CN116894379A - Injection molding quality optimization method - Google Patents
Injection molding quality optimization method Download PDFInfo
- Publication number
- CN116894379A CN116894379A CN202310325988.8A CN202310325988A CN116894379A CN 116894379 A CN116894379 A CN 116894379A CN 202310325988 A CN202310325988 A CN 202310325988A CN 116894379 A CN116894379 A CN 116894379A
- Authority
- CN
- China
- Prior art keywords
- injection molding
- quality
- gen
- data
- quality data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001746 injection moulding Methods 0.000 title claims abstract description 207
- 238000000034 method Methods 0.000 title claims abstract description 108
- 238000005457 optimization Methods 0.000 title claims abstract description 46
- 238000004458 analytical method Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims description 23
- 238000013528 artificial neural network Methods 0.000 claims description 19
- 239000003795 chemical substances by application Substances 0.000 claims description 19
- 230000007547 defect Effects 0.000 claims description 19
- 238000002347 injection Methods 0.000 claims description 18
- 239000007924 injection Substances 0.000 claims description 18
- 238000000137 annealing Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 230000035772 mutation Effects 0.000 claims description 6
- 238000013210 evaluation model Methods 0.000 claims description 4
- 238000002474 experimental method Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 239000000243 solution Substances 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 abstract description 52
- 230000006870 function Effects 0.000 description 8
- 239000004033 plastic Substances 0.000 description 7
- 229920003023 plastic Polymers 0.000 description 7
- 238000001514 detection method Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 239000002994 raw material Substances 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 239000000155 melt Substances 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 241000282373 Panthera pardus Species 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000001816 cooling Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000000465 moulding Methods 0.000 description 2
- 238000002922 simulated annealing Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002425 crystallisation Methods 0.000 description 1
- 230000008025 crystallization Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Biophysics (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Biomedical Technology (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Molecular Biology (AREA)
Abstract
The invention discloses a quality optimization method of injection molding, which comprises the following steps of S1, predicting the quality of the injection molding by using a GASA method; s2, establishing a proxy model to analyze injection molding quality data; s3, centralizing an injection molding piece quality analysis result according to a stacking integrated method, providing an injection molding optimization method, predicting the injection molding piece quality in the injection molding piece production process by using a GASA method, establishing a proxy model to analyze injection molding piece quality data, and centralizing the analysis result by the stacking integrated method to obtain a corresponding injection molding piece quality optimization method.
Description
Technical Field
The invention relates to the field of injection molding technology forming, in particular to an injection molding quality optimization method.
Background
At present, plastic products are widely applied to various fields of national economy and play an increasingly important role, such as aerospace, information and communication, ship, automobile manufacturing, medicine, construction, agriculture, packaging and other fields. Injection molding has become one of the most important processing methods for plastics, and injection molded articles account for about one third of the total amount of plastic articles. Injection molding is a complex, nonlinear process with multiple variables, where factors affecting the quality of the article include the injection molding machine, material properties, molding process parameters, and the mold. The injection molding process parameters play a vital role in the quality of the product, influence the macroscopic states of flow, heat transfer, cooling and the like of the melt in the cavity and the microscopic properties of crystallization, orientation, stress and the like of the plastic, and finally influence the performance and quality of the product. The process parameters mainly comprise melt temperature, mold temperature, filling time, injection pressure, speed/pressure switching point, dwell pressure, dwell time, cooling time, mold opening time and the like. Common defects of the quality of the injection molded product include buckling deformation, weld marks, short shots, pits, sink marks, large shrinkage, flash, insufficient mechanical properties and the like. In order to ensure stable quality of injection molding in the injection molding process, the quality of the injection molding needs to be detected at any time to judge whether the production of the injection molding machine is qualified, but most existing detection methods only carry out quality inspection on injection molding finished products and feed back the quality inspection to operators according to results to carry out corresponding adjustment, so that the quality optimization process of the injection molding is slower, and a new injection molding optimization method is needed.
Chinese patent document CN104809306a discloses a "method for optimizing injection molding process parameters based on multiple quality indexes", comprising the steps of: 1. the multi-quality index injection molding process parameter optimization comprehensive balance method of the comprehensive balance method is to firstly perform visual analysis of single quality indexes on each quality index to obtain a combination of a primary order and a secondary order of influencing factors of each quality index and an optimal process level, and then perform comprehensive comparison and analysis on analysis results of the indexes according to theoretical knowledge and actual experience to obtain a better process scheme. Optimizing: (1) And respectively carrying out visual analysis on each quality index, and finding out the optimal molding process parameters and the primary and secondary influence factors of each index by using a single-target optimization method. (2) Considering the primary and secondary of each factor, comprehensively balancing a plurality of quality indexes, and finding out the production condition which meets each index as much as possible. 2. The multi-quality index injection molding process parameter optimization comprehensive evaluation method of the weighted comprehensive evaluation method is used for the situation that the evaluation index cannot be quantitatively analyzed by using a unified dimension, and the comprehensive evaluation is performed by using a dimensionless score. The influence degree of each quality index on the comprehensive quality of the product is emphasized, the dimensions of each index are inconsistent, and the weighted comprehensive score converts the multi-objective problem into a single objective, so that the comprehensive optimization of the multi-objective problem is realized. Optimizing: (1) mapping each quality index to the [0,1] space uniformly. (2) And according to the analysis result of the orthogonal test and the influence degree of each quality index on the comprehensive quality of the product, weighting according to a percentage system, and determining the weight of each quality index. And (3) weighting the comprehensive score calculation to obtain the comprehensive score. (4) And (3) carrying out mean and variance analysis on the comprehensive scores, determining the influence degree and the influence trend graph of the process parameters on the comprehensive scores, and analyzing to obtain the optimal injection molding process parameter combination. 3. And optimizing the multi-quality index injection molding process parameters of the artificial neural network method. But this patent only can carry out analysis evaluation to the quality of injection molding after injection molding production, can't predict the quality of injection molding before injection molding production, can't realize the quality optimization of injection molding more soon.
Disclosure of Invention
The invention mainly solves the technical problem that the quality of the injection molding can not be optimized rapidly by predicting the quality of the injection molding in advance, provides a quality optimizing method of the injection molding,
the technical problems of the invention are mainly solved by the following technical proposal: s1, predicting injection molding quality data by using a GASA method;
s2, establishing a proxy model to analyze injection molding quality data;
and S3, centralizing the quality analysis result of the injection molding piece according to a stacking integrated method and providing an injection molding optimization method.
In the existing injection molding production process, most of existing production lines are used for detecting the quality of injection molding finished products after production is completed, and determining an optimal scheme for injection molding production according to the quality detection result of the finished products. However, the scheme has a problem that the detection can be carried out after a certain amount of injection molding parts are produced, and the optimization direction and method of the production quality of the injection molding parts are judged according to the production quality result of the injection molding parts after the detection. The whole test feedback and final optimization process needs to consume a long time, so that the optimization efficiency of the injection molding piece is greatly reduced, and in the process of detection, the injection molding piece is continuously produced, and if the detected finished product of the injection molding piece has defects, the injection molding piece produced in the detection process can be wasted due to the defects. Therefore, in the production process of the injection molding by using the GASA method, the quality of the injection molding to be produced is predicted according to the production condition in the injection molding and the melting state of the raw materials, a proxy model is established to analyze the data obtained by prediction, and finally, a stacking integration method is used for centralizing the analysis result and providing a corresponding injection molding optimization scheme according to the analysis result. The whole quality optimization flow can be synchronously carried out in the production process of the injection molding, the injection molding optimization scheme can be provided along with the production flow of the injection molding in real time, the injection molding is always maintained to be produced with higher quality, and meanwhile, the timeliness of the method is further ensured due to synchronous optimization in the production process of the injection molding.
Preferably, in the step S1, an initial target Z is generated by first setting parameters according to the requirements of quality prediction gen The training error of the vector machine is taken as a fitness function, and the training error is brought into an initial target Z gen Calculating fitness of individual i, selecting excellent individual for crossover operation in combination with fitness and generating new target Z' gen Then, a plurality of individuals are selected from the initial population for mutation, and a mutated target Z' is generated gen PerformingComparison of Z after annealing operation gen And Z' gen And obtaining an adaptability difference value delta f, judging whether the delta f is qualified or not according to a Metropolis criterion, and repeating the steps after the judgment is finished to obtain an optimal solution. The GASA method is to fuse simulated annealing operation in a genetic algorithm, the genetic algorithm can solve the complex optimization problem, and the simulated annealing algorithm can realize rapid convergence and local search. When the injection molding piece is in daily production, an initial production state Z of the injection molding piece is established according to the running state of the injection molding machine, related indexes and the state of injection molding raw materials in the injection molding machine gen And predicting the relevant quality data, wherein the relevant parameters of the injection molding machine and the relevant parameters of the injection molding raw materials in the injection molding machine have errors when the injection molding piece is in the production process, and the errors are brought into the initial production state Z gen Obtaining the fitness of individuals i with different parameters and at the same time at Z gen The individual i fitness cross operation of each parameter is carried out to obtain a new production state Z' gen If the parameter individual i is mutated in the production process of the injection molding, the production state Z' can be obtained gen . After annealing the production state of the injection molding, the initial production state Z is compared gen And a new production state Z' gen The production change value delta f can be obtained, whether the production change value delta f changes in a controllable range is judged by using a Metropolis criterion, if the change amplitude of delta f is beyond the standard because of a certain parameter individual i, the parameter changes in a poor direction in the production process of the injection molding part, the injection molding machine or the injection molding raw material is correspondingly adjusted according to the data change value, and if the change amplitude of delta f does not exceed the standard, the existing production state is maintained. Each parameter individual i represents an influencing factor influencing the production quality of injection molding, and during annealing operation, the new state Z 'generated by the change of all parameter individuals i is realized by an iterative loop mode' gen Are all in accordance with the initial state Z gen And comparing to obtain a result.
Preferably, in the step S1, when Δf is smaller than 0, the individual receiving the mutation is represented, and when Δf is equal to or greater than 0, the individual receiving the mutation is represented by a probability p=exp (- - Δf/T) gen ) Receiving the individual, wherein af=fit (Z gen )-fit(Z gen )，T gen Is the temperature attenuation coefficient, T gen ＝kT gen-1 The target after the annealing operation is Z'. gen . Δf is the initial production state Z gen And the new state Z 'generated under the influence of individuals with different parameters i' gen And determining whether to receive the variation parameters i corresponding to different values of Deltaf after annealing operation according to Metropolis criterion.
Preferably, in the step S1, when the error is trained by using the vector machine, the input sample is set to x, the output value is set to y, w is a weight, b is a bias term,for a non-linear mapping of low-dimensional to high-dimensional space, x i And x j Represents the input value, K (x i ，x j ) As a kernel function, a i And->Langerhans multiplier, the expression of the predictive algorithm:
the decision function of the support vector machine can be obtained on the basis that:
in the daily production process of injection molding parts, factors which affect the production quality of the injection molding parts are changed differently when each injection molding part is produced, and some influencing factors have the condition of larger numerical value change amplitude when a certain injection molding part is produced, if the plurality of groups of influencing factor data are adopted, the data prediction result has larger distortion phenomenon, and finally the proposal of the optimization method is influenced, so that influencing factors in the production of all injection molding parts are not used for predicting the quality data of the injection molding parts. Therefore, the invention brings the data of the same influencing factor and different batches into the vector machine to obtain the training error, brings the training error into the initial target to calculate and obtain a relatively accurate fitness of i, is favorable for the delta f obtained by the subsequent annealing operation to be relatively accurate, and is further convenient for obtaining a relatively accurate prediction result. When the vector machine trains the influence factor errors, the corresponding data are brought into the predictive expression and the decision function, and the required training errors can be obtained.
Preferably, the agent model in step S2 includes an agent model of an artificial neural network and an agent model of gaussian process regression, wherein the agent model of gaussian process regression uses average absolute error and error duty ratio, and simultaneously combines root mean square error and goodness-of-fit evaluation model to establish agent models of three indexes to evaluate quality data of injection molding. The GASA method predicts and obtains injection molding quality data according to the change of influencing factors in the injection molding production process, brings the data into a proxy model of Gaussian process regression, evaluates and analyzes the predicted injection molding quality data according to several dimensions of an average absolute error, an error duty ratio root mean square error and a fitting goodness-of-fit evaluation model, and evaluates and analyzes a certain amount of predicted data to accurately grasp the production state of an injection molding machine when the injection molding machine utilizes injection molding raw materials for production, thereby being beneficial to accurately adjusting related influencing factors and ensuring the yield of injection molding production.
Preferably, in the step S2, the proxy model of the artificial neural network builds a model according to the shrinkage mark and the average volume shrinkage rate of the injection molding, and performs training of model parameters according to the two standards of the injection molding, so as to obtain a relatively accurate result to evaluate and analyze the quality of the injection molding. If only the agent model using Gaussian process regression has a single analysis result, whether the analysis result of the model is correct cannot be demonstrated, and the agent model of the artificial neural network is used for evaluating and analyzing the injection molding quality data obtained through prediction. The agent model of the artificial neural network is mainly evaluated and analyzed from the aspects of shrinkage marks and average volume shrinkage of the injection molding, and the agent model of the artificial neural network and the agent model analysis result of Gaussian process regression are comprehensively considered, so that a more accurate analysis result can be further obtained.
Preferably, in the step S3, all the analysis results of the quality data of all the injection molding parts are collected by using a stacking integration method, and all the data are input into a plurality of data sets s= { (X) i ，Y i ) I=1, 2, …, n }, performing corresponding training on each data set s, inputting the training result of the data set into a meta learner, evaluating the quality data of the injection molding according to the result of the meta learner, and giving a corresponding injection molding optimization method. The production yield of the injection molding is generally large, if the quality of the injection molding is analyzed and evaluated by the agent model each time, the corresponding injection molding optimization method is obtained by re-analysis and judgment, the whole injection molding optimization method is longer in the process of extraction, and the injection molding optimization efficiency is prolonged. The stacking integration method can substitute all injection molding quality data analysis results into the data sets for training, each data set contains results generated by one type of influence factors, subsets are classified according to different types of influence factors, the injection molding optimization method for processing the same type of influence factors can be obtained after training, the element learner can record the processing method after processing, if similar conditions are encountered later, the element learner does not need to spend a great deal of time again to obtain the injection molding optimization method according to the analysis results, the previous processing method can be directly traversed and fed back to operators, and the optimization efficiency of injection molding quality is accelerated.
Preferably, in the step S3, a quality index standard needs to be established, the quality of the injection molding is determined according to the standard, and Z is set ij Is the value processed by a non-dimensionality method, X ij The j-th injection defect value, X of the i-th injection experiment max X is the maximum value of defects in injection molding quality data min The minimum value of defects in the injection molding quality data is represented by the non-dimensionalized formula:
the average value of the quality data of the injection molding piece is then:
the standard deviation of the quality data of the injection molding is as follows:
therefore, the weight of the quality data of the injection molding piece can be deduced as follows:
z in the above j Sigma, the mean value of quality data of injection molding j Is the standard deviation of the quality data of the injection molding. And obtaining predicted data of the quality of the injection molding part, analyzing the predicted data to obtain a result, and determining whether the quality of the injection molding part is qualified or not according to the quality data of the injection molding part. The types of defects of injection molding parts in the injection molding process are numerous, and if different influencing factors are judged according to respective standards, the whole quality evaluation process is extremely complex, and the quality judgment time is extremely long. Therefore, when the quality standard is established, a data mean value and a standard deviation are obtained by utilizing a non-quantization formula, the data mean value can better reflect the normal horizontal line of the data, an excessively high or excessively low value cannot appear, the defect value and the average value are compared to obtain the standard deviation, the larger the standard deviation value is to indicate that more data are reflected in the whole system, so that the larger the weight representing the defect is, the opposite is, and the weight of the quality data of the injection molding can be obtained. The square difference of the data, and the weight of the data can clearly feed back whether the quality of the injection molding piece reaches the standard.
The beneficial effects of the invention are as follows: the method is different from the prior art in that the quality of the injection molding can be detected without waiting for the injection molding to finish production, and the quality of the injection molding can be predicted and analyzed during the production period of the injection molding to obtain the optimization method, so that the production quality of the injection molding can be optimized in time, and the optimization efficiency of the injection molding is improved.
Drawings
FIG. 1 is a flow chart of a method of optimizing the quality of an injection molded part according to the present invention.
Detailed Description
The technical scheme of the invention is further specifically described below through examples and with reference to the accompanying drawings.
Examples: in the injection molding quality optimization method of the embodiment, as shown in fig. 1, the steps include S1, performing prediction of injection molding quality data by using a GASA method;
s2, establishing a proxy model to analyze injection molding quality data;
and S3, centralizing the quality analysis result of the injection molding piece according to a stacking integrated method and providing an injection molding optimization method.
In step s1, an initial target Z is generated according to the required setting parameters of quality prediction gen The training error of the vector machine is taken as a fitness function, and the training error is brought into an initial target Z gen Calculating fitness of individual i, selecting excellent individual for crossover operation in combination with fitness and generating new target Z' gen Then, a plurality of individuals are selected from the initial population for mutation, and a mutated target Z' is generated gen Comparison of Z after annealing operation gen And Z' gen And obtaining an adaptability difference value delta f, judging whether the delta f is qualified or not according to a Metropolis criterion, and repeating the steps after the judgment is finished to obtain an optimal solution. When Δf is less than 0, individuals receiving the mutation are represented, and when Δf is greater than or equal to 0, according to the probability p=exp (- Δf/T) gen ) Receiving the individual, wherein Δf=fit (Z' gen )-fit(Z gen )，T gen Is the temperature attenuation coefficient, T gen ＝kT gen-1 The target after the annealing operation is Z'. gen . When the vector machine trains errors, the input sample is set as x, the output value is set as y, w is a weight, b is a bias term,for a non-linear mapping of low-dimensional to high-dimensional space, x i And x j Represents the input value, K (x i ，x j ) As a kernel function, a i And->Langerhans multiplier, the expression of the predictive algorithm:
the decision function of the support vector machine can be obtained on the basis that:
and step S2, the proxy model comprises a proxy model of an artificial neural network and a proxy model of Gaussian process regression, wherein the proxy model of Gaussian process regression utilizes average absolute error and error duty ratio, and simultaneously combines root mean square error and fitting goodness-of-fit evaluation model to establish three index proxy models to evaluate injection molding quality data. And the agent model of the artificial neural network builds a model according to the shrinkage mark and the average volume shrinkage rate of the injection molding part, trains model parameters according to the two standards of the injection molding part, and finally obtains a more accurate result to evaluate the quality of the injection molding part.
In the step S3, the analytical results of all injection molding quality data are collected by using a stacking integration method, and all data are input into a plurality of data sets S= { (X) i ，Y i ) I=1, 2, …, n }, performing corresponding training on each data set s, inputting the training result of the data set into a meta learner, evaluating the quality data of the injection molding according to the result of the meta learner, and giving a corresponding injection molding optimization method. Meanwhile, a quality index standard is required to be established to analyze the data result, the quality of the injection molding part is judged according to the standard, and Z is set ij Is the value processed by a non-dimensionality method, X ij The j-th injection defect value, X of the i-th injection experiment max To the maximum value of defects in the injection molding quality leopard data, X min The minimum value of defects in the injection molding quality leopard data is shown as a dimensionless formula:
the average value of the quality data of the injection molding piece is then:
the standard deviation of the quality data of the injection molding is as follows:
therefore, the weight of the quality data of the injection molding piece can be deduced as follows:
z in the above j Sigma, the mean value of quality data of injection molding j Is the standard deviation of the quality data of the injection molding. There are several factors to consider in the production of injection molded parts, such as: the shape of injection products, the number of mold cavities, a pouring system, a heat exchange system and the like. However, it is difficult to produce injection molded parts of high quality according to the above-mentioned factors. In the injection molding process, defects such as buckling deformation, bubbles and the like of the injection molding product often occur, and the existence of the defects can influence the attractiveness of the injection molding product and the service performance of the injection molding product. The defect of the bubble is difficult to measure in practice, and is only shown in a picture form in simulation software, so that the bubble size data is not directly given. For debugging of process parameters, operators need to continuously try on a plastic injection molding machine, and the method has high cost and low efficiency. Thus, how to make a wagerThe method has a certain research significance in quantifying defects of plastic products and establishing prediction of injection molding quality.
In the injection molding process, different injection molding process parameters can generate different influences on the quality of injection molding, the effect that the relation between each injection molding parameter and the quality of the injection molding is helpful to realize efficient production is studied, in the embodiment, the influence of parameters such as the temperature in an injection molding machine, the dwell time, the melt temperature, the injection time, the conversion volume, the dwell pressure and the like on the quality of the injection molding in the injection molding production process can be assisted by the method of orthogonal test, and the situation that the injection molding comprehensive quality is firstly reduced and then increased along with the increase of the temperature of a mold and the melt temperature is found. In the single factor experiment of the conversion volume, along with the increase of the conversion volume, the injection time, the dwell time and the dwell pressure, the injection quality is increased firstly and then is reduced.
By analyzing the working flow of the plastic injection molding machine, the influence on the injection molding quality can be effectively reduced by adjusting the structure of the injection mold. According to the method and the device for predicting the quality of the injection molding according to the related factors of the injection molding production in the injection molding production process, the prediction process and the injection molding production are almost carried out synchronously, so that the prediction process is required to be high in efficiency and high in accuracy. In the embodiment, the GASA method is used for predicting the quality data of the injection molding finished product produced by the injection molding machine according to the production state of the injection molding. The obtained injection molding quality data are immediately substituted into a proxy model for data analysis, wherein the proxy model comprises a Gaussian process regression proxy model and an artificial neural network proxy model. The Gaussian process regression and the artificial neural network have the following characteristics and differences in proxy model establishment: first, in terms of algorithm accuracy: compared with the Gaussian process regression, the artificial neural network can obtain a regression model based on a training set more accurately, but the situation of fitting can also occur at the same time; when a certain error exists in the data, the prediction performance of the agent model based on the artificial neural network is drastically reduced. Secondly, the algorithm is stable in calculation result due to the Gaussian process regression principle. The training result of the artificial neural network is often influenced by the initial weight, and especially under the conditions of complex model and large system error, the problem of algorithm stability is more remarkable. Furthermore, in terms of computational efficiency: under the selection of a specific kernel function, GP is equivalent to a neural network with a hidden layer when the number of hidden nodes approaches infinity. Therefore, the regression calculation amount based on the Gaussian process is extremely large, the calculation speed is low, and the method is only suitable for small sample data regression. Finally, in the meaning of model analysis: because the artificial neural network is based on data, the calculation process is just like a black box, and the solution and the resolution are not resolved. The Gaussian process regression is different, and the priori and posterior are Gaussian processes, so that the probability significance is achieved. Therefore, in view of the respective advantages of the gaussian process regression proxy model and the artificial neural network proxy model, in the embodiment, the gaussian process regression proxy model and the artificial neural network proxy model are used in a combined mode, and both proxy models work in the process of evaluating and analyzing the data quality of the injection molding. The analysis emphasis points of the two agent models are not completely consistent, the two agent models have respective advantages and disadvantages, and a more accurate analysis result can be obtained by combining the analysis of the quality data of the injection molding obtained by prediction by the two agent models, so that the proposal of a follow-up injection molding optimization scheme is facilitated. After the agent model evaluates and analyzes the quality data of the injection molding, a large number of analysis results are generated, and if the analysis results are obtained each time, the analysis results are judged by artificial identification one by one, so that manpower and material resources are wasted, and meanwhile, the production efficiency is greatly influenced. Therefore, after the agent model analyzes the quality data of the injection molding, a stacking integration method is used to collect and disperse the results of the quality analysis of all injection molding into n subsets
The folded data are used for training n-1 base learners of the first layer, and parameter adjustment of the base learners is realized in the process, so that n trained base models can be obtained. The remaining 1-fold data are predicted using the n trained base models to obtain predicted results P1, P2, …, pn in the form of n/K rows by n columns. Repeating the steps for n-1 times to obtain all the prediction results. And taking the obtained result as the input of the meta learner, training the meta learner, outputting the result as n multiplied by 1 column data, and realizing the parameter optimization of the meta learner in the process. Thus, the complete Stacking integrated learning model after training is obtained. After the stacking integrated method is used for processing the quality analysis result of the injection molding, the injection molding optimization method of different injection molding problems can be known according to the processing result, and after the injection molding optimization method is determined, a worker can adjust corresponding influencing factors according to the injection molding optimization method to optimize the quality of the injection molding.
Claims (8)
1. The injection molding quality optimization method is characterized by comprising the following steps of
S1, predicting quality data of injection molding parts by using a GASA method;
s2, establishing a proxy model to analyze injection molding quality data;
and S3, centralizing the quality analysis result of the injection molding piece according to a stacking integrated method and providing an injection molding optimization method.
2. The method according to claim 1, wherein in step S1, the initial target Z is generated according to the required setting parameters of the quality prediction gen The training error of the vector machine is taken as a fitness function, and the training error is brought into an initial target Z gen Calculating fitness of individual i, selecting excellent individual for crossover operation in combination with fitness and generating new target Z' gen Then, a plurality of individuals are selected from the initial population for mutation, and a mutated target Z' is generated gen Comparison of Z after annealing operation gen And Z' gen And obtaining an adaptability difference value delta f, judging whether the delta f is qualified or not according to a Metropolis criterion, and repeating the steps after the judgment is finished to obtain an optimal solution.
3. The method according to claim 2, wherein in the step S1, when Δf is smaller than 0, the individual receiving the variation is represented, and when Δf is equal to or greater than 0, the probability p=exp (- Δf/T) is calculated gen ) Receiving the individual, wherein Δf=fit (Z' gen )-fit(Z gen ),T gen Is the temperature attenuation coefficient, T gen ＝kT gen-1 Annealing, annealingThe target after operation is Z' "g en 。
4. The injection molding quality optimization method according to claim 1, wherein in the step S1, when the error is trained by using a vector machine, the input sample is set to x, the output value is set to y, w is a weight, b is an offset,for a non-linear mapping of low-dimensional to high-dimensional space, x i And x j Represents the input value, K (x i ,x j ) As a kernel function, a i And->Langerhans multiplier, the expression of the predictive algorithm:
the decision function of the support vector machine can be obtained on the basis that:
5. the injection molding quality optimization method according to claim 1, wherein the proxy model in the step S2 includes a proxy model of an artificial neural network and a proxy model of gaussian process regression, the proxy model of gaussian process regression uses average absolute error and error duty ratio, and simultaneously combines root mean square error and goodness-of-fit evaluation model to establish three index proxy models for evaluation analysis of injection molding quality data.
6. The method for optimizing the quality of the injection molding according to claim 1, wherein the agent model of the artificial neural network in the step S2 is used for establishing a model according to the shrinkage and the average volume shrinkage of the injection molding, training model parameters according to the two standards of the injection molding, and finally obtaining a relatively accurate result to evaluate and analyze the quality of the injection molding.
7. The injection molding quality optimization method according to claim 1, wherein in the step S3, all the injection molding quality data analysis results are collected by using a stacking integration method, and all the data are input into a plurality of data sets s= { (X) i ，Y i ) I=1, 2, ····, n, each data set S is trained accordingly, the results of the dataset training are input to a meta learner, and judging the quality data of the injection molding according to the result of the meta-learner and giving a corresponding injection molding optimization method.
8. The method according to claim 1, wherein in step S3, a quality index standard is established, the quality of the injection molded part is determined according to the standard, and Z is set ij Is the value processed by a non-dimensionality method, X ij The j-th injection defect value, X of the i-th injection experiment max X is the maximum value of defects in injection molding quality data min The minimum value of defects in the injection molding quality data is represented by the non-dimensionalized formula:
the average value of the quality data of the injection molding piece is then:
the standard deviation of the quality data of the injection molding is as follows:
therefore, the weight of the quality data of the injection molding piece can be deduced as follows:
z in the above j Sigma, the mean value of quality data of injection molding j Is the standard deviation of the quality data of the injection molding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310325988.8A CN116894379A (en) | 2023-03-30 | 2023-03-30 | Injection molding quality optimization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310325988.8A CN116894379A (en) | 2023-03-30 | 2023-03-30 | Injection molding quality optimization method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116894379A true CN116894379A (en) | 2023-10-17 |
Family
ID=88313948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310325988.8A Pending CN116894379A (en) | 2023-03-30 | 2023-03-30 | Injection molding quality optimization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116894379A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117493819A (en) * | 2024-01-03 | 2024-02-02 | 江门塚田正川科技有限公司 | Automatic injection molding production equipment and intelligent regulation and control method thereof |
-
2023
- 2023-03-30 CN CN202310325988.8A patent/CN116894379A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117493819A (en) * | 2024-01-03 | 2024-02-02 | 江门塚田正川科技有限公司 | Automatic injection molding production equipment and intelligent regulation and control method thereof |
| CN117493819B (en) * | 2024-01-03 | 2024-03-29 | 江门塚田正川科技有限公司 | Automatic injection molding production equipment and intelligent regulation and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107451101B (en) | Method for predicting concentration of butane at bottom of debutanizer by hierarchical integrated Gaussian process regression soft measurement modeling | |
| CN104699894B (en) | Gaussian process based on real-time learning returns multi-model Fusion Modeling Method | |
| CN113119425B (en) | Injection product quality prediction method based on improved support vector machine | |
| Fei et al. | Practical applications of Taguchi method for optimization of processing parameters for plastic injection moulding: a retrospective review | |
| CN110640982A (en) | Multi-target optimization method for injection molding process parameters of thin-wall injection molding part | |
| CN108520155B (en) | Vehicle behavior simulation method based on neural network | |
| Abdul et al. | Shrinkage prediction of injection molded high density polyethylene parts with taguchi/artificial neural network hybrid experimental design | |
| CN104809306A (en) | Injection molding technology parameter optimizing method based on multiple quality indexes | |
| Zhang et al. | An intelligent system for low-pressure die-cast process parameters optimization | |
| Zhao et al. | Optimization of injection-molding process parameters for weight control: converting optimization problem to classification problem | |
| CN116894379A (en) | Injection molding quality optimization method | |
| CN111768000A (en) | Industrial process data modeling method for online adaptive fine-tuning deep learning | |
| CN114881289A (en) | Fusion casting charge forming process parameter optimization method and system based on digital twinning | |
| CN111310348A (en) | Material constitutive model prediction method based on PSO-LSSVM | |
| CN111079856B (en) | Multi-period intermittent process soft measurement modeling method based on CSJITL-RVM | |
| CN113158473B (en) | Semi-supervised integrated instant learning industrial rubber compound Mooney viscosity soft measurement method | |
| CN109783898A (en) | A kind of intelligent optimization method of injection molding manufacturing technique parameter | |
| Meyes et al. | Interdisciplinary data driven production process analysis for the internet of production | |
| CN110263488A (en) | A kind of industrial Mooney Viscosity of Rubber Mix flexible measurement method based on integrated instant learning | |
| CN112733435A (en) | Whole vehicle size matching deviation prediction method based on multi-model fusion | |
| CN114692677B (en) | Welding defect identification method based on multi-target feature selection | |
| CN113103535A (en) | GA-ELM-GA-based injection molding part mold parameter optimization method | |
| CN114274457B (en) | Injection mold pouring system structural parameter decision method based on Kriging-GA | |
| CN110378035A (en) | It is a kind of that soft-measuring modeling method is hydrocracked based on deep learning | |
| CN110502779A (en) | A kind of Intelligentized design method of the injection forming mold based on fuzzy reasoning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |