An automobile component production and manufacturing process simulation optimization method and system

Abstract

The application is specifically a kind of automobile part production and manufacturing process simulation optimization method and system, relates to the automobile manufacturing process simulation optimization technical field, and includes: based on the standardized process data set, selecting the physical field control equation corresponding to the target manufacturing process, establishing a multi-physical field process simulation model, calibrating the output of the multi-physical field process simulation model by using historical production data, and obtaining the verified simulation model when the deviation value between the simulation output result and the measured result is within the preset precision threshold range. In the application, the multi-physical field process simulation model is established and the calibration verification is performed by using the historical production data, so that the deviation between the simulation model output and the actual production result is controlled, the technical problem that the optimization scheme is difficult to land due to the insufficient precision of the simulation model is solved, and the technical effect of improving the engineering credibility of the simulation model is achieved.

Description

Technical Field

[0001] This invention relates to the field of automotive manufacturing process simulation and optimization technology, and in particular to a method and system for simulating and optimizing the manufacturing process of automotive parts. Background Technology

[0002] The production and manufacturing of automotive parts involves various processes such as stamping, injection molding, welding, and machining. The parameter configuration of each process has a direct impact on product quality, production efficiency, manufacturing costs, and energy consumption.

[0003] The determination of existing process parameters mainly relies on the accumulated experience of engineers and repeated trial production and verification. Under the requirement of multi-objective optimization, engineers gradually approach the feasible solution through manual parameter adjustment and small-batch trial production, while using single-physics simulation software to simulate and evaluate local process procedures.

[0004] The above approach has the following drawbacks: process parameters rely on human experience to determine, resulting in long development cycles and high trial-and-error costs; multiple objectives such as quality, efficiency, cost, and energy consumption are mutually constrained, making it difficult for human experience to obtain the globally optimal process solution among multiple objectives; the simulation model lacks a systematic calibration and verification process, and deviations are easily generated between simulation output and actual production, making it difficult for the optimized solution to be effectively implemented in actual production.

[0005] The aforementioned shortcomings collectively lead to technical problems such as low efficiency in automotive component process development and insufficient reliability of optimization solutions. Summary of the Invention

[0006] The purpose of this invention is to provide a simulation optimization method and system for automotive component manufacturing processes in order to solve the above-mentioned problems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A simulation optimization method for automotive component manufacturing processes includes: Obtain the optimization target parameters and full-dimensional process basic data of the automotive component to be optimized, and perform cleaning and standardization processing on the process basic data to generate a standardized process dataset; Based on the standardized process dataset, the physical field control equations corresponding to the target manufacturing process are selected, a multi-physics process simulation model is established, and the output of the multi-physics process simulation model is calibrated using historical production data. When the deviation between the simulation output result and the measured result is within the preset accuracy threshold range, the verified simulation model is obtained. Sensitivity analysis is performed on each process parameter in the validated simulation model to screen key process parameters whose sensitivity index values ​​are higher than the preset sensitivity threshold. Based on the experimental plan, combined sampling is performed on the key process parameters and simulation calculation is performed using the validated simulation model to generate a response relationship dataset. The agent model is trained based on the response relationship dataset. The optimization objective parameter is used as the optimization objective. The Pareto optimal solution set is solved using a multi-objective optimization algorithm. The candidate solutions in the Pareto optimal solution set are simulated and verified to output the optimal process solution.

[0008] Preferably, the optimization target parameters include at least one of the following: molding defect control index, production cycle time target value, yield rate target value, manufacturing cost constraint value, and energy consumption constraint value; the process basic data includes part 3D CAD model data, material constitutive parameters, equipment performance parameters, process constraint data, on-site production condition data, and quality acceptance standard data; the standardization process includes: converting each parameter to a unified dimension under the International System of Units (SI), performing Z-score standardization on each parameter, and eliminating differences in the dimensions and numerical ranges of different parameters.

[0009] Preferably, the process data cleaning process includes: when the collected value of a certain parameter deviates from the historical average value of the parameter by more than a preset deviation threshold, the record is marked as an outlier and removed; when the collected value of a certain parameter is within the preset deviation threshold range, the record is retained.

[0010] Preferably, the calibration of the output of the multiphysics process simulation model using historical production data includes: acquiring historical mass production data and small-batch trial production data; comparing the simulation output of the multiphysics process simulation model under the same process conditions with the measured results to calculate the deviation value; decomposing the deviation value into systematic deviation components and random deviation components; adjusting the material constitutive parameters and physical parameters in the boundary conditions for the systematic deviation components; adjusting the mesh density and the convergence tolerance parameters of the solver for the random deviation components; repeating the deviation comparison and parameter adjustment when the adjusted deviation value exceeds a preset accuracy threshold; and obtaining the verified simulation model when the adjusted deviation value is within the preset accuracy threshold range.

[0011] Preferably, after performing the sensitivity analysis, the method further includes: performing a grading process on the sensitivity index value to divide the key process parameter into a highly sensitive parameter and a moderately sensitive parameter; when performing combined sampling on the key process parameter based on the experimental plan, a first sampling interval is used for the highly sensitive parameter and a second sampling interval is used for the moderately sensitive parameter, wherein the first sampling interval is smaller than the second sampling interval.

[0012] Preferably, the step of using a multi-objective optimization algorithm to solve for the Pareto optimal solution set includes: defining the objective function as a vector that simultaneously minimizes multiple optimization objective function values ​​corresponding to the key process parameter vector, wherein the multiple optimization objective function values ​​include at least two of minimizing the molding defect index, minimizing the production cycle time, maximizing the yield rate, minimizing the manufacturing cost, and minimizing energy consumption; defining the constraints as the upper and lower limits of equipment capacity, the process feasibility boundary, and the qualified threshold of the quality acceptance standard for each key process parameter; performing mean normalization processing based on the range on each optimization objective parameter value, uniformly mapping each objective value to the same numerical range, and then substituting it into the objective function to perform optimization and solve for the Pareto optimal solution set.

[0013] Preferably, the step of performing simulation reproduction verification on candidate solutions in the Pareto optimal solution set includes: inputting the process parameter values ​​of the candidate solutions into the verified simulation model to perform simulation reproduction calculation and obtain simulation reproduction results; comparing the simulation reproduction results with the prediction results of the surrogate model and calculating the reproduction deviation value; when the reproduction deviation value exceeds a preset reproduction threshold, adjusting the training parameters of the surrogate model and retraining; when the reproduction deviation value is within the preset reproduction threshold range, confirming the candidate solution as valid; performing normalization processing on the multiple optimization target parameter values ​​of each valid candidate solution, assigning weight coefficients to each optimization target, calculating the weighted comprehensive score value of each valid candidate solution, and determining the candidate solution with the best weighted comprehensive score value as the optimal process solution.

[0014] Preferably, after outputting the optimal process scheme, the method further includes: applying random perturbations of a preset amplitude to each key process parameter in the optimal process scheme to generate multiple perturbation schemes; performing simulation calculations on each perturbation scheme using the verified simulation model to obtain the optimization target parameter values ​​of each perturbation scheme; confirming that the optimal process scheme has production robustness when the fluctuation amplitude of the optimization target parameter values ​​of each perturbation scheme relative to the optimal process scheme is within a preset robustness threshold range; and selecting the suboptimal scheme from the Pareto optimal solution set for re-verification when the fluctuation amplitude exceeds the preset robustness threshold.

[0015] Preferably, after outputting the optimal process scheme, a closed-loop iterative correction step is further included: obtaining actual production data of the optimal process scheme in small-batch trial production, comparing the actual production data with the simulation prediction results of the optimal process scheme, and calculating the production deviation value; when the production deviation value exceeds a preset production deviation threshold, using the actual production data to perform reverse correction on the parameters of the verified simulation model, synchronously updating the constraint parameter values ​​related to equipment status and process environment in the optimization constraints, and re-executing sensitivity analysis and optimization solution; when the production deviation value is within the preset production deviation threshold range, fixing the values ​​of each key process parameter in the optimal process scheme as mass production process parameters; during the mass production stage, continuously acquiring real-time operating condition data during the production process, and when the real-time operating condition data deviates from the operating condition benchmark value corresponding to the fixed mass production process parameters by more than a preset operating condition deviation threshold, triggering the re-execution of model calibration and subsequent optimization processes.

[0016] A simulation and optimization system for automotive component manufacturing processes includes: The data acquisition and standardization module is used to acquire the optimization target parameters and full-dimensional process basic data of the automotive parts to be optimized, and to perform cleaning and standardization processing on the process basic data to generate a standardized process dataset. The simulation model establishment and calibration module is used to select the physical field control equations corresponding to the target manufacturing process based on the standardized process dataset, establish a multi-physics process simulation model, and use historical production data to calibrate the output of the multi-physics process simulation model to obtain a verified simulation model. The sensitivity analysis and response sampling module is used to perform sensitivity analysis on each process parameter in the validated simulation model, screen key process parameters, perform combined sampling on the key process parameters based on the experimental plan, and perform simulation calculations using the validated simulation model to generate a response relationship dataset. The multi-objective optimization and verification module is used to train a proxy model based on the response relationship dataset, use the optimization objective parameters as the optimization objective, use a multi-objective optimization algorithm to solve the Pareto optimal solution set, perform simulation reproduction verification on the candidate solutions in the Pareto optimal solution set, and output the optimal process solution.

[0017] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention establishes a multi-physics process simulation model and uses historical production data to perform calibration and verification, thereby controlling the deviation between the simulation model output and the actual production results. This solves the technical problem that insufficient simulation model accuracy makes it difficult to implement optimization schemes, and achieves the technical effect of improving the engineering credibility of the simulation model.

[0018] 2. This invention reduces the parameter dimensionality of the optimization solution and the number of simulation calculations by performing sensitivity analysis on process parameters and screening key process parameters. This solves the technical problem of excessive computational overhead in multi-objective optimization and achieves the technical effect of completing multi-objective optimization within an acceptable computational time. By training a surrogate model to replace the computationally expensive, validated simulation model for optimization iteration, the computational resource consumption during the optimization process is further reduced. By obtaining the Pareto optimal solution set through a multi-objective optimization algorithm and verifying it through simulation, this invention solves the technical problem that human experience is insufficient to obtain the globally optimal process solution among multiple objectives. It achieves the technical effect of obtaining a comprehensive optimal compromise solution among multiple objectives such as quality, efficiency, cost, and energy consumption. Attached Figure Description

[0019] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0021] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0022] Example 1 Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.

[0023] In this embodiment, it includes: A car parts manufacturer (Company A) undertook the mass production of the inner door panel for a certain car model. The inner door panel is made of high-strength steel sheet (code: DP590), 1.2mm thick, with a complex geometry, including deep-drawing areas and multiple flanging features. In initial production, the part frequently experienced cracking and wrinkling defects, resulting in a yield rate of only about 72%, high manufacturing costs, and a lengthy development cycle due to the engineering team's reliance on experience and repeated trial molding. Therefore, Company A initiated a process simulation optimization process with the core objectives of reducing the molding defect rate, improving the yield rate, and controlling energy consumption.

[0024] Step S1: Obtain the optimization target definition and all-factor process data, and generate a standardized process dataset. For the stamping process of the inner door panel, the following optimization targets are first defined: forming defect control indicators (maximum thinning rate not exceeding 25%, wrinkling height not exceeding 0.3mm), yield target value (not less than 92%), production cycle target value (not exceeding 18 times / minute), and energy consumption constraint value (single piece stamping energy consumption not exceeding 0.85kWh). The process system boundary is defined as the stamping forming process within a single-stage drawing die.

[0025] Subsequently, comprehensive process data was collected. CAD models were obtained from the part design database, material constitutive parameters were taken from the DP590 material test report, equipment performance parameters were obtained from the press nameplate and test records, and on-site production condition data were obtained from the data acquisition system, totaling 487 historical condition records. After data cleaning, 19 records with missing values ​​and 11 records with outliers were removed, retaining 457 valid records.

[0026] The retained records undergo dimensional unification and Z-score standardization. Taking the blank holder force parameter as an example, the original collected value is in kN, which is converted to the SI unit N, and then standardized. ; in, These are the original parameter values. This is the historical average of this parameter. This represents the historical standard deviation of this parameter. Taking blank holder force as an example, , The original value of a certain record The standardization result is: ; Table 1. Process Basic Data Collection and Cleaning Results

[0027] Step S2: Establish a multiphysics process simulation model and perform calibration. Based on the constitutive parameters of DP590 material (elastic modulus 206 GPa, yield strength 380 MPa, Poisson's ratio 0.3, anisotropy coefficient) , , A 3D CAD model of the car door inner panel and an elastoplastic mechanical model were selected as the governing equations to establish a multiphysics process simulation model for stamping.

[0028] Selecting the physical field control equations corresponding to the target manufacturing process specifically refers to: Based on the physical nature of the stamping process of the inner door panel, the material undergoes elastic deformation during the stamping process and then enters the plastic flow stage, which involves geometric nonlinearity (large deformation) and contact nonlinearity. Therefore, the elastic-plastic large deformation mechanical model is selected as the core control equation.

[0029] The model consists of the following three types of equations: The equilibrium equation is established based on the principle of virtual work and is expressed as follows: ,in Let Cauchy be the stress tensor. For strain tensor, For boundary loads; In the above formula: Here is the Cauchy stress tensor (unit: Pa). For the strain tensor (dimensionless). It is a virtual strain tensor; This is the surface force vector acting on the boundary (unit: N / m²). This is a displacement vector (unit: m). It is a virtual displacement vector; This represents the volume domain (in m³) of the continuum under the current configuration. This represents the boundary surface of the continuum under the current configuration; the colon ":" indicates the double dot product operation of the tensor, i.e. (Using Einstein's summation convention).

[0030] Geometric equations: Using finite strain theory, the relationship between strain rate and velocity gradient is as follows: ; Physical equations (constitutive model): Hill'48 anisotropic yield criterion is selected. ; Based on the isotropic strengthening model, the material parameters (F, G, H, L, M, N) are determined by the thickness anisotropy coefficient of DP590 steel plate. The result is obtained by reverse calculation.

[0031] Establishing a multiphysics process simulation model for stamping refers specifically to: The above control equations are discretized, and an explicit dynamic analysis step is established in the finite element software. The 3D CAD model of the inner door panel is imported, and the Belytschko-Tsay shell element is used for the sheet metal (the element size is described later). The dies (punch, die, and blank holder) are defined as rigid bodies. Contact pairs (sheet metal-punch, sheet metal-die, sheet metal-blade holder) are set up, and the contact algorithm adopts the penalty function method. The friction model is selected as Coulomb friction (the initial value of the friction coefficient is described later). Blank holder force boundary conditions are applied to the blank holder, and stamping speed boundary conditions are applied to the punch to complete the construction of the multiphysics simulation model.

[0032] Discretize the above governing equations using the finite element method: The equilibrium equations based on the principle of virtual work can be written in the weak form as follows: ; The Galerkin weighted residual method is used to discretize the continuum into a finite number of elements, and the displacement field within each element is obtained through shape functions. Approximately expressed as: ; in Let be the nodal displacement vector. Substituting the above approximation into the virtual work equation, after element assembly, we obtain the global discrete equation system: ; in For the quality matrix, It is the internal force vector (including the stress calculated by the constitutive model). The force vector is the external force vector. The stamping process involves large deformation and contact nonlinearity. An explicit central difference method is used for time integration, and an explicit dynamic analysis step is established in the finite element software.

[0033] Mesh the computational domain, using a 3mm mesh size for areas with greater drawing depth and a 5mm mesh size for other areas, generating approximately 86,400 shell elements. Boundary conditions were set according to the equipment nameplate parameters: maximum press tonnage 2,500kN, slide speed range 5 to 25mm / s, and initial die clearance set to 1.25mm (i.e., 1.04 times the sheet thickness).

[0034] Model calibration was performed using historical mass production data and 12 sets of small-batch trial production data. The simulated maximum thinning rate was compared with the measured results; the initial deviation was 3.5%, exceeding the preset accuracy threshold of 3%. After decomposing the deviation, the systematic deviation component was mainly caused by an excessively low friction coefficient, while the random deviation component was caused by insufficient mesh density. To address the systematic deviation, the friction coefficient was adjusted from 0.12 to 0.15; to address the random deviation, the mesh size was refined from 3mm to 2mm in the deep drawing region. After adjustment and recalculation, the deviation converged to 0.7%, meeting the accuracy threshold requirement, confirming the model passed validation.

[0035] The friction coefficient was adjusted from 0.12 to 0.15; to address random deviations, the mesh size was refined from 3mm to 2mm in the deep drawing region as follows: The deviation between the simulation output and the measured result Decomposed into systematic deviation components and random deviation components .

[0036] Systematic bias is defined as the average difference between multiple simulation and experimental results. The systematic error of the model's physical parameters is reflected; the random deviation is defined as the standard deviation. This reflects the numerical discretization error.

[0037] For systematic deviation components (Initial in this case) This is corrected by adjusting the coefficient of friction. (Coefficient of friction) The correction amount is based on the local sensitivity coefficient of the friction coefficient to the maximum thinning rate simulated. Confirmed. Specifically, two disturbance simulations were performed beforehand around a friction coefficient of 0.12. ), calculate the rate of change of thinning rate, and obtain (That is, for every 0.01 increase in the coefficient of friction, the maximum thinning rate decreases by approximately 0.18 percentage points). In this case, the initial... To eliminate this deviation, the thinning rate needs to be reduced by approximately 3.1%, therefore the required increase in the friction coefficient is... By rounding down and incorporating engineering experience, the coefficient of friction was adjusted from 0.12 to 0.15.

[0038] For random deviation components (Initial in this case) Discretization errors are reduced by refining the mesh size in the deep drawing region. According to the finite element convergence criterion, the mesh size... It has a second-order relationship with the solution accuracy. The original mesh size of 3mm corresponds to a discrete error of 0.9%. The target discrete error is to be controlled within 0.3%. Therefore, the required mesh size is... Considering computational cost, we take... After refining the mesh in the deep drawing region from 3 mm to 2 mm, the randomness deviation was reduced to less than 0.3%. Table 2 Record of Simulation Model Calibration Process

[0039] Step S3: Sensitivity analysis, screening of key process parameters, and generation of response relationship dataset. Sensitivity analysis was performed on each of the four stamping process parameters in the validated simulation model to calculate the sensitivity index value of each parameter to the three optimization objectives of maximum thinning rate, yield rate, and energy consumption per unit. The preset sensitivity threshold was 0.15.

[0040] The sensitivity analysis performed on each of the four stamping process parameters in the validated simulation model included: Local sensitivity analysis (also known as first-order factorial analysis) is used. For the first... Process parameters and the One optimization objective Sensitivity index value Defined as: when When it changes within the preset range The relative rate of change and The ratio of their relative rates of change, i.e.: ; in, For the first The initial reference value of each process parameter (take the median value of the process parameter range). In order to be in The simulation obtained at the first One target value; and They are respectively The parameter values ​​after the preset step size for upward and downward movement (in this case, the step size is 5% of the parameter range); and This is the simulation output under the corresponding conditions.

[0041] For each process parameter The arithmetic mean of its sensitivity values ​​to multiple optimization objectives (maximum thinning rate, yield rate, and unit energy consumption) is obtained. Taking blank holder force as an example: Reference value The step size is 5%, which is ±16.25kN. Take the following values ​​respectively: , Two simulations were performed, and the calculations were obtained. , , Overall sensitivity mean The blank holder force is higher than the preset threshold of 0.15, therefore it is determined to be a key process parameter.

[0042] Sensitivity index value Defined as optimization objective For process parameters The elasticity (the ratio of relative rates of change) is mathematically defined as: ; Where the partial derivative represents when When making infinitesimal changes around the reference value The instantaneous relative rate of change. In practical engineering calculations, since the simulation model is a black-box function, it is impossible to directly calculate the partial derivative analytically. Therefore, the forward-backward difference approximation is used: ; in , , Take 5% of the parameter range. Substituting the above difference approximation into the sensitivity definition, we obtain the difference form of the sensitivity calculation formula. The approximate equality sign indicates that the difference discretization approximates the partial derivatives. This approximation is based on a first-order Taylor expansion, when... It has second-order accuracy when it is small enough.

[0043] Table 3. Results of sensitivity analysis of process parameters

[0044] The combined sensitivity of blank holder force and drawbead resistance is higher than 0.35, classifying them as highly sensitive parameters; the combined sensitivity of stamping speed is 0.34, higher than the threshold of 0.15, classifying them as medium-sensitive parameters; the combined sensitivity of die clearance is 0.09, lower than the threshold, and is therefore excluded. The set of key process parameters is determined to consist of three parameters: blank holder force, drawbead resistance, and stamping speed.

[0045] Based on the Latin hypercube sampling method, a denser sampling interval was used for highly sensitive parameters (blank weight force and drawbead resistance), with 8 sampling levels for each; a sparser sampling interval was used for moderately sensitive parameters (stamping speed), with 4 sampling levels, generating a total of 48 combinations of process parameters. Each combination was input into a validated simulation model for calculation, generating a response relationship dataset (48 records in total). Table 4 lists 6 representative records from this dataset.

[0046] Table 4 Response Relationship Dataset (Representative Records)

[0047] Step S4: Multi-objective optimization solution and output of optimal process scheme Using the response relation dataset (48 records) as training samples, a Kriging surrogate model was trained to fit the mapping relationship between three key process parameters (blank force, drawbead resistance, and stamping speed) and three optimization objectives (maximum thinning rate, yield rate, and energy consumption per piece).

[0048] The objective function is defined as: minimizing the maximum thinning rate. Maximize yield rate (equivalent to minimizing) Minimize energy consumption per unit ,Right now: ; in, The constraints correspond to blank holder force, drawbead resistance, and stamping speed, respectively, and are as follows: ; ; The NSGA-II algorithm was used to optimize the surrogate model and obtain the Pareto optimal solution set (a total of 21 non-dominated solutions). Three candidate solutions were selected from these and input into the validated simulation model for reproduction verification.

[0049] The NSGA-II algorithm was used to optimize the surrogate model and obtain the Pareto optimal solution set (a total of 21 non-dominated solutions). Three candidate solutions were selected from these and input into the validated simulation model for reproduction verification.

[0050] Table 5 Simulation Reproduction Verification Results of Candidate Solutions

[0051] The reproduction deviations of all three candidate solutions were within the preset reproduction threshold (1.5%), thus confirming them as valid candidate solutions. After performing mean normalization based on the range on the three optimization objective values ​​of each solution, weighting coefficients were assigned: (Thinning rate) (Yield rate) (Energy consumption), and Calculate the weighted comprehensive score of each scheme. : ; in, For the first The normalized value of the optimization objective. These are the corresponding weight coefficients. Taking candidate solution B as an example, the normalized objective values ​​are... , (Normalization of yield rate after inversion) ,but:

[0052] Table 6. Overall Scoring Results of Valid Candidate Solutions

[0053] Weighted composite score of candidate solution C The minimum value is determined to be the optimal process scheme, with the corresponding process parameters being: blank holder force 352kN, draw bead resistance 30.0N / mm, and stamping speed 15mm / s.

[0054] Random perturbations were applied to the optimal process scheme, generating 15 perturbation schemes for simulation calculation. The maximum thinning rate fluctuation of each perturbation scheme was within 0.8%, and the yield fluctuation was within 1.2%, all within the preset robustness threshold range, confirming that the optimal process scheme has production robustness.

[0055] Step S5: Closed-loop iterative correction The optimal process scheme (blank force 352kN, drawbead resistance 30.0N / mm, stamping speed 15mm / s) was put into small-batch trial production, producing a total of 200 pieces. Actual production data was collected and compared with simulation prediction results.

[0056] Table 7 Comparison of Small-Batch Trial Production Data and Simulation Predictions

[0057] All three production deviation values ​​were within the preset production deviation threshold range, and the values ​​of each key process parameter in the optimal process scheme were solidified as mass production process parameters. In the subsequent mass production stage, real-time operating data was continuously monitored. When any parameter deviated from the operating condition benchmark value by more than the preset operating condition deviation threshold, the process was triggered to return to step S2 to re-execute the calibration and optimization process.

[0058] Throughout the implementation process, the data flow exhibits a clear, progressive relationship: Step S1 starts with 487 original working condition records, which are cleaned and standardized to output 457 valid records, forming a standardized process dataset; Step S2 uses this dataset and the CAD model as input, and through three rounds of calibration iterations, the simulation deviation is converged from the initial 3.5% to 0.7%, outputting a validated simulation model; Step S3 uses the validated simulation model to perform sensitivity analysis on four process parameters, compressing the parameter dimensions from four to three, and generating 48 response relationship records; Step S4 trains a surrogate model with the 48 response relationship records, obtains 21 Pareto non-dominated solutions through the NSGA-II algorithm, and outputs the optimal process scheme with blank holder force of 352kN, drawbead resistance of 30.0N / mm, and stamping speed of 15mm / s after reproduction verification and weighted scoring; Step S5 puts this scheme into small-batch trial production, and the measured deviations are all within the threshold range, finally solidifying it into mass production process parameters, completing the complete data flow from raw data to a feasible mass production scheme.

[0059] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention. It should be understood that in the various embodiments of this application, the sequence numbers of the above processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0060] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A simulation optimization method for automotive component manufacturing processes, characterized in that, include: Obtain the optimization target parameters and full-dimensional process basic data of the automotive component to be optimized, and perform cleaning and standardization processing on the process basic data to generate a standardized process dataset; Based on the standardized process dataset, the physical field control equations corresponding to the target manufacturing process are selected, a multi-physics process simulation model is established, and the output of the multi-physics process simulation model is calibrated using historical production data. When the deviation between the simulation output result and the measured result is within the preset accuracy threshold range, the verified simulation model is obtained. Sensitivity analysis is performed on each process parameter in the validated simulation model to screen key process parameters whose sensitivity index values ​​are higher than the preset sensitivity threshold. Based on the experimental plan, combined sampling is performed on the key process parameters and simulation calculation is performed using the validated simulation model to generate a response relationship dataset. The agent model is trained based on the response relationship dataset. The optimization objective parameter is used as the optimization objective. The Pareto optimal solution set is solved using a multi-objective optimization algorithm. The candidate solutions in the Pareto optimal solution set are simulated and verified to output the optimal process solution.

2. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, The optimization target parameters include at least one of the following: molding defect control index, production cycle time target value, yield rate target value, manufacturing cost constraint value, and energy consumption constraint value; the process basic data includes part 3D CAD model data, material constitutive parameters, equipment performance parameters, process constraint data, on-site production condition data, and quality acceptance standard data; the standardization process includes: converting each parameter to a unified dimension under the International System of Units (SI), performing Z-score standardization on each parameter, and eliminating differences in the dimensions and numerical ranges of different parameters.

3. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, The basic process data cleaning process includes: when the collected value of a parameter deviates from the historical average value of the parameter by more than a preset deviation threshold, the record is marked as an outlier and removed; when the collected value of a parameter is within the preset deviation threshold range, the record is retained.

4. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, The calibration of the multiphysics process simulation model output using historical production data includes: acquiring historical mass production data and small-batch trial production data; comparing the simulation output results of the multiphysics process simulation model under the same process conditions with the measured results, and calculating the deviation value; the simulation output result is the maximum thinning rate, and the measured result is the measured maximum thinning rate; decomposing the deviation value into systematic deviation components and random deviation components; adjusting the friction coefficient for the systematic deviation component and adjusting the mesh size of the deep drawing region for the random deviation component; when the adjusted deviation value exceeds a preset accuracy threshold, repeating the deviation comparison and parameter adjustment; when the adjusted deviation value is within the preset accuracy threshold range, obtaining the calibrated simulation model.

5. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, After performing sensitivity analysis, the process further includes: performing a grading process on the sensitivity index values ​​to divide the key process parameters into highly sensitive parameters and moderately sensitive parameters; when performing combined sampling on the key process parameters based on the experimental plan, a first sampling interval is used for the highly sensitive parameters and a second sampling interval is used for the moderately sensitive parameters, wherein the first sampling interval is smaller than the second sampling interval.

6. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, The method of finding the Pareto optimal solution set using a multi-objective optimization algorithm includes: defining the objective function as a vector that simultaneously minimizes multiple optimization objective function values ​​corresponding to the key process parameter vector, wherein the multiple optimization objective function values ​​include at least two of the following: minimizing the molding defect index, minimizing the production cycle time, maximizing the yield rate, minimizing the manufacturing cost, and minimizing energy consumption; defining the constraints as the upper and lower limits of equipment capacity, the process feasibility boundary, and the qualified threshold of the quality acceptance standard for each key process parameter; performing mean normalization based on the range on each optimization objective parameter value, uniformly mapping each objective value to the same numerical range, and then substituting it into the objective function to perform optimization and obtain the Pareto optimal solution set.

7. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, The simulation reproduction verification of candidate solutions in the Pareto optimal solution set includes: inputting the process parameter values ​​of the candidate solutions into the verified simulation model to perform simulation reproduction calculations and obtain simulation reproduction results; comparing the simulation reproduction results with the prediction results of the surrogate model and calculating the reproduction deviation value; when the reproduction deviation value exceeds a preset reproduction threshold, adjusting the training parameters of the surrogate model and retraining; when the reproduction deviation value is within the preset reproduction threshold range, confirming the candidate solution as valid; performing normalization processing on the multiple optimization objective parameter values ​​of each valid candidate solution, assigning weight coefficients to each optimization objective, calculating the weighted comprehensive score value of each valid candidate solution, and determining the candidate solution with the optimal weighted comprehensive score value as the optimal process solution.

8. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, After outputting the optimal process scheme, the process further includes: applying random perturbations of a preset amplitude to each key process parameter in the optimal process scheme, generating multiple perturbation schemes, performing simulation calculations on each perturbation scheme using the verified simulation model, and obtaining the optimization target parameter values ​​of each perturbation scheme; when the fluctuation amplitude of the optimization target parameter values ​​of each perturbation scheme relative to the optimal process scheme is within a preset robustness threshold range, the optimal process scheme is confirmed to have production robustness; when the fluctuation amplitude exceeds the preset robustness threshold, the suboptimal scheme in the Pareto optimal solution set is selected for re-verification.

9. The simulation optimization method for automotive component manufacturing process according to claim 1, characterized in that, After outputting the optimal process scheme, a closed-loop iterative correction step is also included: obtaining the actual production data of the optimal process scheme in small-batch trial production, comparing the actual production data with the simulation prediction results of the optimal process scheme, and calculating the production deviation value; when the production deviation value exceeds the preset production deviation threshold, using the actual production data to perform reverse correction on the parameters of the verified simulation model, synchronously updating the constraint parameter values ​​related to equipment status and process environment in the optimization constraints, and re-executing sensitivity analysis and optimization solution; When the production deviation value is within the preset production deviation threshold range, the values ​​of each key process parameter in the optimal process scheme are fixed as mass production process parameters. During the mass production stage, real-time operating condition data is continuously acquired during the production process. When the real-time operating condition data deviates from the operating condition benchmark value corresponding to the fixed mass production process parameters by more than a preset operating condition deviation threshold, the model calibration and subsequent optimization process are re-executed.

10. A simulation and optimization system for automotive component manufacturing processes, and a simulation and optimization method for automotive component manufacturing processes according to any one of claims 1-9, characterized in that, include: The data acquisition and standardization module is used to acquire the optimization target parameters and full-dimensional process basic data of the automotive parts to be optimized, and to perform cleaning and standardization processing on the process basic data to generate a standardized process dataset. The simulation model establishment and calibration module is used to select the physical field control equations corresponding to the target manufacturing process based on the standardized process dataset, establish a multi-physics process simulation model, and use historical production data to calibrate the output of the multi-physics process simulation model to obtain a verified simulation model. The sensitivity analysis and response sampling module is used to perform sensitivity analysis on each process parameter in the validated simulation model, screen key process parameters, perform combined sampling on the key process parameters based on the experimental plan, and perform simulation calculations using the validated simulation model to generate a response relationship dataset. The multi-objective optimization and verification module is used to train a proxy model based on the response relationship dataset, use the optimization objective parameters as the optimization objective, use a multi-objective optimization algorithm to solve the Pareto optimal solution set, perform simulation reproduction verification on the candidate solutions in the Pareto optimal solution set, and output the optimal process solution.

Images