Method, device and equipment for predicting compression limit load of damaged laminated plate
By constructing a data sample library and training a neural network model, the problems of high cost and long time consumption in predicting the compressive limit load of damaged composite laminates in existing technologies have been solved, and fast and economical load prediction and performance evaluation have been achieved.
Patent Information
- Application Number
- CN202410969390.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-19
- Publication Date
- 2024-11-08
AI Technical Summary
Existing finite element methods are costly and time-consuming in solving for the compressive limit load of damaged composite laminates, making it difficult to quickly assess the remaining mechanical properties and provide guidance on repair strategies.
A neural network model is trained using a data sample library. By acquiring damage data of laminated plates, the neural network model is used to predict the compressive limit load. This includes establishing a three-dimensional finite element model, material failure determination and parameter degradation model, constructing a data sample library, and using the neural network for training and evaluation.
It enables rapid prediction of the compressive limit load of damaged laminates, reduces costs and time consumption, and provides support for determining the repair tolerance of composite laminates.
Smart Images

Figure CN118917186A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of load prediction, and in particular to a method, device and equipment for predicting the compressive limit load of a damaged laminate. Background Art
[0002] Carbon fiber resin matrix composite laminates are widely used in aerospace engineering structures due to their excellent mechanical properties. Composite materials will inevitably suffer some damage during their service. It is usually necessary to evaluate the residual mechanical properties based on the degree of damage of the composite laminates, determine the damage tolerance and select appropriate patching and repair strategies.
[0003] However, the use of traditional test methods or finite element calculation methods to evaluate mechanical properties requires a lot of time and economic costs, and cannot quickly evaluate the residual mechanical properties of damaged composite laminates and provide guidance for subsequent repair strategies. It is necessary to develop a rapid prediction method for the ultimate bearing capacity of damaged composite laminates, monitor the residual mechanical properties of damaged composite laminates in real time, and provide data support for subsequent damage tolerance judgment and patch repair strategy selection. Due to the complexity of composite compression failure, establishing a numerical method that fully and accurately simulates the compression process of composite laminates usually requires a lot of time to correct and fit the model, and the failure theory is complex and has many influencing factors, making it difficult to establish a unified prediction model.
[0004] In summary, the existing finite element method for solving the compressive limit load of damaged composite laminates has the problems of high cost and long time consumption. Summary of the invention
[0005] The purpose of the present application is to provide a method, device and equipment for predicting the compressive limit load of damaged laminates, which can reduce the cost and time of predicting the compressive limit load of damaged laminates and realize rapid prediction of the compressive limit load of damaged laminates.
[0006] To achieve the above objectives, this application provides the following solutions:
[0007] In a first aspect, the present application provides a method for predicting the compressive limit load of a damaged laminate, comprising:
[0008] Acquire damage data of a target laminate; the target laminate is a damaged laminate whose compressive limit load is to be predicted; the damage data includes: coordinates of the center position of the damage and the damage depth;
[0009] Inputting damage data of the target laminate into a compressive limit load prediction model to obtain the compressive limit load of the target laminate;
[0010] Among them, the compression limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compression limit loads; the compression limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software.
[0011] Optionally, the method for determining the compression limit load prediction model specifically includes:
[0012] Obtain damage data of laminates with different damage types;
[0013] Finite element software is secondary developed by writing Python scripts to batch calculate the compressive limit loads of laminates with different damage types; wherein, the calculation process of the compressive limit load of a laminate of any damage type is as follows: a three-dimensional finite element model of the laminate of the damage type is established; based on the three-dimensional finite element model, the compressive failure process of the laminate of the damage type is simulated by using material failure judgment criteria and material parameter degradation model to determine the failure mode and material parameter degradation mode of the laminate of the damage type; the compressive limit load of the laminate of the damage type is determined according to the failure mode and the material parameter degradation mode;
[0014] Construct a data sample library based on the damage data of laminates with different damage types and the corresponding compressive limit loads;
[0015] Dividing the data sample library into a training set and a test set according to a set ratio;
[0016] Using the training set to train the neural network model to obtain a trained neural network model;
[0017] The test set is used to perform performance evaluation on the trained neural network model to determine the compression limit load prediction model.
[0018] Optionally, the material failure determination criteria include:
[0019] If satisfied Then the failure mode is determined to be fiber tensile failure;
[0020] If satisfied Then the failure mode is determined to be fiber compression failure;
[0021] If satisfied Then the failure mode is determined to be matrix tensile failure;
[0022] If satisfied Then the failure mode is determined to be matrix compression failure;
[0023] If satisfied Then the failure mode is determined to be matrix-fiber shear failure;
[0024] If satisfied Then the failure mode is determined to be delamination tensile failure;
[0025] If satisfied Then the failure mode is determined to be layered compression failure;
[0026] Among them, σ1 is the normal stress in the fiber direction; σ2 is the normal stress in the matrix direction; σ3 is the normal stress in the normal direction; τ 12 is the shear stress component in the fiber direction; τ 13 is the shear stress component in the matrix direction; τ 23 is the shear stress component in the normal direction; X T is the tensile strength in the fiber direction; X C is the compressive strength in the fiber direction; Y T is the tensile strength in the matrix direction; Y C is the compressive strength in the matrix direction; Z T is the tensile strength in the normal direction; Z C is the compressive strength in the normal direction; S 12 S is the shear strength of the material in the fiber direction; 13 S is the shear strength value of the material in the matrix direction; 23 is the shear strength value of the material in the normal direction.
[0027] Optionally, the material parameter degradation model includes:
[0028] If the failure mode is fiber tensile failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.07 times the initial value; where E 11 is the elastic modulus in the fiber direction; E 22 is the elastic modulus in the matrix direction; E 33 is the elastic modulus in the normal direction; v 12 is the Poisson's ratio in the fiber direction; v 13 is the Poisson's ratio in the matrix direction; v 23 is the Poisson's ratio in the normal direction; G 12 is the shear elastic modulus in the fiber direction; G 13 is the shear elastic modulus in the matrix direction; G 23is the shear elastic modulus in the normal direction;
[0029] If the failure mode is fiber compression failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.14 times the initial value;
[0030] If the failure mode is matrix tensile failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.2 times the initial value;
[0031] If the failure mode is matrix compression failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.4 times the initial value;
[0032] If the failure mode is matrix-fiber shear failure, the material parameter degradation mode is: G 12 、v 12 Reduce to 0;
[0033] If the failure mode is delamination tensile failure or delamination compression failure, the material parameter degradation mode is: E 33 、v 13 、v 23 , G 13 , G 23 Reduced to 0.
[0034] Optionally, the training set is used to train the neural network model to obtain a trained neural network model, specifically including:
[0035] The training set is input into the neural network model, and the error back propagation algorithm and the Adam optimization algorithm are used to train the hyperparameters in the neural network model with the goal of minimizing the mean square error loss function to obtain a trained neural network model.
[0036] Optionally, using the test set to perform performance evaluation on the trained neural network model to determine the compressive limit load prediction model specifically includes:
[0037] Inputting the test set into the trained neural network model, using the score function and the root mean square error as performance evaluation indicators, and adjusting the hyperparameters in the trained neural network model;
[0038] The adjusted neural network model is determined as the compression limit load prediction model.
[0039] Optionally, the neural network model is a long short-term memory network model.
[0040] Optionally, before using the training set to train the neural network model to obtain the trained neural network model, the method further includes:
[0041] The data in the training set and the test set are normalized.
[0042] In a second aspect, the present application provides a device for predicting the compressive limit load of a damaged laminate, comprising:
[0043] A damage data acquisition module is used to acquire damage data of a target laminate; the target laminate is a damaged laminate whose compressive limit load is to be predicted; the damage data includes: coordinates of the center position of the damage and the damage depth;
[0044] A compression limit load prediction module, used for inputting damage data of a target laminate into a compression limit load prediction model to obtain the compression limit load of the target laminate;
[0045] Among them, the compression limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compression limit loads; the compression limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software.
[0046] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for predicting the compressive limit load of a damaged laminate.
[0047] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0048] The present application provides a method, device and equipment for predicting the compressive limit load of a damaged laminate. The compressive limit load prediction model is obtained by training a neural network model with a data sample library. The compressive limit load prediction model is used to predict the compressive limit load of the damaged laminate. Compared with the method of using a finite element calculation method to calculate the compressive limit load, the method can reduce the cost and time of predicting the compressive limit load of the damaged laminate, and achieve rapid prediction of the compressive limit load of the damaged laminate, providing a large amount of material for engineering performance evaluation of aircraft composite components, and further providing support for the judgment of the repair tolerance of composite laminates. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0050] Figure 1 This is an application environment diagram of a method for predicting the compressive limit load of a damaged laminate in one embodiment of the present application;
[0051] Figure 2 A schematic flow chart of a method for predicting the compressive limit load of a damaged laminate provided in one embodiment of the present application;
[0052] Figure 3 A detailed flow chart of a method for determining a compression limit load prediction model;
[0053] Figure 4 A schematic diagram of the complete process of the actual application of the method for predicting the compressive limit load of a damaged laminate provided by another embodiment of the present application;
[0054] Figure 5 A schematic diagram of a finite element model of a damaged composite laminate provided in another embodiment of the present application;
[0055] Figure 6 A schematic diagram of a recurrent neural network structure provided in another embodiment of the present application;
[0056] Figure 7 A schematic diagram of a neural network structure without hidden states provided in another embodiment of the present application;
[0057] Figure 8 A schematic diagram of a neural network model training process provided by another embodiment of the present application;
[0058] Fig. 9A comparison diagram of the predicted value of the compression limit load and the finite element simulation result provided in another embodiment of the present application;
[0059] Fig.10 A structural diagram of a device for predicting the compressive limit load of a damaged laminate provided in one embodiment of the present application;
[0060] Fig.11 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0061] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0062] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0063] The method for predicting the compressive limit load of a damaged laminate provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the damage data of the target laminate (that is, the damage data of the damaged laminate whose compressive limit load is to be predicted) to the server 104, and the server 104 receives the damage data of the target laminate. For the damage data of the target laminate, the server 104 inputs the damage data of the target laminate into the compressive limit load prediction model to obtain the compressive limit load of the target laminate; wherein, the compressive limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compressive limit loads; the compressive limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software. The server 104 can feedback the obtained compressive limit load of the target laminate to the terminal 102.
[0064] In addition, in some embodiments, the method for predicting the compressive limit load of a damaged laminate can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly predict the compressive limit load based on the damage data of the target laminate, or the server 104 can obtain the damage data of the target laminate from the data storage system and predict the compressive limit load based on the damage data of the target laminate.
[0065] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.
[0066] In an exemplary embodiment, Figure 2 As shown, a method for predicting the compressive limit load of a damaged laminate is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 to 202.
[0067] in:
[0068] Step 201, obtaining damage data of a target laminate; the target laminate is a damaged laminate whose compressive limit load is to be predicted; the damage data includes: coordinates of the center position of the damage and the damage depth.
[0069] Step 202, inputting the damage data of the target laminate into a compressive limit load prediction model to obtain the compressive limit load of the target laminate; wherein the compressive limit load prediction model is obtained by training a neural network model using a data sample library.
[0070] The data sample library includes: damage data of laminates with different damage types and corresponding compressive limit loads; the compressive limit loads in the data sample library are determined by simulating the compression failure process of laminates with different damage types using finite element software.
[0071] Implementing the above steps 201 to 202 can reduce the cost and time consumption of predicting the compressive limit load of the damaged laminate, and achieve rapid prediction of the compressive limit load of the damaged laminate.
[0072] In another exemplary embodiment of the present application, Figure 3As shown, the method for determining the compression limit load prediction model in the above step 202 is replaced by the following steps 301 to 305:
[0073] Step 301, obtaining damage data of laminates with different damage types.
[0074] Step 302, secondary development of finite element software by writing Python scripts to batch calculate the compressive limit loads of laminates with different damage types.
[0075] Among them, the calculation process of the compressive limit load of a laminate with any damage type is: establish a three-dimensional finite element model of the laminate with this damage type; based on the three-dimensional finite element model, use the material failure judgment criterion and the material parameter degradation model to simulate the compressive failure process of the laminate with this damage type, and determine the failure mode and material parameter degradation mode of the laminate with this damage type; determine the compressive limit load of the laminate with this damage type according to the failure mode and the material parameter degradation mode.
[0076] Step 303: construct a data sample library according to the damage data of laminates with different damage types and the corresponding compressive limit loads (as output labels).
[0077] Step 304: divide the data sample library into a training set and a test set according to a set ratio.
[0078] Step 305: Use the training set to train the neural network model to obtain the trained neural network model; use the test set to perform performance evaluation on the trained neural network model to determine the compression limit load prediction model.
[0079] In another exemplary embodiment of the present application, in the above step 202, the three-dimensional Hashin failure criterion is selected as the material failure determination criterion, and the material failure determination criterion includes:
[0080] If satisfied (as the failure discriminant), the failure mode is determined to be fiber tensile failure (σ1>0).
[0081] If satisfied The failure mode is determined to be fiber compression failure (σ1<0).
[0082] If satisfied The failure mode is determined to be matrix tensile failure (σ2>0).
[0083] If satisfied The failure mode is determined to be matrix compression failure (σ2<0).
[0084] If satisfied The failure mode is determined to be matrix-fiber shear failure (σ1<0).
[0085] If satisfied The failure mode is determined to be delamination tensile failure (σ3>0).
[0086] If satisfied The failure mode is determined to be delamination compression failure (σ3<0).
[0087] Among them, σ1 is the normal stress in the fiber direction; σ2 is the normal stress in the matrix direction; σ3 is the normal stress in the normal direction; τ 12 is the shear stress component in the fiber direction; τ 13 is the shear stress component in the matrix direction; τ 23 is the shear stress component in the normal direction; X T is the tensile strength in the fiber direction; X C is the compressive strength in the fiber direction; Y T is the tensile strength in the matrix direction; Y C is the compressive strength in the matrix direction; Z T is the tensile strength in the normal direction; Z C is the compressive strength in the normal direction; S 12 S is the shear strength of the material in the fiber direction; 13 S is the shear strength value of the material in the matrix direction; 23 is the shear strength value of the material in the normal direction.
[0088] The Camanho degradation model is selected as the material parameter degradation model, and the material parameter degradation model includes:
[0089] If the failure mode is fiber tensile failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.07 times the initial value; where E 11 is the elastic modulus in the fiber direction; E 22 is the elastic modulus in the matrix direction; E 33 is the elastic modulus in the normal direction; v 12 is the Poisson's ratio in the fiber direction; ν 13 is the Poisson's ratio in the matrix direction; v 23 is the Poisson's ratio in the normal direction; G 12 is the shear elastic modulus in the fiber direction; G 13is the shear elastic modulus in the matrix direction; G 23 is the shear elastic modulus in the normal direction.
[0090] If the failure mode is fiber compression failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.14 times the initial value.
[0091] If the failure mode is matrix tensile failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.2 times the initial value.
[0092] If the failure mode is matrix compression failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.4 times the initial value.
[0093] If the failure mode is matrix-fiber shear failure, the material parameter degradation mode is: G 12 、v 12 Reduced to 0.
[0094] If the failure mode is delamination tensile failure or delamination compression failure, the material parameter degradation mode is: E 33 、v 13 , ν 23 , G 13 , G 23 Reduced to 0.
[0095] In the calculation process, in order to prevent the occurrence of singular stiffness matrix, the material properties that degenerate to 0 are replaced by degenerating to a very small value, which is taken as 1.0×10 -6 .
[0096] In another exemplary embodiment of the present application, the neural network model uses a long short-term memory network model (long short-term memory, LSTM). The constructed LSTM uses a single hidden layer and 100 hidden units, and the learning rate is 0.01; a nonlinear function is used to introduce nonlinear characteristics into the network. The nonlinear function is called an activation function. The most commonly used activation functions are: Rectified linear unit (ReLU) function, squeezing function, Tanh (hyperbolic tangent) function. This embodiment selects the ReLU function as the activation function. The ReLU function only retains positive elements and discards all negative elements by setting the response and activation values to 0. It is a simple nonlinear transformation that can enhance the learning ability of the network; the error back propagation algorithm and the Adam optimization algorithm are introduced to optimize the neural network model; the score function Score and the root mean square error RMSE are used as the performance evaluation indicators of the model.
[0097] Therefore, the above step 305 specifically includes: inputting the training set into the neural network model, using the error back propagation algorithm and the Adam optimization algorithm, training the hyperparameters in the neural network model with the goal of minimizing the mean square error loss function MSELoss, and obtaining the trained neural network model.
[0098] The test set is input into the trained neural network model, the score function Score and the root mean square error RMSE are used as performance evaluation indicators, and the hyperparameters in the trained neural network model are adjusted; the adjusted neural network model is determined as the compression limit load prediction model.
[0099] In this embodiment, the ReLU function is defined as follows:
[0100] ReLU(A)=max(A,0);
[0101] Among them, A represents the input of the ReLU function.
[0102] The error back propagation algorithm uses a gradient descent method, where the loss function selected is the mean square error loss function MSELoss, which is defined as follows:
[0103]
[0104] In the formula, x represents the model prediction value, y represents the actual value, n represents the total number of samples in the training set, and y i is the actual value corresponding to sample i, f(x i ) is the model prediction value corresponding to sample i, x i Represents the input of the neural network model.
[0105] The Adam optimization algorithm formula is as follows:
[0106]
[0107] Where: m t is the first-order moment estimate of the gradient corresponding to time t; β1 is a hyperparameter, generally taken as 0.9; v t is the gradient second-order moment estimate corresponding to time t; β2 is a hyperparameter, generally taken as 0.99; m t-1 is the first-order moment estimate of the gradient corresponding to time t-1; v t-1 is the second-order moment estimate of the gradient corresponding to time t-1; L represents the loss function. If the loss function adopts the mean square error loss function, L is MSELoss; θ represents the parameter vector of the model. In order to update the parameters normally, the following deviation correction needs to be made:
[0108]
[0109] Where: and are the first-order and second-order estimates of the modified gradient corresponding to time t; represents β1 to the power of t, Represents β2 to the power of t. The formula for updating the neural network parameters by the Adam algorithm is as follows:
[0110]
[0111] Where: α is the learning rate of the neural network model; ε is a constant to prevent zero division error, which is taken as 10 -8 θ t represents the parameter vector of the model at time t; θ t+1 Represents the parameter vector of the updated model at time t+1.
[0112] The score function Score and root mean square error RMSE formula are as follows:
[0113]
[0114] Among them, y i is the actual value corresponding to sample i, f(x i ) is the model prediction value corresponding to sample i, x i Represents the input of the neural network model, and N represents the total number of samples in the test set.
[0115] In another exemplary embodiment of the present application, in step 304, the data sample library is divided into a training set and a test set according to a set ratio. Specifically, the data samples in the data sample library are shuffled, and data are randomly selected at a ratio of 80%-20% as the training set and the test set of the model, and the data in the training set and the test set are normalized; the data set normalization processing method is as follows:
[0116]
[0117] Where: For x j The maximum value in the column. For x j The minimum value in the column, x i,j Represents the value of the i-th row and j-th column.
[0118] The above-mentioned embodiment has the following advantages: first, finite element analysis is performed based on the existing method for predicting the compressive limit load of damaged composite laminates, and a data set of damage parameters corresponding to the compressive limit load is established; second, a long short-term memory network model is established, and the constructed model is optimized by introducing appropriate algorithms and technologies to improve the learning ability and generalization of the model; third, the long short-term memory network model is introduced into the prediction of the compressive limit load to achieve rapid prediction of the compressive limit load.
[0119] Taking damaged carbon fiber resin-based composite laminates as an example, based on the long short-term memory network model, a complete process of the practical application of the compression limit load prediction method of damaged laminates is introduced in detail.
[0120] like Figure 4 As shown in FIG. 1 , the process of real-time prediction using the compression limit load prediction method for damaged laminates includes:
[0121] Step SS1: A progressive damage analysis method for the compressive limit load of a composite laminate with scratch damage is established in the finite element software, and the compressive limit load is obtained by numerically simulating the compression failure process of the damaged composite laminate structure.
[0122] Specifically, a finite element model is established according to the given geometric parameters of the composite laminate, and a UMAT subroutine is written to define material properties, failure judgment criteria and material parameter degradation model to perform finite element calculations and obtain the compressive limit load of the damaged composite laminate.
[0123] Step SS2: The finite element software is redeveloped by writing Python scripts to batch calculate the compressive limit loads of composite laminates with different damage types and create a data sample library.
[0124] Step SS3: Build a long short-term memory neural network model, determine the hyperparameters and loss function of the neural network model, and use the root mean square error RMSE and score function Score as model performance evaluation indicators.
[0125] Step SS4: Divide the data sample library into a training set and a test set at a ratio of 80%-20%, and normalize the training set and the test set.
[0126] Step SS5: Select the data samples in the training set to train the long short-term memory neural network model until the stopping criteria are met, and use the test set to verify the accuracy and generalization ability of the model.
[0127] Firstly, a progressive damage analysis method for the compressive limit load of a composite laminate with scratch damage is established in the finite element software. The compressive limit load is obtained by numerically simulating the compression failure process of the damaged composite laminate structure. The three-dimensional finite element model of the damaged composite laminate is as follows: Figure 5 As shown in Figure 5 The defect in the central area of the three-dimensional finite element model is the damaged area. Displacement load and fully constrained static conditions are applied to the upper and lower ends of the damaged composite laminate in the Y direction.
[0128] Secondly, by writing Python scripts to redevelop the finite element software, the compressive limit loads of composite laminates with different damage types were batch calculated and a data sample library was created.
[0129] Third, build a long short-term memory neural network model, determine the hyperparameters and loss function of the neural network model, and use the root mean square error RMSE and score function Score as model performance evaluation indicators; divide the data sample library into training set and test set at a ratio of 80%-20%, and normalize the training set and test set; select data samples in the training set to train the long short-term memory neural network model until the stopping criteria are met, and use the test set to verify the accuracy and generalization ability of the model.
[0130] Recurrent neural networks can predict the information of the next time series based on the information of the previous time. In essence, it introduces the concept of hidden state in the fully connected neural network. By storing the hidden variables in the current time step in the hidden state and introducing new weight parameters, it can achieve accurate prediction of the information of the next time. Figure 6 and Figure 7 As shown in Figure 6The structure of the recurrent neural network is shown. The recurrent neural network can predict the information of the next time series based on the information of the previous time. Its essence is the concept of hidden state introduced in the fully connected neural network. By storing the hidden variables in the current time step in the hidden state and introducing new weight parameters, accurate prediction of the information at the next time step can be achieved. Among them, x\o are the input\output layer of the model respectively, h is the parameter of the middle hidden layer of the model, W is the introduced hidden state weight parameter, and H t-1 , H t , H t+1 are the intermediate hidden layer parameters at time t-1, time t, and time t+1; Figure 7 shows the structure of a neural network without hidden states, where k1-k4 are input layer parameters, h1-h n is the hidden layer parameter, and F is the final output layer. Since the latent variable model has the phenomenon of long-term information preservation and short-term input loss, problems such as gradient disappearance or gradient explosion often occur during neural network training. Therefore, the long short-term memory network is introduced based on RNN. The long short-term memory network can be regarded as a new architecture composed of multiple gated unit loop nesting.
[0131] The long short-term memory network structure mainly includes three gated unit structures: forget gate, input gate and output gate. The forget gate is used to filter out redundant information in the previous time step, the input gate is used to select useful information from the input information, and the output gate is used to control the output information of the network structure. Assume x t is the input at time t, h t is the hidden state at time t, C t-1 is the storage unit at time t-1, w f 、w i 、w C 、w o is the weight value in the structure, b f , b i , b C , b o is the deviation, σ and tanh are sigmod and tanh functions respectively, and the long short-term memory network can be expressed as follows:
[0132] f t =σ(w f [h t-1 ,x t ]+b f )
[0133] i t =σ(w i [h t-1 ,x t ]+b i )
[0134]
[0135] o t =σ(w o [h t-1 ,x t ]+b o )
[0136] h t =o t *tanh(C t )
[0137] Since hyperparameters such as the learning rate, number of training rounds, and number of hidden layers in the long short-term memory network have a great influence on the model performance, the number of hidden layers has a smaller impact on the model. Even if there is only one hidden layer, complex nonlinear functions can be fitted through sufficient neurons. Therefore, the number of hidden layers of the multilayer perceptron neural network model established in this embodiment is 1.
[0138] Introduce nonlinear transformation, use element-wise nonlinear function to transform hidden variables, and then use them as input to the next fully connected layer. Nonlinear function is called activation function. This embodiment uses ReLU function as activation function to introduce nonlinear characteristics into the network and enhance the learning ability of the network.
[0139] The model is trained based on the data in the divided training set, and the trained model is evaluated and verified using the test set. The model training and verification flow chart is as follows: Figure 8 As shown in , the predicted compressive limit load value obtained in this embodiment is compared with the finite element simulation result as shown in Fig. 9 shown.
[0140] The above embodiments solve the problems of high cost and long time consumption in solving the compressive limit load of damaged composite laminates using the existing finite element method. First, a three-dimensional finite element model of a composite laminate with scratch damage is established, and a UMAT subroutine is written to define material properties, failure criteria and material parameter degradation model. A progressive damage analysis of the damaged composite laminate is performed in the finite element software to obtain the limit load and failure mode. Secondly, the finite element software is secondary developed by writing Python scripts to batch calculate the compressive limit load of composite laminates with different damage types and create data samples. Finally, the long short-term memory neural network model is initialized, the model hyperparameters and loss function are determined, and 80% of the data samples are extracted as a training set to train the model until the stopping criteria are met. The remaining 20% of the data samples are used to verify the accuracy and generalization ability of the model. The model established by this method can realize the rapid prediction of the compressive limit load of damaged composite laminates, with high prediction accuracy and calculation rate.
[0141] Based on the same inventive concept, the embodiment of the present application also provides a device for realizing the above-mentioned prediction of the compressive limit load of the damaged laminate. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more devices for predicting the compressive limit load of damaged laminates provided below can refer to the limitations of the method for predicting the compressive limit load of damaged laminates above, and will not be repeated here.
[0142] In an exemplary embodiment, Fig.10 As shown, a compression limit load prediction device for a damaged laminate is provided, comprising:
[0143] The damage data acquisition module 1011 is used to acquire damage data of a target laminate; the target laminate is a damaged laminate whose compressive limit load is to be predicted; the damage data includes: coordinates of the center position of the damage and the damage depth.
[0144] The compression limit load prediction module 1012 is used to input the damage data of the target laminate into the compression limit load prediction model to obtain the compression limit load of the target laminate.
[0145] Among them, the compression limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compression limit loads; the compression limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software.
[0146] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store damage data of the target laminate. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting the compressive limit load of a damaged laminate is implemented.
[0147] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0148] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0149] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0150] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0151] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0152] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0153] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, etc., but is not limited thereto.
[0154] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0155] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A method for predicting the compressive limit load of damaged laminates. It is characterized in that include: Acquire damage data of target laminate; The target laminate is a damaged laminate whose compressive limit load is to be predicted; The damage data includes: coordinates of the center position of the damage and the damage depth; Inputting damage data of the target laminate into a compressive limit load prediction model to obtain the compressive limit load of the target laminate; Among them, the compression limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compression limit loads; the compression limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software.
2. The method for predicting the compressive limit load of a damaged laminate according to claim 1, It is characterized in that The method for determining the compression limit load prediction model specifically includes: Obtain damage data of laminates with different damage types; Finite element software is secondary developed by writing Python scripts to batch calculate the compressive limit loads of laminates with different damage types; wherein, the calculation process of the compressive limit load of a laminate of any damage type is as follows: a three-dimensional finite element model of the laminate of the damage type is established; based on the three-dimensional finite element model, the compressive failure process of the laminate of the damage type is simulated by using material failure judgment criteria and material parameter degradation model to determine the failure mode and material parameter degradation mode of the laminate of the damage type; the compressive limit load of the laminate of the damage type is determined according to the failure mode and the material parameter degradation mode; Construct a data sample library based on the damage data of laminates with different damage types and the corresponding compressive limit loads; Dividing the data sample library into a training set and a test set according to a set ratio; Using the training set to train the neural network model to obtain a trained neural network model; The test set is used to perform performance evaluation on the trained neural network model to determine the compression limit load prediction model.
3. The method for predicting the compressive limit load of a damaged laminate according to claim 2, It is characterized in that The material failure determination criteria include: If satisfied Then the failure mode is determined to be fiber tensile failure; If satisfied Then the failure mode is determined to be fiber compression failure; If satisfied Then the failure mode is determined to be matrix tensile failure; If satisfied Then the failure mode is determined to be matrix compression failure; If satisfied Then the failure mode is determined to be matrix-fiber shear failure; If satisfied The failure mode is determined to be delamination tensile failure; If satisfied Then the failure mode is determined to be layered compression failure; Among them, σ 1 is the normal stress in the fiber direction; σ 2 is the normal stress in the matrix direction; σ 3 is the normal stress in the normal direction; τ 12 is the shear stress component in the fiber direction; τ 13 is the shear stress component in the matrix direction; τ 23 is the shear stress component in the normal direction; X T is the tensile strength in the fiber direction; X C is the compressive strength in the fiber direction; Y T is the tensile strength in the matrix direction; Y C is the compressive strength in the matrix direction; Z T is the tensile strength in the normal direction; Z C is the compressive strength in the normal direction; S 12 S is the shear strength of the material in the fiber direction; 13 S is the shear strength value of the material in the matrix direction; 23 is the shear strength value of the material in the normal direction.
4. The method for predicting the compressive limit load of a damaged laminate according to claim 2, It is characterized in that The material parameter degradation model includes: If the failure mode is fiber tensile failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.07 times the initial value; where E 11 is the elastic modulus in the fiber direction; E 22 is the elastic modulus in the matrix direction; E 33 is the elastic modulus in the normal direction; v 12 is the Poisson's ratio in the fiber direction; v 13 is the Poisson's ratio in the matrix direction; v 23 is the Poisson's ratio in the normal direction; G 12 is the shear elastic modulus in the fiber direction; G 13 is the shear elastic modulus in the matrix direction; G 23 is the shear elastic modulus in the normal direction; If the failure mode is fiber compression failure, the material parameter degradation mode is: E 11 、E 22 、E 33 、v 12 、v 13 、v 23 , G 12 , G 13 , G 23 Reduced to 0.14 times the initial value; If the failure mode is matrix tensile failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.2 times the initial value; If the failure mode is matrix compression failure, the material parameter degradation mode is: E 22 , G 12 , G 23 Reduced to 0.4 times the initial value; If the failure mode is matrix-fiber shear failure, the material parameter degradation mode is: G 12 、v 12 Reduce to 0; If the failure mode is delamination tensile failure or delamination compression failure, the material parameter degradation mode is: E 33 、v 13 , ν 23 , G 13 , G 23 Reduced to 0.
5. The method for predicting the compressive limit load of a damaged laminate according to claim 2, It is characterized in that The training set is used to train the neural network model to obtain a trained neural network model, which specifically includes: The training set is input into the neural network model, and the error back propagation algorithm and the Adam optimization algorithm are used to train the hyperparameters in the neural network model with the goal of minimizing the mean square error loss function to obtain a trained neural network model.
6. The method for predicting the compressive limit load of a damaged laminate according to claim 2, It is characterized in that The performance of the trained neural network model is evaluated using the test set to determine the compression limit load prediction model, specifically including: Inputting the test set into the trained neural network model, using the score function and the root mean square error as performance evaluation indicators, and adjusting the hyperparameters in the trained neural network model; The adjusted neural network model is determined as the compression limit load prediction model.
7. The method for predicting the compressive limit load of a damaged laminate according to claim 1, It is characterized in that The neural network model is a long short-term memory network model.
8. The method for predicting the compressive limit load of a damaged laminate according to claim 2, It is characterized in that Before using the training set to train the neural network model to obtain the trained neural network model, the method further includes: The data in the training set and the test set are normalized.
9. A device for predicting the compressive limit load of damaged laminates. It is characterized in that include: A damage data acquisition module, used to acquire damage data of a target laminate; The target laminate is a damaged laminate whose compressive limit load is to be predicted; The damage data includes: coordinates of the center position of the damage and the damage depth; A compression limit load prediction module, used for inputting damage data of a target laminate into a compression limit load prediction model to obtain the compression limit load of the target laminate; Among them, the compression limit load prediction model is obtained by training the neural network model using a data sample library; the data sample library includes: damage data of laminates with different damage types and corresponding compression limit loads; the compression limit load in the data sample library is determined by simulating the compression failure process of laminates with different damage types using finite element software.
10. A computer device, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting the compressive limit load of a damaged laminate according to any one of claims 1 to 8.
Citation Information
Cited By
Prediction method and device for ultimate bearing capacity of composite material wallboard assembly
CN120995894A
