A composite material fatigue delamination prediction method, device, medium and product
By combining nondestructive testing and neural networks with finite element models, a data-driven fatigue cohesion model is constructed, which solves the problems of high testing costs and complex mechanism modeling of traditional Paris formulas, and realizes efficient and accurate prediction of fatigue delamination of composite materials.
Patent Information
- Application Number
- CN202410486912.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-22
- Publication Date
- 2026-02-06
- Estimated Expiration
- 2044-04-22
AI Technical Summary
In the study of fatigue damage in composite laminates, the traditional Paris formula test is costly, time-consuming, and has limited applicability. Its mechanism modeling is complex and its accuracy is low, resulting in low efficiency and poor accuracy in fatigue delamination prediction.
Non-destructive testing technology is used to obtain layered damage information. A data-driven fatigue cohesion model is constructed by combining the finite element model and neural network. The neural network model is used to train and predict the layered damage propagation, simplifying the calculation process and reducing the dependence on the Paris formula.
It improves the accuracy and efficiency of fatigue delamination prediction in composite materials, reduces costs, simplifies model structure, and avoids the limitations and complexity of traditional methods.
Smart Images
Figure CN118228557B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of composite material fatigue delamination prediction, in particular to a neural network composite material fatigue delamination prediction method and device based on in-situ sample training, medium and product. BACKGROUND
[0002] Fiber reinforced plastic (FRP) is light in weight, high in tensile strength, corrosion resistant, and has excellent fatigue resistance. Because it can meet the technical requirements of long service life, high reliability and light weight for aircraft component materials due to its many advantages, it is widely used in aerospace field, and its application in aircraft has become one of the important indicators for evaluating the advancement of aircraft.
[0003] It is well known that FRP is often subjected to tensile fatigue load in engineering structures, and irreversible damage and significant decline in performance are inevitable. In addition, the complex actual working environment also accelerates the accumulation of fatigue damage, leading to premature fatigue failure of FRP. Therefore, a full understanding of fatigue damage behavior is the key to improving the reliability, safety and life of FRP materials and structures. Delamination is a common problem in composite laminates fatigue damage, and has always been a hot research topic. When subjected to cyclic loading, fatigue delamination may start and propagate at low loads, leading to structural failure.
[0004] Numerical simulation technology can realize the simulation of delamination of complex structures at low cost, and has become a necessary means for the study of composite material delamination problems.
[0005] In the existing research on delamination damage of composite laminates, the traditional Paris formula test is often used to establish a fatigue propagation prediction mechanism model based on the Paris formula and damage fracture mechanics framework, and then to predict fatigue delamination. This technical scheme has the following limitations:
[0006] 1) The traditional Paris formula determines the empirical coefficients C and m in the Paris formula through experimental measurement. This usually requires fatigue loading tests at different stress amplitudes and measurement of crack propagation. However, fatigue loading tests have high cost, long time consumption, and the Paris formula is an empirical formula, which is usually limited to specific materials, loading methods and environmental conditions, and has large dispersion.
[0007] 2) In most studies, the structure of the mechanism model is complex, especially in fatigue damage calculation, a large amount of numerical processing is required, the accuracy is not high, and the fatigue calculation in the model depends on the Paris formula, which further affects the accuracy and efficiency of numerical analysis.
[0008] Therefore, there is an urgent need for a low-cost, short-time and wide-range method that can accurately and efficiently predict the delamination of composite materials. SUMMARY
[0009] The purpose of the present application is to provide a composite material fatigue delamination prediction method, device, medium and product, which uses machine learning technology to replace complex calculation, constructs a simple structure data-driven fatigue cohesive force model, and directly uses the delamination damage basic information obtained by non-destructive testing technology for neural network model training, improves the accuracy and stability of fatigue delamination prediction, reduces the cost, and improves the prediction efficiency.
[0010] To achieve the above purpose, the present application provides the following scheme:
[0011] In a first aspect, the present application provides a composite material fatigue delamination prediction method, comprising:
[0012] Performing non-destructive testing on a service composite material structure to obtain delamination damage information, wherein the delamination damage information includes the actual service time of the service composite material structure and corresponding delamination front position information;
[0013] Establishing a finite element model of the service composite material structure, and simulating the delamination damage expansion of the service composite material structure under cyclic loading according to the finite element model;
[0014] According to the delamination front position information, the energy release rate, damage mode mixing ratio, total damage amount and first fatigue damage variable accumulation rate in the early delamination damage expansion accumulation stage are obtained in the failed neural network cohesive force unit at the corresponding position, wherein the neural network cohesive force unit is arranged in the delamination occurrence area of the finite element model, the failed neural network cohesive force unit is used to take the energy release rate, the damage mode mixing ratio and the total damage amount as input, and output the second fatigue damage variable accumulation rate by using the neural network model, and the first fatigue damage variable accumulation rate is calculated according to the second fatigue damage variable accumulation rate and the actual service time;
[0015] All the energy release rate, damage mode mixing ratio, total damage amount and first fatigue damage variable accumulation rate are taken as in-situ samples to train the neural network model;
[0016] According to the trained neural network model in the failed neural network cohesive force unit, the subsequent delamination damage expansion is predicted.
[0017] Optionally, the energy release rate, the damage mode mixing ratio and the total damage amount are taken as inputs, a second fatigue damage variable cumulative rate is output by using a neural network model, and the first fatigue damage variable cumulative rate is calculated according to the second fatigue damage variable cumulative rate and the actual service time, specifically comprising:
[0018] Initializing a fatigue damage variable cumulative rate;
[0019] According to the initialized fatigue damage variable cumulative rate, the total damage amount at the current time is calculated;
[0020] The energy release rate, the damage mode mixing ratio and the total damage amount obtained at the current time are taken as inputs, and the second fatigue damage variable cumulative rate is output by using a neural network model;
[0021] According to the second fatigue damage variable cumulative rate, a virtual service time at the current time is calculated;
[0022] According to the virtual service time and the actual service time, the second fatigue damage variable is corrected to obtain the first fatigue damage variable cumulative rate at the current time;
[0023] For the next time, the first fatigue damage variable cumulative rate at the last time is taken as a new initialized fatigue damage variable cumulative rate, and the step of "according to the initialized fatigue damage variable cumulative rate, the total damage amount at the current time is calculated" is returned.
[0024] Optionally, the expression of the second fatigue damage variable cumulative rate is:
[0025]
[0026] Wherein, dD f / dN is the second fatigue damage variable cumulative rate; D f is the fatigue damage variable; N is the virtual service time simulated by the finite element model; f is ΔG / G C , φ, D tot is the mapping function of dD f / dN; ΔG is the change amount of the energy release rate; G C is the critical energy release rate; φ is the damage mode mixing ratio; D tot is the total damage amount.
[0027] Optionally, the expression of the total damage amount is:
[0028]
[0029] Wherein, D tot is the total damage amount; D sis a static damage variable; dD f / dN is a second fatigue damage variable accumulation rate; D f is a fatigue damage variable; N is a virtual service time simulated by a finite element model.
[0030] Optionally, an expression of the virtual service time is:
[0031]
[0032] wherein, N fail is the virtual service time; dD f / dN is a second fatigue damage variable accumulation rate.
[0033] Optionally, the neural network model is a single-hidden-layer neural network, including one hidden layer and one output layer.
[0034] In a second aspect, the present application provides a computer device, comprising a memory, a processor, a computer program stored in the memory and executable in the processor, wherein the processor executes the computer program to implement the steps of the composite material fatigue delamination prediction method according to any one of the preceding aspects.
[0035] In a third aspect, the present application provides a computer readable storage medium, having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the composite material fatigue delamination prediction method according to any one of the preceding aspects.
[0036] In a fourth aspect, the present application provides a computer program product, comprising a computer program, wherein the computer program is executable by a processor to implement the steps of the composite material fatigue delamination prediction method according to any one of the preceding aspects.
[0037] According to the embodiments of the present application, the following technical effects are achieved:
[0038] The application provides a composite material fatigue delamination prediction method, device, medium and product. The prediction method directly obtains delamination damage basic information by using nondestructive testing technology, obtains simulation data in an early delamination damage expansion and accumulation stage by using a finite element model, corrects the simulation data by using the delamination damage basic information, obtains in-situ samples in the early delamination damage expansion and accumulation stage, and then is used for neural network model training, and then the trained neural network fatigue cohesion model in the cohesion unit is used to predict subsequent expansion of delamination damage of the same composite material structure, so that a more accurate and efficient solution for composite material fatigue delamination prediction is provided. Compared with a prediction method for establishing a fatigue expansion prediction mechanism model based on a Paris formula and a damage fracture mechanics framework and then predicting fatigue delamination, the application solves the problems of difficult fatigue damage calculation, low fatigue delamination prediction calculation efficiency and low accuracy. BRIEF DESCRIPTION OF DRAWINGS
[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0040] Figure 1 A flowchart of a composite material fatigue delamination prediction method provided for embodiment 1 of the present application is shown in the figure.
[0041] Figure 2 A schematic diagram of a delamination front position in embodiment 1 of the present application is shown in the figure.
[0042] Figure 3 An internal structure diagram of a computer device is shown in the figure. DETAILED DESCRIPTION
[0043] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0044] The purpose of the present application is to provide a composite material fatigue delamination prediction method, device, medium and product. By using machine learning technology to replace complex calculation, a simple structure data-driven fatigue cohesion model is constructed, and delamination damage basic information obtained by using nondestructive testing technology is directly used for neural network model training, so as to improve the accuracy and stability of fatigue delamination prediction, reduce the cost, and improve the prediction efficiency.
[0045] In order to make the above-mentioned purposes, features and advantages of the present application more apparent, further detailed description of the present application will be given below in conjunction with the accompanying drawings and specific embodiments.
[0046] Example 1
[0047] In the study of composite delamination problems, the test standard for composite laminate delamination has been quite mature. However, in engineering practice, due to factors such as complex structure, equipment limitation, test cost and test period, it is usually impossible to predict the delamination propagation behavior of the structure through tests. Mature and effective numerical simulation technology can realize the delamination simulation of complex structures at a lower cost, and becomes a necessary means for the study of composite delamination problems.
[0048] The numerical simulation study of the delamination propagation behavior of composite laminates has gone through three stages: in the early stage, stress was mainly used as the standard, and methods such as point stress criterion and average stress criterion were proposed. Later, the development of fracture mechanics provided a new idea for delamination research, resulting in methods based on stress concentration factor, methods based on J integral, methods based on virtual crack technology and extended finite element method. Virtual crack closure technique considers that the energy consumed by the crack surface is equal to the energy released by the same crack closure. In recent years, the emergence of continuous medium damage mechanics provides a theoretical basis for the application of cohesive zone model method in delamination numerical simulation. The cohesive zone model is essentially a phenomenological model based on the energy point of view, which can effectively avoid the crack tip singularity phenomenon in fracture mechanics methods, and is widely used in crack propagation analysis. In the research, the cohesive zone model is usually applied by using the cohesive zone element provided in ABAQUS and the user-defined subroutine. By embedding the cohesive zone element between the ordinary elements, the damage is limited to the cohesive zone element layer to accurately capture the crack propagation, and the damage calculation of the cohesive zone element calls the VUMAT subroutine module, in which researchers can define the constitutive behavior, damage model and fracture criterion of the material, etc. This provides great flexibility for simulating various complex material behaviors.
[0049] An existing study on "generating samples based on Paris formula for neural network training, and then predicting fatigue delamination" has the following main steps:
[0050] 1. Test the material according to the Paris formula, and perform fatigue loading at different stress levels to obtain the relationship between crack propagation rate and energy release rate. Through fitting the test data, the parameters C and m in the Paris formula are obtained.
[0051] 2. Construct a fatigue cohesive zone model based on the mechanism, compile the VUMAT subroutine, embed it in ABAQUS to realize the finite element application of the fatigue cohesive zone model, and perform numerical prediction of composite fatigue delamination.
[0052] The technical solution has the following limitations:
[0053] 1) The traditional Paris formula determines the empirical coefficients C and m in the Paris formula through experimental measurement. This usually requires fatigue loading tests at different stress amplitudes and measurement of crack propagation. However, fatigue loading tests have high costs and long time consumption for sample preparation, testing, and detection. Moreover, the Paris formula is an empirical formula, and its application scope is usually limited to specific materials, loading methods, and environmental conditions, with large dispersion.
[0054] 2) In most studies, the structure of the mechanism model is complex, especially in fatigue damage calculation, which requires a large amount of numerical processing, and the accuracy is not high. Moreover, the fatigue calculation in the model relies on the Paris formula, which further affects the accuracy and efficiency of numerical analysis.
[0055] To solve the problems of difficult fatigue delamination damage calculation, low efficiency and low accuracy of fatigue delamination prediction in the prior art, the embodiment proposes a composite material fatigue delamination prediction method based on non-destructive testing samples, aiming to accurately and efficiently realize composite material fatigue delamination prediction. The composite material fatigue delamination prediction method comprises:
[0056] (1) Non-destructive testing is performed on the service composite material structure to obtain delamination damage information, wherein the delamination damage information includes the actual service time of the service composite material structure and corresponding delamination front position information.
[0057] (2) A finite element model of the service composite material structure is established, and the delamination damage propagation of the service composite material structure under cyclic loading is simulated according to the finite element model;
[0058] (3) According to the delamination front position information, the energy release rate, damage mode mixing ratio, total damage amount, and first fatigue damage variable accumulation rate in the early delamination damage propagation accumulation stage are obtained in the failed neural network cohesive element at the corresponding position, wherein the neural network cohesive element is arranged in the delamination occurrence area of the finite element model, the failed neural network cohesive element is used to input the energy release rate, the damage mode mixing ratio, and the total damage amount, output the second fatigue damage variable accumulation rate by using the neural network model, and calculate the first fatigue damage variable accumulation rate according to the second fatigue damage variable accumulation rate and the actual service time.
[0059] (4) All the energy release rate, damage mode mixing ratio, total damage amount, and first fatigue damage variable accumulation rate are used as in-situ samples to train the neural network model.
[0060] (5) predicting the subsequent propagation of delamination damage based on the trained neural network model in the failed neural network cohesion unit.
[0061] This embodiment is based on the existing neural network fatigue cohesion model, which considers using non-destructive testing technology to collect samples from the early accumulation stage of fatigue damage for training, and then using the trained neural network model to predict the subsequent damage propagation of the same structure, providing a more accurate and efficient solution for composite material fatigue delamination prediction.
[0062] In order to make the person skilled in the art more clearly understand the above-mentioned composite material fatigue delamination prediction process of the embodiment, the following is specifically described.
[0063] As shown in Figure 1 The composite material fatigue delamination prediction method comprises:
[0064] Step 1, using non-destructive testing technology to obtain the basic information of delamination damage from the served composite material structure, the basic information refers to the delamination front position in the composite material structure under a certain service time, which includes the actual service time of the composite material structure and the corresponding delamination front position coordinate information.
[0065] The composite material structure is subjected to cyclic external forces during service, and the delamination interface inside the material is subjected to alternating stress loading, resulting in fatigue damage at the interface, damage accumulation, and ultimately leading to delamination damage propagation. Delamination damage usually leads to a decrease in the mechanical properties of the material, and the overall stiffness, strength and load-carrying capacity of the composite material will decrease, affecting the stability and reliability of the structure. Delamination damage usually manifests as peeling, separation or fracture between different plies inside the composite material. Near the delamination interface of the composite material, fatigue loading will cause the generation and propagation of microcracks. These cracks usually propagate along the delamination interface and eventually lead to fatigue failure of the material.
[0066] Since delamination damage generally occurs inside the composite material structure, the delamination behavior between the internal plies and the delamination position cannot be directly observed by the naked eye, and non-destructive testing technology is needed to obtain specific delamination damage information to obtain more accurate results.
[0067] The non-destructive testing methods for composite materials can be classified into ultrasonic, X-ray, infrared, electronic speckle pattern interferometry, terahertz, eddy current, resistance method, acoustic vibration method, etc. Among them, the ultrasonic non-destructive testing methods are various, including ultrasonic microscopy, array ultrasonic testing method, ultrasonic guided wave testing method, laser ultrasonic testing method and ultrasonic Lamb wave testing method, etc. It is the most widely used testing technology in the field of composite material testing. However, for some special structural defects in composite structures, if real-time observation of the position, size, shape and other information of the defects is required, the X-ray testing method has obvious advantages, which can be directly observed on the display, and is very convenient and intuitive.
[0068] In this embodiment, the delamination front position of the composite structure under a certain service time is obtained by using X-ray CT technology, so as to identify and evaluate the delamination damage.
[0069] Step 2, a finite element model of the service composite structure is established, and a neural network cohesive element is arranged in the delamination occurrence area. The delamination occurrence area corresponds to the expansion area of the delamination front of the composite structure during service in step 1. The neural network cohesive element is used to simulate the expansion process of the delamination damage, and the neural network model is used to calculate the fatigue damage variable accumulation rate.
[0070] Specifically, the initial fatigue damage variable accumulation rate is given, so as to promote the fatigue delamination of the finite element model to expand in the neural network cohesive element.
[0071] Step 3, according to the delamination front position of the actual service structure obtained by non-destructive testing, data is obtained from the failure units at the corresponding position of the finite element model.
[0072] The delamination front position of the object structure after a certain service time obtained by non-destructive testing is found in the delamination occurrence area in the finite element model, and the energy release rate, damage mode mixing ratio and total damage variable are obtained from the failure neural network cohesive element at the position.
[0073] According to the characteristics of fatigue damage accumulation, the neural network model takes the energy release rate, damage mode mixing ratio and total damage variable of the cohesive element as input, and outputs the fatigue damage variable accumulation rate (i.e. the second fatigue damage variable accumulation rate referred to above). The neural network can be represented as:
[0074]
[0075] where dD f / dN is the second fatigue damage variable accumulation rate; D f is the fatigue damage variable; N is the virtual service time simulated by the finite element model; f is ΔG / G C , φ, D totThe mapping function of dD f is the change of energy release rate; G C is the critical energy release rate, which is obtained by experimental test; φ is the damage mode mixing ratio; D tot is the total damage variable.
[0076] The expression of φ is:
[0077]
[0078] wherein, G I and G II are the energy release rates of pure mode I and pure mode II respectively, which can be obtained by integrating the traction-displacement relationship of the crack tip element. The expression of the change of energy release rate ΔG is:
[0079] ΔG = (1-R)(G Ι + G ΙΙ ) (3);
[0080] wherein, R = F min / F max is the load ratio. F max and F max are the maximum and minimum values of the cyclic fatigue load. The total damage variable of the neural network cohesive element is defined as D tot :
[0081]
[0082] wherein, D s is the static damage variable, and N is the virtual service time simulated by the finite element model. The bearing capacity of the cohesive element is reduced by D tot , so as to simulate the expansion of delamination damage.
[0083] The purpose of calculating the total damage variable is to determine whether the neural network cohesive element fails, so as to ensure the accuracy of the calculation of the virtual service time of the failed element.
[0084] When D tot equals 1, the neural network cohesive element completely fails, which is equivalent to delamination formation. The interlaminar stress during the failure of the element is calculated as:
[0085] σ = (1-D tot )σ max (5);
[0086] wherein, σ max is the interlaminar strength of the structure. Since the static damage variable D s will not reach 1 under cyclic load, when formula (4) is , D totA value equal to 1 indicates that the cohesive unit within the neural network has completely failed. Therefore, in the finite element model, the virtual failure service life of the failed unit during delamination can be determined:
[0087]
[0088] Since there is no fatigue damage accumulation at the beginning of fatigue damage (i.e., the initial fatigue damage variable accumulation rate is 0), the total damage variable is only the static damage variable. The total damage variable can be obtained according to formula (4), and then the fatigue damage variable accumulation rate can be obtained according to formula (1). In the finite element analysis process of fatigue damage accumulation over time, for the next analysis step, the total damage variable calculated using formula (4) uses the fatigue damage variable accumulation rate calculated in the previous analysis step. Based on the total damage variable obtained in this step, the fatigue damage variable accumulation rate is obtained again using formula (1), and so on.
[0089] Without correction based on non-destructive testing data from actual service structures, an initial cumulative rate of fatigue damage variables can be defined. This drives the evolution of hierarchical expansion, achieving, for example... Figure 2 (a) shows the layered extended data of the finite element simulation, where the mesh part consists of failed cohesive elements and the non-mesh part on the left consists of unfailed cohesive elements.
[0090] Each failed cohesive element can be determined according to equation (6) and the initial... Calculate a virtual failure service time N. pesudo N here pesudo It is the virtual service duration during the layering process in the finite element model, corresponding to N in the theoretical formula. fail .
[0091] In the delamination region of the finite element model, locate the delamination leading edge position detected by nondestructive testing from the actual service structure, such as... Figure 2 As shown in (b), the energy release rate, damage mode mixing ratio, total damage variable, and virtual failure service time N are then extracted from the failed cell at that location. pesudo At this point, the data extracted from the failure elements of the finite element model are corrected based on the actual service time obtained from the detection, which shows the leading edge of the delamination of the actual service structure extends to the corresponding location. This is because the detection is performed during the actual service period N. real It was done later, therefore, theoretically Figure 2 (b) The service life corresponding to the blue unit is N. real N here real The actual service time during the formation of the stratification is represented by N in the formula. fail Therefore, it can be based on the actual service life N real The virtual lifetime N of the finite element modelpesudo The initial pseudo-fatigue damage variable accumulation rate The correction factor is:
[0092]
[0093] Further, the value is corrected to obtain the correct fatigue damage variable accumulation rate
[0094]
[0095] Therefore, the above steps are, step 4: according to the object structure layering expansion obtained by non-destructive testing, the real service time length at the position and the virtual failure service time length of the failure unit in the finite element model when the layering is formed, the initial fatigue damage variable accumulation rate is corrected, and the real fatigue damage variable accumulation rate is obtained.
[0096] The energy release rate, damage mode mixing ratio, total damage variable and corrected fatigue damage variable accumulation rate extracted from the finite element model form the in-situ sample.
[0097] Step 5: use the in-situ sample obtained from the early stage of layering damage expansion based on non-destructive testing to train the neural network model, and then use the trained neural network fatigue cohesion model to predict the further expansion of the layering.
[0098] The trained neural network embeds the weight and bias parameters into the cohesion model. The cohesion model uses a single hidden layer neural network as the constitutive law, takes the energy release rate, damage mode mixing ratio and total damage variable of the cohesion unit as input, and directly outputs the predicted fatigue damage variable accumulation rate to realize layering prediction.
[0099] The single hidden layer neural network contains one hidden layer and one output layer, and the calculation formula of the neural network is:
[0100] z1=W1·θ+b1 (9);
[0101] Where z1 is an intermediate state value, the input is W1∈R 3×n is the weight matrix, b1∈R n is the bias, n is the number of neurons in the hidden layer, and Sigmoid is used as the activation function, so the output of the hidden layer is:
[0102]
[0103] The output of the neural network can be obtained as:
[0104]
[0105] where W2∈R n×1 is a weight matrix; b2 is the bias of the output layer.
[0106] For the training of the neural network, the mean square error is selected as the loss function, and the adaptive linear momentum with a learning rate of 0.001 is used as the optimization algorithm. All sample groups are divided into two subgroups, 80% of which are used for training, and the remaining 20% are used for verification. The total training period is set to 5000. After the training of the neural network model is completed, the weight and bias parameter files are output.
[0107] The neural network fatigue cohesive model is realized by the ABAQUS user-defined subroutine VUMAT for finite element application. The finite element model of the structure is established, and the cyclic load is applied. The subroutine reads the external file of the weight and bias parameters of the trained neural network model, calculates the fatigue damage using the neural network formula, drives the hierarchical propagation, and realizes the fatigue hierarchical prediction based on the in-situ sample of non-destructive testing.
[0108] The present embodiment directly uses the non-destructive testing technology to collect samples from the early accumulation stage of fatigue damage for training using the neural network fatigue cohesive model framework in the prior art, and then predicts the subsequent damage development of the same structure, avoiding the need for a large number of fatigue test experiments, greatly improving the efficiency of composite material fatigue damage research, and reducing the research cost.
[0109] The key to the method for predicting fatigue delamination proposed in the present embodiment is how to use non-destructive testing technology to obtain the samples required for neural network training, which determines the superiority of the neural network fatigue cohesive model based on in-situ samples of non-destructive testing compared with the traditional cohesive model, which is embodied in:
[0110] 1. When obtaining the samples required for training the neural network model, the basic information of delamination damage obtained directly by using non-destructive testing technology is combined with the finite element simulation data, and the simulation sample data is corrected by the damage information obtained by non-destructive testing to obtain correct neural network training samples. Instead of correcting the simulation data by calculating the fatigue growth rate through the empirical Paris formula, the dependence of the neural network fatigue cohesive model in the prior art on the empirical Paris formula can be completely eliminated, the problem of obtaining the empirical parameters of the Paris formula is solved, the errors caused by the limitations and dispersion of the Paris formula are avoided, and the time and economic cost of determining the Paris formula parameters through a large number of fatigue test experiments are also reduced.
[0111] 2、The embodiment embeds the weight and bias parameters of the trained neural network model into the cohesion model when modeling the mechanism, and the calculation part of the fatigue damage is replaced by the neural network instead of the traditional cohesion formula calculation, which simplifies the model and improves the calculation efficiency. In the training step of the neural network model, for the acquisition of samples, the in-situ samples of the target specimen for fatigue delamination research are directly obtained by using non-destructive testing technology, and then the target specimen is further layered for analysis, and there is no need to generate samples by using other models and to modify the Paris formula, so that the accurate prediction of fatigue delamination can be realized more simply and efficiently.
[0112] Embodiment 2
[0113] A computer device, comprising: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executing the computer program to implement the steps of the composite material fatigue delamination prediction method in embodiment 1.
[0114] Embodiment 3
[0115] A computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the composite material fatigue delamination prediction method in embodiment 1.
[0116] Embodiment 4
[0117] A computer program product comprising a computer program, the computer program being executed by a processor to implement the steps of the composite material fatigue delamination prediction method in embodiment 1.
[0118] Embodiment 5
[0119] A computer device, which can be a database, and its internal structure diagram can be as shown in Figure 3 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. 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. 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 transactions to be processed. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement the composite material fatigue delamination prediction method in embodiment 1.
[0120] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions.
[0121] A person 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 a computer program guiding related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided by 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 memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.
[0122] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.
[0123] The principles and implementations of the present application are described in the specific examples, and the above examples are only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the present application.
Claims
1. A method of predicting fatigue delamination in a composite material, the method comprising: The method comprises: nondestructive testing of a service composite structure to obtain delamination damage information, wherein the delamination damage information comprises actual service duration of the service composite structure and corresponding delamination front position information; establishing a finite element model of the service composite structure, and simulating delamination damage propagation of the service composite structure under cyclic loading according to the finite element model; obtaining, according to the delamination front position information, energy release rate, damage mode mixing ratio, total damage amount and first fatigue damage variable accumulation rate of an early delamination damage propagation accumulation stage in a failed neural network cohesive element at a corresponding position, wherein the neural network cohesive element is arranged in a delamination occurrence area of the finite element model, the failed neural network cohesive element is used to take the energy release rate, the damage mode mixing ratio and the total damage amount as inputs, output a second fatigue damage variable accumulation rate by using a neural network model, and calculate the first fatigue damage variable accumulation rate according to the second fatigue damage variable accumulation rate and the actual service duration; specifically comprising: initializing fatigue damage variable accumulation rate; calculating the total damage amount at the current time according to the initialized fatigue damage variable accumulation rate; taking the energy release rate, the damage mode mixing ratio and the total damage amount obtained at the current time as inputs, outputting the second fatigue damage variable accumulation rate by using a neural network model; calculating a virtual service duration at the current time according to the second fatigue damage variable accumulation rate; correcting the second fatigue damage variable according to the virtual service duration and the actual service duration to obtain the first fatigue damage variable accumulation rate at the current time; for the next time, taking the first fatigue damage variable accumulation rate at the previous time as a new initialized fatigue damage variable accumulation rate, and returning to the step of "calculating the total damage amount at the current time according to the initialized fatigue damage variable accumulation rate"; the expression of the second fatigue damage variable accumulation rate is: wherein dD f / dN is the second fatigue damage variable accumulation rate; D f is the fatigue damage variable; N is the virtual service length simulated by the finite element model; f is ΔG / G C , D tot is the mapping function of dD f / dN; ΔG is the change in energy release rate; G C is the critical energy release rate; is the damage mode mixing ratio; D tot is the total damage amount; taking all the energy release rates, the damage mode mixing ratios, the total damage amounts and the first fatigue damage variable accumulation rates as in-situ samples to train the neural network model; predicting subsequent delamination damage propagation according to the trained neural network model in the failed neural network cohesive element; According to the actual service length and the virtual service length, an initial pseudo fatigue damage variable accumulation rate is corrected to obtain a correct fatigue damage variable accumulation rate is corrected to obtain a correct fatigue damage variable accumulation rate The specific formula of the correction factor is where N pesudo is the virtual service length, and N real is the actual service length.
2. The method of claim 1, wherein, the expression of the total damage amount is: where D tot is the total damage; D s is the static damage variable; dD f is the second fatigue damage variable accumulation rate; D f is the fatigue damage variable; and N is the virtual service life simulated by the finite element model.
3. The method of claim 1, wherein, the expression of the virtual service duration is: wherein N fail is the virtual service length; dD f is the second fatigue damage variable cumulative rate.
4. The method of claim 1, wherein, the neural network model is a single-hidden-layer neural network comprising one hidden layer and one output layer.
5. A computer apparatus comprising: A memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the composite material fatigue delamination prediction method of any one of claims 1-4.
6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the composite material fatigue delamination prediction method of any one of claims 1-4.
7. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the composite material fatigue delamination prediction method of any one of claims 1-4.