A method and system for optimizing injection molding process parameters based on Gaussian function
Through the injection molding process parameter optimization method based on Gaussian function, the Kriging agent model and multi-objective constraint optimization algorithm are used to solve the defect problem that the CAE model cannot be optimized, and the injection molding production efficiency and product quality are improved.
Patent Information
- Application Number
- CN202411241344.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-05
AI Technical Summary
During the injection molding process, defects caused by the inability to effectively optimize the CAE model, and multiple mold tests are usually required to obtain the optimal process parameters, resulting in waste of production and delays in mold production period.
The injection molding process parameter optimization method based on Gaussian function is adopted, and the injection molding process parameters are gradually optimized until the computing resources are exhausted.
It improves the injection molding production efficiency and yield rate, obtains more accurate final injection molding parameters, reduces waste in the production process, and improves product quality.
Smart Images

Figure CN118769496B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of industrial automation, and in particular relates to a method and system for optimizing injection molding process parameters based on Gaussian functions. Background Art
[0002] With the development of China's basic industries, plastic products play an important role in various fields of production and life, including the electronics industry, aerospace, information communication and packaging. Injection molding is a multi-input nonlinear process. Factors such as injection molding process parameters, material properties and molds will affect the quality of injection molded products. The injection molding process mainly includes six stages: mold closing, filling, pressure holding, cooling, mold opening and demolding. During the injection molding process, the location and number of gates can make up for the deficiencies in mold and product design to a certain extent, thereby improving product quality. However, process parameters such as injection pressure, time and temperature have a relatively greater impact on product quality.
[0003] At present, in the actual injection molding production process, better process parameters are usually obtained through repeated experiments. This is not only highly dependent on the practical experience of experts and a long experimental development cycle, but also leads to high optimization costs and is difficult to adapt to the fast pace of R&D and production. Computer-aided engineering (CAE) analysis can effectively alleviate these deficiencies, enabling staff to find better injection molding process parameters in a shorter time, thereby achieving fast and efficient mold trials and obtaining better quality products.
[0004] Although CAE-assisted analysis can effectively improve the efficiency of parameter design, it still relies on the work experience of professionals. Most defects of injection molding products can be solved in the virtual mold trial stage, but there are still a small number of defects that cannot be optimized by CAE. It usually takes multiple mold trials to obtain the best process parameters, which can easily lead to large production waste and delays in mold production. Therefore, further optimization of injection molding process parameters is of great significance to improving injection molding production efficiency and yield rate. Summary of the invention
[0005] The purpose of this application is to provide a method and system for optimizing injection molding process parameters based on Gaussian functions, so as to solve the technical problem that a small part of the above defects cannot be optimized through CAE, and usually multiple mold trials are required to obtain the optimal process parameters.
[0006] According to one aspect of the present application, a method for optimizing injection molding process parameters based on a Gaussian function is proposed, the method comprising the following steps:
[0007] S1. Acquire a data set, where the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters;
[0008] S2, using Latin hypercube sampling to extract the preprocessed data set as the training sample set, and fitting the Kriging proxy model based on Gaussian process regression;
[0009] S3, extract the remaining data sets as test sample sets, evaluate the prediction accuracy of the Kriging proxy model, and use the Kriging proxy model with an error result less than the error threshold as the injection molding production process proxy model;
[0010] S4. Design a multi-objective constraint optimization algorithm to globally optimize the injection molding production process proxy model and find the optimal process parameter combination;
[0011] S5. Expand the optimal process parameter combination and the quality parameters corresponding to the optimal process parameter combination to the training sample set, update the Kriging proxy model, reposition to step S3, continue iterative optimization until the computing resources are exhausted, and output the optimal dominant solution set as the final optimized injection molding process parameters.
[0012] In the above technical scheme, this application uses Kriging proxy model construction, model evaluation, multi-objective optimization and optimization parameter update model to ultimately effectively optimize the key process parameters in the injection molding process, solves the problem that the CAE model cannot effectively optimize production parameters for defects, and further improves the quality of products.
[0013] Furthermore, step S1 includes determining key injection molding process parameters and quality parameters, the injection molding process parameters include initial injection molding parameters of actual production and / or initial experimental parameters obtained through orthogonal experimental design, and the quality parameters corresponding to the injection molding process parameters include product quality parameters of actual production and / or simulated quality parameters obtained based on CAE simulation tests. Orthogonal experimental design is used to ensure that all possible combinations are covered and the interaction between parameters can be fully reflected, and CAE simulation tests are used to improve the credibility of the initialization data through simulation experiments.
[0014] Further, step S2 includes the following sub-steps:
[0015] S21, preprocessing the data set;
[0016] S22, using Latin hypercube sampling to extract the preprocessed data set to obtain a training sample set, the training sample set includes several groups of process parameter sample points And each set of process parameter sample points Corresponding quality parameters ;
[0017] S23, each group of injection molding process parameter sample points As a single decision vector, that is, each set of injection molding process parameter sample points Corresponding quality parameters For the target value corresponding to a single decision vector, a Kriging surrogate model is fitted, and the relevant parameters of the Kriging surrogate model are determined using the maximum likelihood estimation method.
[0018] In the above technical solution, data preprocessing, sample point acquisition and Kriging proxy model fitting are performed to construct a Kriging proxy model based on Gaussian process regression. The data set obtained by Latin hypercube sampling can better reflect the changing relationship between different parameters and can achieve better results when used for model training.
[0019] Furthermore, the S3 steps include:
[0020] S31. Randomly select a required number of groups of process parameter sample points from the remaining data sets. And each set of process parameter sample points Corresponding actual quality parameters As a test sample set, the actual quality parameters Including actual product quality parameters and / or simulated quality parameters obtained based on CAE simulation tests;
[0021] S33, using the Kriging proxy model constructed in step 1 to test several groups of process parameter sample points of the sample set Make predictions and obtain sample points for each set of process parameters Corresponding prediction quality parameters ;
[0022] S34. Calculate actual quality parameters and prediction quality parameters In response to the error result being less than the error threshold, the Kriging proxy model is transformed into a multi-objective constrained optimization injection molding production process proxy model, and the error threshold ranges from 0 to 5%.
[0023] Furthermore, the expression of the injection molding production process proxy model is:
[0024]
[0025] in, Represented as the optimized quality parameter set, Denoted as the first quality parameter, Expressed as Inequality constraints, Expressed as An equality constraint.
[0026] Further, step S4 includes the following sub-steps:
[0027] S41, initialize and determine the population size of the sample data set , randomly generated in the parameter space The initial population Pop1 and The initial population of individuals is Pop2;
[0028] S42, using the Kriging proxy model to evaluate the fitness of each individual, where the individuals in the Pop1 population are unconstrainedly calculated and then the binary tournament selection method is used to generate Individuals that meet the constraints are added to MatingPool1. After further calculating the degree of violation of each constraint by each individual in Pop2 population, the binary tournament selection method is used to generate Individuals that meet the constraints join MatingPool2;
[0029] S43, using the genetic evolution operator to obtain the offspring Offspring1 from MatingPool1, and using the differential evolution operator to obtain the offspring Offspring2 from MatingPool2;
[0030] S44, merge Pop2, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the truncated archive strategy to select and update Pop1; merge Pop1, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the non-dominated sorting strategy to select and update Pop2;
[0031] S45, in response to the current iteration satisfying the convergence condition, outputting the current population Pop2 as the optimal solution set of Pareto, that is, the optimal process parameter combination;
[0032] or
[0033] In response to the current iteration not satisfying the convergence condition, the best individual is selected in step S44 as the original population of the next generation, and the process returns to step S42 for the next iteration.
[0034] Furthermore, S42 and Respectively expressed as:
[0035]
[0036]
[0037] In the formula, is the size of the input population, is the maximum number of evolutionary iterations, is the number of current evolution iterations.
[0038] Furthermore, the new generation crowding distance in the non-dominated sorting strategy is expressed as:
[0039]
[0040]
[0041] in, is represented as crowding distance, It is expressed as the proportion of feasible solutions in the population, It is expressed as the fitness value of the individual.
[0042] In the above technologies, the binary tournament can effectively evaluate individuals in the population and select individuals with higher fitness values and satisfying constraints to enter the next generation; through genetic evolution operators and differential evolution operators, a diverse offspring population can be generated, providing more exploration space for the optimization algorithm. The truncated archive strategy maintains the excellent genes of the population by simply sorting the fitness values and selecting the highest individuals. The non-dominated sorting strategy can effectively select and retain excellent individuals in the population, while taking into account the fitness values and constraints. The above strategies can improve the effect of the optimization algorithm.
[0043] Furthermore, the injection molding process parameters include at least two of material temperature, mold temperature, locking pressure, injection stage 1 pressure, injection stage 2 pressure, gate position, injection time, and holding time.
[0044] Further, the quality parameter includes at least two of the number of pores, the length of the weld line, and the amount of warpage.
[0045] In the second aspect, the present application proposes an injection molding process parameter optimization system based on Gaussian function, the system comprising:
[0046] A data set acquisition module is configured to acquire a data set, wherein the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters;
[0047] The model building module is configured to extract the preprocessed data set as the training sample set using Latin hypercube sampling, and fit the Kriging proxy model based on Gaussian process regression;
[0048] The multi-objective model module is configured to extract the remaining data sets as the test sample set, evaluate the prediction accuracy of the Kriging proxy model, and use the Kriging proxy model with an error result less than the error threshold as the injection molding production process proxy model;
[0049] The parameter optimization module is configured to design a multi-objective constraint optimization algorithm to globally optimize the injection molding production process proxy model and solve the optimal process parameter combination;
[0050] The model updating module is configured to expand the optimal process parameter combination and the quality parameters corresponding to the optimal process parameter combination to the training sample set, update the Kriging proxy model, relocate to the multi-objective model module, continue iterative optimization, until the computing resources are exhausted, and output the optimal dominant solution set as the final optimized injection molding process parameters.
[0051] Compared with the prior art, the beneficial results of the present invention are:
[0052] (1) The present invention abandons the traditional empirical mold trial and CAE simulation mold trial methods, introduces the concept of artificial intelligence, and adopts an injection molding process parameter optimization method based on Gaussian process regression. The proxy model is trained through the data set, and the evaluation coefficient is selected to verify the model accuracy. The multi-objective constraint optimization algorithm is designed to solve the model and obtain the optimized injection molding parameters. This method not only improves the efficiency of obtaining the final optimized parameters, but also has a significant optimization effect. The final injection molding parameters obtained are more accurate and can fully meet the requirements of the injection molding process.
[0053] (2) In order to improve design efficiency and yield rate, the present application obtains more accurate injection molding process parameters while accelerating the speed of obtaining the final optimized parameters, thereby meeting the injection molding requirements, reducing waste in the production process, and improving product quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated into and constitute a part of this specification. The accompanying drawings illustrate the embodiments and are used together with the description to explain the principles of the present invention. It will be easy to recognize other embodiments and many expected advantages of the embodiments because they become better understood by reference to the following detailed description. The elements of the drawings are not necessarily to scale with each other. The same reference numerals refer to corresponding similar parts.
[0055] Figure 1 is a flow chart of a method for optimizing injection molding process parameters based on a Gaussian function according to an embodiment of the present application;
[0056] Figure 2 is a schematic flow chart of an injection molding process parameter optimization method based on a Gaussian function according to an embodiment of the present application;
[0057] Figure 3 is a schematic diagram of an optimization algorithm framework of an injection molding process parameter optimization method based on a Gaussian function according to an embodiment of the present application;
[0058] Figure 4is a schematic diagram of an electrical wiring harness connection connector structure according to an embodiment of the present application;
[0059] Figure 5 It is a framework diagram of an injection molding process parameter optimization system based on Gaussian function according to an embodiment of the present application. DETAILED DESCRIPTION
[0060] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0061] refer to Figure 1 , Figure 1 A flowchart of the injection molding process parameter optimization method based on Gaussian function of the present application is shown. As shown in the figure, the method includes the following steps:
[0062] S101. Acquire a data set, where the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters.
[0063] In some specific embodiments, it includes determining key injection molding process parameters and quality parameters, the injection molding process parameters include initial injection molding parameters of actual production and / or initial experimental parameters obtained through orthogonal experimental design, and the quality parameters corresponding to the injection molding process parameters include product quality parameters of actual production and / or simulated quality parameters obtained based on CAE simulation test. Specifically, the initial injection molding parameters and the corresponding quality parameter data set are obtained through orthogonal experimental design and CAE simulation calculation and the parameter range is determined. The initial injection molding process parameters include at least two of the material temperature, mold temperature, locking pressure, injection stage one pressure, injection stage two pressure, gate position, injection time, and holding time, and the quality parameters include at least two of the number of pores, weld line length, and warpage.
[0064] Specifically, when less data sets are collected from actual production, step S101 also includes determining key process parameters and quality parameters, determining parameter ranges and levels, designing orthogonal experimental tables, and conducting experiments and data collection. Identify and select key process parameters that affect the quality of injection molding, such as material temperature, mold temperature, clamping force, injection pressure, injection speed, holding time, cooling time, etc., and determine key quality parameters such as warpage, weld line length, number of pores, etc. Determine the parameter range and level, and determine a reasonable range of values and level number for each key process parameter. Use empirical data and preliminary experimental results to determine the parameter range to ensure that all possible process conditions are covered. According to the schemes in the orthogonal experimental table, conduct experiments or CAE simulations one by one, and record the process parameters and corresponding quality parameters under each experimental scheme. Ensure that the experimental conditions are consistent and reduce the impact of external factors on the results.
[0065] S102. Use Latin hypercube sampling to extract the preprocessed data set as a training sample set, and fit the Kriging proxy model established based on Gaussian process regression.
[0066] In some specific embodiments, the data set is preprocessed to obtain a training sample set and a Kriging proxy model of the injection molding process is fitted by Gaussian process regression, which specifically includes the following steps:
[0067] S1021, preprocessing the data set;
[0068] S1022. Use Latin hypercube sampling to extract the preprocessed data set to obtain a training sample set, which includes several groups of process parameter sample points. And each set of process parameter sample points Corresponding quality parameters ;
[0069] S1023, each group of injection molding process parameter sample points As a single decision vector, that is, each set of injection molding process parameter sample points Corresponding quality parameters For the target value corresponding to a single decision vector, fit the Kriging surrogate model and use the maximum likelihood estimation method to determine the relevant parameters of the Kriging surrogate model .
[0070] Specifically, when the actual production data set is sufficient to cover the training sample set and the test sample set, the preprocessed data set can be divided into the training sample set and the test sample set, and then steps S1021 to S1023 are executed. If the actual production data set is small, the number of sample points to be generated is determined according to the experimental design and model requirements. The number of sample points should be sufficient to cover all areas of the parameter space to improve the fitting accuracy of the model. The Latin hypercube sampling method is used to extract several groups of process parameter sample points from the parameter space. , obtain the product quality parameters corresponding to the sample points through CAE simulation .Will and The combination is used as the initial sample to determine the relevant model parameters And fit the Kriging proxy model. Specifically, for the sample points and quality parameters Standardization is performed to make the data have the same dimension, and the standardized sample points and quality parameters As training data, fit the Kriging proxy model. Among them, a set of process parameter sample points is a decision vector, a process parameter in the group is a decision variable, and a series of decision variables constitute a solution, namely a decision vector. A solution is calculated by the trained model to obtain the target value, namely the corresponding quality parameter combination, and the relevant parameters of the model are determined by the maximum likelihood estimation method. . is the regression coefficient, for The polynomial of The error of pseudo-random distribution, then the expression of Kriging surrogate model is:
[0071]
[0072] S103, extracting the remaining data sets as test sample sets, evaluating the prediction accuracy of the Kriging proxy model, and using the Kriging proxy model whose error result is less than the error threshold as the injection molding production process proxy model.
[0073] In some specific embodiments, when there are many actual production data sets, a required number of groups of process parameter sample points are randomly selected from the remaining data sets. And each set of process parameter sample points Corresponding actual quality parameters , using the Kriging proxy model constructed in step S102 to test the sample set Make predictions and obtain sample points for each set of process parameters Corresponding prediction quality parameters , calculate the actual quality parameters and prediction quality parameters The error results are used to evaluate the prediction accuracy of the Kriging proxy model. In response to the Kriging proxy model whose error results are less than the error threshold, it is transformed into an injection molding production process proxy model for multi-objective constrained optimization. The error threshold ranges from 0 to 5%.
[0074] In some specific embodiments, when the actual production data set is small, the initial experimental parameters of the orthogonal experimental design are extracted as the process parameter sample points of the test sample set. CAE simulation is performed to evaluate the prediction accuracy of the Kriging surrogate model and convert it into a global constrained multi-objective optimization problem, which includes the following steps:
[0075] S1031': Randomly select several groups of process parameter sample points from the remaining sample points .
[0076] S1032': Simultaneously calculate several groups of process parameter sample points Perform CAE simulation and Kriging proxy model prediction to obtain the actual quality parameters of each test sample and prediction quality parameters , where the actual quality parameter Calculate the actual quality parameters for the simulated quality parameters of the CAE simulation test and prediction quality parameters The error result is determined according to the need, and the error range is generally and If the error is controlled within 5%, the model accuracy can be considered qualified. After the model accuracy is qualified, it is transformed into a multi-objective constrained optimization problem.
[0077] Furthermore, the global constrained multi-objective optimization problem is transformed from the Kriging proxy model with qualified accuracy. The mathematical expression of the problem is:
[0078]
[0079] In a specific embodiment, wherein Represented as the optimized quality parameter set, Expressed as the first quality parameter, such as warpage, Expressed as an inequality constraint, such as the injection temperature range, Expressed as an equality constraint, such as feed volume.
[0080] S104. Design a multi-objective constraint optimization algorithm to globally optimize the injection molding production process proxy model, optimize the process parameters, and obtain the optimal process parameter combination.
[0081] In a specific embodiment, several process parameters are selected as optimization targets, and a multi-objective constraint optimization algorithm is designed. Specifically, the following steps are included:
[0082] S1041. Initialize and determine the population size of the sample data set. , an initial population Pop1 of N individuals and an initial population Pop2 of N individuals are randomly generated in the parameter space, where each individual is a decision vector, and the decision vector is composed of 2-6 decision variables among material temperature, mold temperature, clamping force, injection pressure, injection speed, holding time, and cooling time.
[0083] S1042, using the Kriging proxy model to evaluate the fitness value of each individual, where Pop1 performs unconstrained calculations, and Pop2 calculates the degree of violation of each individual to each constraint condition in addition to calculating the fitness value; performing binary tournament selection within the two populations, respectively, to obtain two parent pools MatingPool1 and MatingPool2 for generating the next generation of individuals, wherein Individuals that meet the constraints join MatingPool1. Individuals that meet the constraints join MatingPool2.
[0084] Specifically, the constructed Kriging agent model is used to evaluate the fitness value of each individual and generate a fitness list. For each individual, the degree of violation of each constraint is calculated and a violation degree list is generated. Pop1 is selected for an unconstrained binary tournament to generate The parent pool MatingPool1 of each individual performs a constrained binary tournament selection on Pop2 to generate The parent pool MatingPool2 of individuals. Individuals with lower constraint violations are given priority. If the violation degrees are the same, individuals with higher fitness values are selected. Through the binary tournament, individuals in the population can be effectively selected, and individuals with higher fitness values that meet the population requirements can be selected to enter the next generation, thereby improving the effect of the optimization algorithm.
[0085] in, and Respectively expressed as:
[0086]
[0087]
[0088] In the formula, is the size of the input population, is the maximum number of evolutionary iterations, is the number of current evolution iterations.
[0089] S1043, using the genetic evolution operator to obtain the offspring Offspring1 from MatingPool1, and using the differential evolution operator to obtain the offspring Offspring2 from MatingPool2;
[0090] Specifically, the genetic evolution operator includes randomly selecting a pair of parent individuals from MatingPool1 to perform crossover operations and mutation operations (randomly changing certain genes), etc., to generate two offspring individuals, and repeating the above steps until a sufficient number of offspring individuals are generated. The differential evolution operator includes randomly selecting three different parent individuals from MatingPool2, generating mutant individuals through differential mutation operations, generating test individuals through crossover operations, comparing the fitness values of the test individuals with those of the parent individuals, and selecting the better individuals as offspring. Repeat the above steps until a sufficient number of offspring individuals are generated. Through the genetic evolution operator and the differential evolution operator, a diverse offspring population can be generated, providing more exploration space for the optimization algorithm, thereby improving the optimization effect.
[0091] S1044, merge Pop2, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the truncated archive strategy to select and update Pop1; merge Pop1, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the non-dominated sorting strategy to select and update Pop2;
[0092] Specifically, the population Pop2 is merged with the offspring Offspring1 and Offspring2 to form a new set, and the fitness value of each individual after the merger is calculated using the Kriging proxy model. The N individuals with the highest fitness values are selected from the merged population to update the population Pop1. The truncated archive strategy maintains the excellent genes of the population by simply sorting the fitness value and selecting the highest individual. The population Pop1 is merged with the offspring Offspring1 and Offspring2 to form a new set, and the fitness value of each individual after the merger is calculated using the Kriging proxy model. Non-dominated sorting is performed to generate multiple Pareto fronts. For each individual, the degree of constraint violation is calculated. According to the non-dominated sorting and the degree of constraint violation, individuals with higher fitness values and lower violation degrees are selected to update the population Pop2. The above steps can effectively select and retain excellent individuals in the population, while taking into account both fitness values and constraints, thereby improving the effect of the optimization algorithm.
[0093] Among them, the new generation crowding distance in the non-dominated sorting strategy is expressed as:
[0094]
[0095]
[0096] in, is represented as crowding distance, It is expressed as the proportion of feasible solutions in the population, It is expressed as the fitness value of the individual.
[0097] S1045, in response to the current iteration satisfying the convergence condition, such as exhaustion of computing resources, outputting the current population Pop2 as the optimal solution set of Pareto;
[0098] or
[0099] In response to the current iteration not meeting the convergence condition, the best individual is selected in step S44 as the original population of the next generation, and the process returns to step S1042 for the next round of iteration.
[0100] Specifically, common convergence conditions include: reaching the maximum number of iterations, exhaustion of computing resources, and little change in the fitness value of the population. At the end of each iteration, check the current convergence condition. If the convergence condition is met, the non-dominated solution in the current population is the Pareto optimal solution set, and the Pareto optimal solution set can be output in the form of tables, graphs, etc., which is convenient for analysis and display for further analysis or application. If the convergence condition is not met, select the best individual as the original population of the next generation, continue to execute step S1042, evaluate the fitness of the new population, perform binary tournament selection and offspring generation, until the convergence condition is met.
[0101] S105, expanding the optimal process parameters and the quality parameters corresponding to the optimal process parameters to the training sample set, updating the Kriging proxy model, repositioning to step S103, continuing iterative optimization until the computing resources are exhausted, and outputting the optimal dominant solution set as the final optimized injection molding process parameters.
[0102] In a specific embodiment, the optimized process parameters and the corresponding quality parameters are added to the training data set, the injection molding process Kriging proxy model is updated, and the process is directed to step S103 until the iterative optimization reaches the maximum number of iterations, that is, the computing resources are exhausted, wherein the maximum number of iterations is set to 300-400 times, and the output optimal dominant solution set is obtained as the final optimized injection molding process parameters.
[0103] Continue to refer Figure 2 , Figure 2 The schematic diagram of the process flow of the injection molding process parameter optimization method based on Gaussian function according to the present application is shown, and the injection molding process parameter optimization method includes the following steps:
[0104] Step 201: Obtain an initial parameter data set, obtain initial injection molding process parameters and their corresponding quality parameter data set through orthogonal experimental design and CAE simulation calculation, and determine the parameter range.
[0105] Step 202: pre-process the data set and use Gaussian process regression to fit the Kriging proxy model of the injection molding process.
[0106] Step 203: Check the model accuracy, extract a sample point set and perform CAE simulation to evaluate the prediction accuracy of the Kriging proxy model, and determine whether the model accuracy is qualified. If "yes", execute step 204, if "no", return to execute step 202.
[0107] Step 204: convert it into a global constrained multi-objective optimization problem, design a multi-objective optimization algorithm, select several process parameters as optimization targets, and design a multi-objective constrained optimization algorithm.
[0108] Step 205: Solve the multi-objective optimization problem through an algorithm, and solve the transformed global multi-objective constrained optimization problem through an optimization algorithm to obtain optimized process parameters and output them.
[0109] Step 206: Add the solution to the data set, add the optimized process parameters and corresponding quality parameters to the training data set, update the Kriging proxy model, and re-evaluate the prediction accuracy of the model.
[0110] Continue to refer Figure 3 , Figure 3 The optimization algorithm framework flow chart of the injection molding process parameter optimization method based on the Gaussian function of the present application is shown, and the following steps are:
[0111] Step 301: Initialize the population, determine the population size according to the complexity of the optimization problem and the computing resources, and randomly generate the initial population.
[0112] Step 302: Evaluate individual values, use the Kriging proxy model to evaluate the target value of each individual, that is, calculate the quality parameter corresponding to each individual. Calculate the degree of violation of each individual for each constraint (such as material temperature range, mold temperature range, etc.), and select individuals according to the constraint.
[0113] Step 303: Generate offspring by evolutionary computation, and update the offspring population by using genetic evolution operator and differential evolution operator respectively.
[0114] Step 304: Environment selection, merging the offspring population with the parent population to form a new population, and performing environment selection on the new population using the truncated archive strategy and the non-dominated sorting strategy respectively.
[0115] Step 305: Determine the convergence condition and whether the maximum number of iterations has been reached. If yes, output the optimal solution set to end the operation. If not, select the optimal individual as the original population of the next generation and return to step 302 for the next round of iteration.
[0116] refer to Figure 4 , Figure 4 The schematic diagram of the structure of the electrical wiring harness connector according to the present application is shown. The injection molded product of this embodiment is an electrical wiring harness connector. The product material is PA66+25GF, the shape is a rectangular parallelepiped, and the size is . The average wall thickness is 0.6mm, the thickest is 0.87mm, the thinnest is 0.36mm, and the volume V1=573.6mm3. The product is a typical grid-type plastic part. The main body has 6 grids. The upper buckle is unevenly stressed, easy to break, and the thickness of the curved surface is uneven. The product is used for electrical wiring harness connection and plug-in, and has certain strength and appearance requirements, especially the main body cannot have invisible cracks. The pores should be as few and small as possible, and cannot be distributed in the grid and buckle positions to avoid premature breakage and affect the performance of the product.
[0117] Based on orthogonal test method and CAE simulation calculation, the initial injection molding process parameters and corresponding quality are determined. The training data set is obtained through Latin hypercube sampling, and the Kriging surrogate model is fitted. The remaining part is used to test the accuracy of the Kriging surrogate model prediction and converted into a constrained multi-objective optimization problem.
[0118] Optionally, in this embodiment, the DACE toolbox is used in MATLAB to establish a Kriging proxy model, and the genetic algorithm GA is selected as the test algorithm. In order to verify its accuracy, another part of the data set is used to analyze the accuracy of the Kriging proxy model prediction by CAE simulation and model prediction, with the number of pores M and the amount of warpage T as quality parameters. The CAE analysis value is compared with the proxy model analysis value.
[0119] Optionally, in this embodiment, the number of pores M and the amount of warpage T are selected as optimization targets, and a multi-objective evolutionary optimization algorithm is designed.
[0120] S41: Initialize two initial populations Pop1 and Pop2 with a population size of N.
[0121] S42: The target values and constraint violation degrees of all individuals in the initial population are evaluated through the Kriging surrogate model, and the individuals that meet the constraints are added to the two parent pools MatingPool1 and MatingPool2 for archiving.
[0122] S43: The first parent pool MatingPool1 is updated by the genetic evolution operator to obtain the offspring population Offspring1, and the second parent pool MatingPool2 is updated by the differential evolution operator to obtain the offspring population Offspring2.
[0123] S44: Combine the two child populations with the parent population to form two new populations, use the truncated archive strategy for environmental selection for the first population, and use the non-dominated sorting strategy for environmental selection for the second population.
[0124] S45: Check whether the convergence condition is met. If the convergence condition is met, the best individual in Pop2 is output as the Pareto optimal solution set; otherwise, the best individual selected in S44 is used as the original population of the next generation and directed to S42.
[0125] Furthermore, the final optimized injection molding parameters obtained by solving the trained Kriging proxy model using the designed multi-objective optimization algorithm are matched with the screw rotation control of the injection molding machine used in actual injection molding to obtain the rotation control scheme of each section of the screw.
[0126] Further references Figure 5 As an implementation of the above method, in the second aspect, the present application provides an embodiment of an injection molding process parameter optimization system 500 based on a Gaussian function. Figure 1 Corresponding to the method embodiment shown, the system can be specifically applied to various electronic devices. The system 500 includes a data set acquisition module 501, a model construction module 502, a multi-objective model module 503, a parameter optimization module 504 and a model update module 505 which are interconnected and communicated with each other, wherein:
[0127] A data set acquisition module 501 is configured to acquire a data set, wherein the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters;
[0128] The model building module 502 is configured to extract the preprocessed data set as a training sample set using Latin hypercube sampling, and fit the Kriging proxy model established based on Gaussian process regression;
[0129] The multi-objective model module 503 is configured to extract the remaining data sets as test sample sets, evaluate the prediction accuracy of the Kriging proxy model, and use the Kriging proxy model with an error result less than the error threshold as the injection molding production process proxy model;
[0130] The parameter optimization module 504 is configured to design a multi-objective constraint optimization algorithm to perform global optimization on the injection molding production process proxy model and solve the optimal process parameter combination;
[0131] The model updating module 505 is configured to expand the optimal process parameter combination and the quality parameters corresponding to the optimal process parameter combination to the training sample set, update the Kriging proxy model, relocate to the multi-objective model module, continue iterative optimization, until the computing resources are exhausted, and output the optimal dominant solution set as the final optimized injection molding process parameters.
[0132] Although the principles of the present invention are described in detail above in conjunction with the preferred embodiments of the present invention, those skilled in the art should understand that the above embodiments are merely explanations of the exemplary implementations of the present invention, and are not intended to limit the scope of the present invention. The details in the embodiments do not constitute limitations on the scope of the present invention, and any obvious changes such as equivalent transformations, simple replacements, etc. based on the technical solution of the present invention, without departing from the spirit and scope of the present invention, fall within the protection scope of the present invention.
Claims
1. A method for optimizing injection molding process parameters based on Gaussian function, characterized in that: The method comprises: S1. Acquire a data set, wherein the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters; S2, using Latin hypercube sampling to extract the preprocessed data set as a training sample set, and fitting a Kriging proxy model established based on Gaussian process regression; S3. Extract the remaining data sets as test sample sets, evaluate the prediction accuracy of the Kriging proxy model, and use the Kriging proxy model with an error result less than the error threshold as the injection molding production process proxy model. The expression of the injection molding production process proxy model is: in, Represented as the optimized quality parameter set, Denoted as the first quality parameter, Expressed as Inequality constraints, Expressed as equality constraints; S4, designing a multi-objective constraint optimization algorithm to perform global optimization on the injection molding production process proxy model to solve the optimal process parameter combination; S5, expanding the optimal process parameter combination and the quality parameters corresponding to the optimal process parameter combination to the training sample set, updating the Kriging proxy model, repositioning to the S3 step, continuing iterative optimization until the computing resources are exhausted, and outputting the optimal dominant solution set as the final optimized injection molding process parameters; Wherein, the S4 step includes the following sub-steps: S41, initialize and determine the population size of the sample data set , randomly generated in the parameter space The initial population Pop1 and The initial population of individuals is Pop2; S42, using the Kriging proxy model to evaluate the fitness of each individual, where the individuals in the Pop1 population are unconstrainedly calculated and then the binary tournament selection method is used to generate Individuals that meet the constraints are added to MatingPool1. After further calculating the degree of violation of each constraint by each individual in Pop2 population, the binary tournament selection method is used to generate Individuals that meet the constraints join MatingPool2; S43, using the genetic evolution operator to obtain the offspring Offspring1 from MatingPool1, and using the differential evolution operator to obtain the offspring Offspring2 from MatingPool2; S44, merge Pop2, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the truncated archive strategy to select and update Pop1; merge Pop1, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the non-dominated sorting strategy to select and update Pop2; S45, in response to the current iteration satisfying the convergence condition, outputting the current population Pop2 as the optimal solution set of Pareto, that is, the optimal process parameter combination; or In response to the current iteration not satisfying the convergence condition, the best individual selected in step S44 is used as the original population of the next generation, and the process returns to step S42 for the next round of iteration.
2. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The S1 step includes determining key injection molding process parameters and quality parameters, wherein the injection molding process parameters include initial injection molding parameters of actual production and / or initial experimental parameters obtained through orthogonal experimental design, and the quality parameters corresponding to the injection molding process parameters include product quality parameters of actual production and / or simulated quality parameters obtained based on CAE simulation testing.
3. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The S2 step includes the following sub-steps: S21, preprocessing the data set; S22, using Latin hypercube sampling to extract the preprocessed data set to obtain a training sample set, wherein the training sample set includes several groups of process parameter sample points And each set of process parameter sample points Corresponding quality parameters ; S23, each group of injection molding process parameter sample points As a single decision vector, that is, each group of injection molding process parameter sample points Corresponding quality parameters The target value corresponding to a single decision vector is fitted to the Kriging surrogate model, and the relevant parameters of the Kriging surrogate model are determined by the maximum likelihood estimation method. .
4. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The S3 step includes: S31, randomly selecting a required number of groups of process parameter sample points from the remaining data sets And each set of process parameter sample points Corresponding actual quality parameters As a test sample set, the actual quality parameters Including actual product quality parameters and / or simulated quality parameters obtained based on CAE simulation tests; S33, using the Kriging proxy model constructed in step S2 to analyze several groups of process parameter sample points of the test sample set Make predictions and obtain sample points for each set of process parameters Corresponding prediction quality parameters ; S34, calculating the actual quality parameter and the prediction quality parameter The error result is converted into the injection molding production process proxy model of multi-objective constraint optimization in response to the Kriging proxy model having the error result less than the error threshold, and the error threshold ranges from 0 to 5%.
5. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The S42 and Respectively expressed as: In the formula, is the size of the input population, is the maximum number of evolutionary iterations, is the number of current evolution iterations.
6. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The new generation crowding distance in the non-dominated sorting strategy is expressed as: in, is represented as crowding distance, It is expressed as the proportion of feasible solutions in the population, It is expressed as the fitness value of the individual.
7. The method for optimizing injection molding process parameters based on Gaussian function according to claim 1, characterized in that: The process parameters include at least two of material temperature, mold temperature, locking pressure, injection stage one pressure, injection stage two pressure, gate position, injection time, and holding time; the quality parameters include at least two of the number of pores, weld line length, and warpage.
8. An injection molding process parameter optimization system based on Gaussian function, characterized in that: The system comprises: A data set acquisition module, configured to acquire a data set, wherein the data set includes injection molding process parameters and quality parameters corresponding to the injection molding process parameters; A model building module is configured to extract the preprocessed data set using Latin hypercube sampling as a training sample set, and fit a Kriging proxy model established based on Gaussian process regression; The multi-objective model module is configured to extract the remaining data sets as test sample sets, evaluate the prediction accuracy of the Kriging proxy model, and use the Kriging proxy model with an error result less than the error threshold as the injection molding production process proxy model. The expression of the injection molding production process proxy model is: in, Represented as the optimized quality parameter set, Denoted as the first quality parameter, Expressed as Inequality constraints, Expressed as equality constraints; A parameter optimization module is configured to design a multi-objective constraint optimization algorithm to globally optimize the injection molding production process proxy model and solve the optimal process parameter combination; A model updating module is configured to expand the optimal process parameter combination and the quality parameters corresponding to the optimal process parameter combination to the training sample set, update the Kriging proxy model, relocate to the multi-objective model module, continue iterative optimization until the computing resources are exhausted, and output the optimal dominant solution set as the final optimized injection molding process parameters; The parameter optimization module configuration implements the following steps: S41, initialize and determine the population size of the sample data set , randomly generated in the parameter space The initial population Pop1 and The initial population of individuals is Pop2; S42, using the Kriging proxy model to evaluate the fitness of each individual, where the individuals in the Pop1 population are unconstrainedly calculated and then the binary tournament selection method is used to generate Individuals that meet the constraints are added to MatingPool1. After further calculating the degree of violation of each constraint by each individual in Pop2 population, the binary tournament selection method is used to generate Individuals that meet the constraints join MatingPool2; S43, using the genetic evolution operator to obtain the offspring Offspring1 from MatingPool1, and using the differential evolution operator to obtain the offspring Offspring2 from MatingPool2; S44, merge Pop2, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the truncated archive strategy to select and update Pop1; merge Pop1, Offspring1, and Offspring2, calculate the fitness through the Kriging proxy model, and use the non-dominated sorting strategy to select and update Pop2; S45, in response to the current iteration satisfying the convergence condition, outputting the current population Pop2 as the optimal solution set of Pareto, that is, the optimal process parameter combination; or In response to the current iteration not satisfying the convergence condition, the best individual selected in step S44 is used as the original population of the next generation, and the process returns to step S42 for the next round of iteration.
Citation Information
Patent Citations
Multi-objective optimization method for injection molding parameters
CN110377948A
Multi-target optimization method for injection molding process parameters of thin-walled injection molded part
CN110640982A
Injection molding process parameter optimization method for transparent complex multi-cavity plastic part
CN112069610A
Thin-wall plastic part injection molding process parameter multi-objective optimization method
CN112101630A
Proxy model assisted antenna design optimization method based on variable spatial data constraint
CN116595867A
