A design method and manufacturing method for composite profiled part customization
By using asymmetric layup design and neural network prediction model, the problem of curing deformation of composite parts during the molding process was solved, enabling the production of composite irregular parts with specified surfaces from planar molds, reducing costs and improving surface accuracy and assembly accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AEROSPACE INST OF ADVANCED MATERIALS & PROCESSING TECH
- Filing Date
- 2022-10-27
- Publication Date
- 2026-05-19
AI Technical Summary
During the molding process of fiber-reinforced resin matrix composite parts, the prominent anisotropy makes deformation control difficult, resulting in curing deformation of composite components after demolding, which affects product performance and assembly difficulty.
By employing an asymmetric layup design combined with finite element simulation and artificial neural network prediction models, composite material irregular parts with specified surfaces are produced from a planar mold by simulating the curing deformation behavior of composite materials. Deformation prediction and reverse design are then performed using finite element software and neural network models.
This method enables the production of composite material irregular parts with specified surfaces from planar molds, reducing costs, improving surface accuracy and assembly accuracy, and reducing part weight, thus avoiding the weight increase problem caused by adding more layers in traditional methods.
Smart Images
Figure CN115659806B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of composite material application and manufacturing technology, and in particular to a design method and manufacturing method for customizing the surface of composite material irregular parts. Background Technology
[0002] Resin-based composite materials are novel materials composed of reinforcing and matrix materials. These materials retain the main characteristics of their original components while acquiring properties not found in the original components through a composite effect. They are widely used in aerospace, automotive engineering, shipbuilding, wind power, and sporting goods industries. However, during the molding process of fiber-reinforced resin-based composite parts, their prominent anisotropy makes deformation control difficult, leading to curing deformation after demolding. Curing deformation is a significant defect in the autoclave molding process of fiber-reinforced resin-based composites, causing assembly difficulties, residual stress, or directly affecting product performance. To date, methods such as avoiding asymmetric layups or mold compensation are commonly used to reduce or avoid undesirable curing deformation in composite parts. However, simply increasing the number of layers to achieve layup symmetry and thus reduce curing deformation actually increases the weight of the composite structure, thereby negating the advantages of high performance.
[0003] To address the aforementioned issues, it is essential to provide a custom design for the profile of composite material irregular parts and a molding method that matches this design. By employing an asymmetric layup design, the potential advantages of composite material parts can be maximized, thereby molding composite material irregular components with specified profile requirements. Summary of the Invention
[0004] In order to overcome the problem that existing technologies cannot produce curved parts with specified shapes from planar molds, this invention provides a design method for customizing the surface of composite material irregular parts. It utilizes the curing deformation behavior of composite material parts to obtain composite material irregular parts with specified surface requirements from planar molds, and proposes a matching molding and manufacturing method.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] The first aspect of this invention provides a design method for customizing the surface of composite material irregular parts, comprising the following steps:
[0007] 1) Establish a model of the composite material flat plate, which includes the length and width information of the composite material flat plate;
[0008] 2) In the finite element software, the flat plate model is meshed;
[0009] 3) For the flat plate model, a composite material card is established. This card is a set of parameters consisting of the corresponding mechanical property parameters and thermal expansion coefficients of the composite single-layer plate in the three directions of in-plane axial direction, in-plane transverse direction, and thickness direction.
[0010] 4) Establish composite material layup information based on the card, including the layup area, layup angle, and layup direction of each layer; and establish a link between the composite material card and the composite material layup information;
[0011] 5) Establish the constraint boundary conditions and temperature load conditions for the flat plate model; where steps 2), 3), and 5) are parallel operations with no specific order.
[0012] 6) Based on the above-described mesh, cards, layup information, constraint boundary conditions, and temperature load conditions, simulate the curing deformation behavior of the flat plate model, establish a curing deformation model of the composite material flat plate, and calculate the final residual stress and deformation state dataset of the flat plate model.
[0013] 7) Based on the dataset obtained in step 6), establish a knowledge base for the curing deformation of composite flat plate parts;
[0014] 8) Construct a neural network prediction model for the curing deformation of composite flat plate parts, and use the dataset from the knowledge base in step 7) to train and validate the neural network prediction model;
[0015] 9) Predict the deformation state and maximum deformation corresponding to batch material parameter combinations using the neural network prediction model, and obtain material parameter combinations with target deformation through a large-scale screening method to complete the design. Each set of material parameter combinations corresponds to a composite material card, representing a type of composite material.
[0016] Preferably, in step 1), the composite material is a continuous fiber reinforced thermosetting resin-based composite material, wherein the fiber is selected from carbon fiber, glass fiber, or aramid fiber, and the resin is selected from high-temperature epoxy resin, reheat epoxy resin, bismaleimide resin, polyimide resin, or cyano resin.
[0017] Preferably, the finite element software mentioned in step 2) is Abaqus software.
[0018] Preferably, the mesh described in step 2) is a shell element mesh, and the shell element is an S4R linear reduced integral element.
[0019] Preferably, the mechanical performance parameters mentioned in step 3) include the elastic model, shear modulus, and Poisson's ratio.
[0020] Preferably, the constraint boundary conditions in step 5) include constraint position and constraint direction; the temperature load conditions include loading position, loading temperature and initial temperature, wherein the loading position is the entire area of the composite material flat plate model, and the loading temperature is the resin curing temperature of the composite material.
[0021] Preferably, the knowledge base described in step 7) is operated using a Python script batch command.
[0022] Preferably, in step 8), the dataset of the knowledge base is preprocessed, and then feature selection learning is performed on the preprocessed dataset to obtain the influence weight of each material parameter on the maximum deformation. Material parameters with a large influence weight on the maximum deformation rate are selected to form a training dataset. The training dataset is used to train and validate the neural network prediction model to obtain a validated neural network prediction model.
[0023] Preferably, the feature selection learning adopts an embedding selection method based on the L1 regularization algorithm.
[0024] Preferably, the neural network prediction model adopts an artificial neural network regression model, specifically the Scikit-learn software.
[0025] Preferably, the training and validation of the neural network prediction model employs a grid search and 5-fold cross-validation method.
[0026] Preferably, the large-scale screening method in step 9) is as follows: first, all possible candidate materials are generated in the design space, and then the neural network prediction model is used to test them one by one; after the candidate materials are generated, their properties can be easily and directly evaluated using the neural network prediction model.
[0027] A second aspect of the present invention provides a method for customizing the surface of a composite material irregular part, comprising the following steps:
[0028] 1) Obtain a combination of material parameters using the method provided in the first aspect of the present invention, and select the corresponding composite material and layup information based on the combination of material parameters;
[0029] 2) Clean the mold, lay the selected composite material on the mold according to the above layup information, and then bag it.
[0030] 3) After bagging, the prepreg is pressurized and heated to fully cure it. Then, after demolding, composite material parts are obtained.
[0031] Preferably, the bagging operation uses a release film, breathable felt, release cloth and sealing strip to seal the prepreg in a vacuum bag, and installs a vacuum nozzle to perform a vacuuming operation on the prepreg, so that the prepreg is completely isolated from the air.
[0032] Preferably, the pressurization and heating operations are carried out in an autoclave or an oven, the heating operation is selected from the curing temperature curve corresponding to the prepreg, and the pressurization operation is selected from the pressure curve corresponding to the prepreg.
[0033] Preferably, after the demolding operation, the composite material irregular part needs to be deburred.
[0034] The above-described technical solution of the present invention has the following advantages:
[0035] This invention provides a design method for customizing the surface profile of composite material irregular-shaped parts. Utilizing the curing deformation behavior of composite material parts, it obtains composite material irregular-shaped components with specified surface requirements from a planar mold and proposes a matching molding method. This invention applies artificial neural network learning to the prediction and reverse design of composite material curing deformation. By combining finite element simulation analysis and artificial neural network learning, it designs composite material irregular-shaped structural parts with specified surface requirements. Based on this design result, reverse design can be performed on the curing deformation. According to the known target deformation amount, a combination of material parameters with that target deformation amount is selected, and the corresponding material is selected based on the selected material parameter combination, thereby realizing the production of composite material irregular-shaped structural parts with specific shapes from a planar mold. Compared with traditional forward curing deformation control methods, this invention's design method does not require specialized curved surface molds, resulting in lower costs. Simultaneously, the manufactured parts have significant advantages such as greater weight reduction, higher surface accuracy, and better assembly precision. Attached Figure Description
[0036] The accompanying drawings are provided for illustrative purposes only, and the proportions of the components in the drawings may not be consistent with the actual product.
[0037] Figure 1 This is a design flowchart for customizing the surface of a composite material irregular part according to an embodiment of the present invention.
[0038] Figure 2 This is a schematic diagram of an artificial neural network model in an embodiment of the present invention.
[0039] Figure 3 This is a manufacturing process diagram for customizing the surface of a composite material irregular part according to an embodiment of the present invention.
[0040] Figures 4A-4B This is a schematic diagram of the shape change during the theoretical manufacturing process of this invention.
[0041] In the figure: 1-flat mold; 2-composite prepreg; 3-composite structural component with specified profile requirements.
[0042] Figures 5A-5B This is a picture of the actual manufactured product in an embodiment of the present invention. Detailed Implementation
[0043] To make the technical solution of the present invention more obvious and understandable, specific embodiments are provided and described in detail below with reference to the accompanying drawings.
[0044] This embodiment discloses a novel design method and manufacturing method for customizing the surface of composite material irregular parts. The purpose is to overcome the problem that existing technologies cannot produce curved surface parts with specified shapes from planar molds. By utilizing the curing deformation behavior of composite material parts, composite material irregular parts with specified surface requirements can be obtained from planar molds.
[0045] Figure 1 This is a flowchart illustrating the design process for customizing the profile of a composite material part used in this embodiment of the invention, comprising the following steps:
[0046] Step S1: Create a model of the composite material flat plate, which includes the length and width information of the composite material flat plate.
[0047] Step S2: In the finite element software Abaqus, the flat plate model is meshed; the mesh uses shell element mesh, which is an S4R linear reduced integral element.
[0048] Step S3: For the flat plate model, create composite material cards, one card corresponding to one material. The card information includes a set of parameters including the mechanical properties and coefficient of thermal expansion of the composite single-layer plate in the three directions of in-plane axial direction, in-plane transverse direction, and thickness direction. In terms of material selection, the composite material is a continuous fiber reinforced thermosetting resin matrix composite material. The fibers are carbon fiber, glass fiber, aramid fiber, etc., and the resins are high-temperature epoxy resin, reheat epoxy resin, bismaleimide resin, polyimide resin, cyano resin, etc.
[0049] Step S4: Establish composite material layup information based on the card. This layup information includes the layup area, layup angle, and layup direction of each layer. The fiber layup sequence of the composite flat sheet is [90 / 90 / 45 / -45], and the size is 300×150mm. 2 And establish a link between the composite material card and the composite material layup information.
[0050] Step S5: Establish the constraint boundary conditions and temperature load conditions for the flat plate model. Regarding the selection of boundary conditions and initial conditions, the constraint boundary conditions include constraint location and constraint direction, and the temperature load conditions include loading location, loading temperature, and initial temperature.
[0051] Step S6: Based on the above-described mesh, cards, layup information, constraint boundary conditions, and temperature load conditions, simulate the curing deformation behavior of the flat plate model, establish a curing deformation model of the composite material flat plate, and calculate a large-scale dataset of the final residual stress and deformation state of the flat plate model.
[0052] Step S7: Establish a knowledge base for the curing deformation of composite flat plates. This knowledge base is formed using a large-scale dataset as described in Step S6. For software selection, the knowledge base is operated using Python script batch commands. Regarding dataset selection, sampling data is first generated uniformly, and then different material parameter samples are randomly selected using a random permutation algorithm to form the material input parameters for subsequent simulations, generating approximately 3400 sets of data. During the knowledge base establishment process, to improve computational efficiency, these 3400 simulation cases are automatically created and submitted for calculation using Python scripts and batch calculation files, and the corresponding finite element analysis results are automatically extracted using Python scripts.
[0053] Step S8: Establish a neural network prediction model for the curing deformation of composite flat plates. The neural network prediction model is trained and validated using the dataset obtained in Step S7. Regarding dataset selection, the large-scale dataset obtained from the knowledge base in Step S7 is processed to obtain a new dataset. Feature selection learning is then performed on this dataset to obtain the influence weights of each material parameter on the maximum deformation. Material parameters with significant influence weights on the maximum deformation rate are selected to form a training dataset. This training dataset contains six input features (i.e., the aforementioned material parameters with significant influence weights) and one output feature. The input features are Young's modulus in the fiber direction, Young's modulus perpendicular to the fiber direction in the plane, shear modulus in the plane, coefficient of thermal expansion in the fiber direction in the plane, coefficient of thermal expansion perpendicular to the fiber direction in the plane, and curing temperature. The output feature is the maximum curing deformation. Specifically, in model selection, this embodiment uses an artificial neural network regression model, such as... Figure 2 As shown in the figure (x represents the input feature, and y represents the output feature), the model was trained using Scikit-learn software. For model validation, the neural network prediction model was trained and validated using grid search and 5-fold cross-validation. During training, the structure and hyperparameters of the artificial neural network model were optimized to obtain the best prediction accuracy.
[0054] In this embodiment, the optimal neural network model obtained is Regression Network 1 (with 90 hidden layer neurons, an optimization algorithm of 'L-BFGS', an activation function of 'Tanh', and a regularization penalty coefficient α of 10). -7 R shows good performance on the test set. 2 The accuracy is above 99%, and the RMSE value is below 6%. Based on this regression network 1, predictions were made for two groups of new materials, and the results show that the error between the neural network model's predictions and the finite element analysis results is less than 2%.
[0055] Step S9: Reverse design of the profile customization. The neural network prediction model predicts the deformation state and maximum deformation of all possible material combinations, and a large-scale screening method is used to obtain possible material combinations with the target deformation. The large-scale screening method involves first generating all possible candidate materials in the design space, and then testing each one using the optimal neural network model. After generating candidate materials, their properties can be easily and directly evaluated using the optimal neural network model.
[0056] In this embodiment, firstly, a new virtual dataset (containing more than 4.4 million materials) is created using loop code. Then, taking the best neural network model (regression network 1) as an example, the maximum curing deformation of different materials is quickly calculated. The six input features (material parameters) and output features (maximum curing deformation) given in the virtual dataset are sorted. Then, the input features whose output features are the same as the target maximum curing deformation are selected.
[0057] In this embodiment, the target maximum curing deformation was set to 120 mm. It was found that approximately 33,000 material groups met this requirement within the range of 120 ± 0.5 mm. To narrow down the candidate materials, this embodiment comprehensively considered factors such as mechanical properties and curing process conditions. Referring to the material properties of T800 carbon fiber / high-temperature epoxy resin composites, the number of candidate materials was 24. Among the candidate materials with the specified maximum curing deformation obtained through reverse engineering, a T800 / epoxy composite material with similar performance was selected from commercially available materials. The in-plane fiber-direction Young's modulus, in-plane perpendicular-fiber-direction Young's modulus, in-plane shear modulus, in-plane fiber-direction thermal expansion coefficient, in-plane perpendicular-fiber-direction thermal expansion coefficient, and curing temperature of this material were 163 GPa, 8.7 GPa, 4.21 GPa, and 2.22 × 10⁻⁶, respectively. -6 1 / ℃, 38.67×10 -6 1 / ℃ and 180℃.
[0058] This embodiment also discloses a manufacturing method for customizing the surface of composite material irregular-shaped parts, and uses this manufacturing method to manufacture the aforementioned reverse design results, such as... Figure 3The flowchart shown includes the following steps:
[0059] 1) Obtain the material parameter combination using the above design method, and select the corresponding composite material and layup information based on the material parameter combination;
[0060] 2) First, clean the mold: This embodiment uses a flat mold, which has low mold cost.
[0061] Secondly, prepreg laying: Prepreg is laid on a flat mold according to the layup information described above. The prepreg is a possible material combination for the target deformation obtained in the above design method. In this embodiment, the prepreg is the T800 / epoxy composite material obtained in the above design, and the layup information and flat plate size are [90 / 90 / 45 / -45] and 300×150mm, respectively. 2 .
[0062] Next, bagging and sealing: The prepreg is bagged and sealed after being laid. The bagging operation uses a release film, breathable felt, release cloth and sealing strip to seal the prepreg in a vacuum bag, and a vacuum nozzle is installed to vacuum the prepreg to completely isolate the prepreg from the air.
[0063] Next, temperature curing: The prepreg is pressurized and heated in an autoclave or oven to cause a curing reaction, resulting in a fully cured composite flat panel. The temperature operation is performed using the curing temperature profile provided by the prepreg supplier (180°C in this embodiment), and the pressure operation is performed using the pressure profile provided by the prepreg supplier (0.6 MPa in this embodiment).
[0064] Finally, demolding: The composite material part is demolded. After demolding, the composite material irregular part is deburred to obtain the composite material irregular part with the specified surface requirements.
[0065] This manufacturing method is also a test method to verify whether the above design method is correct. That is, by manufacturing according to the parameter data output by the design method, we can observe whether the actual manufactured product is consistent with the product theoretically designed by the design method. Figures 4A-4B This is a schematic diagram illustrating the shape change during the theoretical manufacturing process, i.e., from Figure 4A The initial layering of the flat part in the flat mold is eventually formed into a flat part. Figure 4B The irregularly shaped parts with curved surfaces. Figures 5A-5B This is a true image of the product after actual demolding in this embodiment. Figure 5A For product 3D model, Figure 5BThe side view of the product shows that it has a curved shape. The results indicate that the maximum curing deformation in the thickness direction is 114 mm. In this embodiment, the target maximum curing deformation is set at 120 mm. Comparing this, the error in the high-throughput custom design of this novel composite material irregular part is approximately 8.01%.
[0066] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Those skilled in the art can modify or make equivalent substitutions to the technical solutions of the present invention. The scope of protection of the present invention is defined by the claims.
Claims
1. A design method for customizing the surface profile of composite material irregular parts, characterized in that, Includes the following steps: Establish a model of the composite material flat plate, which includes the length and width information of the composite material flat plate; In the finite element software, the flat plate model is meshed; For the flat plate model, a composite material card is established. This card is a set of parameters consisting of the corresponding mechanical property parameters and thermal expansion coefficients of the composite single-layer plate in the three directions of in-plane axial, in-plane transverse and thickness directions. Based on the card, composite material layup information is established, which includes the layup area, layup angle, and layup direction of each layer; and a link is established between the composite material card and the composite material layup information. Establish the constraint boundary conditions and temperature load conditions for the flat plate model; Based on the above-described mesh, card, layup information, constraint boundary conditions, and temperature load conditions, the curing deformation behavior of the flat plate model is simulated to establish a curing deformation model of the composite material flat plate. The final residual stress and deformation state dataset of the flat plate model is calculated from this model. Based on the final residual stress and deformation state dataset of the flat plate model, a knowledge base for the curing deformation of composite flat plate parts is established. A neural network prediction model for the curing deformation of composite flat plate parts is constructed, and the neural network prediction model is trained and validated using the dataset of the knowledge base. The neural network prediction model predicts the deformation state and maximum deformation of batch material parameter combinations, and a large-scale screening method is used to obtain material parameter combinations with target deformation to complete the design. One set of material parameter combinations corresponds to a composite material card, representing a composite material.
2. The design method as described in claim 1, characterized in that, The composite material used in the composite flat plate model is a continuous fiber reinforced thermosetting resin-based composite material, wherein the fiber is selected from carbon fiber, glass fiber, or aramid fiber, and the resin is selected from high-temperature epoxy resin, reheat epoxy resin, bismaleimide resin, polyimide resin, or cyano resin.
3. The design method as described in claim 1, characterized in that, The finite element software is Abaqus; the mesh is a shell element mesh, which is an S4R linear reduced integral element.
4. The design method as described in claim 1, characterized in that, The mechanical performance parameters include the elastic model, shear modulus, and Poisson's ratio; the constraint boundary conditions include constraint location and constraint direction; the temperature load conditions include loading location, loading temperature, and initial temperature, wherein the loading location is the entire area of the composite material flat plate model, and the loading temperature is the resin curing temperature of the composite material.
5. The design method as described in claim 1, characterized in that, The knowledge base is operated using Python script batch commands.
6. The design method as described in claim 1, characterized in that, Before utilizing the data in the knowledge base, the neural network prediction model preprocesses the dataset in the knowledge base, then performs feature selection learning on the preprocessed dataset to obtain the influence weights of each material parameter on the maximum deformation. Material parameters with larger influence weights on the maximum deformation rate are selected to form a training dataset. These material parameters include in-plane Young's modulus in the fiber direction, in-plane Young's modulus perpendicular to the fiber direction, in-plane shear modulus, in-plane coefficient of thermal expansion in the fiber direction, in-plane coefficient of thermal expansion perpendicular to the fiber direction, and curing temperature. This training dataset is used to train and validate the neural network prediction model, resulting in a validated neural network prediction model. The feature selection learning employs an embedding selection method based on the L1 regularization algorithm.
7. The design method as described in claim 1, characterized in that, The neural network prediction model adopts an artificial neural network regression model; the training and validation of the neural network prediction model adopts grid search and 5-fold cross-validation methods.
8. The design method as described in claim 1, characterized in that, The large-scale screening method is as follows: first, all possible candidate materials are generated in the design space, and then each one is tested using a neural network prediction model; after the candidate materials are generated, their properties can be easily and directly evaluated using a neural network prediction model.
9. A manufacturing method for customizing the surface of a composite material irregular-shaped part, characterized in that, Includes the following steps: A combination of material parameters is obtained by using the method described in any one of claims 1-8, and the corresponding composite material and layup information are selected based on the combination of material parameters. Clean the mold, apply the selected composite material to the mold according to the above layup information, and then bag it. After bagging, the prepreg is pressurized and heated to fully cure it. Then, it is demolded to obtain composite material shaped parts.
10. The manufacturing method as described in claim 9, characterized in that, The bagging operation uses a release film, breathable felt, release cloth, and sealing strip to seal the prepreg in a vacuum bag, and a vacuum nozzle is installed to evacuate the prepreg, completely isolating it from the air. The pressurization and heating operations are carried out in an autoclave or oven. The heating operation is performed using the curing temperature curve corresponding to the prepreg, and the pressurization operation is performed using the pressure curve corresponding to the prepreg. After the demolding operation, the composite material parts need to be deburred.