CN-120145625-B - Injection molding process parameter multi-objective optimization method based on high-precision prediction model

CN120145625BCN 120145625 BCN120145625 BCN 120145625BCN-120145625-B

Abstract

The invention discloses an injection molding process parameter multi-objective optimization method based on a high-precision prediction model, which is characterized in that based on simulation test results, data are enhanced through MIC and GAN to improve the quality of simulation data, then super-parameter tuning is performed by utilizing a mixed framework of KOA and IVYA, the interpretability of a XGBoost model is enhanced by adopting a SHAP method, an optimized process parameter searching range is provided, and finally the Pareto front of a quality target is searched by using MOCGO, and an optimal process parameter combination is determined. The test result shows that the process parameter combination optimized by the method of the invention reduces the warp deformation and the volume shrinkage by 35.7% and 10.1%, can effectively reduce the warp deformation and the volume shrinkage of the molded plastic part, further obtains high-quality plastic part products, and can provide theoretical basis and data support for obtaining the optimal process parameter combination of injection molding.

Inventors

ZHU LIUYU
FAN XIYING
GUO YONGHUAN
JI GUANGZHONG

Assignees

江苏师范大学

Dates

Publication Date: 20260508
Application Date: 20250117

Claims (6)

1. The multi-objective optimization method for the injection molding process parameters based on the high-precision prediction model is characterized by comprising the following steps of: Step1, taking the buckling deformation and the volume shrinkage as quality optimization targets, selecting technological parameters affecting the two targets as optimization design variables, determining a value range of the technological parameters according to recommended values of the technological parameters, taking the value range of the technological parameters as a test design space, and performing injection molding simulation based on Moldflow to obtain a quality optimization target test result; step2, preprocessing the process parameter data through MIC and GAN based on simulation test results; step3, super parameters of the XGBoost model are optimized by utilizing a KOA-IVYA mixed optimization algorithm, and an MIC-GAN-KOA-IVYA-XGBoost injection molding quality prediction model is constructed; Step4, analyzing an MIC-GAN-KOA-IVYA-XGBoost injection molding quality prediction model by a SHAP interpretation method, and defining the influence degree of each process parameter on a prediction result and providing an optimized process parameter search range by generating a characteristic importance graph; Step5, performing multi-objective optimization on the optimized process parameter range by using MOCGO and MIC-GAN-KOA-IVYA-XGBoost prediction models to obtain a Pareto front of a quality optimization target; step6, calculating the comprehensive score of each solution in the Pareto front of the obtained quality optimization target through an AHP-EW-GM method, and sequencing to determine the optimal combination of the process parameters.
2. The multi-objective optimization method for injection molding process parameters based on a high-precision prediction model according to claim 1, wherein in Step2, when preprocessing process parameter data by MIC, the specific steps are as follows: ① Generating variable scatter diagram and meshing process parameters And quality target The values of (2) are represented in a two-dimensional space through a scatter diagram, and the data are subjected to grid division; ② Calculating mutual information of each grid, namely, when calculating the mutual information under the given grid division, replacing probability distribution of variables with frequency of scattered point distribution in the grid, and measuring process parameters through the mutual information And quality target Dependency relationship between them; ③ The maximum value of the mutual information is obtained by calculating the mutual information values under all grid division and selecting the maximum value; ④ MIC is calculated at a plurality of parameters representing different processes Line number and representing different quality targets Under the column number combination of (2), calculating the maximum mutual information value by using the following formula: Wherein: Is mutual information; Normalization adjustment is performed on grid division; ⑤ And screening important process parameters, namely after the MIC values of all the process parameters on the quality targets are calculated, comparing the MIC values with a preset threshold value, and selecting the process parameters with the MIC values larger than or equal to the threshold value, wherein the process parameters are regarded as having stronger correlation.
3. The multi-objective optimization method for injection molding process parameters based on a high-precision prediction model according to claim 2, wherein in Step2, when the process parameter data is preprocessed by GAN, the specific steps are as follows: ① Initialization generator and arbiter network, setting generator Distinguishing device Is set to be a constant; ② Setting training parameters, setting generator Random noise of input(s) Dimension, batch size, total training wheel number, learning rate, optimizer, momentum parameter; ③ Defining an overall optimization objective, wherein the overall optimization objective expression is as follows: Wherein: representation discriminator For real data The probability of true is determined; representation generator According to random noise The generated counterfeit sample; ④ Training cycle by random noise Input generator Generating initial synthetic data and inputting the initial synthetic data together with the real data into a discriminator Classifying and judging device Optimizing self parameters by calculating a loss function to enhance the ability to distinguish between real data and generated data, while generating According to the discriminant Optimizing self parameters, gradually approaching the generated data to the real data until the discriminator The real data and the generated data cannot be effectively distinguished, and the synthesized data is finally output; ⑤ When the discriminator When the generated data cannot be distinguished, adopting a k-means clustering algorithm to perform clustering analysis on the original data and the generated data, and performing dimension reduction on the data by a t-SNE method to perform visual analysis.
4. The multi-objective optimization method for injection molding process parameters based on a high-precision prediction model according to claim 3, wherein in Step3, the mixing optimization algorithm using KOA-IVYA is specifically as follows: ① KOA global search: The individual location update formula for KOA is as follows: Wherein: Is that Updated positions after the iteration; Is that The position after the iteration; is a sign of changing the search direction; Is a celestial body The speed required to reach the new location; is a function of attraction between the current location and the location of the best solution; is a number randomly generated on the basis of normal distribution; is the position of the current best solution; ② IVYA local development: IVYA the individual location update formula is as follows: Wherein: Is a gain vector for increasing randomness; ③ Dynamic population adjustment: The formula for the variation of population size by iteration number is as follows: Wherein: KOA population size; Is IVYA population sizes; is the initial total number of individuals; the current iteration times; is the maximum number of iterations.
5. The method for multi-objective optimization of injection molding process parameters based on a high-precision prediction model as claimed in claim 4, wherein in Step5, when the multi-objective optimization is performed by MOCGO, candidate solution positions are generated by 4 different modes of generating completely randomly based on the optimal solution positions, based on the optimal solution and the mean population, based on the mean population and the optimal solution, Generating a new candidate solution based on the current global optimal solution: generating a new solution around the average solution position: generating a new solution around the optimal solution location: Wherein: is a randomly generated scaling factor, wherein 1,2,3; is a randomly generated integer 1 or 2, controlling the relative influence between the leader and the mean population, wherein 1,2,3,4,5,6; is the current optimal solution; is the average position of the population; Introducing random variation on solution positions, wherein the formula is as follows: Wherein: is a random number generator with uniform distribution, and the generated value is at the upper and lower boundaries And Between them; and evaluating the fitness of the generated potential solution through an objective function, comparing the potential solution with other solutions in a dominance relation, reserving non-dominance solutions, and forming a Pareto front after iteration is finished.
6. The multi-objective optimization method for injection molding process parameters based on a high-precision prediction model according to claim 5, wherein Step6 is specifically as follows: ① According to expert judgment, constructing a pairwise comparison matrix between indexes by using the AHP, and giving out a subjective weight ratio W AHP of each index; ② Calculating an objective weight ratio W EW by using the data distribution information through EW, and quantifying the importance of each index; ③ The subjective weight and the objective weight are fused through the GM, and the advantages of the subjective weight and the objective weight are combined to generate a final weight ratio WGM; ④ And scoring each scheme of the solution set based on the final weight, and sequencing the solutions on the Pareto front to select the technological parameter combination with optimal comprehensive performance.

Description

Injection molding process parameter multi-objective optimization method based on high-precision prediction model Technical Field The invention relates to an injection molding process parameter multi-objective optimization method based on a high-precision prediction model, in particular to an injection molding process parameter multi-objective optimization method based on MIC-GAN-KOA-IVYA-XGBoost and MOCGO high-precision prediction models, and belongs to the technical field of injection molding processing. Background In the technical field of injection molding processing in modern manufacturing industry, optimization of process parameters is important for improving product quality, reducing rejection rate and improving production efficiency. Warp deformation and volume shrinkage are two of the most common and critical quality problems in injection molding processes, warp deformation can cause deviation of product shape, and volume shrinkage affects dimensional accuracy, and control of the two directly relates to reliability and service life of the product. Therefore, how to achieve multi-objective optimization under complex process conditions, balancing the minimization of these two drawbacks has become a focus of extensive attention. Simulation techniques such as finite element analysis and Moldflow analysis are widely applied to injection molding process simulation to help predict warp deformation and volume shrinkage, and generally, firstly, the injection molding process of a thin-wall plastic part is simulated and analyzed through Moldflow software to analyze cooling effect and flow performance, then a mathematical model is constructed based on simulation data to construct the relation between a quality target and process parameters, and finally, a process parameter optimal combination is obtained through a search optimization algorithm. For example, cao Y et al construct an Adaptive Network Fuzzy Inference System (ANFIS) between warp deflection and process parameters, and obtain a process parameter optimization scheme by using a Genetic Algorithm (GA), sun Zheng et al construct a mathematical proxy model between warp deflection, volume shrinkage and process parameters based on a Gradient Enhanced Kriging (GEK) model, search a Pareto solution set by adopting a multi-objective differential evolution (MODE) algorithm, determine a process parameter optimization scheme according to a weighting coefficient, and Liu X et al construct a mathematical proxy model between warp deflection, volume shrinkage and process parameters by using a GA-optimized Extreme Learning Machine (ELM) model, and determine a final process parameter optimization scheme by applying a multi-objective firefly algorithm and combining a GRA-TOPSI multi-objective decision method. In the study of practical injection molding problems, how to extract key process parameters from simulation data remains a challenge, and a number of interrelated goals and constraints often require comprehensive consideration and trade-offs, such as warp and volume shrinkage defects, which also increase volume shrinkage due to extended cooling time while minimizing warp by decreasing cooling rate. Multi-objective optimization is a method of solving optimization problems involving multiple conflicting objectives with the objective of finding a reasonable tradeoff between multiple objectives, generating a set of Pareto fronts of multi-objective optimal values. Current research on multi-objective optimization generally converts multi-objective problems into single-objective problems through the concept of weighted combination, and common methods include gray correlation analysis (GRA), approach to ideal solution ordering (TOPSIS), entropy weight and Fuzzy Comprehensive Evaluation (FCE). However, from the perspective of multiple targets, all targets are often constrained, and improvements to one target tend to be at the expense of the other. Thus, for a multi-objective optimization problem, a large number of non-dominant solutions (Pareto solutions) are typically generated. In the multi-objective optimization research of injection molding process parameters at present, researchers often determine the final Pareto optimal solution according to engineering experience and repeated experiments, and how to determine the optimal weighing scheme to achieve the optimal comprehensive purpose is still a problem in the industry at present. Disclosure of Invention Aiming at the problems, the invention provides the multi-objective optimization method for the injection molding process parameters based on the high-precision prediction model, which can effectively reduce the buckling deformation and the volume shrinkage of the molded plastic part, further obtain a high-quality plastic part product and provide theoretical basis and data support for obtaining the optimal process parameter combination of injection molding. In order to achieve the above purpose, the injection molding proc