow stress of the current analysis step is calculated based on the final output data; wherein the output data consists of the derivative of plastic strain, the derivative of plastic strain rate and the derivative of temperature.
[0066] In this embodiment, the analysis step numbering starts from 0, which is the initial analysis step, number 1 is the first analysis step, and so on. The hidden layers and incremental steps are both numbered from 1, so the first hidden layer corresponds to the first analysis step and the first incremental step.
[0067] In the embodiment of the present invention, the hidden layer is chain-derived in the following manner:
[0068]
[0069] In the formula, is the output data, W (k) is the weight matrix of the kth hidden layer, is the element in the weight term matrix, is the product of the elements, called the Hadamard product, and tanh is the activation function. It is an intermediate variable and has no real meaning. is the input data for the current analysis step, is the bias matrix of the kth hidden layer.
[0070] in,
[0071]
[0072] Where k is the number of incremental steps in finite element analysis; n is the number of neural network input nodes; and m is the number of neural network output nodes.
[0073] Then superimpose it with the previous hidden layer to get the final output data:
[0074]
[0075] Since the flow stress is to be output in the end, it is necessary to define the output data By input The output data consists of the derivative of plastic strain, the derivative of plastic strain rate and the derivative of temperature.
[0076]
[0077] Finally, the flow stress of the current analysis step is calculated based on the final output data:
[0078]
[0079] Where Δt is the time increment, η is the Taylor-Quinney coefficient that defines the conversion of plastic work into heat energy, and C p is the specific heat coefficient, and ρ is the material density.
[0080] Then, the flow stress of each analysis step output by each hidden layer is denormalized to obtain the von Mises equivalent flow stress:
[0081]
[0082] Therefore, the trained neural network model finally outputs the von Mises equivalent flow stress of each incremental step, and then denormalizes it to obtain the final von Mises equivalent flow stress, which is the target parameter used to construct the constitutive model of the composite material.
[0083] Regarding step 104:
[0084] In this step, all matrix products are explicitly written in the FORTRAN subroutine using a Python interface loop, the main part of the built-in constitutive law is used to time-integrate the final von Mises equivalent flow stress in a given time increment, and the user's FORTRAN subroutine is written to complete the implementation of the flow law to obtain the constitutive model of the composite material.
[0085] The constitutive model can be directly used to predict the nonlinear behavior of the composite material under thermal load. The constitutive model is verified using a test set. The prediction results of the constitutive model can be found in Figure 4 , Figure 5 and Figure 6 It can be seen that the prediction accuracy of the composite material constitutive model based on the neural network in the embodiment of the present invention is high and is not restricted by the specific temperature-stress-strain curve shape.
[0086] Please refer to Figure 7 The embodiment of the present invention provides a prediction device for nonlinear behavior of composite materials under thermal load based on a neural network, which is used to implement the steps of any method embodiment in the specification, and the device includes:
[0087] The analysis unit 701 is used to perform finite element model analysis on the composite material specimen based on the temperature-stress-strain curve obtained from the tensile thermal load test, extract the true temperature, stress, strain and stress increment, strain increment and temperature increment of the specimen at each analysis step, and generate a data set;
[0088] A training unit 702 is used to train a pre-built neural network using a data set to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model;
[0089] The construction unit 703 is used to write a material constitutive subroutine based on the target parameters and construct a composite material constitutive model based on a neural network.
[0090] In one embodiment of the present invention, the analysis unit 701 is used to perform:
[0091] Establish a uniaxial tensile test of composite materials under thermal load, record the true stress and strain of the material at each temperature, and obtain the temperature-stress-strain curve;
[0092] A finite element model is established based on the composite material specimen, and the temperature-stress-strain curve is input for finite element calculation, so as to apply temperature-stress load combinations to the finite element model in sequence; wherein, a uniformly spaced stress load is applied to one end of the finite element model;
[0093] For each temperature-stress load combination, perform:
[0094] According to the finite element calculation results, the reaction force and displacement of the finite element model are extracted, and the stress and strain at each analysis step under the current temperature condition are calculated;
[0095] Calculate the stress increment, strain increment and temperature increment of the finite element model at each analysis step under the current temperature-stress load combination, and generate a training sample;
[0096] The data set is generated using the stress increment, strain increment and temperature increment of the finite element model at each analysis step under each temperature-stress load combination.
[0097] In one embodiment of the present invention, the neural network model in the training unit 702 is trained in the following manner:
[0098] Get the training set obtained by splitting the data set;
[0099] Based on the strain increment of each training sample in each analysis step in the training set, the plastic strain rate in each analysis step is generated, and the natural logarithm of the ratio of the plastic strain rate to the reference strain rate is used to preprocess the plastic strain rate in each analysis step;
[0100] The stress increment of each training sample in each analysis step in the training set is used as a label, and each training sample is normalized to generate input data;
[0101] Initialize the neural network, the output layer and the hidden layer are initialized separately, and the bias items are initialized using random distribution;
[0102] The initialized neural network is trained and optimized using the input data and labels of each training sample in the training set until a neural network model that meets expectations is obtained.
[0103] In one embodiment of the present invention, the plastic strain rate in the training unit 702 is preprocessed by the following formula:
[0104]
[0105] In the formula, is the plastic strain rate, is the reference strain rate, σ1 is the stress obtained after preprocessing in the current analysis step;
[0106] The input data is obtained as follows:
[0107]
[0108] In the formula, ε p is the plastic strain of the current analysis step, T is the temperature of the current analysis step, is the input data of the current analysis step in the training sample, which consists of three parts: x1, x2, and x3. x1, x2, and x3 are normalized values.
[0109] In one embodiment of the present invention, the neural network model in the training unit 702 includes an input layer, a plurality of hidden layers and an output layer; wherein the number of hidden layers is equal to the number of incremental steps in the finite element model analysis, the number of incremental steps is the number of analysis steps minus one, the hidden layers and the incremental steps are in a one-to-one correspondence, and the first hidden layer corresponds to the first analysis step;
[0110] The hidden layer is used to perform chain derivation of the output data based on the input data, weight terms and bias terms of the current analysis step, and superimpose with the previous hidden layer to obtain the final output data. Finally, the flow stress of the current analysis step is calculated based on the final output data; wherein the output data consists of the derivative of plastic strain, the derivative of plastic strain rate and the derivative of temperature.
[0111] In one embodiment of the present invention, the hidden layer in the training unit 702 is chain-derived in the following manner:
[0112]
[0113] In the formula, is the output data, W (k) is the weight matrix of the kth hidden layer, is the element in the weight term matrix, is the product of the elements, called the Hadamard product, and tanh is the activation function. It is an intermediate variable and has no real meaning. is the input data for the current analysis step, is the bias matrix of the kth hidden layer.
[0114] It should be noted that the prediction device for the nonlinear behavior of composite materials under thermal load based on neural network provided in the above embodiment is only illustrated by the division of the above functional units. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the device can be divided into different functional units to complete all or part of the functions described above. In addition, the above device embodiment and the method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.
[0115] The embodiment of the present application also provides a computer device, please refer to Figure 8 The computer device includes a processor and a memory, wherein at least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the prediction method of the nonlinear behavior of the composite material under thermal load based on the neural network provided in the above-mentioned method embodiments.
[0116] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the neural network-based prediction method for nonlinear behavior of composite materials under thermal loads provided in the above-mentioned method embodiments.
[0117] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes any of the neural network-based methods for predicting nonlinear behavior of composite materials under thermal loads in the above embodiments.
[0118] For the convenience of description, the above system or device is described as being divided into various modules or units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.
[0119] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application or some parts of the embodiments.
[0120] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0121] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.
Claims
1. A method for predicting the nonlinear behavior of composite materials under thermal load based on neural network, characterized in that: The method comprises: Based on the temperature-stress-strain curve obtained from the tensile thermal load test, the finite element model analysis of the composite specimen is carried out to extract the true temperature, stress, strain and stress increment, strain increment and temperature increment of the specimen at each analysis step to generate a data set; Using the data set to train a pre-built neural network, so as to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model; Based on the target parameters, a material constitutive subroutine is written to construct a composite material constitutive model based on a neural network.
2. The method according to claim 1, characterized in that Based on the temperature-stress-strain curve obtained from the tensile heat load test, the composite specimen is subjected to finite element model analysis, and the true temperature, stress, strain, stress increment, strain increment and temperature increment of the specimen at each analysis step are extracted to generate a data set, including: Establish a uniaxial tensile test of composite materials under thermal load, record the true stress and strain of the material at each temperature, and obtain the temperature-stress-strain curve; A finite element model is established based on the composite material specimen, and a temperature-stress-strain curve is input to perform finite element calculation, so as to sequentially apply a temperature-stress load combination to the finite element model; wherein a uniformly spaced stress load is applied to one end of the finite element model; For each temperature-stress load combination, perform: According to the finite element calculation results, the reaction force and displacement of the finite element model are extracted, and the stress and strain at each analysis step under the current temperature condition are calculated; Calculating the stress increment, strain increment and temperature increment of the finite element model at each analysis step under the current temperature-stress load combination to generate a training sample; A data set is generated using the stress increment, strain increment, and temperature increment of the finite element model at each analysis step under each temperature-stress load combination.
3. The method according to claim 1, characterized in that The neural network model is trained in the following way: Obtaining a training set obtained by splitting the data set; Based on the strain increment of each training sample in each analysis step in the training set, the plastic strain rate in each analysis step is generated, and the plastic strain rate in each analysis step is preprocessed by using the natural logarithm of the ratio of the plastic strain rate to the reference strain rate; The stress increment of each training sample in each analysis step in the training set is used as a label, and each training sample is normalized to generate input data; Initialize the neural network, the output layer and the hidden layer are initialized separately, and the bias items are initialized using random distribution; The initialized neural network is trained and optimized using the input data and labels of each training sample in the training set until a neural network model that meets expectations is obtained.
4. The method according to claim 3, characterized in that The plastic strain rate is preprocessed by the following formula: In the formula, is the plastic strain rate, is the reference strain rate, σ1 is the stress obtained after preprocessing in the current analysis step; The input data is obtained as follows: In the formula, ε p is the plastic strain of the current analysis step, T is the temperature of the current analysis step, is the input data of the current analysis step in the training sample, which consists of three parts: x1, x2, and x3. x1, x2, and x3 are normalized values.
5. The method according to claim 4, characterized in that The neural network model includes an input layer, a plurality of hidden layers and an output layer; wherein the number of the hidden layers is equal to the number of incremental steps in the finite element model analysis, the number of incremental steps is the number of analysis steps minus one, the hidden layers and incremental steps are in a one-to-one correspondence, and the first hidden layer corresponds to the first analysis step; The hidden layer is used to perform chain derivation on the output data based on the input data, weight term and bias term of the current analysis step, and superimpose with the previous hidden layer to obtain the final output data, and finally calculate the flow stress of the current analysis step based on the final output data; wherein the output data consists of the derivative of plastic strain, the derivative of plastic strain rate and the derivative of temperature.
6. The method according to claim 5, characterized in that The hidden layer is chain-derived as follows: In the formula, is the output data, W (k) is the weight matrix of the kth hidden layer, is the element in the weight term matrix, is the product of the elements, called the Hadamard product, and tanh is the activation function. It is an intermediate variable and has no real meaning. is the input data for the current analysis step, is the bias matrix of the kth hidden layer.
7. A prediction device for nonlinear behavior of composite materials under thermal load based on neural network, used to implement the steps of any method described in claims 1 to 6, characterized in that: The device comprises: An analysis unit is used to perform finite element model analysis on the composite specimen based on the temperature-stress-strain curve obtained from the tensile thermal load test, extract the true temperature, stress, strain and stress increment, strain increment and temperature increment of the specimen at each analysis step, and generate a data set; A training unit, used to train a pre-built neural network using the data set, so as to obtain target parameters for constructing a constitutive model of the composite material from the trained neural network model; The construction unit is used to write a material constitutive subroutine based on the target parameters and construct a composite material constitutive model based on a neural network.
8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-6.
9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.
10. A computer program product, characterized in that The method comprises a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.