Ribbed member creep aging process method and molded surface prediction method

Abstract

The invention discloses a ribbed member creep aging process method and a molded surface prediction method. The process method comprises the following steps: acquiring multiple groups of dependent variable-time data of a sample stage under different creep aging process conditions; establishing a creep aging whole-process material constitutive model; correcting the constitutive model according to an actual experiment and a sample finite element simulation result; key features of the ribbed component are obtained; a neural network is used for constructing a mapping relation among the process conditions, the discrete feature points of the components and the forming molded surfaces; the neural network structure is corrected according to a creep age forming result; and optimizing the process parameters and the component structure. According to the method, most of experimental data sources are finite element simulation, simulation is actually verified, and the reliability of finite element simulation and neural network prediction can be effectively guaranteed; according to the method, process parameter optimization and forming molded surface point cloud key position prediction are realized, the creep aging process suitable for component forming can be quickly obtained, and the molded surface of the formed component can be quickly predicted by adopting the process.

Description

Technical Field

[0001] The invention relates to the technical field of creep aging forming, and in particular to a creep aging process method for a ribbed component and a forming surface prediction method. Background Art

[0002] The actual load-bearing conditions of the rocket body structure are complex, and its design must ensure sufficient strength and stiffness, and reduce weight as much as possible to improve the payload and overall performance. The application of ribbed components can provide additional strength and stiffness without significantly increasing the weight, thereby meeting these requirements. Creep aging forming technology is an advanced forming technology that meets the high performance requirements and low-cost manufacturing of ribbed thin-walled components. The key to using creep aging forming technology to achieve component shape collaborative manufacturing is how to shorten the process research and development cycle, so as to quickly obtain a suitable creep aging process and predict the component profile after the process is formed, and guide the actual production of creep aging of ribbed components. In the prior art, the specific creep aging process of the component is determined mainly by using finite element simulation to obtain the forming profile under the action of different process parameters, and then further compare the target profile with the forming profile error, and continuously iterate the three parameters of aging temperature, aging time and loading pressure, so that the component forming accuracy meets the standard, thereby obtaining the final creep aging process. This method not only has low overall research and development efficiency and a long exploration cycle, but also has high calculation costs using finite element simulation.

[0003] In response to the above problems, scholars in the industry have conducted some relevant research. For example, the invention patent with the authorization announcement number CN115828423B discloses a method for predicting the rebound of the outer plate of marine equipment based on a dual-branch deep learning algorithm. This method obtains the outer plate rebound prediction model based on the convolutional neural network to predict the rebound of the outer plate forming. However, the three-dimensional morphology of the outer plate is input into the convolutional neural network, and the training process is slow, the debugging cycle is long, and the equipment requirements are high. For another example, the invention patent with the authorization announcement number CN109508488B discloses a method for predicting the process parameters of shot peening based on the optimization of the BP neural network by a genetic algorithm. This method uses the genetic algorithm to optimize the weights and thresholds of the BP neural network to predict the process parameters of shot peening, but it does not input the part features into the neural network and the data sources are all actual experiments, which is costly, the small amount of data is prone to overfitting, and the process development cycle is long.

[0004] Therefore, there is an urgent need to explore a method that can quickly obtain the optimal creep aging process parameters suitable for the forming of ribbed components and obtain the target springback profile of the component under the process parameters. Summary of the invention

[0005] The purpose of the present invention is to provide a creep aging forming method for ribbed components to solve the problems of low research and development efficiency, long exploration cycle and high cost of the existing creep aging process methods mentioned in the background technology.

[0006] To achieve the above object, the present invention provides a creep aging process for a reinforced member, comprising the following steps:

[0007] Step 1: Conduct a specimen-level creep aging forming test to obtain the strain-time data of the specimen under multiple groups of different creep aging process parameters;

[0008] Step 2: Based on the strain-time data obtained in step 1, establish the material constitutive model for the entire creep aging process;

[0009] Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters on the specimens; and carry out basic creep aging experiments under the same process parameters on the specimens; modify the parameters of the constitutive model according to the results of the basic experiments and simulations;

[0010] Step 4: Establish a spatial coordinate system, and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system;

[0011] Step 5: Identify the key structural features of the component and discretize the identified key structural features, thereby achieving dimensionality reduction from the three-dimensional morphology of the component to discrete points;

[0012] Step 6: Based on step 5, the discrete feature points of the component are obtained by the optimal Latin hypercube sampling method;

[0013] Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging under different process conditions and discrete feature points of different components;

[0014] Step 8: Use a neural network to construct a mapping relationship between process conditions, discrete feature points of components and forming surfaces of components; use the process conditions and the initial positions of discrete feature points of components as input layers, and use the positions of discrete feature points of components after forming as output layers to construct a neural network model, and then judge the forming accuracy of the components and whether there are defects on the forming surfaces from the positions of the discrete feature points of the components after forming;

[0015] Step 9: Modify the neural network model constructed in step 8 according to the actual creep aging forming results of the component;

[0016] Step 10: Use optimization algorithms to optimize the creep aging process parameters and component structure of the component, and finally obtain the optimal creep aging process parameters;

[0017] In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

[0018] Furthermore, in step 1, the number of groups of creep aging process parameters is 6 to 18, the aging temperature is 140-210° C., the aging stress is 240-420 MPa, and the aging time is 6-12 h.

[0019] Furthermore, the constitutive model in step 2 includes the strain changes in the loading stage, the heating stage, the cooling stage, and the heat preservation and loading stage, which is used to realize the prediction of the strain-time change of the component in the whole process.

[0020] Furthermore, the key structural features of the component in step 5 include rib position, rib height and rib thickness.

[0021] Furthermore, the optimization algorithm in step 10 is an optimization algorithm based on Levy flight strategy, chaos mapping, random walk strategy or sine-cosine optimization strategy.

[0022] The present invention also provides a method for predicting the creep aging forming surface of a ribbed component. The prediction method uses the forming process method as described above to obtain the optimal creep aging process parameters, and then brings the optimal creep aging process parameters and structural characteristics back into the neural network model established in step 8, thereby obtaining the forming surface of the target component.

[0023] The present invention also provides another creep aging process method for a reinforced member, comprising the following steps:

[0024] Step 1: Conduct a specimen-level creep aging forming test to obtain the strain-time data of the specimen under multiple groups of different creep aging process parameters;

[0025] Step 2: Based on the strain-time data obtained in step 1, establish the constitutive model of the material in the whole creep aging process;

[0026] Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters on the specimens; and carry out basic creep aging experiments under the same process parameters on the specimens; modify the parameters of the constitutive model according to the results of the basic experiments and simulations;

[0027] Step 4: Establish a spatial coordinate system, and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system;

[0028] Step 5: Identify the key structural features of the component and discretize the identified key structural features, thereby achieving dimensionality reduction from the three-dimensional morphology of the component to discrete points;

[0029] Step 6: Based on step 5, the discrete feature points of the component are obtained by the optimal Latin hypercube sampling method;

[0030] Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging under different process conditions and discrete feature points of different components;

[0031] Step 8: Use a neural network to construct a mapping relationship between process conditions, discrete feature points of components and component forming surfaces; use the process conditions and the initial positions of discrete feature points of components as the output layer, and use the positions of discrete feature points of components after forming as the input layer to construct a neural network, and directly input the desired surface point cloud position, and then output the corresponding creep aging process parameters;

[0032] In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

[0033] The invention also provides another method for predicting the creep aging forming surface of a ribbed component. The prediction method uses the creep aging process parameters of the component obtained by the forming process method as described above, and then corrects the neural network structure constructed in step 8 according to the representative actual creep aging forming results of the component, and then brings the corrected result back into the neural network model established in step 8, so as to obtain the forming surface of the target component.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The creep aging process method for ribbed components of the present invention modifies the parameters of the constitutive model according to the results of basic experiments and simulations; most of the experimental data comes from finite element simulations and the simulations are actually verified, which can effectively ensure the reliability of finite element simulations and neural network predictions; the structural features of the components are discretized to achieve dimensionality reduction from three-dimensional morphology to discrete data points, greatly reducing the amount of calculation. The present invention calls the neural network prediction model twice to achieve process parameter optimization and prediction of key positions of the forming surface point cloud, which can quickly obtain a creep aging process suitable for component forming, and the process can quickly predict the surface of the component after forming.

[0036] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0038] Figure 1 It is a schematic diagram of a process flow of a creep aging forming process for a ribbed member of the present invention;

[0039] Figure 2 is a schematic diagram of a neural network mapping relationship constructed in an embodiment of the present invention;

[0040] Figure 3 is a schematic diagram of the distribution positions of discrete characteristic points obtained after the component is formed and its springback profile after forming in an embodiment of the present invention;

[0041] Figure 4 It is a schematic diagram of the predicted surface accuracy of the creep aging forming surface prediction method of a ribbed component in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be described in detail below in conjunction with the various embodiments shown in the accompanying drawings, but it should be noted that these embodiments are not limitations of the present invention, and any equivalent transformations or substitutions in functions, methods, or structures made by ordinary technicians in the field based on these embodiments are all within the scope of protection of the present invention.

[0043] See also Figure 1 This embodiment provides a creep aging process method for a reinforced component, comprising the following steps:

[0044] Step 1: Conduct a specimen-level creep aging forming test to obtain strain-time data of multiple different specimens under multiple groups of different creep aging process parameters; wherein the number of groups of creep aging process parameters is 6 to 12, the aging temperature is 140-210°C, the aging stress is 240-420MPa, and the aging time is 6-12h. Preferably, the number of groups of creep aging process parameters is 9, the aging temperatures are 150°C, 180°C and 210°C, and the aging stresses are 240MPa, 280MPa and 320MPa, respectively.

[0045] Step 2: Based on the strain-time data obtained in step 1, establish a constitutive model of the material for the entire creep aging process; the constitutive model includes the strain changes in the loading stage, heating stage, cooling stage, and insulation and loading stage, which is used to realize the prediction of the strain-time changes of the components during the entire process.

[0046] Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters for multiple different samples; and carry out basic creep aging experiments under the same process parameters for multiple different samples; modify the parameters of the constitutive model according to the results of basic experiments and simulations.

[0047] Step 4: Establish a spatial coordinate system and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system.

[0048] Step 5: Identify the key structural features of the component, including rib position, rib height, and rib thickness, etc.; and discretize the identified key structural features to achieve dimensionality reduction from the three-dimensional morphology of the component to discrete points.

[0049] Step 6: Based on step 5, the discrete feature points of the component are obtained through the optimal Latin hypercube sampling method.

[0050] Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging of multiple sets of discrete feature points of different process conditions and different components; In this step, carry out component-level creep aging forming simulations of multiple sets of orthogonal process conditions to obtain the corresponding surface point cloud data after creep aging of discrete feature points of the corresponding components. Finite element simulations with different process parameters can be submitted through batch processing, and the post-processing results can also be automatically extracted.

[0051] Step 8: Use a neural network to construct a mapping relationship between creep aging process conditions, discrete feature points of components and component forming surfaces; use the creep aging process conditions of components and the initial positions of discrete feature points of components as the input layer, and the positions of discrete feature points of components after forming as the output layer to construct a neural network model. The positions of discrete feature points of components after forming can be used to determine the forming accuracy of the component and whether defects such as wrinkles appear on the forming surface.

[0052] Step 9: According to the actual creep aging forming results of representative components, the neural network model constructed in step 8 is modified to improve the prediction accuracy; this step uses the actual creep aging forming surface of representative components to optimize the database and neural network.

[0053] Step 10: Use optimization algorithm to optimize the creep aging process parameters and structure of the component; Levy flight is a medium-short distance mixed random search method that conforms to Levy distribution, which has excellent global search ability, takes the neural network prediction error as the fitness function, and optimizes the process parameters with the help of the optimization algorithm based on Levy flight strategy optimization. In this step, the global search ability can be improved by not only using Levy flight strategy, but also introducing chaos mapping, random walk strategy, sine-cosine optimization strategy, etc., which can also achieve the purpose of the invention.

[0054] In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

[0055] The present invention also provides a method for predicting the creep aging forming surface of a ribbed component. The prediction method adopts the forming method as described above. After obtaining the optimal creep aging process parameters, the optimal creep aging process parameters and structural characteristics are fed back into the neural network model established in step 8 to obtain the forming surface of the target component.

[0056] The embodiment of the present invention also provides another creep aging forming process method for a ribbed component, comprising the following steps:

[0057] Step 1: Conduct a specimen-level creep aging forming test to obtain the specimen-level strain-time data under multiple groups of different creep aging process parameters;

[0058] Step 2: Based on the strain-time data obtained in step 1, establish the constitutive model of the material in the whole creep aging process;

[0059] Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters on the specimens; and carry out basic creep aging experiments under the same process parameters on the specimens; modify the parameters of the constitutive model according to the results of the basic experiments and simulations;

[0060] Step 4: Establish a spatial coordinate system, and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system;

[0061] Step 5: Identify the key structural features of the component and discretize the identified key structural features, thereby achieving dimensionality reduction from the three-dimensional morphology of the component to discrete points;

[0062] Step 6: Based on step 5, the discrete feature points of the component are obtained by the optimal Latin hypercube sampling method;

[0063] Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging under different process conditions and discrete feature points of different components; perform multiple groups of component-level creep aging forming simulations under different process conditions orthogonal to obtain the corresponding surface point cloud data after creep aging of discrete feature points of the corresponding components. Finite element simulations with different process parameters can be submitted through batch processing, and the post-processing results can also be automatically extracted.

[0064] Step 8: Use a neural network to construct a mapping relationship between component creep aging process parameter conditions, component discrete feature points and component forming surface; use component creep aging process parameter conditions and component discrete feature point initial positions as the output layer, and use the discrete feature point positions of the component after forming as the input layer to construct a neural network, and directly input the desired surface point cloud position to output the corresponding process parameters.

[0065] In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

[0066] The present invention also provides another method for predicting the creep aging forming surface of a ribbed component. The prediction method adopts the creep aging process parameters of the component obtained by the other creep aging process method of the ribbed component as mentioned above, and then corrects the neural network structure constructed in step 8 according to the representative actual creep aging forming results of the component, and then brings the corrected result back into the neural network model established in step 8, so as to obtain the forming surface of the target component.

[0067] Example

[0068] The embodiment of the present invention takes the high-rib wall panel as an example, and the modeling parameters are as follows:

[0069]

[0070]

[0071] The above modeling parameters are used to construct Figure 2 Neural network mapping relationship.

[0072] A batch of sample points of the flat plate state of the high-rib siding are predicted. By integrating the data of the three modeling parameters of the discrete point x coordinate after rebound, the discrete point rebound amount, and the discrete point z coordinate, the x coordinate, y coordinate, and z coordinate of the discrete sample points are corresponded, and the point cloud data of the discrete sample points after rebound is formed. Finally, the rebound profile of the siding is generated by fitting the point cloud data; Figure 3As shown, (a) is the distribution position of discrete feature points obtained after the component is formed, which can be equivalent to the point cloud position after springback; (b) is the springback profile obtained by fitting the point cloud data.

[0073] The creep aging forming profile prediction method of the ribbed component of the present invention is used to predict the profile accuracy of the wall panel with an aging temperature of 180°C, an aging time of 1.3h, and a mold radius of R2150. Figure 4 .Depend on Figure 4 It can be seen that this method has high prediction accuracy. The deviation between the springback prediction result and the actual forming result in most areas is ±0.8mm. The prediction effect is evaluated by RMSE. The calculation formula is as follows. The calculated RMSE error is 0.724mm, and the prediction effect is good.

[0074]

[0075] Where y is the calculated value of the surface springback of the neural network, i is the surface springback value of the actual component, is the profile deviation value, and p represents the number of profile points taken.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A creep aging process for a reinforced component, characterized in that: The steps include: Step 1: Conduct a specimen-level creep aging forming test to obtain the strain-time data of the specimen under multiple groups of different creep aging process parameters; Step 2: Based on the strain-time data obtained in step 1, establish the material constitutive model for the whole creep aging process; Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters on the specimens; and carry out basic creep aging experiments under the same process parameters on the specimens; modify the parameters of the constitutive model according to the results of the basic experiments and simulations; Step 4: Establish a spatial coordinate system, and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system; Step 5: Identify the key structural features of the component and discretize the identified key structural features, thereby achieving dimensionality reduction from the three-dimensional morphology of the component to discrete points; Step 6: Based on step 5, the discrete feature points of the component are obtained by the optimal Latin hypercube sampling method; Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging under different process conditions and discrete feature points of different components; Step 8: Use a neural network to construct a mapping relationship between process conditions, discrete feature points of components, and component forming surfaces; A neural network model is constructed with process conditions and the initial positions of discrete feature points of components as input layers, and the positions of discrete feature points of components after forming as output layers. Then, the forming accuracy of components and whether there are defects on the forming surface are determined from the positions of discrete feature points of components after forming. Step 9: Modify the neural network model constructed in step 8 according to the actual creep aging forming results of the component; Step 10: Use optimization algorithms to optimize the creep aging process parameters and component structure of the component, and finally obtain the optimal creep aging process parameters; In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

2. A creep aging process for reinforced components according to claim 1, characterized in that: In the step 1, the number of groups of creep aging process parameters is 6 to 18, the aging temperature is 140-210° C., the aging stress is 240-420 MPa, and the aging time is 6-12 h.

3. The creep aging process of a reinforced component according to claim 1, characterized in that: The constitutive model in step 2 includes the strain changes in the loading stage, the heating stage, the cooling stage, and the heat preservation and loading stage, and is used to realize the prediction of the strain amount-time change of the component in the whole process.

4. The creep aging process of a reinforced component according to claim 1, characterized in that: The key structural features of the component in step 5 include rib position, rib height and rib thickness.

5. The creep aging process of a reinforced component according to claim 1, characterized in that: The optimization algorithm in step 10 is an optimization algorithm based on Levy flight strategy, chaos mapping, random walk strategy or sine-cosine optimization strategy.

6. A method for predicting the creep aging forming surface of a reinforced component, characterized in that: After the prediction method obtains the optimal creep aging process parameters by the process method described in any one of claims 1 to 5, the optimal creep aging process parameters and structural characteristics are fed back into the neural network model established in step 8 to obtain the forming surface of the target component.

7. A creep aging process for a reinforced component, characterized in that: The following steps are involved: Step 1: Conduct a specimen-level creep aging forming test to obtain the strain-time data of the specimen under multiple groups of different creep aging process parameters; Step 2: Based on the strain-time data obtained in step 1, establish the constitutive model of the material in the whole creep aging process; Step 3: Embed the constitutive model established in step 2 into the finite element simulation software, and simulate the whole process of creep aging forming with multiple groups of different process parameters on the specimens; and carry out basic creep aging experiments under the same process parameters on the specimens; modify the parameters of the constitutive model according to the results of the basic experiments and simulations; Step 4: Establish a spatial coordinate system, and measure the overall dimensions of multiple different reinforced components in the spatial coordinate system; Step 5: Identify the key structural features of the component and discretize the identified key structural features, thereby achieving dimensionality reduction from the three-dimensional morphology of the component to discrete points; Step 6: Based on step 5, the discrete feature points of the component are obtained by the optimal Latin hypercube sampling method; Step 7: Based on the correction results of step 3 and the discrete feature points obtained in step 6, perform component-level orthogonal finite element simulation to obtain the corresponding surface point cloud data after creep aging under different process conditions and discrete feature points of different components; Step 8: Use a neural network to construct a mapping relationship between process conditions, discrete feature points of components, and component forming surfaces; The neural network is constructed with the process conditions and the initial positions of the discrete feature points of the component as the output layer, and the positions of the discrete feature points of the component after forming as the input layer. The desired surface point cloud positions are directly input, and the corresponding creep aging process parameters are output. In the above steps, steps 1 to 3 are performed in parallel with steps 4 to 6, or steps 1 to 3 are performed followed by steps 4 to 6, or steps 4 to 6 are performed first and then steps 1 to 3; the execution of step 7 is premised on the completion of steps 1 to 3 and steps 4 to 6.

8. A method for predicting the creep aging forming surface of a reinforced component, characterized in that: The prediction method adopts the creep aging process parameters of the component obtained by the process method described in claim 7 above, and then corrects the neural network structure constructed in step 8 according to the representative actual creep aging forming results of the component, and then brings the corrected results back into the neural network model established in step 8, so as to obtain the forming surface of the target component.

Images