An automobile condenser processing parameter optimization method

Abstract

The present application belongs to the technical field of mechanical parameter design optimization, and discloses a kind of automobile condenser processing parameter optimization method, including stamping part processing data acquisition step, stamping part characteristic parameter construction step, forming quality machine learning prediction step, processing parameter optimization step, optimization parameter verification correction step and processing parameter feedback update step, through stamping part processing data acquisition and characteristic parameter construction, the material of automobile condenser stamping part, die, equipment, processing parameter and forming quality data are converted into processing characteristic matrix, the structured expression of multiple features is realized;Through the forming quality machine learning prediction step, the prediction model is obtained by training, the cracking, wrinkling, springback and size deviation risk are output, and the nonlinear mapping between the processing parameters and the forming quality is established;Through the processing parameter optimization, verification correction and feedback update step, the optimal parameter combination is searched within the constraint range, and is supplemented to the historical sample library after verification.

Description

Technical Field

[0001] This invention belongs to the field of mechanical parameter design and optimization technology, and in particular relates to a method for optimizing the processing parameters of automotive condensers. Background Technology

[0002] Automotive condensers typically include heat sinks, connecting brackets, mounting plates, frame plates, reinforcing ribs, and mounting hole structures, some of which require stamping. Stamped parts for automotive condensers often feature thin walls, narrow edges, flanges, multiple holes, and reinforcing ribs. Their stamping quality is influenced by material properties, die condition, equipment condition, and processing parameters. In actual processing, there is a coupling relationship between blank holder force, stamping speed, die clearance, lubrication, and sheet metal positioning deviation. Improper setting of any parameter can lead to cracking, wrinkling, springback deviation, dimensional deviation, abnormal edge burr height, or abnormal flatness deviation. Therefore, a processing parameter optimization method based on machine learning is needed to model, predict, and optimize automotive condenser processing data to reduce the number of manual trial stampings, improve the efficiency of processing parameter determination, and enhance the stability of the forming quality of stamped parts for automotive condensers.

[0003] (1) The existing methods for adjusting the processing parameters of automotive condensers mainly rely on manual experience and repeated trial punches. It is difficult to consider the complex relationship between material strength characteristics, sheet thickness characteristics, die clearance characteristics, blank holder force characteristics, stamping speed characteristics, lubrication status characteristics and forming defect characteristics at the same time, which makes it difficult to predict cracking risk, wrinkling risk, springback deviation and dimensional deviation in advance.

[0004] (2) Existing automotive condenser processing parameter optimization methods usually adjust a single parameter or a single defect, lacking collaborative optimization of blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation, and also lacking optimization parameter verification and correction and processing parameter feedback update mechanisms, making it difficult for the optimal processing parameter combination to adapt to material batch changes, die wear status and equipment fluctuation status. Summary of the Invention

[0005] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a method for optimizing automotive condenser processing parameters. Through data acquisition and feature parameter construction of stamped parts, material data, mold data, stamping equipment data, processing parameter data, and forming quality data of automotive condenser stamped parts are transformed into a stamped part processing feature matrix, achieving a structured expression of multiple features. A forming quality prediction model is trained through a forming quality machine learning prediction step, outputting cracking risk, wrinkling risk, springback deviation, and dimensional deviation, establishing a nonlinear mapping between processing parameters and forming quality. Through processing parameter optimization, optimized parameter verification and correction, and processing parameter feedback and update steps, the optimal combination of processing parameters is searched within constraints, and after verification, it is added to the historical processing sample library to update the forming quality prediction model, forming a closed loop of parameter optimization, verification and correction, and model iteration. This invention solves the problems of traditional process parameter adjustment relying on experience, high trial stamping costs, and lack of continuous optimization capabilities.

[0006] The technical solution adopted in this invention is as follows: a method for optimizing the processing parameters of an automotive condenser, including a stamping part processing data acquisition step, a stamping part feature parameter construction step, a forming quality machine learning prediction step, a processing parameter optimization step, an optimization parameter verification and correction step, and a processing parameter feedback and update step.

[0007] The stamping data acquisition step collects material data, mold data, stamping equipment data, processing parameter data, and forming quality data for stamping parts used in automotive condensers. The step of constructing the characteristic parameters of the stamped parts is based on material data, mold data, stamping equipment data, processing parameter data and forming quality data. It extracts material strength characteristics, sheet thickness characteristics, mold clearance characteristics, blank holder force characteristics, stamping speed characteristics, lubrication state characteristics and forming defect characteristics, and constructs a processing characteristic matrix of the stamped parts. The forming quality machine learning prediction step is based on the stamping part processing feature matrix training to obtain a forming quality prediction model. The forming quality prediction model is then used to predict the cracking risk, wrinkling risk, springback deviation and dimensional deviation under different processing parameter combinations, and generate forming quality prediction results. The process parameter optimization step constructs a process parameter optimization target based on the forming quality prediction result, and searches within the constraints of blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation to obtain the optimal combination of process parameters; The optimization parameter verification and correction step inputs the optimal processing parameter combination into the stamping simulation model or trial stamping verification process to verify the forming quality of the stamped parts for automotive condensers, obtain the verification results, and correct the optimal processing parameter combination based on the verification results to obtain the corrected optimal processing parameter combination. The processing parameter feedback update step receives the verification results and the corrected optimal processing parameter combination output by the optimization parameter verification and correction step, adds the verification results and the corrected optimal processing parameter combination to the historical processing sample library, and updates the forming quality prediction model.

[0008] Furthermore, the stamping part processing data acquisition steps include material data acquisition, mold data acquisition, stamping equipment data acquisition, processing parameter data acquisition, and forming quality data acquisition.

[0009] Furthermore, the stamping part feature parameter construction steps include material strength feature construction, geometry-sensitive feature construction, process control feature construction, and stamping part processing feature matrix construction.

[0010] Furthermore, the forming quality machine learning prediction step includes training sample construction, forming quality prediction model training, defect risk classification, sensitive parameter identification, and forming quality prediction result output.

[0011] Furthermore, the process parameter optimization step includes setting optimization variables, constructing parameter constraint ranges, constructing multi-objective optimization functions, parameter search, and determining the optimal combination of process parameters.

[0012] Furthermore, the optimization parameter verification and correction steps include simulation verification, trial verification, verification result generation, parameter correction, and verification result output.

[0013] Furthermore, the processing parameter feedback update step includes sample supplementation, sample labeling, updating of the historical processing sample library, and model update.

[0014] The beneficial effects achieved by the present invention using the above solution are as follows: (1) This invention transforms the material data, mold data, stamping equipment data, processing parameter data and forming quality data of automotive condenser stamping parts into a stamping part processing feature matrix through the stamping part processing data acquisition step and the stamping part feature parameter construction step. This realizes the structured expression of material strength features, sheet thickness features, mold clearance features, blank holder force features, stamping speed features, lubrication state features and forming defect features, solving the problems of data dispersion, unclear feature correlation and parameter adjustment dependence on experience in the traditional processing parameter adjustment process.

[0015] (2) The present invention obtains a forming quality prediction model based on the stamping part processing feature matrix through a forming quality machine learning prediction step, and predicts the cracking risk, wrinkling risk, springback deviation and dimensional deviation under different processing parameter combinations through the forming quality prediction model, realizing the nonlinear mapping between processing parameter combinations and forming quality, and solving the problem that the traditional trial stamping method is difficult to judge the forming defect risk and dimensional deviation in advance.

[0016] (3) This invention searches for the optimal combination of processing parameters within the constraints of blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation through the processing parameter optimization step, the optimization parameter verification and correction step and the processing parameter feedback and update step. The optimal processing parameter combination is corrected by the verification results. The verification results and the corrected optimal processing parameter combination are added to the historical processing sample library to update the forming quality prediction model. This realizes the closed-loop optimization of processing parameter optimization, verification and correction and model update, and solves the problem that traditional parameter optimization methods lack verification feedback and continuous update capabilities. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall process of optimizing automotive condenser processing parameters proposed in this invention. Figure 2 This is a flowchart of the forming quality machine learning prediction and processing parameter optimization proposed in this invention; Figure 3 This is a flowchart of the optimization parameter verification and correction and processing parameter feedback update proposed in this invention. Detailed Implementation

[0018] Example 1, see Figures 1-3 The present invention provides a method for optimizing the processing parameters of an automotive condenser, including a stamping part processing data acquisition step, a stamping part feature parameter construction step, a forming quality machine learning prediction step, a processing parameter optimization step, an optimization parameter verification and correction step, and a processing parameter feedback and update step. The stamping part processing data acquisition step collects material data, mold data, stamping equipment data, processing parameter data, and forming quality data of the stamping part for automotive condenser, and sends the collected data to the stamping part feature parameter construction step; The stamping part feature parameter construction step is based on material data, mold data, stamping equipment data, processing parameter data and forming quality data. It extracts material strength features, sheet thickness features, mold clearance features, blank holder force features, stamping speed features, lubrication state features and forming defect features, constructs a stamping part processing feature matrix, and sends the stamping part processing feature matrix to the forming quality machine learning prediction step. The forming quality machine learning prediction step is based on the stamping part processing feature matrix training to obtain the forming quality prediction model. The forming quality prediction model predicts the cracking risk, wrinkling risk, springback deviation and dimensional deviation under different processing parameter combinations, generates the forming quality prediction result, and sends the forming quality prediction result to the processing parameter optimization step. The processing parameter optimization step constructs the processing parameter optimization target based on the forming quality prediction result, searches within the constraints of blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation, obtains the optimal processing parameter combination, and sends the optimal processing parameter combination to the optimization parameter verification and correction step. The optimization parameter verification and correction step inputs the optimal processing parameter combination into the stamping simulation model or trial stamping verification process to verify the forming quality of the stamped parts for automotive condensers, obtain the verification results, and correct the optimal processing parameter combination based on the verification results to obtain the corrected optimal processing parameter combination. The processing parameter feedback update step receives the verification results and the corrected optimal processing parameter combination output from the optimization parameter verification and correction step. It then adds the verification results and the corrected optimal processing parameter combination to the historical processing sample library and updates the forming quality prediction model. Ultimately, it achieves closed-loop optimization of automotive condenser processing parameters, including collection, feature construction, quality prediction, parameter optimization, verification and correction, and feedback update.

[0019] Example 2: This example is based on all the above examples. The stamping part processing data acquisition step is used to obtain basic data during the processing of stamping parts for automotive condensers, providing a data source for subsequent machine learning modeling; the specific operation is as follows: Material data acquisition; the stamping part processing data acquisition steps collect the material grade, yield strength, tensile strength, elongation, anisotropy coefficient, hardening index, and sheet thickness of the sheet metal used for stamping parts for automotive condensers; for the same batch of sheet metal, the stamping part processing data acquisition steps record the sheet metal batch number, incoming material direction, and surface condition; for different batches of sheet metal, the stamping part processing data acquisition steps record the range of mechanical property fluctuations to avoid the subsequent forming quality prediction model mistaking different batches of materials for the same input state; Die data acquisition; the stamping part processing data acquisition steps collect punch fillet radius, die fillet radius, die clearance, drawbead height, drawbead position, locating pin position, and die surface roughness; stamping parts for automotive condensers typically have thin walls, narrow edges, flanges, and multi-hole structures. Changes in local die clearance directly affect cracking and springback in the edge area of ​​the stamping part. Therefore, the die data acquisition steps record the corresponding die structure data for the edge flange area, mounting hole area, and reinforcing rib area respectively; Data acquisition for stamping equipment; the data acquisition steps for stamping parts processing include collecting press tonnage, slide stroke, slide speed curve, bottom dead center position deviation, blank holder response status, and number of stamping operations; if the stamping equipment experiences slide speed fluctuations or bottom dead center position deviations during continuous production, the data acquisition steps for stamping parts processing will record the corresponding data as abnormal equipment status samples. Processing parameter data acquisition; the steps for collecting stamping part processing data include collecting blank holder force, stamping speed, lubrication amount, sheet metal positioning deviation, stamping temperature, and feeding cycle time; among them, blank holder force and stamping speed are used as core optimization parameters, lubrication amount and sheet metal positioning deviation are used as auxiliary optimization parameters, and stamping temperature and feeding cycle time are used as production state constraint parameters. Forming quality data acquisition; the stamping part processing data acquisition step collects the location of cracks, wrinkles, springback deviation, dimensional deviation, edge burr height, and flatness deviation of the stamping parts; for defective stamping parts, the stamping part processing data acquisition step binds and stores the defect location with the corresponding processing parameters; for qualified stamping parts, the stamping part processing data acquisition step records the corresponding processing parameter combination as a positive sample. By performing the above operations, the stamping part processing data acquisition step can form processing samples covering materials, molds, equipment, processes, and quality results, providing a complete data foundation for subsequent machine learning optimization.

[0020] Example 3: Based on all the above examples, the stamping part feature parameter construction step is used to convert the raw data obtained from the stamping part processing data acquisition step into structured features adapted to the machine learning model, avoiding unstable model predictions caused by inconsistent dimensions, data sparsity, and weak feature correlation of the raw data; the specific operation is as follows: Material strength characteristics are constructed; the stamping part characteristic parameter construction steps extract the material formability characteristics based on yield strength, tensile strength, elongation, anisotropy coefficient and hardening index; when the material yield strength increases and the elongation decreases, the stamping part characteristic parameter construction steps increase the crack sensitivity weight; when the anisotropy coefficient fluctuates greatly, the stamping part characteristic parameter construction steps increase the springback sensitivity weight. Geometric Sensitive Feature Construction; The stamping part feature parameter construction step extracts geometric sensitive features based on the flange height, hole distance, edge width, reinforcing rib depth, and local fillet radius of the stamping part for automotive condensers; For narrow edge areas, dense hole areas, and flange transition areas, the stamping part feature parameter construction step marks the corresponding areas as forming sensitive areas; Process control feature construction; the stamping part feature parameter construction steps extract process control features based on blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation, and establish a correspondence between process control features and forming quality data; if the cracking risk corresponding to different stamping speeds under the same blank holder force is large, the stamping part feature parameter construction steps take the combination relationship between stamping speed and blank holder force as interactive features; Construction of the feature matrix for stamping parts; the steps for constructing the feature parameters of stamping parts are to combine material strength features, sheet thickness features, die clearance features, blank holder force features, stamping speed features, lubrication state features, and forming defect features into a single processing sample feature vector, represented as: in, It is the processing feature vector of the i-th group of processed samples. It is the material strength characteristic of the i-th group of processed samples; It represents the sheet thickness characteristic of the i-th group of processed samples; The die clearance characteristics of the i-th group of processed samples It represents the blank holder force characteristic of the i-th group of processed samples; It represents the stamping speed characteristic of the i-th group of processed samples; The lubrication state characteristics of the i-th group of processed samples These are the forming defect characteristics of the i-th group of processed samples; The stamping part feature parameter construction step arranges multiple sets of processing sample feature vectors according to processing batches and trial stamping order to generate a stamping part processing feature matrix, and then sends the stamping part processing feature matrix to the forming quality machine learning prediction step. By performing the above operations, the stamping part feature parameter construction step can transform scattered processing data into structured data that is trainable, predictable, and optimizeable, thereby improving the ability of the forming quality prediction model to express the complex forming state of stamping parts for automotive condensers.

[0021] Example 4: This example is based on all the above examples. The forming quality machine learning prediction step is to address the problem that cracking, wrinkling, springback, and dimensional deviations during the processing of stamped parts for automotive condensers are affected by the combined influence of materials, molds, equipment, and process parameters. Traditional experience-based parameter tuning is insufficient to accurately judge the quality results of different combinations of processing parameters. A forming quality prediction model is established based on the stamped part processing feature matrix to predict the forming defect risk and dimensional deviations after changes in processing parameters. The specific operation is as follows: Training sample construction; After receiving the stamping part processing feature matrix, the forming quality machine learning prediction step labels the historical processing samples as qualified samples, cracked samples, wrinkled samples, springback out-of-tolerance samples, and dimensional deviation out-of-tolerance samples; For stamping parts with multiple defects, the forming quality machine learning prediction step uses a multi-label labeling method to record the defect categories; For areas with a small number of trial stamping samples, the forming quality machine learning prediction step supplements them with simulation samples of adjacent processing parameter combinations; Forming quality prediction model training; the forming quality machine learning prediction step inputs the stamping part processing feature matrix into the machine learning model, which uses gradient boosting tree, random forest, support vector regression, or deep neural network; when the structure of stamping parts for automotive condensers is complex and the parameters are significantly coupled, a deep neural network is used to establish a nonlinear mapping relationship between material features, die features, process parameters, and forming quality; the forming quality prediction model outputs cracking risk, wrinkling risk, springback deviation, and dimensional deviation, expressed as: in, This is the predicted forming quality result of the i-th group of processed samples; θ is the forming quality prediction model; θ is the model parameter of the forming quality prediction model. It is the processing feature vector of the i-th group of processed samples; Defect risk classification; the forming quality machine learning prediction step classifies the processing parameter combinations into qualified parameter combinations, cracking risk parameter combinations, wrinkling risk parameter combinations, springback deviation parameter combinations, and dimensional deviation deviation parameter combinations based on the forming quality prediction results. If the cracking risk is higher than the cracking threshold, the forming quality machine learning prediction step marks the corresponding processing parameter combination as a cracking risk parameter combination; if the wrinkling risk is higher than the wrinkling threshold, the forming quality machine learning prediction step marks the corresponding processing parameter combination as a wrinkling risk parameter combination; if the springback deviation or dimensional deviation exceeds the tolerance range, the forming quality machine learning prediction step marks the corresponding processing parameter combination as a dimensional accuracy risk parameter combination. Sensitive parameter identification; the forming quality machine learning prediction step identifies the most sensitive parameters that have the greatest impact on forming quality based on changes in model output; if a small change in blank holder force leads to a significant change in the risk of wrinkling, then blank holder force is identified as a wrinkling sensitive parameter; if an increase in stamping speed leads to a rapid increase in the risk of cracking, then stamping speed is identified as a cracking sensitive parameter; if a change in die clearance leads to an increase in springback deviation, then die clearance is identified as a springback sensitive parameter. The forming quality prediction result is output; the forming quality machine learning prediction step combines the classification results of cracking risk, wrinkling risk, springback deviation, dimensional deviation, and defect risk, as well as the sensitive parameter identification results, into the forming quality prediction result, and sends the forming quality prediction result to the processing parameter optimization step.

[0022] Regarding parameter adjustments: Step 1: Defect label threshold tuning; If the actual cracked sample during the trial stamping process is predicted to be a qualified parameter combination, the cracking threshold is reduced by 0.02 each time to improve the sensitivity of the forming quality machine learning prediction step to cracking defects. If slight surface ripples are misjudged as a combination of wrinkling risk parameters, the wrinkling threshold is increased by 0.02 each time to reduce false alarms. Step 2: Model complexity parameter tuning; If the forming quality prediction model fits the training samples well but has a large prediction error for new batches of sheet metal, then reduce the number of model layers or tree depth and increase the weight of material batch features. If the forming quality prediction model cannot distinguish the quality differences between similar combinations of processing parameters, then increase the number of model layers or trees to enhance the nonlinear fitting ability. Step 3: Sample weight parameter tuning; If the number of cracked samples is small and the false negative rate is high, the weight of cracked samples is increased by 10% each time, so that the forming quality prediction model pays more attention to the risk of cracking. If the proportion of qualified samples is too high, causing the model to tend to output qualified results, then the weight of qualified samples should be reduced and the weight of defective samples should be increased. Step 4: Sensitive parameter screening and tuning; If the sensitive parameters identified by the model are inconsistent with the trial stamping experience, the data quality of the corresponding processing sample should be re-examined, and the corresponding parameter perturbation sample should be added. If multiple parameters affect the same defect simultaneously, parameter interaction features are added to enable the forming quality machine learning prediction step to identify the combined effects between blank holder force and stamping speed, and between die clearance and springback deviation.

[0023] By performing the above operations, this solution addresses the problems of relying on repeated trial stamping based on manual experience, insufficient ability to predict defect risks, and difficulty in quantifying parameter sensitivity in the traditional processing of stamped parts for automotive condensers. It establishes a nonlinear mapping relationship between processing features and forming quality through a machine learning prediction step for forming quality. This allows for the prediction of cracking risk, wrinkling risk, springback deviation, and dimensional deviation corresponding to different combinations of processing parameters in advance, providing a calculable basis for optimizing processing parameters.

[0024] Example 5: This example is based on all the above examples. The processing parameter optimization step aims to address the problem that in the stamping process of automotive condenser stamped parts, the blank holder force, stamping speed, die clearance, lubrication, and sheet metal positioning deviation are interdependent, and adjusting a single parameter can easily lead to other forming quality issues. Based on the forming quality prediction results output by the forming quality machine learning prediction step, a multi-objective processing parameter optimization objective is constructed, and the optimal combination of processing parameters is searched within the range of quality constraints and process constraints. The specific operation is as follows: Optimize variable settings; the processing parameter optimization steps set blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation as optimization variables; blank holder force affects material flow and wrinkling risk, stamping speed affects material strain rate and cracking risk, die clearance affects springback deviation and edge quality, lubrication amount affects material flow uniformity, and sheet positioning deviation affects hole accuracy and edge size deviation; The parameter constraint range is constructed; the processing parameter optimization step constructs the parameter constraint range based on the stamping equipment capacity, die structure limitations, material forming limits, and quality tolerance requirements; if the blank holder force exceeds the upper limit allowed by the equipment, the processing parameter optimization step will exclude the corresponding parameter combination; if the stamping speed exceeds the material's allowable forming speed range, the processing parameter optimization step will mark the corresponding parameter combination as unusable; if the die clearance exceeds the die assembly allowable range, the processing parameter optimization step will exclude the corresponding parameter combination. Multi-objective optimization function construction; the processing parameter optimization steps aim to reduce cracking risk, reduce wrinkling risk, reduce springback deviation, reduce dimensional deviation, and improve production cycle stability. The processing parameter optimization objectives are constructed as follows: in, P is the objective function for optimizing the processing parameters; P is the combination of processing parameters. There is a risk of cracking; It is a risk of wrinkling; It is a rebound deviation; It is a dimensional deviation; It is a production cycle penalty item It is the cracking risk weighting coefficient It is the wrinkle risk weighting coefficient; It is the rebound deviation weighting coefficient; It is the dimensional deviation weighting coefficient; It is the production cycle weighting coefficient; Parameter search; the processing parameter optimization step uses Bayesian optimization, genetic algorithm or particle swarm optimization to search for the optimal combination of processing parameters within the parameter constraints; for each set of candidate processing parameter combinations generated, the processing parameter optimization step calls the forming quality machine learning prediction step to calculate the corresponding forming quality prediction result, and calculates the comprehensive score according to the processing parameter optimization objective function; when the candidate processing parameter combination simultaneously meets the requirements of cracking risk, wrinkling risk, springback deviation and dimensional deviation, the processing parameter optimization step adds the candidate processing parameter combination to the feasible parameter set; The optimal combination of processing parameters is determined; the processing parameter optimization step selects the processing parameter combination with the lowest objective function value from the set of feasible parameters as the optimal processing parameter combination; if the objective function values ​​of multiple candidate processing parameter combinations are close, the processing parameter combination with more stable production cycle, smaller change in blank holder force, and less mold adjustment is selected first. Regarding parameter adjustments: Step 1: Adjusting the crack risk weight; If edge cracking still occurs during the trial stamping verification, the cracking risk weight is increased by 0.05 each time, so that the processing parameter optimization step prioritizes reducing the cracking risk; If the processing parameters are too conservative, resulting in insufficient material flow and wrinkling, then the cracking risk weight should be reduced and the wrinkling risk weight should be increased. Step 2: Adjusting the wrinkle risk weight; If wrinkling occurs in the flanged or edge area, the wrinkling risk weight is increased by 0.05 each time, so that the blank holder force is increased or the lubrication amount is adjusted in the processing parameter optimization step. If excessive suppression of wrinkling risk leads to increased cracking risk, then reduce the wrinkling risk weight and rebalance the cracking risk weight. Step 3: Adjusting the rebound deviation weighting parameters; If the springback of the trial stamped part exceeds the tolerance range, the springback deviation weight is increased by 0.05 each time, and the die clearance and stamping speed are checked simultaneously. If the springback deviation meets the requirements but the risk of cracking increases, then reduce the weight of the springback deviation to avoid excessive pursuit of size compensation that could affect molding safety. Step 4: Adjusting production cycle weights; If the optimal combination of processing parameters results in a stamping speed that is too low and affects production capacity, then the production cycle weight is increased by 0.03 each time. If an excessively fast production cycle leads to an increased risk of defects, the production cycle weight should be reduced so that the optimization steps for processing parameters prioritize meeting the forming quality requirements.

[0025] By performing the above operations, this solution addresses the problems of strong parameter coupling, numerous manual trial and error attempts, and difficulty in balancing quality and production targets in the traditional adjustment of processing parameters for stamped parts used in automotive condensers. It incorporates machine learning predictions into a multi-objective optimization process through a parameter optimization step, searching for the optimal combination of processing parameters within parameter constraints. This allows for the coordinated optimization of blank holder force, stamping speed, die clearance, lubrication, and sheet metal positioning deviation, thereby improving the forming quality and processing efficiency of stamped parts for automotive condensers.

[0026] Example 6: Based on all the above examples, this example optimizes the parameter verification and correction step by performing simulation verification and trial stamping verification on the optimal processing parameter combination output by the processing parameter optimization step. The verification results are then used to correct the optimal processing parameter combination, resulting in a corrected optimal processing parameter combination. This avoids direct mass production using only machine learning predictions, which could lead to deviations in actual forming quality. The specific operation is as follows: Simulation verification; The optimal processing parameter combination is input into the stamping simulation model to simulate and calculate the material flow state, thickness reduction rate, equivalent plastic strain, springback deviation, and dimensional deviation of the stamped parts for automotive condensers, and the simulation verification results are obtained. If the simulation verification results show that the local thickness reduction rate exceeds the safety threshold, the optimization parameter verification and correction step marks the corresponding area as a cracking risk area and feeds it back to the processing parameter optimization step to readjust the blank holder force or stamping speed. The trial stamping verification and optimization parameter verification and correction steps apply the optimal combination of processing parameters to the trial stamping process, checking the crack location, wrinkle location, springback deviation, dimensional deviation, edge burr height, and flatness deviation of the test stamped part to obtain the trial stamping verification results. If the trial stamping verification results show that the forming quality of the test stamped part meets the quality requirements, the optimization parameter verification and correction steps determine the optimal combination of processing parameters as the recommended combination of processing parameters. If the trial stamping verification results show that the test stamped part has defects, the optimization parameter verification and correction steps correct the optimal combination of processing parameters according to the defect type. Verification results are generated; the optimization parameter verification and correction steps are based on the simulation verification results and the trial stamping verification results to generate verification results, which include material flow state, thickness reduction rate, equivalent plastic strain, springback deviation, dimensional deviation, crack location, wrinkling location, edge burr height and flatness deviation; Parameter correction; the optimization parameter verification and correction steps, based on the cracking, wrinkling, springback deviation, and dimensional deviation in the verification results, correct the stamping speed, blank holder force, lubrication amount, die clearance, springback compensation, and sheet metal positioning deviation to obtain the optimal combination of processing parameters after correction. If cracking occurs in the verification results, the optimization parameter verification and correction steps reduce the stamping speed or adjust the blank holder force; if wrinkling occurs in the verification results, the optimization parameter verification and correction steps increase the blank holder force or adjust the lubrication amount; if springback deviation occurs in the verification results, the optimization parameter verification and correction steps adjust the die clearance or increase springback compensation; if dimensional deviation occurs in the verification results, the optimization parameter verification and correction steps correct the sheet metal positioning deviation. The verification results are output; the optimization parameter verification and correction step outputs the verification results and the corrected optimal processing parameter combination, and sends the verification results and the corrected optimal processing parameter combination to the processing parameter feedback update step.

[0027] By performing the above operations, the optimized parameter verification and correction steps can perform simulation verification and trial verification of the machine learning optimization results. Based on the simulation verification results and trial verification results, verification results are generated, and the optimal processing parameter combination is corrected through the verification results, so that the optimal processing parameter combination has a simulation and trial verification basis before entering mass production.

[0028] Example 7: This example is based on all the above examples. The processing parameter feedback update step is used to supplement the historical processing sample library with the verification results output from the optimization parameter verification and correction step and the corrected optimal processing parameter combination, and to continuously update the forming quality prediction model; the specific operation is as follows: Sample supplementation; the processing parameter feedback update step records the material data, mold data, stamping equipment data, processing parameter data, forming quality data, verification results, and the corrected optimal processing parameter combination corresponding to each simulation verification and trial stamping verification as a new processing sample; Sample marking; the processing parameter feedback update step marks the newly added processing samples according to the test stamping results corresponding to the corrected optimal processing parameter combination; if the corrected optimal processing parameter combination obtains a qualified stamped part in the test stamping, the processing parameter feedback update step marks the corresponding newly added processing sample as a valid optimized sample; if the corrected optimal processing parameter combination still has defects, the processing parameter feedback update step marks the corresponding newly added processing sample as a sample to be corrected. The historical processing sample library is updated; the processing parameter feedback update step filters newly added processing samples, retains material batch change samples, defect risk samples, qualified parameter samples, parameter boundary samples, and samples to be corrected, and updates the historical processing sample library; Model update; the processing parameter feedback update step uses the updated historical processing sample library to retrain or incrementally update the forming quality prediction model, so that the forming quality prediction model can learn the forming quality changes corresponding to new material batches, new mold states and new processing parameter combinations.

[0029] By performing the above operations, the processing parameter feedback update step can accumulate the simulation verification results, trial stamping verification results, verification results, and the corrected optimal processing parameter combination in the processing parameter optimization process into a continuously updated data foundation. This allows the forming quality prediction model to be continuously optimized as production data increases, improving the adaptability of the automotive condenser processing parameter optimization method to different material batches, mold wear conditions, and equipment fluctuation conditions.

Claims

1. A method for optimizing the processing parameters of an automotive condenser, characterized in that: It includes steps for collecting stamping part processing data, constructing stamping part feature parameters, predicting forming quality through machine learning, optimizing processing parameters, verifying and correcting optimized parameters, and updating processing parameters. The stamping data acquisition step collects material data, mold data, stamping equipment data, processing parameter data, and forming quality data for stamping parts used in automotive condensers. The step of constructing the characteristic parameters of the stamped parts is based on material data, mold data, stamping equipment data, processing parameter data and forming quality data. It extracts material strength characteristics, sheet thickness characteristics, mold clearance characteristics, blank holder force characteristics, stamping speed characteristics, lubrication state characteristics and forming defect characteristics, and constructs a processing characteristic matrix of the stamped parts. The forming quality machine learning prediction step is based on the stamping part processing feature matrix training to obtain a forming quality prediction model. The forming quality prediction model is then used to predict the cracking risk, wrinkling risk, springback deviation and dimensional deviation under different processing parameter combinations, and generate forming quality prediction results. The process parameter optimization step constructs a process parameter optimization target based on the forming quality prediction result, and searches within the constraints of blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation to obtain the optimal combination of process parameters; The optimization parameter verification and correction step inputs the optimal processing parameter combination into the stamping simulation model or trial stamping verification process to verify the forming quality of the stamped parts for automotive condensers, obtain the verification results, and correct the optimal processing parameter combination based on the verification results to obtain the corrected optimal processing parameter combination. The processing parameter feedback update step receives the verification results and the corrected optimal processing parameter combination output by the optimization parameter verification and correction step, adds the verification results and the corrected optimal processing parameter combination to the historical processing sample library, and updates the forming quality prediction model.

2. The method for optimizing automotive condenser processing parameters according to claim 1, characterized in that: The steps for constructing the characteristic parameters of the stamped part specifically include: Material strength characteristics are constructed; material formability characteristics are extracted based on yield strength, tensile strength, elongation, anisotropy coefficient, and hardening index. Geometric sensitive features are constructed based on the flange height, hole distance, edge width, reinforcing rib depth, and local fillet radius of stamped parts for automotive condensers. Process control features are constructed; based on blank holder force, stamping speed, die clearance, lubrication amount and sheet positioning deviation, process control features are extracted, and a correspondence is established between process control features and forming quality data; Construction of stamping part processing feature matrix: Material strength features, sheet thickness features, die clearance features, blank holder force features, stamping speed features, lubrication state features, and forming defect features are combined into processing sample feature vectors. Multiple sets of processing sample feature vectors are arranged according to processing batch and trial stamping sequence to generate stamping part processing feature matrix.

3. The method for optimizing automotive condenser processing parameters according to claim 2, characterized in that: The forming quality machine learning prediction step specifically includes: Training sample construction; historical processing samples are labeled as qualified samples, cracked samples, wrinkled samples, springback exceeding tolerance samples, and dimensional deviation exceeding tolerance samples; Training the forming quality prediction model: Input the stamping part processing feature matrix into the machine learning model to establish a nonlinear mapping relationship between material features, mold features, process parameters and forming quality, and obtain the forming quality prediction model; Defect risk classification: The forming quality prediction result is obtained through the forming quality prediction model, and the processing parameter combination is divided into qualified parameter combination, cracking risk parameter combination, wrinkling risk parameter combination, springback error parameter combination and dimensional deviation error parameter combination according to the forming quality prediction result. Sensitive parameter identification: Identify the sensitive parameters that have the greatest impact on forming quality based on the changes in the output of the forming quality prediction model; The forming quality prediction results are output; the forming quality prediction results are composed of the classification results of cracking risk, wrinkling risk, springback deviation, dimensional deviation, and defect risk, as well as the sensitive parameter identification results, and the forming quality prediction results are sent to the processing parameter optimization step.

4. The method for optimizing automotive condenser processing parameters according to claim 3, characterized in that: The process parameter optimization step specifically includes: Optimize variable settings; set blank holder force, stamping speed, die clearance, lubrication amount, and sheet metal positioning deviation as optimization variables; Parameter constraint range construction: The parameter constraint range is constructed based on the stamping equipment capacity, die structure limitations, material forming limits, and quality tolerance requirements. Multi-objective optimization function construction: With the objectives of reducing cracking risk, reducing wrinkling risk, reducing springback deviation, reducing dimensional deviation, and improving production cycle stability, processing parameter optimization objectives are constructed. Parameter search: Using Bayesian optimization, genetic algorithm or particle swarm optimization, candidate processing parameter combinations are searched within the parameter constraints, and a comprehensive score of the candidate processing parameter combinations is calculated based on the forming quality prediction results; The optimal combination of processing parameters is determined; the combination of processing parameters with the lowest objective function value is selected from the set of feasible parameters as the optimal combination of processing parameters, and the optimal combination of processing parameters is sent to the optimization parameter verification and correction step.

5. The method for optimizing automotive condenser processing parameters according to claim 4, characterized in that: The optimization parameter verification and correction steps specifically include: Simulation verification: Input the optimal combination of processing parameters into the stamping simulation model, simulate and calculate the material flow state, thickness reduction rate, equivalent plastic strain, springback deviation and dimensional deviation of the stamped parts for automotive condensers, and obtain the simulation verification results; Trial stamping verification: Apply the optimal combination of processing parameters to the trial stamping process, check the cracking location, wrinkling location, springback deviation, dimensional deviation, edge burr height and flatness deviation of the test stamping part, and obtain the trial stamping verification results; Verification results are generated; based on simulation verification results and trial stamping verification results, verification results are generated. Parameter correction: Based on the cracking, wrinkling, springback deviation and dimensional deviation in the verification results, the stamping speed, blank holder force, lubrication amount, die clearance, springback compensation and sheet positioning deviation are corrected to obtain the optimal combination of processing parameters after correction. Output the verification results; output the verification results and the corrected optimal combination of processing parameters, and send the verification results and the corrected optimal combination of processing parameters to the processing parameter feedback update step.

6. The method for optimizing automotive condenser processing parameters according to claim 5, characterized in that: The processing parameter feedback and update step specifically includes: Sample supplementation: Record the material data, mold data, stamping equipment data, processing parameter data, forming quality data, verification results, and corrected optimal processing parameter combinations corresponding to each simulation verification and trial stamping verification as new processing samples; Sample labeling: Samples that obtain qualified stamped parts in trial stamping with the corrected optimal processing parameter combination are labeled as valid optimized samples, and samples that still have defects after the corrected optimal processing parameter combination are labeled as samples to be corrected. The historical processing sample library is updated; newly added processing samples are screened, and samples with material batch changes, defect risk, qualified parameters, parameter boundary, and samples to be corrected are retained, and the historical processing sample library is updated. Model update: Retrain or incrementally update the forming quality prediction model using the updated historical processing sample library.

Images