CN115470669A - Composite material structure curing deformation prediction method and device and storage medium - Google Patents
Composite material structure curing deformation prediction method and device and storage medium Download PDFInfo
- Publication number
- CN115470669A CN115470669A CN202211047205.6A CN202211047205A CN115470669A CN 115470669 A CN115470669 A CN 115470669A CN 202211047205 A CN202211047205 A CN 202211047205A CN 115470669 A CN115470669 A CN 115470669A
- Authority
- CN
- China
- Prior art keywords
- deformation
- curing
- structural
- composite material
- laying angle
- 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
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
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- 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
- G06F2119/08—Thermal analysis or thermal optimisation
-
- 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)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Geometry (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Algebra (AREA)
- Moulding By Coating Moulds (AREA)
Abstract
The invention relates to a method, a device and a storage medium for predicting the structural curing deformation of a fiber reinforced composite material, wherein the method comprises the following steps: constructing a sample set based on a finite element numerical simulation method, and randomly determining a laying angle to obtain the sample set; carrying out data preprocessing on the sample set data; dividing a training set and a test set; building a structural solidification deformation prediction model based on a convolutional neural network, wherein the structural solidification deformation prediction model has the input of a laying angle color image and the output of a structural solidification deformation cloud picture; giving initial network structure parameters, hyper-parameters and loss functions of the convolutional neural network; performing regression learning on the structure solidification deformation prediction model based on the training set; saving the model; and verifying the prediction effect of the model based on the test set. Compared with the prior art, the method can effectively and quickly predict the curing deformation of the continuous fiber reinforced composite material structure, and solves the problem of difficult prediction of the curing deformation caused by various layering forms.
Description
Technical Field
The invention relates to the field of composite material structure design, in particular to a method and a device for predicting solidification deformation of a fiber reinforced composite material structure based on deep learning and a storage medium.
Background
Carbon fiber reinforced Composites (CFRPs) are a branch of advanced composites. Because of its excellent specific strength, specific modulus, strength and durability, it is widely used in aerospace manufacturing. Among these CFRPs have been used for the aft structure of a350, wing skin, struts, stringers and fuselage skin and the control surface, aft and nose structure of B787. Composite structures for aerospace use are known to be cured during the high temperature and pressure cycles of autoclave. This is a complex thermochemical process in which the matrix resin of the composite undergoes a crosslinking reaction to achieve a glassy state that provides the desired mechanical properties of the structure. Due to the mismatch of the thermo-mechanical properties between the fibers and the matrix, the curing shrinkage of the matrix during the cross-linking process, and the interaction between the tool and the component, the resin curing process often causes dimensional changes and residual stress accumulation in the final structure, and the problems of rebound deformation, buckling deformation and the like can occur after the part is cured. Solidification distortion is one of the important problems in the structural design stage, and the distortion causes large dimensional errors when manufacturing large-sized structures such as fuselages, wings, empennages, control surfaces and the like of large-sized passenger planes. The size deviation of the part can form an assembly gap at the joint surface during assembly, so that assembly stress is generated, and the service life, the performance and the reliability of a composite material part are influenced.
The current method for researching the curing deformation prediction of the fiber reinforced composite material is developed mainly by a finite element method based on several different constitutive models. In the thermo-chemical analysis, the distribution of the temperature and the curing degree field of the model is obtained based on a heat conduction equation and a curing equation. In the force-displacement analysis, several different constitutive models are based on:
1. the mechanical property of the composite material in the curing process is taken as a linear elastic model of a constant;
2. the CHILE model considers the correlation between the mechanical property and the temperature and the curing degree in the curing process of the composite material;
3. a viscoelastic model that takes into account stress relaxation or creep behavior of the resin that occurs in a high-temperature state;
4. allowing the Path-dependent model to be replaced with Path-dependent state variables (strain, degree of cure and temperature).
Despite these numerical methods, they are very powerful and provide insight into the complex mechanisms of the curing process based on a combination of information about the degree of cure, glass transition temperature, temperature and deformation. However, these models have high requirements on input data, and some performance data such as relaxation time and conversion factor require a large amount of experimental tests and are not suitable for rapid prediction. In the aerospace manufacturing industry, development of aerospace components typically relies on the experience of researchers and tedious experimentation to explore appropriate design parameters, making development inefficient and consuming a significant amount of time, money, and resources.
Disclosure of Invention
The invention aims to provide a method, a device and a storage medium for predicting the structural curing deformation of a fiber reinforced composite material based on deep learning, which overcome the defect of long calculation time of a finite element model, use the finite element model to generate a sample set in batch, and establish a mapping relation between a composite material laying angle and a given structural curing deformation cloud picture of a component through a convolutional neural network so as to realize the rapid prediction of the structural curing deformation cloud picture of the component.
The purpose of the invention can be realized by the following technical scheme:
a method for predicting the structural curing deformation of a fiber reinforced composite material based on deep learning comprises the following steps:
step 1) constructing an initial sample set of a continuous fiber reinforced composite structure based on a finite element numerical simulation method of thermal-chemical-force-displacement field sequential coupling, and randomly determining a laying angle corresponding to each sample in the initial sample set to obtain a sample set, wherein the data of the sample set comprises the laying angle, a node number, a node coordinate and a solidification deformation numerical value corresponding to the node coordinate;
step 2) carrying out data preprocessing on the sample set data, converting the laying angle into a laying angle color image, and matching the curing deformation value with the coordinate to obtain a structural curing deformation cloud picture;
step 3) dividing the preprocessed sample set into a training set and a testing set;
step 4), establishing a structural solidification deformation prediction model based on a convolutional neural network, wherein the structural solidification deformation prediction model is input into a laying angle color image and output into a structural solidification deformation cloud picture;
step 5) giving initial network structure parameters, hyper-parameters and loss functions of the convolutional neural network;
step 6) carrying out regression learning on the structure solidification deformation prediction model based on the training set, training samples by using a batch random gradient descent method, calculating model loss and precision, and iterating network structure parameters until the model is converged to obtain a mapping relation between a laying angle color image and a structure solidification deformation cloud map, and finishing training;
step 7), storing the trained structure curing deformation prediction model;
and 8) verifying the prediction effect of the trained structure curing deformation prediction model based on the test set.
The step 1) comprises the following steps:
step 1-1) based on finite element analysis software, giving thermal boundary of a fiber reinforced composite material component and thermal-chemical performance parameters of the material, and carrying out heat transfer analysis of a temperature field;
step 1-2) utilizing a sequential coupling method, giving constraints of a model and mechanical parameters and expansion coefficients of materials in finite element analysis software, and determining the elastic modulus E in the resin curing process based on a CHILE model m And shear modulus G m The performance parameters of the composite material are obtained by combining the mesomechanics mixing law of the composite material;
step 1-3) taking the heat transfer analysis result as a predefined field load, carrying out force-displacement analysis, and obtaining a node number, a node coordinate and a curing deformation value corresponding to the node coordinate by adopting finite element simulation calculation;
step 1-4) determining the number N of samples according to requirements;
step 1-5) randomly selecting a preset number of paving angles from the given paving angles as input angles of all samples in a sample set to generate N paving angle input quantities;
and 1-6) establishing a sample set based on the laying angle, the node number, the node coordinate and the solidification deformation value corresponding to the node coordinate.
The control equation of the temperature field is as follows:
where ρ is c Denotes the composite density, C c The specific heat capacity of the composite material is shown,
denotes the rate of temperature rise, k denotes the anisotropic heat transfer coefficient of the composite material,
representing the rate of heat generation upon curing of the resin:
wherein, alpha represents the degree of curing,
denotes the curing rate, c f Representing the fiber volume content, p r Denotes the resin density, H R Representing the total exotherm of the resin curing reaction.
The step 3) is specifically as follows: the method comprises the steps of normalizing the laying angle and the node coordinate in a sample set, converting the laying angle into a three-channel laying angle color image, converting the laying angle color image into a two-dimensional image matrix form, generating a structural solidification deformation cloud picture by using a function based on the normalized node coordinate and the solidification deformation numerical value, and converting the structural solidification deformation cloud picture into the two-dimensional image matrix form.
The initial network structure parameters comprise a neural network structure, convolution kernel size and filter number.
The loss function is the mean absolute percent error MAPE:
wherein the content of the first and second substances,
is a tag value, y (i) For the prediction value, m is the number of training set samples.
And 6) stopping iteration in the step 6) under the condition of loss function convergence, and in the iteration process, adjusting the size of a convolution kernel, the number of filters and a hyper-parameter of the structure curing deformation prediction model to enable the convergence value of the loss function to be smaller than a pre-configured threshold value.
And if the loss function is not converged within the pre-configured time, increasing the number of samples participating in training and re-training.
A deep learning based fiber reinforced composite structure solidification deformation prediction apparatus comprising a memory, a processor, and a program stored in the memory, the processor implementing the method when executing the program.
A storage medium having stored thereon a program which, when executed, implements the method as described above.
Compared with the prior art, the invention has the following beneficial effects:
(1) The method obtains the prediction result by establishing the mapping relation between the laying angle and the curing deformation value based on the convolutional neural network, does not need to obtain performance data which is difficult to obtain, does not need to accurately set parameters through manual experience, can obtain the curing deformation cloud picture of the continuous fiber reinforced composite material structure within a few seconds, has high calculation speed and high accuracy, and overcomes the defect of low calculation efficiency of the conventional finite element analysis method.
(2) The invention needs smaller training set samples and has better robustness for sample data outside the training set.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic view of a finite element model;
FIG. 3 is a schematic temperature profile of a curing process;
FIG. 4 is a graph comparing results of finite element analysis with experimental results;
FIG. 5 is a schematic diagram of a convolutional neural network;
FIG. 6 is a graph of loss of the structure solidification deformation prediction Model when it is trained under different hyper-parameters, wherein (a) is a loss function of Model1, (b) is a loss function of Model2, (c) is a loss function of Model3, (d) is a loss function of Model4, (e) is a loss function of Model5, and (f) is a loss function of Model 6;
FIG. 7 is a comparison of the lay angle input, finite element analysis results and the prediction results of the structural solidification deformation prediction model.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
A method for predicting the structural curing deformation of a fiber reinforced composite material based on deep learning is shown in figure 1 and comprises the following steps:
step 1) constructing an initial sample set of a continuous fiber reinforced composite structure based on a finite element numerical simulation method of thermal-chemical-force-displacement field sequential coupling, and randomly determining a laying angle corresponding to each sample in the initial sample set to obtain a sample set, wherein the data of the sample set comprises the laying angle, a node number, a node coordinate and a solidification deformation numerical value corresponding to the node coordinate;
step 1-1), establishing a finite element model, writing DISP and HETVAL subprograms, giving thermal boundary of a fiber reinforced composite material component and thermal-chemical performance parameters of the material in ABAQUS, and carrying out heat transfer analysis of a temperature field;
the governing equation of the temperature field is:
where ρ is c Denotes the composite density, C c The specific heat capacity of the composite material is shown,
denotes the rate of temperature rise, k denotes the anisotropic heat transfer coefficient of the composite material,
indicating the rate of heat generation during resin curing.
The resin cure exotherm is defined by the HETVAL subroutine and is calculated by the following procedure:
wherein, alpha represents the degree of curing,
denotes the curing rate, c f Denotes the fiber volume content, p r Denotes the resin density, H R Representing the overall exotherm of the resin curing reaction, for example, the curing reaction kinetics of AS4/3501-6 can be written AS follows;
wherein the content of the first and second substances,
wherein A is i Is Arrhenius constant, E i The activation energy of the autocatalysis model is shown, R is a universal gas constant, A is a frequency factor, and T is the system temperature.
(1-c f )ρ r H R ＝ρ c H c
Where ρ is c Is the composite density, H c Is the total exotherm during the curing of the composite;
ρ f is the fiber density.
Step 1-2) utilizing a sequential coupling method, giving constraints of a model and mechanical parameters and expansion coefficients of materials in finite element analysis software, and determining the elastic modulus E in the resin curing process based on a CHILE model m And shear modulus G m The performance parameters of the composite material are obtained by combining the mesomechanics mixing law of the composite material;
the composite material performance parameter solving calculation formula is as follows:
wherein, E m ，
The elastic modulus of the AS4/3501-6 resin at the present stage and the elastic modulus of the resin when the resin is completely cured are respectively;
wherein G is m Is the shear modulus, upsilon, of the resin at the present stage m Is the poisson's ratio of the resin;
wherein the content of the first and second substances,
is the elastic constant of the composite material;
wherein the content of the first and second substances,
the thermal expansion coefficients of the composite material, the resin and the fiber respectively;
and 1-3) taking the heat transfer analysis result as a predefined field load, carrying out force-displacement analysis, and obtaining a node number, a node coordinate and a solidification deformation value corresponding to the node coordinate by adopting finite element simulation calculation.
After the steps are completed, preparing a composite material laminated plate sample with the same size and layer as the finite element model by a vacuum bag pressing method, comparing the solidification deformation of the composite material laminated plate sample and the finite element model, and verifying the effectiveness of the finite element analysis model.
Step 1-4) determining the number N of samples according to the requirement;
step 1-5) compiling a Python script, randomly selecting a preset number of laying angles from given laying angles as input angles of all samples in a sample set, and generating N laying angle input quantities; writing Python scripts to rewrite inp files of a finite element model, and generating N inp files with laying angles to submit and calculate; compiling Python scripts, extracting node numbers and node coordinates of the topmost node in the odb files of the N models, and writing the node numbers and the node coordinates into a table respectively;
and 1-6) establishing a sample set based on the laying angle, the node number, the node coordinate and the solidification deformation value corresponding to the node coordinate.
And 2) carrying out data preprocessing on the sample set data, carrying out normalization processing on the laying angle and the node coordinate in the sample set, converting the laying angle into a three-channel laying angle color image, converting the laying angle color image into a two-dimensional image matrix form, generating a structure curing deformation cloud picture by using a function based on the normalized node coordinate and the curing deformation numerical value, and converting the structure curing deformation cloud picture into the two-dimensional image matrix form.
Step 3) dividing the preprocessed sample set into a training set and a testing set;
step 4) establishing a structural solidification deformation prediction model based on a convolutional neural network, wherein the structural solidification deformation prediction model is input into a laying angle color image and output into a structural solidification deformation cloud picture;
step 5) giving a neural network structure, a convolution kernel size, the number of filters, a hyperparameter and a loss function of the convolution neural network;
the loss function is the mean absolute percent error MAPE:
wherein the content of the first and second substances,
is a tag value, y (i) For the prediction value, m is the number of training set samples.
Step 6) carrying out regression learning on the structure solidification deformation prediction model based on the training set, training samples by using a batch random gradient descent method, calculating model loss and precision, and iterating network structure parameters until the model is converged to obtain a mapping relation between a laying angle color image and a structure solidification deformation cloud map, and finishing training;
and in the iteration process, if the value of the loss function is larger, the size of a convolution kernel of the structure curing deformation prediction model, the number of filters and the hyper-parameter are increased to enable the convergence value of the loss function to be smaller than a pre-configured threshold value.
If the loss function is not converged within the pre-configured time, increasing the number N of samples participating in the training in the step 1-4), and performing the training again.
Step 7), storing the trained structure curing deformation prediction model;
and 8) verifying the prediction effect of the trained structure curing deformation prediction model based on the test set.
The above functions, if implemented in the form of software functional units and sold or used as a separate product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present invention or a part thereof which substantially contributes to the prior art may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In one example, an AS4/3501-6 prepreg system was used, and the material parameters used in the simulation are shown in Table 1.
TABLE 1 AS4/3501-6 material parameters used in the finite element model.
As shown in fig. 2, a rectangular laminate with a geometry of 300 x 150 x 0.5mm was built up, dividing the geometry into four layers, each layer having a thickness of 0.125mm, creating a finite element model with 2888 elements. The temperature boundary conditions were applied to the lower panel by the DISP subroutine using the manufacturer's recommended cure process temperature curve in fig. 3.
12 samples were prepared in the above stacking sequence by a vacuum bag process, with lay-up angles of (a) [0/0/-45/45], (b) [90/90/-45/45], (c) [0/0/0/-45], (d) [0/-45/-45/0], (e) [ 45/-45/45/45/45 ], (f) [0/0/45/0], (g) [0/0/90/90], (h) [90/45/90/-45], (i) [45/90/0/0], (j) [90/90/45/0], (k) [45/0/45/90] and (l) [ -45/90/45/45 ], respectively. The size of the panel is consistent with that of the finite element, and the final thickness is 0.5 +/-0.08 mm. The curing process temperature profile used was the MRCPT profile shown in fig. 3, where the initial temperature was 25 ℃ and the heating/cooling rate was 2.5 ℃/min. The results of the experiment and the finite element analysis are consistent with each other as shown in fig. 4.
The number of samples N is determined to be 1200. Using the constructed thermodynamic model, a data set was constructed by numerical simulation. Using the random. Choice method of python language, 4 laying angles of the laminated board fibers are randomly selected from twelve angles of-75 °, -60 °, -45 °, -30 °, -15 °, 0 °, 15 °, 30 °, 45 °, 60 °, 75 ° and 90 °, and 1200 inp files are generated for submitting ABAQUS calculation. After the batch submission calculations of these models were completed, the data was extracted using the obdAccess library and the abaquusconstants library built into ABAQUS. And calculating node serial numbers, initial coordinates and average displacements of the top 780 nodes at the termination time, and writing the node serial numbers, the initial coordinates and the average displacements into a table as a sample set.
The sample set is divided into a training set and a testing set, which respectively account for 70% and 30% of the total number of samples. And carrying out normalization processing on the laying angle and the node coordinates, converting the laying angle and the solidification deformation value into an image matrix for input, and constructing and training a neural network by adopting a half model in consideration of symmetry.
A neural network structure is provided, the structure of which is shown in fig. 5. The cost function chosen in this example is The Mean Absolute Percentage Error (MAPE). The hyper-parameters of the model are shown in table 2.
TABLE 2 hyper-parameters for training convolutional neural networks
The loss function is plotted during the iteration of the training, as shown in fig. 6. And evaluating the training effect of the model according to the curve convergence condition. It can be seen from FIG. 6 that Model5 performs best, reaches the desired accuracy threshold, and does not require an increase in training set size.
Using the test set samples, the finite element analysis results were compared, as shown in fig. 7, where the lay angle for each sample was as follows: (a) [ -60/60/-75/60], (b) [ -45/15/-15/0], (c) [ -45/-30/90/45], (d) [60/45/60/-30], (e) [45/-75/-60/90], (f) [15/60/90/-75], (g) [0/-60/-30/15] and (h) [90/75/15/0]. According to fig. 7, the fact that the structural curing deformation prediction model based on the convolutional neural network has a good prediction effect on the structural curing deformation cloud picture is verified.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions that can be obtained by a person skilled in the art through logic analysis, reasoning or limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. The method for predicting the structural curing deformation of the fiber reinforced composite material based on deep learning is characterized by comprising the following steps of:
step 1) constructing an initial sample set of a continuous fiber reinforced composite structure based on a finite element numerical simulation method of thermal-chemical-force-displacement field sequential coupling, and randomly determining a laying angle corresponding to each sample in the initial sample set to obtain a sample set, wherein the data of the sample set comprises the laying angle, a node number, a node coordinate and a solidification deformation numerical value corresponding to the node coordinate;
step 2) carrying out data preprocessing on the sample set data, converting the laying angle into a laying angle color image, and matching the curing deformation numerical value with the coordinate to obtain a structural curing deformation cloud picture;
step 3) dividing the preprocessed sample set into a training set and a testing set;
step 4), establishing a structural solidification deformation prediction model based on a convolutional neural network, wherein the structural solidification deformation prediction model is input into a laying angle color image and output into a structural solidification deformation cloud picture;
step 5) giving initial network structure parameters, hyper-parameters and loss functions of the convolutional neural network;
step 6) carrying out regression learning on the structure solidification deformation prediction model based on the training set, training samples by using a batch random gradient descent method, calculating model loss and precision, and iterating network structure parameters until the model is converged to obtain a mapping relation between a laying angle color image and a structure solidification deformation cloud map, and finishing training;
step 7), storing the trained structure curing deformation prediction model;
and 8) verifying the prediction effect of the trained structure curing deformation prediction model based on the test set.
2. The method for predicting the structural curing deformation of the fiber reinforced composite material based on the deep learning of claim 1, wherein the step 1) comprises the following steps:
step 1-1) based on finite element analysis software, giving thermal boundary of a fiber reinforced composite material component and thermal-chemical performance parameters of the material, and carrying out heat transfer analysis of a temperature field;
step 1-2) utilizing a sequential coupling method, giving constraints of a model and mechanical parameters and expansion coefficients of materials in finite element analysis software, and determining the elastic modulus E in the resin curing process based on a CHILE model m And shear modulus G m The performance parameters of the composite material are obtained by combining the mesomechanics mixing law of the composite material;
step 1-3) taking the heat transfer analysis result as a predefined field load, carrying out force-displacement analysis, and obtaining a node number, a node coordinate and a curing deformation value corresponding to the node coordinate by adopting finite element simulation calculation;
step 1-4) determining the number N of samples according to requirements;
step 1-5) randomly selecting a pre-configured number of paving angles from the given paving angles as input angles of all samples in a sample set to generate N paving angle input quantities;
and 1-6) establishing a sample set based on the laying angle, the node number, the node coordinate and the solidification deformation value corresponding to the node coordinate.
3. The method for predicting the structural curing deformation of the fiber reinforced composite material based on the deep learning as claimed in claim 2, wherein the control equation of the temperature field is as follows:
where ρ is c Denotes the composite density, C c The specific heat capacity of the composite material is shown,
denotes the rate of temperature rise, k denotes the anisotropy of the compositeThe thermal coefficient of the material is as follows,
representing the rate of heat generation upon curing of the resin:
wherein, alpha represents the degree of curing,
denotes the curing rate, c f Denotes the fiber volume content, p r Denotes the resin density, H R Representing the total exotherm of the resin curing reaction.
4. The method for predicting the structural curing deformation of the fiber reinforced composite material based on the deep learning as claimed in claim 1, wherein the step 3) is specifically as follows: normalizing the laying angle and the node coordinate in the sample set, converting the laying angle into a three-channel laying angle color image, converting the laying angle color image into a two-dimensional image matrix form, generating a structural curing deformation cloud picture by using a function based on the normalized node coordinate and the curing deformation numerical value, and converting the structural curing deformation cloud picture into the two-dimensional image matrix form.
5. The method of claim 1, wherein the initial network structure parameters comprise neural network structure, convolution kernel size, and filter number.
6. The method according to claim 5, wherein the loss function is the mean absolute percent error MAPE:
wherein the content of the first and second substances,
is a tag value, y (i) For the prediction value, m is the number of training set samples.
7. The method for predicting the structural solidification deformation of the fiber reinforced composite material based on the deep learning of claim 5, wherein the condition for terminating the iteration in the step 6) is that the loss function is converged, and in the iteration process, the convergence value of the loss function is smaller than a preconfigured threshold value by adjusting the size of a convolution kernel, the number of filters and a hyper-parameter of the structural solidification deformation prediction model.
8. The method of claim 7, wherein if the loss function does not converge within a pre-configured time, the number of samples involved in the training is increased, and the training is repeated.
9. A deep learning based fibre-reinforced composite structure cure deformation prediction apparatus comprising a memory, a processor, and a program stored in the memory, wherein the processor when executing the program implements the method of any of claims 1-8.
10. A storage medium having a program stored thereon, wherein the program, when executed, implements the method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211047205.6A CN115470669A (en) | 2022-08-29 | 2022-08-29 | Composite material structure curing deformation prediction method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211047205.6A CN115470669A (en) | 2022-08-29 | 2022-08-29 | Composite material structure curing deformation prediction method and device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115470669A true CN115470669A (en) | 2022-12-13 |
Family
ID=84369513
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211047205.6A Pending CN115470669A (en) | 2022-08-29 | 2022-08-29 | Composite material structure curing deformation prediction method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115470669A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115630558A (en) * | 2022-12-21 | 2023-01-20 | 南京航空航天大学 | Method for predicting assembly deformation of composite material component |
| CN116108762A (en) * | 2023-04-13 | 2023-05-12 | 南京航空航天大学 | Large composite material component assembly deformation prediction method by using force sensor |
| CN116312898A (en) * | 2023-05-11 | 2023-06-23 | 中国电子科技集团公司信息科学研究院 | Method and device for identifying mechanical parameters of composite material and training identification model of composite material |
| CN116663374A (en) * | 2023-07-28 | 2023-08-29 | 北京理工大学 | Structural deformation prediction method and device for needled porous nanocomposite |
| CN117725707A (en) * | 2024-02-08 | 2024-03-19 | 北京理工大学 | Method and device for predicting solidification deformation of grid structural member |
| CN117727408A (en) * | 2024-02-08 | 2024-03-19 | 北京理工大学 | Curing deformation forecasting and optimizing method for composite material grid structure |
-
2022
- 2022-08-29 CN CN202211047205.6A patent/CN115470669A/en active Pending
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115630558A (en) * | 2022-12-21 | 2023-01-20 | 南京航空航天大学 | Method for predicting assembly deformation of composite material component |
| CN116108762A (en) * | 2023-04-13 | 2023-05-12 | 南京航空航天大学 | Large composite material component assembly deformation prediction method by using force sensor |
| CN116108762B (en) * | 2023-04-13 | 2023-06-27 | 南京航空航天大学 | Large composite material component assembly deformation prediction method by using force sensor |
| CN116312898A (en) * | 2023-05-11 | 2023-06-23 | 中国电子科技集团公司信息科学研究院 | Method and device for identifying mechanical parameters of composite material and training identification model of composite material |
| CN116312898B (en) * | 2023-05-11 | 2023-08-25 | 中国电子科技集团公司信息科学研究院 | Method and device for identifying mechanical parameters of composite material and training identification model of composite material |
| CN116663374A (en) * | 2023-07-28 | 2023-08-29 | 北京理工大学 | Structural deformation prediction method and device for needled porous nanocomposite |
| CN116663374B (en) * | 2023-07-28 | 2023-10-03 | 北京理工大学 | Structural deformation prediction method and device for needled porous nanocomposite |
| CN117725707A (en) * | 2024-02-08 | 2024-03-19 | 北京理工大学 | Method and device for predicting solidification deformation of grid structural member |
| CN117727408A (en) * | 2024-02-08 | 2024-03-19 | 北京理工大学 | Curing deformation forecasting and optimizing method for composite material grid structure |
| CN117727408B (en) * | 2024-02-08 | 2024-04-16 | 北京理工大学 | Curing deformation forecasting and optimizing method for composite material grid structure |
| CN117725707B (en) * | 2024-02-08 | 2024-04-19 | 北京理工大学 | Method and device for predicting solidification deformation of grid structural member |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115470669A (en) | Composite material structure curing deformation prediction method and device and storage medium | |
| Berthelot | Transverse cracking and delamination in cross-ply glass-fiber and carbon-fiber reinforced plastic laminates: Static and fatigue loading | |
| Jareteg et al. | Variation simulation for composite parts and assemblies including variation in fiber orientation and thickness | |
| Davidson et al. | Probabilistic defect analysis of fiber reinforced composites using kriging and support vector machine based surrogates | |
| Kolios et al. | Evaluation of the reliability performance of failure criteria for composite structures | |
| Kumari et al. | Static behavior of arbitrarily supported composite laminated cylindrical shell panels: an analytical 3D elasticity approach | |
| Fan et al. | A deep learning method for fast predicting curing process-induced deformation of aeronautical composite structures | |
| Åkermo et al. | Influence of interply friction on the forming of stacked prepreg | |
| Luo et al. | Rapid prediction and inverse design of distortion behaviors of composite materials using artificial neural networks | |
| Jagannathan et al. | Probabilistic strength based matrix crack evolution in multi-directional composite laminates | |
| Gholami et al. | Mechanical and failure analysis of thick composites under hygrothermal conditions by a novel coupled hygro-thermo-mechanical multiscale algorithm | |
| Milazzo et al. | Buckling and post-buckling of variable stiffness plates with cutouts by a single-domain Ritz method | |
| Shokrieh et al. | Modeling residual stresses in composite materials | |
| Campagna et al. | A non-linear Ritz method for progressive failure analysis of variable angle tow composite laminates | |
| Ma et al. | Effect of embedded periodic fiber placement gap defects on the microstructure and bistable behavior of thermoplastic composite laminates | |
| US20230153489A1 (en) | Systems and methods for semi-discrete modeling of delamination migration in composite laminate materials | |
| Patel et al. | Multiscale analysis of notched fiber reinforced laminates | |
| CRUZ et al. | Effect of thermal residual stresses on buckling and post-buckling properties of laminated composites perimetrally reinforced | |
| Pradhan et al. | Effect of material anisotropy and curing stresses on interface delamination propagation characteristics in multiply laminated FRP composites | |
| Mari et al. | Residual strength of wound composite pressure vessels subjected to fire exposure | |
| El Said et al. | Multiscale modelling of laminated composite structures with defects and features | |
| Pineda et al. | Multiscale model for progressive damage and failure of laminated composites using an explicit finite element method | |
| Abidin et al. | Validation of experimental hybrid natural/synthetic composite laminate specimen using finite element analysis for UAV wing application | |
| Mezeix et al. | Parameter study of tool-laminate interface through simulation for composite manufacturing using autoclave process | |
| Maksimović et al. | Structural Analysis and Optimization of Layered Composite Structures: Numerical and Experimental Investigations |
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 |