CN114611389A - Artificial intelligence-based efficient composite material failure simulation method - Google Patents
Artificial intelligence-based efficient composite material failure simulation method Download PDFInfo
- Publication number
- CN114611389A CN114611389A CN202210210663.0A CN202210210663A CN114611389A CN 114611389 A CN114611389 A CN 114611389A CN 202210210663 A CN202210210663 A CN 202210210663A CN 114611389 A CN114611389 A CN 114611389A
- Authority
- CN
- China
- Prior art keywords
- failure
- composite material
- angle
- model
- transverse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000002131 composite material Substances 0.000 title claims abstract description 71
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000004088 simulation Methods 0.000 title claims abstract description 24
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 8
- 230000006378 damage Effects 0.000 claims abstract description 34
- 230000000977 initiatory effect Effects 0.000 claims abstract description 16
- 238000003062 neural network model Methods 0.000 claims abstract description 11
- 238000000342 Monte Carlo simulation Methods 0.000 claims abstract description 5
- 238000010845 search algorithm Methods 0.000 claims description 11
- 238000007906 compression Methods 0.000 claims description 6
- 230000006835 compression Effects 0.000 claims description 6
- 239000000203 mixture Substances 0.000 claims description 6
- 238000013145 classification model Methods 0.000 claims description 4
- 238000009472 formulation Methods 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract description 24
- 238000004364 calculation method Methods 0.000 abstract description 11
- 238000004458 analytical method Methods 0.000 abstract description 7
- 210000002569 neuron Anatomy 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000004913 activation Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 239000000835 fiber Substances 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 239000000758 substrate Substances 0.000 description 3
- 238000009795 derivation Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 210000004205 output neuron Anatomy 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 101100005554 Mus musculus Ccl20 gene Proteins 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/26—Composites
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The invention provides an artificial intelligence-based efficient composite material failure simulation method, and belongs to the technical field of composite material structure failure analysis. The method comprises the following steps: carrying out non-dimensionalization on a failure criterion formula to obtain independent variables; generating a data set of independent parameter variables by adopting a Monte-Carlo method, and selecting a training data set of a neural network model based on a failure criterion; respectively constructing a neural network model for predicting the damage initiation and the crack angle of the composite material, and training; adopting an artificial neural network to preliminarily judge the damage initiation and the possible failure position of the composite material, and then further searching the failure specific position; and constructing a failure criterion subprogram in finite element software ABAQUS to simulate the composite material failure. The method ensures the prediction accuracy and precision, improves the calculation efficiency, and can realize the high-efficiency simulation of the composite material structure failure behavior by adopting Puck and LaRC05 failure criteria.
Description
Technical Field
The invention belongs to the technical field of composite material structure failure analysis, and relates to an artificial intelligence-based composite material failure high-efficiency simulation method.
Background
Composite materials are important structural materials with great potential in a variety of fields due to their excellent properties. However, the complex failure mechanism of the composite material structure prevents the full exertion of the mechanical properties. The numerical simulation method plays an important role in the design and strength analysis of the composite material structure, the structural efficiency can be obviously improved, and the cost and the design period are reduced. In the simulation of composite materials, the failure initiation criterion is of great significance for the accurate prediction of mechanical behavior.
Currently, multi-generation composite failure criteria have emerged, including the earlier maximum stress/strain criteria and the zea-hil, zea-wu criteria, etc., which then take into account the combined stress effects. With the deep research on failure mechanisms, failure criteria based on the failure mechanisms are provided, such as Hashin, Puck, LaRC series failure criteria and the like. The Puck and LaRC05 failure criteria are excellent in performance, and the failure envelope prediction is more accurate. In addition, these two failure criteria may give the direction of the substrate crack, providing a tool for simulation of the substrate crack. However, to determine the failure index and the fracture angle, a large number of iterative calculations are required, which severely limits their application in composite structure simulations.
In order to improve the computational efficiency, researchers have proposed various search algorithms to improve the computational efficiency. Various search algorithms are proposed based on the golden section search algorithm. By analyzing the characteristics of the failure index curve for the Puck failure criteria, a Select Range Golden Section Search (SRGSS) algorithm (ref.1: SCHIRMAIER F, WEILAND J, KARGER L, et al. A. New effectiveness and reliable optimization to determination of the details of the same for Puck's 3D matrix failure criterion for UD compositions [ J ]. compositions science and technology,2014,100: 19-25.) has been proposed. A Classified Local Golden Section Search (CLGSS) algorithm is proposed for the LaRC05 failure criterion (ref.2: Wang X, Guan Z, et al. an access and easy to implementation method for predicting matrix crack and plasticity of compositions with an effect Search for LaRC05 crack [ J ]. compositions Part A: Applied Science and efficiency, 2020,105808.). Under the condition of ensuring the calculation accuracy, the calculation time is reduced to about 20 percent of the original calculation time by the algorithms, and the calculation efficiency is obviously improved. In addition, many search algorithms based on other methods are provided, which have advantages in accuracy and efficiency.
Artificial Neural Networks (ANN), a deep learning method emerging in recent years, have theoretically been shown to fit any numerical formula, being good at dealing with repetitive tasks and internal relationships that are not yet clear. In recent years, artificial neural network models are gradually applied to the research of composite material mechanics. However, when the artificial neural network is used for simulating the finite element analysis of the composite material failure, a larger error still exists in the judgment of the composite material structure failure result, and the method is not suitable for practical application.
In the Puck and LaRC05 failure criteria, the determination of the failure index and the fracture angle needs a large amount of iterative calculation, and the method is very suitable for predicting by using an ANN model so as to improve the efficiency of performing composite material structure failure simulation by using the failure criteria.
Disclosure of Invention
Aiming at the problems that in the prior art, a large amount of iterative computation is needed for predicting composite material failure, the computation is long in time consumption, the application of the composite material in composite material structure simulation is limited, the accuracy is not enough when an artificial neural network is adopted for carrying out composite material failure finite element analysis, and the requirements needed by the composite material structure simulation cannot be met, the invention provides an artificial intelligence-based composite material failure initiation and fracture angle efficient simulation method, the neural network replaces a large amount of iterative computation, the computation time is reduced, the prediction precision is guaranteed, and the efficient simulation of composite material structure failure behaviors by adopting Puck and LaRC05 failure criteria is realized.
The invention provides an artificial intelligence-based composite material failure high-efficiency simulation method, which comprises the following steps:
(1) and carrying out non-dimensionalization on the failure criterion formula to obtain independent variables. Wherein the failure criterion is dimensionless by dividing the transverse normal stress and the transverse shear stress by the transverse tensile strength and dividing the longitudinal shear stress by the longitudinal shear strength. The obtained independent parameters will be used as input for the artificial neural network. K independent parameters are obtained, and K is a positive integer.
(2) A training data set of independent variables is generated. And generating a data set of independent variables by adopting a Monte-Carlo method, and selecting the data set as a training data set of the neural network.
(3) And (4) constructing two neural network models which are respectively used for predicting the damage initiation and the crack angle of the composite material. Wherein, the damage starting model is a two-classification model, and the crack angle model is a multi-classification model. The model is trained and validated using a training data set.
(4) Establishing a method for determining the damage initiation and angle of the composite material. And establishing a composite material damage initiation and crack angle determination method based on the trained neural network model and the golden section search algorithm. Acquiring the stress state of the composite material, carrying out non-dimensionalization on the stress according to the step 1 to obtain seven independent parameter values, inputting the seven independent parameter values into a trained damage initial model, judging whether damage occurs or not through the damage initial model, inputting the obtained seven independent parameter values into the trained crack angle model when the composite material is judged to be invalid, predicting the approximate position of the crack angle through a crack angle damage neural network, and determining the final crack angle through a golden section search algorithm for two angle ranges with the maximum probability in the result of predicting the crack angle.
(5) And constructing a failure criterion subprogram in finite element software ABAQUS to simulate the composite material failure. And (3) reconstructing a damage starting model and a crack angle model in a failure criterion subprogram, importing the damage starting model and the crack angle model trained in the step (3) into the subprogram, and realizing the method for determining the damage starting and the crack angle of the composite material in the step (4) in the subprogram.
In the (1), two independent parameters alpha are newly introduced for the LaRC05 failure criterionφAnd betaTCDetermining seven independent parameter variables: sigma22、σ33、τ23、τ12、τ13、αφAnd betaTC(ii) a Wherein σ22、σ33Representing the normal stress component on the corresponding plane of the coordinate system of the composite material; tau is23、τ12、τ13Representing the shear stress component on the corresponding plane of the composite material coordinate system; parameter alphaφ=tan(φ0),βTC=YT/YC,φ0For transverse compression failure angle, YTDenotes transverse tensile Strength, YCThe transverse compressive strength is indicated.
And (2) selecting independent parameter data corresponding to the failure indexes within the range of (0.8,1.1) to form a training data set.
In the step (3), the input of the damage starting model is K independent parameter values, and a classification parameter representing whether the composite material is failed or not is output; the input of the crack angle model is K independent parameter values, 10 classification parameters are output, and each classification parameter corresponds to an angle range.
Compared with the prior art, the invention has the advantages and positive effects that:
(1) in the Puck and LaRC05 failure criteria, the determination of failure indexes and fracture angles needs a large amount of iterative computation, the method replaces a large amount of iterative computation through a neural network, the computation efficiency is obviously improved, and the efficiency of performing composite material structure failure simulation by adopting the Puck and LaRC05 failure criteria is improved;
(2) the method of the invention constructs two neural network models, judges whether the composite material is damaged or not through the damage starting model, preliminarily judges the approximate position of the crack through the crack angle model, and acquires the accurate damage position by combining the searching method, thereby improving the calculation efficiency and ensuring the damage prediction precision of the composite material; meanwhile, the neural network model established by the method has certain independence, can be used together with a plurality of search methods, and can obviously improve the calculation efficiency.
Drawings
FIG. 1 is a flow chart of the composite material failure simulation based on the artificial neural network provided by the invention.
FIG. 2 is a schematic diagram of an artificial neural network architecture employed in the present invention;
FIG. 3 is a training curve and a loss curve of a crack angle neural network according to the present invention;
FIG. 4 is a flow chart of a method for efficiently determining damage initiation and crack angle of a composite material according to the present invention;
FIG. 5 is a graphical illustration of a typical failure indicator curve having a plurality of local maxima;
FIG. 6 is a comparison of substrate crack paths simulated using the original method and the proposed method in accordance with the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
According to the artificial intelligence-based composite material failure initiation and fracture angle efficient simulation method, firstly, the possibility and possible failure positions of composite material failure are preliminarily judged by adopting an artificial neural network, then the failure specific positions are further searched, the prediction accuracy and precision are ensured, meanwhile, the calculation efficiency is improved, and the efficient simulation of composite material structure failure behaviors by adopting Puck and LaRC05 failure criteria can be realized.
The embodiment of the invention relates to a composite material failure high-efficiency simulation method based on an artificial neural network, which is realized by aiming at a LaRC05 failure criterion. The overall flow of this example is shown in fig. 1, and the following steps are specifically described below with reference to the examples.
Step 1, carrying out non-dimensionalization derivation according to failure criteria, and determining the number of independent variables.
The embodiment of the invention takes the LaRC05 failure criterion as an example and carries out non-dimensionalized derivation. The coordinate system of the composite material is a 1-2-3 coordinate system, wherein the direction 1 is along the longitudinal direction of the fiber, the direction 2 is along the transverse direction of the fiber, and the direction 3 is the vertical direction of the fiber. The LaRC05 failure criteria may be expressed as follows:
wherein f ismatThe value is a failure index, when the value is less than 1, the failure is represented, and when the value is more than or equal to 1, the failure is represented; theta is an included angle between the failure surface and the thickness direction of the laminated plate; tau isnT(theta) represents a transverse shear stress, taunL(θ) represents a longitudinal shear stress; sigman(θ) represents the positive stress on the failure face; sTDenotes transverse shear strength, SLRepresents the longitudinal shear strength; y isTDenotes transverse tensile Strength, YCRepresents the transverse compressive strength; mu.sTDenotes the transverse shear friction coefficient, μLRepresents the longitudinal shear friction coefficient; sigma22、σ33Representing the positive stress on the corresponding plane of the composite material coordinate system; tau is23、τ12、τ13Representing the shear stress on the corresponding plane of the composite material coordinate system; phi is a0The angle between the failure plane and the thickness direction of the composite material under transverse compression is generally 50-55 degrees.
The invention aims at five stress components in the failure criterion and adopts different strengths to carry out dimensionless transformation. Transverse positive stress sigman(theta) and transverse shear stress taunT(theta) divided by transverse tensile strength YTLongitudinal shear stress divided by longitudinal shear strength SL. In addition, for the convenience of analysis, two independent parameters alpha are providedφAnd betaTCRespectively and transverse compression failure angle phi0And transverse tensile-compressive strength ratio(YT/YC) It is related. By deriving a non-dimensionalizing formula and seven independent variables that define the LaRC05 failure criteria. The dimensionless formulation of the LaRC05 failure criterion is as follows:
wherein the content of the first and second substances,the transverse shear stress, the longitudinal shear stress and the normal stress on the fracture surface are obtained after non-dimensionalization. The way in which the present invention performs dimensionless can be seen from equation (5).
Through the above non-dimensionalization process, seven independent variables σ can be determined22、σ33、τ23、τ12、τ13、αφAnd betaTCMeanwhile, the normalization of the numerical value is ensured, so that the method is suitable for various materials.
The PUCK criterion is dimensionless in the same manner as above, and the dimensionless formula is as follows:
wherein f ismatIs a failure index;is transverse shear stress, longitudinal shear stress and normal stress on a fracture surface obtained through non-dimensionalization; theta is an included angle between the failure surface and the thickness direction of the laminated plate; s. theLRepresents the longitudinal shear strength; y isTDenotes transverse tensile Strength, YCRepresents the transverse compressive strength; superscripts t, c represent tensile and compressive states, respectively; rtAnd RcIs an intermediate parameter, Rt,cIs RtOr Rc;γTLIs the ratio of transverse tensile strength to longitudinal shear strength; sin for medical use2ψ、cos2Psi is an intermediate parameter;is thatOrTo representOrAndrepresenting the slope information of the failure envelope at the intersection of tension and compression.
embodiments of the present invention continue with the LaRC05 failure criterion.
And 2, generating a data set of independent parameter variables by adopting a Monte-Carlo method based on a non-dimensionalization formula.
The specific process for generating the data set in the embodiment of the invention comprises the following steps: generation of random numbers of seven independent variables in MATLAB based on the Monte-Carlo method, wherein the shear stress component τ23、τ12、τ13Has a value range of (-2,2) and a positive stress component sigma22、σ33Has a value range of (-10,2), alphaφHas a value range of (0.2,0.3), betaTCHas a value range of (50, 55). In practical application, stress states near the failure time of the composite material are mainly concerned, so that generated data are screened, and a failure index f is selectedmatThe data at (0.8,1.1) is taken as data set data. In the embodiment of the invention, 100 ten thousand groups of data meeting the requirements are generated by adopting MATLAB. Each training sample includes seven independent variable values, and the samples are labeled as theta and a failure indicator.
And 3, constructing an artificial neural network model and training.
According to the embodiment of the invention, an open source artificial neural network library Keras is selected for neural network modeling and training, and neural network models are respectively established aiming at the damage initiation and crack angles of the composite material.
In the embodiment of the invention, the damage initiation model is a four-layer network, which comprises an output layer, two hidden layers and an output layer, wherein the two hidden layers and the output layer are respectively provided with 7 neurons, 20 neurons, 40 neurons and 1 neuron, the hidden layer activation function is 'ReLU', and the output layer activation function is 'Sigmoid'. A schematic diagram of the neural network of the injury initiation model is shown in fig. 2. The input of the damage initiation model is seven independent variable values, the output is a classification parameter, when the value of the classification parameter is 0, the composite material does not fail, and when the value of the classification parameter is 1, the composite material fails.
In the embodiment of the invention, the crack angle model is also a four-layer network, the four-layer neurons are respectively 7, 30, 90 and 10 neurons, wherein 10 output neurons respectively correspond to [0,18 ], [18,36) and … [162,180 ]; the hidden layer activation function is "Tanh" and the output layer activation function is "Softmax". The input of the crack angle model is seven independent variable values, the output is ten classification parameters which correspond to 10 output neurons, each neuron corresponds to an angle range, and the sum of the 10 classification parameter values is 1. If one of the output classification parameters is equal to 1, the crack angle falls within the range.
In the embodiment of the invention, the two established neural networks both adopt an Adam algorithm as an optimizer, and the learning rate is set to be 0.05. And (3) training by adopting the data set obtained in the step (2). The training curve and the loss curve of the crack angle model are shown in fig. 3. According to the embodiment of the invention, the prediction accuracy of the two trained models is over 95%.
And 4, efficiently determining the damage initiation and the crack angle of the composite material.
The invention establishes an efficient determination method based on a trained neural network model and a gold search algorithm, as shown in fig. 4. First, after the stress state of the composite material is obtained, the stress state is subjected to non-dimensionalization according to a non-dimensionalization formula. And then, inputting the obtained 7 parameters into a damage starting model for damage prediction, if no damage exists, directly ending, and if the damage exists, predicting a failure angle. And (4) inputting the 7 parameters corresponding to the stress state into a crack angle model, and predicting the range of the crack angle. Considering the condition that the composite material failure index has a plurality of extreme values, the method selects two largest angle ranges in the crack angle model prediction result to respectively solve the local maximum values, and determines the maximum values through comparison. After the angle range is determined, the embodiment of the invention adopts the golden section algorithm to solve, and finally outputs the failure index and the crack angle obtained by calculation.
As shown in fig. 5, there are multiple local maxima for the failure index curve. At the moment, two angle ranges with the maximum probability in the prediction results of the crack angle model are selected, local maximum values in the two angle ranges are determined by adopting a golden section search algorithm for the two angle ranges respectively, and finally the global maximum value and the specific angle are determined by comparing the two local maximum values. The obtained global maximum value is the failure index, and the specific angle is the position where the crack angle appears.
And 5, realizing an efficient search algorithm in finite element software ABAQUS, and verifying the accuracy and the efficiency of the method by simulating typical failures. The method specifically comprises the following steps: efficient simulation is achieved by a neural network algorithm that implements the lesion initiation model and the crack angle model in the ABAQUS subroutine. Firstly, outputting the weight matrix of the trained model, and importing the weight matrix into a Fortran subprogram. The structure of the neural network is then realized by the "matmul", "tanh" and "exp" functions in Fortran. The efficient determination method of the invention is realized by programming. The failure simulation of the composite material under transverse shear load is adopted to verify the model, the XFEM method is adopted to simulate the crack path of the composite material, and the simulated crack path is shown in figure 5.
As can be seen from the comparison of the left and right images in FIG. 5, the method of the present invention can accurately predict the crack path of the composite material. Through simulation time-consuming analysis, it was found that the time-consuming using the method of the present invention was about 78s, which was only 4% of the original method, which was about 1922 s. The comparison result shows the accuracy and the high efficiency of the method provided by the invention.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the technical scope of the present invention.
Claims (7)
1. An artificial intelligence-based composite material failure efficient simulation method is characterized by comprising the following steps:
step 1, dividing transverse normal stress and transverse shear stress by transverse tensile strength, dividing longitudinal shear stress by longitudinal shear strength, carrying out non-dimensionalization, and obtaining independent parameters from a non-dimensionalized failure criterion, wherein K is the number of the obtained independent parameters, and K is a positive integer;
step 2, generating a data set of independent parameter variables by adopting a Monte-Carlo method, and selecting a training data set of the neural network model based on a failure criterion;
step 3, constructing two neural network models which are respectively used for predicting the damage initiation and the crack angle of the composite material; the damage starting model is a two-classification model, and the crack angle model is a multi-classification model; training two neural network models by using a training data set;
inputting K independent parameter values into the damage starting model, and outputting a classification parameter representing whether the composite material fails; the input of the crack angle model is K independent parameter values, 10 classification parameters are output, and each classification parameter corresponds to an angle range;
step 4, acquiring a stress state of the composite material, carrying out non-dimensionalization on the stress according to the step 1 to obtain K independent parameter values, inputting the K independent parameter values into a trained damage initial model, inputting the K independent parameter values into the trained crack angle model when judging that the composite material fails, predicting the range of a crack angle, and determining the final crack angle by adopting a golden section search algorithm for two angle ranges with the maximum probability in the result of predicting the crack angle;
and 5, constructing a failure criterion subprogram in the finite element software ABAQUS, reconstructing a damage starting model and a crack angle model in the subprogram, importing the damage starting model and the crack angle model trained in the step 3 into the subprogram, and realizing the method for determining the damage starting and the angle of the composite material in the step 4 in the subprogram.
2. The method of claim 1, wherein in step 1, two parameters α are newly introduced for LaRC05 failure criterionφAnd betaTCDetermining K as 7 independent parameter variables: sigma22、σ33、τ23、τ12、τ13、αφAnd betaTC(ii) a Wherein σ22、σ33Representing the normal stress component on the corresponding plane of the composite material coordinate system; tau is23、τ12、τ13Representing the shear stress component on the corresponding plane of the composite material coordinate system; parameter alphaφ=tan(φ0),βTC=YT/YC,φ0For transverse compression failure angle, YTDenotes transverse tensile Strength, YCThe transverse compressive strength is indicated.
3. The method according to claim 2, wherein in step 1, the LaRC05 failure criteria are dimensionless to obtain:
wherein, fmatIs a failure index; theta is an included angle between the failure surface and the thickness direction of the laminated plate; tau isnT(theta) represents a transverse shear stress, taunL(θ) represents a longitudinal shear stress; sigman(θ) represents a positive stress on the failure face; sTDenotes transverse shear strength, SLRepresents the longitudinal shear strength; mu.sTDenotes the transverse shear friction coefficient, μLRepresents the longitudinal shear friction coefficient;
4. The method according to claim 1, wherein in step 1, two parameters β are introduced for the PUCK failure criterionTCAnd gammaTLDetermining K as 11 independent parameter variables: sigma22、σ33、τ23、τ12、τ13、 βTCAnd gammaTL(ii) a Wherein σ22、σ33Representing the normal stress component on the corresponding plane of the composite material coordinate system; tau is23、τ12、τ13Representing the shear stress component on the corresponding plane of the composite material coordinate system; parameter gammaTL=YT/SL,βTC=YT/YC,YTDenotes transverse tensile Strength, YCDenotes the transverse compressive strength, SLRepresents the longitudinal shear strength;andis the slope information of the failure envelope at the stretching and compression junction.
5. The method according to claim 2 or 3, wherein in step 2, the shear stress component τ is generated as training data for the independent parameters of the LaRC05 failure criterion23、τ12、τ13Has a value range of (-2,2) and a positive stress component sigma22、σ33Has a value range of (-10,2), alphaφHas a value range of (0.2,0.3), betaTCHas a value range of (50, 55).
6. The method according to claim 1 or 2, wherein in step 2, the independent parameter data corresponding to the failure index in the range of (0.8,1.1) is selected to form a training data set.
7. The method according to claim 1 or 2, wherein in the step 4, two local maximum values of the failure index are determined by using golden section search algorithm for two angle ranges respectively, then a global maximum value of the failure index is determined by comparing the two local maximum values, and a crack angle corresponding to the global maximum value and the failure index are output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210210663.0A CN114611389A (en) | 2022-03-04 | 2022-03-04 | Artificial intelligence-based efficient composite material failure simulation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210210663.0A CN114611389A (en) | 2022-03-04 | 2022-03-04 | Artificial intelligence-based efficient composite material failure simulation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114611389A true CN114611389A (en) | 2022-06-10 |
Family
ID=81861787
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210210663.0A Pending CN114611389A (en) | 2022-03-04 | 2022-03-04 | Artificial intelligence-based efficient composite material failure simulation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114611389A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115906570A (en) * | 2022-11-22 | 2023-04-04 | 大连理工大学 | Composite material open-hole laminated plate progressive damage prediction method based on data driving |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111368389A (en) * | 2019-10-11 | 2020-07-03 | 暨南大学 | Method for predicting failure strength of composite material laminated plate |
WO2020143284A1 (en) * | 2019-01-11 | 2020-07-16 | 华东理工大学 | Abaqus-based multiaxial creep fatigue prediction method |
CN111709174A (en) * | 2020-06-18 | 2020-09-25 | 江南大学 | Composite material laminated plate strength analysis method based on failure surface theory |
CN112818582A (en) * | 2021-03-24 | 2021-05-18 | 上海邦麟复合材料科技有限公司 | Fiber reinforced composite material performance judgment method and application thereof |
CN113706511A (en) * | 2021-08-31 | 2021-11-26 | 佛山市南海区广工大数控装备协同创新研究院 | Composite material damage detection method based on deep learning |
CN114117840A (en) * | 2021-10-28 | 2022-03-01 | 中国运载火箭技术研究院 | Structural performance prediction method based on simulation and test data hybrid drive |
-
2022
- 2022-03-04 CN CN202210210663.0A patent/CN114611389A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020143284A1 (en) * | 2019-01-11 | 2020-07-16 | 华东理工大学 | Abaqus-based multiaxial creep fatigue prediction method |
CN111368389A (en) * | 2019-10-11 | 2020-07-03 | 暨南大学 | Method for predicting failure strength of composite material laminated plate |
CN111709174A (en) * | 2020-06-18 | 2020-09-25 | 江南大学 | Composite material laminated plate strength analysis method based on failure surface theory |
CN112818582A (en) * | 2021-03-24 | 2021-05-18 | 上海邦麟复合材料科技有限公司 | Fiber reinforced composite material performance judgment method and application thereof |
CN113706511A (en) * | 2021-08-31 | 2021-11-26 | 佛山市南海区广工大数控装备协同创新研究院 | Composite material damage detection method based on deep learning |
CN114117840A (en) * | 2021-10-28 | 2022-03-01 | 中国运载火箭技术研究院 | Structural performance prediction method based on simulation and test data hybrid drive |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115906570A (en) * | 2022-11-22 | 2023-04-04 | 大连理工大学 | Composite material open-hole laminated plate progressive damage prediction method based on data driving |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112131760B (en) | CBAM model-based prediction method for residual life of aircraft engine | |
CN109376242B (en) | Text classification method based on cyclic neural network variant and convolutional neural network | |
Al-Assaf et al. | Fatigue life prediction of unidirectional glass fiber/epoxy composite laminae using neural networks | |
CN112800172B (en) | Code searching method based on two-stage attention mechanism | |
Zhang et al. | Prediction of FRP-concrete interfacial bond strength based on machine learning | |
Yoshimura et al. | New regularization by transformation for neural network based inverse analyses and its application to structure identification | |
CN106096723A (en) | A kind of based on hybrid neural networks algorithm for complex industrial properties of product appraisal procedure | |
CN114611389A (en) | Artificial intelligence-based efficient composite material failure simulation method | |
CN115982141A (en) | Characteristic optimization method for time series data prediction | |
CN114239397A (en) | Soft measurement modeling method based on dynamic feature extraction and local weighted deep learning | |
CN111695260B (en) | Material performance prediction method and system | |
CN110263125B (en) | Service discovery method based on extreme learning machine | |
CN112231455A (en) | Machine reading understanding method and system | |
CN117034699A (en) | Metal fatigue S-N curve prediction method based on tensile test | |
Wang et al. | Data-driven inverse method with uncertainties for path parameters of variable stiffness composite laminates | |
CN115392434A (en) | Depth model reinforcement method based on graph structure variation test | |
CN113835739A (en) | Intelligent prediction method for software defect repair time | |
Papadrakakis et al. | Advances in structural optimization | |
CN111199779A (en) | Virtual drug screening method and device based on molecular docking | |
CN114065628B (en) | Auxiliary laser protective coating material selection and design method and system | |
CN117236374A (en) | Layering interpretation method based on fully developed material graph neural network | |
CN116959647A (en) | Method for predicting mechanical properties of micro-nano composite material based on finite element analysis | |
CN115132279A (en) | Depth residual error shrinkage network-based multi-omics and phenotype association prediction method | |
Wang et al. | Research and application of BP algorithm based on genetic algorithm in system performance bottleneck analysis | |
CN117828744A (en) | BP neural network-based wing structure optimization design method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |