Methods, equipment and media for optimizing process parameters in injection molding

Abstract

The present application provides a method, device and medium for optimizing process parameters in an injection molding process, and belongs to the technical field of plastic product production. The method for optimizing process parameters in the injection molding process includes: collecting production data of the injection molding process, and the production data includes production samples and multiple process parameters; using the PCA method to extract the principal components corresponding to the production samples, projecting the production data onto the principal components for dimensionality reduction, and obtaining a principal component matrix; using the principal component matrix to calculate the comprehensive contribution score of each process parameter in the multiple process parameters on the principal component, screening the process parameters according to the comprehensive contribution score, and obtaining a contribution process parameter matrix; inputting the contribution process parameter matrix into a quality prediction model constructed based on XGBoost, and predicting the corresponding prediction quality index of the output; and using the prediction quality index to optimize the process parameters. The present application can solve the problem that the dimensions of the process parameters in the prior art are too high, resulting in information redundancy and variable collinearity, which affects the XGBoost modeling effect.

Description

Technical Field

[0001] The present application belongs to the technical field of plastic product production, and specifically relates to a method, equipment and medium for optimizing process parameters in an injection molding process. Background Art

[0002] Injection molding is one of the most common processes for producing plastic products, widely used in industries such as automotive, electronics, home appliances, and medical. This process involves setting multiple parameters, such as melt temperature, injection speed, hold time, and cooling time. These parameters are complexly coupled and significantly impact the quality of plastic products.

[0003] Traditional process optimization often relies on manual experience or trial-and-error methods to adjust process parameters. This is not only inefficient but also difficult to achieve stable control of quality indicators. With the development of intelligent manufacturing, data-driven approaches are gradually being applied to injection molding processes. This method collects historical data and establishes a mapping relationship between parameters and product quality to achieve process optimization and intelligent prediction.

[0004] As a data-driven method, the extreme gradient boosting method XGBoost has been widely used in industrial modeling scenarios due to its excellent nonlinear modeling capabilities and feature selection capabilities. This algorithm has also been applied to injection molding scenarios. For example, CN119538702A provides a multi-objective optimization method for injection molding process parameters. First, a simulation result data set is obtained based on simulation analysis and spatial design experiments. Then, the NGO optimization algorithm is used to optimize the XGBoost model, and an NGO-XGBoost injection molding quality optimization model is constructed. The NGO-XGBoost model is then interpreted and analyzed to obtain the importance and positive and negative effects of each optimized design variable on the predicted quality optimization target. Multi-objective optimization is performed to obtain the quality optimization target frontier; finally, the optimal combination of process parameters is determined in the frontier of the instruction optimization target.

[0005] However, in injection molding scenarios, process parameters are extremely dimensional, leading to issues such as information redundancy and variable collinearity. Directly using raw parameters for modeling with XGBoost increases model complexity and the risk of overfitting, impacting modeling effectiveness. Even using the NGO optimization algorithm to optimize the XGBoost model can only optimize, but not reduce, the complex dimensionality of the process parameters. This limits the performance and application of injection molding process modeling. Summary of the Invention

[0006] This application aims to provide a process parameter optimization solution for injection molding. It aims to solve the problem that the process parameter dimensions in injection molding scenarios are extremely high, resulting in information redundancy and variable collinearity, which affect the modeling effect of XGBoost.

[0007] According to a first aspect of the present application, the present application provides a method for optimizing process parameters in an injection molding process, comprising:

[0008] Collecting production data of the injection molding process, wherein the production data includes production samples and corresponding multiple process parameters;

[0009] Use principal component analysis (PCA) to extract the principal components corresponding to the production samples, project the production data onto the principal components for dimensionality reduction, and obtain the principal component matrix;

[0010] Using the principal component matrix, calculate the comprehensive contribution score of each process parameter on the principal component among multiple process parameters, and filter multiple process parameters according to the comprehensive contribution score to obtain the contribution process parameter matrix;

[0011] The contribution process parameter matrix is ​​input into the quality prediction model built based on XGBoost, and the corresponding predicted quality indicators are predicted and output;

[0012] Optimize process parameters using predictive quality metrics.

[0013] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the step of collecting production data of the injection molding process includes:

[0014] Collect production data from the injection molding process, where the production data includes production samples and corresponding process parameters;

[0015] Constructing a production data matrix using production samples and corresponding multiple process parameters, wherein the production data matrix includes all collected production data;

[0016] All production data in the production data matrix are standardized to obtain a standardized data matrix corresponding to the production data matrix.

[0017] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the steps of using principal component analysis (PCA) to extract principal components corresponding to production samples, projecting production data onto principal components for dimensionality reduction, and obtaining a principal component matrix include:

[0018] Construct the covariance matrix corresponding to the standardized data matrix;

[0019] Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues ​​corresponding to the process parameters in the principal component direction, as well as the eigenvectors corresponding to the eigenvalues;

[0020] Sort the eigenvalues ​​from large to small, according to the relationship between the cumulative equation contribution rate and the variance contribution rate threshold:

[0021]

[0022] Select the first k eigenvalues ​​that satisfy the size relationship, and use the eigenvectors corresponding to the first k eigenvalues ​​as the principal components corresponding to the production sample; where, represents the eigenvalue, Indicates the equation contribution rate threshold;

[0023] The first k eigenvalues ​​are used to form a feature matrix, and the standardized data matrix is ​​combined with the feature matrix to construct a principal component matrix to project the production data onto the principal components for dimensionality reduction.

[0024] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the steps of using a principal component matrix to calculate the comprehensive contribution score of each of the multiple process parameters on the principal component, screening the multiple process parameters according to the comprehensive contribution score, and obtaining the contribution process parameter matrix include:

[0025] Using the principal component matrix, calculate the loading of each of the multiple process parameters on the principal component;

[0026] Use the load combined with the cumulative equation contribution rate to calculate the comprehensive contribution score of each process parameter among multiple process parameters;

[0027] Sort by comprehensive contribution scores from high to low, and select a predetermined number of process parameters as contributing process parameters;

[0028] Using the contributing process parameters, a contributing process parameter matrix is ​​constructed.

[0029] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the step of inputting the contribution process parameter matrix into the quality prediction model constructed based on XGBoost and predicting and outputting the corresponding predicted quality index includes:

[0030] Use XGBoost to build a quality prediction model; the quality prediction model includes a base learner and an objective function, and the base learner is a regression tree;

[0031] The penalty coefficient of the regression tree corresponding to the tree structure complexity, the regularization coefficient and the prediction weight of the leaf node are used to construct the regularization term of the quality prediction model;

[0032] Use the regularization term combined with the loss function to construct the objective function;

[0033] Construct a training set using historical data and corresponding preset quality labels;

[0034] Use principal component analysis to process historical data and obtain the contribution process parameter matrix corresponding to the historical data;

[0035] Use the contribution process parameter matrix and preset quality labels corresponding to historical data to train and validate the quality prediction model;

[0036] During the training and validation process, the parameters of the quality prediction model are tuned until the objective function converges;

[0037] When the objective function converges, the contribution process parameter matrix is ​​input into the quality prediction model, and the predicted quality index is output.

[0038] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the step of optimizing process parameters using the predicted quality index includes:

[0039] After using the quality prediction model to predict the predicted quality indicators corresponding to the output contributing process parameters, the predicted quality indicators are used to estimate the quality status corresponding to the contributing process parameters, and a fluctuation warning corresponding to the predicted quality indicators is provided;

[0040] as well as,

[0041] Use different contributing process parameters to form process parameter combinations, and use the quality prediction model to output the predicted quality indicators corresponding to different process parameter combinations;

[0042] Reverse optimize the injection molding process according to the predicted quality indicators corresponding to different process parameter combinations.

[0043] Preferably, in the above-mentioned method for optimizing process parameters in the injection molding process, the steps of reversely optimizing the injection molding process according to the predicted quality indicators corresponding to different process parameter combinations include:

[0044] Set target quality indicators;

[0045] Use optimization algorithms combined with predicted quality indicators to infer the approximate optimal process parameter combination from different process parameter combinations;

[0046] Use target quality indicators and predicted quality indicators, and meet the following conditions according to the indicators:

[0047]

[0048] Select the optimal parameter combination from the approximately optimal parameter combinations; is the predicted quality index corresponding to the process parameter combination, Target quality indicator.

[0049] Preferably, the above-mentioned method for optimizing process parameters in the injection molding process further comprises, after the step of optimizing the process parameters using the predicted quality index:

[0050] Combined with the process setting constraints of the injection molding process, the recommended process parameter solutions are screened according to the predicted quality indicators output by the quality prediction model;

[0051] According to the recommended process parameter plan, the process parameters of the injection molding site are adjusted using the relevant control system, and the production status information of the injection molding site is read in real time, where the production status information includes production data.

[0052] According to the second aspect of the present application, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, the method for optimizing process parameters in the injection molding process provided by any of the above technical solutions is implemented.

[0053] According to the third aspect of the present application, the present application also provides a computer storage medium on which a computer program is stored. When the computer program is executed, the method for optimizing process parameters in the injection molding process provided by any of the above technical solutions is implemented.

[0054] The technical solution of this application has at least the following technical effects:

[0055] The process parameter optimization scheme for the injection molding process provided by this application is based on the collection of production data generated during the injection molding process. This original production data has the problem of high parameter dimensionality. To address the problems of high parameter dimensionality, variable redundancy and strong collinearity in traditional injection molding processes, the technical scheme of this application first reduces the dimensionality of the process parameters through principal component analysis (PCA), extracts the principal components and calculates the contribution of each process parameter to the principal components. Then, representative process parameters are selected as contributing process parameters, and a contributing process parameter matrix is ​​constructed as the input feature set. The quality indicators are then modeled and predicted using the XGBoost regression model to improve modeling accuracy and generalization ability. The constructed quality prediction model can be used to generate the optimal parameter combination scheme, provide control suggestions based on existing process constraints, and achieve closed-loop optimization and quality control of the injection molding process. This scheme has the advantages of effective dimensionality reduction, efficient modeling, strong interpretability and controllable parameter adjustment. It can solve the problems of high process parameter dimensionality, information redundancy and variable collinearity caused by the existing technology. The technical scheme of this application can even be applied to various industrial manufacturing scenarios such as injection molding, die casting and 3D printing, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0057] Figure 1A schematic flow chart of a method for optimizing process parameters in an injection molding process provided in an embodiment of the present application;

[0058] Figure 2 for Figure 1 A schematic flow chart of a method for collecting production data provided by the illustrated embodiment;

[0059] Figure 3 for Figure 1 A schematic flow chart of a method for reducing the dimension of production data provided by the illustrated embodiment;

[0060] Figure 4 for Figure 1 A schematic flow chart of a method for calculating a comprehensive contribution score of a process parameter provided in the illustrated embodiment;

[0061] Figure 5 for Figure 1 A schematic flow chart of a method for predicting a quality indicator of a process parameter provided by the illustrated embodiment;

[0062] Figure 6 for Figure 1 A schematic flow chart of a process parameter optimization method provided by the illustrated embodiment;

[0063] Figure 7 for Figure 6 A schematic flow chart of a method for reverse optimization of a process provided by the illustrated embodiment;

[0064] Figure 8 A schematic flow chart of a method for screening a recommended process parameter solution provided in an embodiment of the present application;

[0065] Figure 9 A schematic flow chart of a method for optimizing process parameters in a second injection molding process provided in an embodiment of the present application;

[0066] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] In order to more clearly illustrate the overall concept of the present application, a detailed description is given below in an illustrative manner in conjunction with the accompanying drawings.

[0068] The following description sets forth many specific details to facilitate a thorough understanding of the present application. However, the present application may also be implemented in other ways than those described herein, and therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below. It should be noted that the embodiments of the present application and the features of each embodiment may be combined with each other unless there is a conflict.

[0069] In this application, unless otherwise clearly specified and limited, a first feature "above" or "below" a second feature may be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples.

[0070] The existing technology has the following defects:

[0071] As a data-driven method, extreme gradient boosting (XGBoost) has gained widespread application in industrial modeling scenarios due to its outstanding nonlinear modeling and feature selection capabilities. The algorithm has also found some application in injection molding scenarios. However, the extremely high dimensionality of process parameters in injection molding scenarios can easily lead to problems such as information redundancy and variable collinearity. Directly using raw parameters for modeling with XGBoost increases model complexity and the risk of overfitting, thus impacting modeling effectiveness. Even when optimizing the XGBoost model using the NGO optimization algorithm, only process parameter combinations can be selected, making it difficult to reduce the complex dimensionality of the process parameters. This limits the performance and application of injection molding process modeling.

[0072] In order to overcome the above technical difficulties, the following embodiment of the present application provides a solution for optimizing process parameters in the injection molding process. By performing PCA dimensionality reduction processing on the original process parameters of the injection molding process, the main components are extracted to simplify the feature space structure, and the feature contribution of each main component is analyzed. On this basis, process parameters with high influence (i.e., high contribution) among the main components are selected as input features for training of the XGBoost model, thereby achieving accurate prediction of injection molding quality and optimized control of process parameters. Through the above method, the problems of high process parameter dimensionality, variable redundancy, strong feature collinearity and low modeling efficiency existing in the prior art can be solved.

[0073] To achieve the above purpose, see Figure 1 , Figure 1 A schematic diagram of a process parameter optimization method in an injection molding process provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the process parameter optimization method in the injection molding process includes:

[0074] S110: Collecting production data of the injection molding process, wherein the production data includes production samples and corresponding multiple process parameters.

[0075] In an embodiment of the present application, multiple batches of production data are specifically obtained from an injection molding device or a data acquisition system, and the production data includes a large number of production samples. Each sample data includes m process parameters, such as injection speed, holding pressure, mold opening time, cooling time, etc. There are n production samples in total, so that the original production data constitutes an m×n production data matrix. Subsequently, the production data needs to be standardized to obtain a standardized data matrix corresponding to the production data. S120: Use principal component analysis PCA to extract the principal components corresponding to the production samples, project the production data onto the principal components for dimensionality reduction, and obtain a principal component matrix. Using principal component analysis PCA can reduce redundancy and collinearity between input features, thereby achieving dimensionality reduction of production data.

[0076] Principal component analysis (PCA) is an efficient data dimensionality reduction technique that can project high-dimensional data into a few principal component spaces, thereby reducing data dimensions while retaining the main information and improving modeling efficiency. In the embodiment of the present application, China uses PCA to extract the principal components of production samples, which can project production data onto the principal components, thereby reducing the dimensionality of process parameters. In addition, PCA is combined with the XGBoost model to screen key variables based on the principal component contribution and use them for model training, thereby giving full play to the model's feature extraction and learning capabilities, and improving the performance and practicality of intelligent modeling of injection molding processes.

[0077] Although PCA can be used for modeling, its linear combination form is not conducive to the interpretation of the physical meaning of the parameters. To balance dimensionality reduction and interpretability, the embodiment of the present application infers important process parameters from the original process parameters based on the principal component contribution.

[0078] S130: Using the principal component matrix, calculate the comprehensive contribution score of each process parameter in the plurality of process parameters on the principal component, filter the plurality of process parameters according to the comprehensive contribution score, and obtain a contribution process parameter matrix.

[0079] The embodiment of the present application uses the principal component matrix obtained above to calculate the comprehensive contribution score of each process parameter on the principal component, and screens out the process parameters that have the main contribution to production quality according to the comprehensive contribution score, obtaining a process parameter matrix. Using this process parameter matrix in combination with XGBoost for quality prediction, it is possible to obtain the contribution degree of different process parameter combinations to production quality. This method is conducive to the interpretation of the physical meaning of the parameters, can take into account both dimensionality reduction and interpretability, and derive process parameters with high contribution.

[0080] S140: Input the contribution process parameter matrix into the quality prediction model built based on XGBoost, and predict the corresponding prediction quality index.

[0081] The embodiment of the present application uses XGBoost to construct a quality prediction model. By inputting the contribution process parameter matrix into the quality prediction model, the production quality corresponding to different contribution process parameter combinations can be predicted. The above-mentioned contribution process parameters are the dimensionality reduction of the original process parameters. Therefore, inputting them into the quality prediction model constructed by XGBoost can reduce the model complexity, reduce the problem of model overfitting, and improve the modeling ability.

[0082] S150: Optimize process parameters using predicted quality indicators.

[0083] The present embodiment uses a quality prediction model to predict key quality indicators (KQIs) such as product weight, dimensional accuracy, or surface defect rate based on input process parameter data, and uses these predicted quality indicators to optimize process parameters. Optimization methods include forward prediction and process optimization. Specifically, forward prediction involves inputting process parameters and then model-predicting quality indicators, providing quality status estimates and fluctuation warnings. Process optimization involves predicting the corresponding output given a given input parameter combination, assisting in comparative analysis of the quality impacts of different parameter combinations.

[0084] In summary, the process parameter optimization method in the injection molding process provided by this application is based on the collection of production data generated during the injection molding process. The original production data has the problem of high parameter dimensionality. In response to the problems of high parameter dimensionality, variable redundancy and strong collinearity in traditional injection molding processes, the technical solution of this application first reduces the dimensionality of the process parameters through principal component analysis (PCA), extracts the principal components and calculates the contribution of each process parameter to the principal components, and then selects representative process parameters as contributing process parameters, constructs a contributing process parameter matrix as the input feature set. Subsequently, the XGBoost regression model is used to model and predict quality indicators, improving modeling accuracy and generalization ability. The constructed quality prediction model can be used to generate the optimal parameter combination solution, provide control suggestions based on existing process constraints, and achieve closed-loop optimization and quality control of the injection molding process. This solution has the advantages of effective dimensionality reduction, efficient modeling, strong interpretability and controllable parameter adjustment, and can solve the problems of high process parameter dimensionality, information redundancy and variable collinearity caused by the existing technology. The technical solution of this application can even be applied to various industrial manufacturing scenarios such as injection molding, die casting and 3D printing, and has good application prospects.

[0085] Among them, as a preferred embodiment, Figure 2 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S110: collecting production data of the injection molding process includes:

[0086] S111: Collecting production data in the injection molding process, wherein the production data includes a production sample and corresponding multiple process parameters.

[0087] Process parameters include melt temperature, injection speed, mold temperature, holding pressure, holding time, cooling time, etc. In the original production data, there is a complex coupling relationship between these process parameters, which has a significant impact on the quality of the product.

[0088] S112: Constructing a production data matrix using the production samples and the corresponding multiple process parameters, wherein the production data matrix includes all collected production data.

[0089] S113: Standardize all production data in the production data matrix to obtain a standardized data matrix corresponding to the production data matrix.

[0090] The technical solution provided in the embodiment of the present application obtains multiple batches of production data from the injection molding equipment or data acquisition system. The production data includes a large number of production samples and process parameters corresponding to the production samples. Among them, for the input parameters, each sample data contains process parameters, usually m≥10, such as injection speed, holding pressure, mold opening time, cooling time, etc. The production data of each batch includes a large number of production samples. samples, the original input data constitutes a Production data matrix .

[0091] In addition, this application requires labeling the production data with output indicators such as product weight, dimensional accuracy, and surface defect rate. This quality labeling, combined with the aforementioned production data, can serve as the training and validation sets for training and validating the quality prediction model built with XGBoost.

[0092] The embodiment of the present application performs standardization processing on the input production data, and converts each feature (i.e., process parameter) into a standardization formula with zero mean and unit variance as follows:

[0093]

[0094] i represents the sample and m represents the process parameters.

[0095] in,

[0096] : The j-th eigenvalue of the i-th sample, that is, the j-th process parameter;

[0097] : The sample mean of the j-th feature;

[0098] : The sample standard deviation of the jth feature;

[0099] The above production data matrix obtained after standardization corresponds to the following standardized data matrix:

[0100]

[0101] Furthermore, the process parameters in the raw production data have complex dimensions. Directly inputting them into the model for training will significantly impact model performance and applicability. To reduce redundancy and collinearity between input features (i.e., process parameters), after obtaining the standardized data matrix corresponding to the aforementioned production data, this application also requires parameter dimensionality reduction and principal component extraction based on PCA.

[0102] Specifically, as a preferred embodiment, Figure 3 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S120: using principal component analysis (PCA) to extract principal components corresponding to production samples, projecting production data onto principal components for dimensionality reduction, and obtaining a principal component matrix includes the following steps:

[0103] S121: Construct the covariance matrix corresponding to the standardized data matrix.

[0104] The covariance matrix corresponding to the standardized data matrix is ​​of the following form:

[0105]

[0106] in,

[0107] : standardized data matrix;

[0108] : The transposed matrix of

[0109] : sample size;

[0110] : Covariance matrix between features.

[0111] S122: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​corresponding to the process parameters in the principal component direction, and eigenvectors corresponding to the eigenvalues.

[0112] After obtaining the covariance matrix, this application needs to perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue and the corresponding eigenvector .

[0113]

[0114] in,

[0115] : The eigenvalue of the kth principal component.

[0116] : The eigenvector of the kth principal component.

[0117] It should be noted that the above eigenvalues ​​represent process parameters, and the eigenvectors represent vectorized forms of the process parameters.

[0118] S123: Sort the eigenvalues ​​from large to small according to the relationship between the cumulative equation contribution rate and the variance contribution rate threshold:

[0119]

[0120] Select the first k eigenvalues ​​that satisfy the size relationship, and use the eigenvectors corresponding to the first k eigenvalues ​​as the principal components corresponding to the production sample; where, represents the eigenvalue, Indicates the equation contribution rate threshold;

[0121] S124: Use the first k eigenvalues ​​to form a feature matrix, and use the standardized data matrix combined with the feature matrix to construct a principal component matrix to project the production data onto the principal component for dimensionality reduction.

[0122] The calculation formula for the proportion of variance explained by the eigenvalue of the kth principal component in PCA (that is, the proportion of the ath principal component in the total data variance) is as follows:

[0123]

[0124] in,

[0125] : No. The eigenvalue corresponding to the principal component represents the variance of the principal component;

[0126] : The sum of the eigenvalues ​​of all principal components, representing the total variance; p is j Serial number.

[0127] :No. The contribution ratio of each principal component to the overall variance.

[0128] Then among all the principal components, the cumulative contribution rate meets the preset threshold (e.g. 90%):

[0129]

[0130] Assuming that the number of eigenvalues ​​corresponding to the principal components that meet the above preset threshold is at least k, then the eigenvalues ​​corresponding to the first k principal components are selected. , generally ranging from 30% to 100%. In addition, the eigenvectors corresponding to the k eigenvalues ​​are the principal components of the production samples.

[0131] The principal component matrix is ​​constructed using the first k eigenvalues ​​as follows:

[0132]

[0133] Among them, Z represents the above-mentioned standardized data matrix, and the above-mentioned eigenvectors are not considered here. , because adding eigenvectors may lead to secondary sorting and confusion.,The principal component matrix can be used to screen the load combinations (i.e.,,combinations of process parameters) in the subsequent calculation of the comprehensive,contribution score.

[0134] Although the principal component matrix obtained by PCA analysis can be used for modeling, its linear combination form is not conducive to the interpretation of the physical meaning of the parameters. In order to balance dimensionality reduction and interpretability, the embodiment of this application proposes a method for inferring important original parameters based on the principal component contribution:

[0135] Specifically, as a preferred embodiment, Figure 4 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S130: using the principal component matrix, calculating the comprehensive contribution score of each process parameter in the principal component of multiple process parameters, screening multiple process parameters according to the comprehensive contribution score, and obtaining a contribution process parameter matrix, specifically includes:

[0136] S131: Using the principal component matrix, calculate the load of each process parameter on the principal component among the multiple process parameters.

[0137] This application requires calculating the contribution of process parameters to the principal components, i.e., the load, so as to select the concentrated process parameter combination that has the greatest contribution to production quality based on the load. The specific calculation method is as follows:

[0138]

[0139] in, represents the weight of the jth original process parameter in the direction of the ith principal component in the principal component matrix M. The principal component here is selected from the principal component matrix M. In addition, the embodiment of the present application needs to calculate the loads of all m original process parameters on the principal components.

[0140] S132: Use the load combined with the cumulative equation contribution rate to calculate the comprehensive contribution score of each process parameter among multiple process parameters.

[0141] According to the cumulative equation contribution rate, the calculation formula for the comprehensive contribution score of the original process parameters is:

[0142]

[0143] in, It represents the eigenvalue corresponding to the kth principal component. According to the above calculation formula, the comprehensive contribution score of the jth process parameter is the product of the cumulative contribution rate and the load.

[0144] S133: Sort by comprehensive contribution score from high to low, and select a predetermined number of process parameters as contributing process parameters.

[0145] S134: Use the contribution process parameters to construct a contribution process parameter matrix.

[0146] The embodiment of this application needs to combine the above comprehensive contribution scores to perform feature selection. Sort by order from largest to smallest, select the top d process parameters with the highest scores as contributing process parameters, and construct the contributing process parameter matrix as follows:

[0147]

[0148] n represents the number of samples, and d represents the number of the above-mentioned contributing process parameters.

[0149] In summary, the technical solution provided by the embodiments of this application uses the comprehensive contribution scores of process parameters on the principal components to select the top d process parameters with the highest contributions as contributing process parameters, thereby constructing a contributing process parameter matrix. This matrix reflects the process combinations of these contributing process parameters, thereby facilitating the subsequent quality prediction model to predict the production quality resulting from these different contributing process parameter combinations. This allows the linear combination method of PCAD to be interpreted in a physical sense, achieving both dimensionality reduction and interpretability of the process parameters.

[0150] In addition, after obtaining the contribution process parameter matrix, the embodiment of the present application can use the contribution process parameters and the above-mentioned quality labels to construct a regression XGBoost regression model.

[0151] Specifically, as a preferred embodiment, Figure 5 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S140: inputting the contribution process parameter matrix into the quality prediction model constructed based on XGBoost, and predicting and outputting the corresponding predicted quality index, includes:

[0152] S141: Use XGBoost to build a quality prediction model; wherein the quality prediction model includes a base learner and an objective function, and the base learner is a regression tree.

[0153] XGBoost is a high-performance machine learning algorithm based on the gradient boosting framework. Its core architecture combines decision tree ensembles, regularized optimization, and parallel computing. As an additive model, XGBoost minimizes the objective function by iteratively adding decision trees (base learners). Its architecture can be divided into the following parts:

[0154] Base learner (weak learner): CART regression tree (default).

[0155] Objective function: contains loss function and regularization term.

[0156] Optimization method: Gradient Boosting combined with second-order Taylor expansion.

[0157] System optimization: Weighted Quantile Sketch, Sparsity-aware, etc.

[0158] S142: Use the penalty coefficient of the regression tree corresponding to the tree structure complexity, the regularization term coefficient and the prediction weight of the leaf node to construct the regularization term of the quality prediction model.

[0159] S143: Use the regularization term combined with the loss function to construct the objective function.

[0160] The objective function of the quality prediction model is as follows:

[0161]

[0162] in:

[0163] is the loss function, which measures the deviation between the true value and the predicted value; Represents the true value of the quality corresponding to the i-th production sample; is the predicted value of the i-th sample; is a regularization term used to control the complexity of the model. The expression is:

[0164] ,

[0165] The penalty coefficient that controls the complexity of the tree structure,

[0166] In this expression, represents the regularization coefficient of the leaf node weight,

[0167] It is j The prediction weight of leaf nodes,

[0168] T is the number of regression trees in the model.

[0169] S144: Construct a training set using historical data and corresponding preset quality labels. The historical data refers to historical production data of the injection molding process, including a large number of production samples and process parameters.

[0170] S145: Use principal component analysis to process the historical data to obtain a contribution process parameter matrix corresponding to the historical data. The processing steps for obtaining the contribution process parameter matrix are as described above and will not be repeated here.

[0171] S146: Use the contribution process parameter matrix corresponding to the historical data and the preset quality labels to train and verify the quality prediction model.

[0172] For quality prediction model training and parameter optimization, a cross-validation approach is used during the training phase to evaluate the performance of different process parameter combinations X* and select the optimal parameters. The training data set is divided into multiple subsets, which are trained and validated alternately to ensure the model's good generalization capabilities.

[0173] The training objective of the quality prediction model is to minimize the validation error of the model, which is the loss function mentioned above. The validation error indicators include mean square error (MSE) and mean absolute error (MAE):

[0174]

[0175] S147: Parameters of the quality prediction model are tuned during the training and validation process until the objective function converges.

[0176] After minimizing the model's validation error, the model needs to be adjusted, using methods such as grid search or Bayesian optimization to traverse or intelligently search for parameter combinations.

[0177] Among them, the quality prediction model built based on XGBoost has the following main adjustment parameters:

[0178] learning_rate: learning rate, which controls the model update step size;

[0179] max_depth: the maximum depth of each tree;

[0180] subsample: sampling ratio, which controls the data subset used in each iteration;

[0181] reg_alpha, reg_lambda: L1 and L2 regularization coefficients, controlling model complexity;

[0182] n_estimators: The number of trees, which determines the number of weak learners.

[0183] S148: When the objective function converges, the contribution process parameter matrix is ​​input into the quality prediction model, and the predicted quality index is output.

[0184] By training and optimizing the parameters of the quality prediction model, the model accuracy and generalization ability can be improved. Specifically, the XGBoost model after feature screening eliminates irrelevant interference features while maintaining key variables, effectively avoiding overfitting and significantly improving the model's generalization ability and prediction accuracy for new data. In addition, parameter tuning can guide actual production parameter adjustment. The output results of the quality prediction model can be used to identify key parameters that affect quality, and through reverse reasoning, assist production personnel in parameter adjustment and process optimization, which helps to reduce defective product rates, improve product consistency, and achieve closed-loop control in intelligent manufacturing.

[0185] After the model training and parameter tuning of the quality prediction model are completed, it is necessary to deploy the model and predict key quality indicators. Specifically, as a preferred embodiment, Figure 6 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S150: optimizing process parameters using predicted quality indicators includes:

[0186] S151: After using the quality prediction model to predict and output the predicted quality indicators corresponding to the contributing process parameters, a predetermined quality monitoring model is used to analyze the predicted quality indicators, estimate the quality status corresponding to the contributing process parameters, and provide fluctuation warnings corresponding to the predicted quality indicators. A trained predetermined quality monitoring model can be selected and input into the predicted quality indicators. The model estimates the contribution of each contributing process parameter to product quality, the quality status differences caused by different values ​​of each contributing process parameter, and generates a fluctuation chart of the predicted quality indicators. This provides early warnings for production data where the predicted quality indicators do not meet the standards.

[0187] as well as,

[0188] S152: Using different contributing process parameters to form process parameter combinations, and using a quality prediction model to output predicted quality indicators corresponding to the different process parameter combinations.

[0189] S153: Reversely optimize the injection molding process according to the predicted quality indicators corresponding to different process parameter combinations.

[0190] The technical solution provided in the embodiment of the present application can first achieve model output. The quality prediction model after training can predict key quality indicators such as product weight, dimensional accuracy or surface defect rate for the input data (i.e. the above-mentioned contributing process parameters X*).

[0191] The present application embodiment supports the following two types of applications:

[0192] Forward prediction: Input parameters → Model predicts quality indicators, providing quality status estimation and fluctuation warning;

[0193] Process optimization: Given a certain input parameter combination, predict its corresponding output, assist in comparative analysis of the quality impact of different parameter combinations, support operation optimization, and reverse optimize the process parameters of the injection molding process.

[0194] Specifically, as a preferred embodiment, Figure 7 As shown, in the above-mentioned method for optimizing process parameters in the injection molding process, step S153: according to the predicted quality indicators corresponding to different process parameter combinations, the steps of reversely optimizing the injection molding process include:

[0195] S1531: Set target quality indicators.

[0196] S1532: Use optimization algorithms combined with predicted quality indicators to infer the approximately optimal process parameter combination from different process parameter combinations;

[0197] S1533: Use the target quality indicator and the predicted quality indicator, and according to the indicators meet the conditions:

[0198]

[0199] Select the optimal parameter combination from the approximately optimal parameter combinations; is the predicted quality index corresponding to the process parameter combination, Target quality indicator.

[0200] The technical solution provided in the embodiments of this application is to realize the reverse deduction of feasible process solutions from target quality, and can be combined with the trained quality prediction model to build a reverse optimization module. When the target quality index is known, an optimization algorithm (such as particle swarm optimization, genetic algorithm or L-BFGS that minimizes the target error) is used in combination with the trained model to reversely deduce the optimal or approximate process parameter combination. , to meet the above conditions.

[0201] In addition, after obtaining the predicted quality indicators, model deployment and parameter control suggestions can be generated. Specifically, as a preferred embodiment, Figure 8 As shown, the above-mentioned method for optimizing process parameters in the injection molding process further includes, after step S150: optimizing process parameters using predicted quality indicators:

[0202] S160: In combination with the process setting constraints of the injection molding process, a recommended process parameter solution is screened out according to the predicted quality indicators output by the quality prediction model.

[0203] S170: According to the recommended process parameter plan, use the relevant control system to adjust the process parameters of the injection molding site, and read the production status information of the injection molding site in real time, wherein the production status information includes production data.

[0204] The technical solution provided in the embodiment of the present application, after reverse reasoning to obtain feasible input parameters, combines process setting constraints (such as injection speed range or upper and lower limits of holding pressure, etc.) to screen recommended parameter solutions that meet process feasibility (such as process setting constraints determined by expert experience), saves the trained model and deploys it to a computer connected to the injection molding machine control system, reads production status information in real time, and finally outputs the following content: 1. Recommended parameter combination (optimal value or multiple groups of candidate values); 2. Parameter impact ranking or sensitivity analysis results; 3. Quality prediction value and its confidence interval; 4. Parameters obtained by reverse reasoning of the model, which can be directly used for production process setting or closed-loop control feedback.

[0205] in addition, Figure 9 Schematic diagram of the process parameters optimization method in the second injection molding process provided in the embodiment of the present application. Figure 9 As shown, the process parameter optimization method includes:

[0206] S201: Injection molding data collection.

[0207] S202: Data preprocessing.

[0208] S203: PCA analysis.

[0209] S204: Original feature screening based on principal component contribution analysis.

[0210] S205: PCA-XGBoost generates contributing process parameters.

[0211] S206: Model construction.

[0212] S207: Model training and optimization.

[0213] S208: Pre-training model. There are two reasoning methods for generating models: forward reasoning and reverse reasoning.

[0214] S209: Forward reasoning.

[0215] S210: Prediction quality indicator.

[0216] S211: Reverse reasoning.

[0217] S212: Optimizers such as Bayesian optimization or grid search are used to optimize the parameters.

[0218] S213: Process parameters / parameter combinations.

[0219] In summary, the technical solutions provided in the above embodiments of the present application, by combining principal component analysis (PCA) with the extreme gradient boosting algorithm (XGBoost), propose a modeling method suitable for injection molding process parameter optimization and quality prediction, which has the following significant beneficial effects:

[0220] (1) Dimensionality reduction compresses redundant features to improve modeling efficiency

[0221] By performing PCA dimensionality reduction on high-dimensional injection molding parameters, the redundancy and multicollinearity problems between input features are effectively reduced, the model input structure is simplified, and the computational efficiency and model stability are improved.

[0222] (2) Parameter selection is more physically interpretable

[0223] This application does not directly use principal components as modeling inputs, but rather infers the importance of the original variables based on the principal component contributions, and selects representative original parameters so that the model retains the process significance of the parameters while maintaining its predictive ability, facilitating process control and personnel understanding.

[0224] (3) Improve model accuracy and generalization ability

[0225] After feature screening, the XGBoost model eliminates irrelevant interference features while maintaining key variables, effectively avoiding overfitting and significantly improving the model's generalization ability and prediction accuracy for new data.

[0226] (4) Can guide actual production parameter adjustment

[0227] The model output results can be used to identify key parameters that affect quality, and through reverse reasoning, assist production personnel in parameter adjustment and process optimization, which helps to reduce defective product rates, improve product consistency, and achieve closed-loop control in intelligent manufacturing.

[0228] (5) The method is highly versatile and applicable to a variety of industrial scenarios

[0229] The modeling process of this application has good versatility and is not only applicable to injection molding processes, but can also be extended to other complex manufacturing processes with multi-parameter control problems, such as die casting, 3D printing, forging and other industrial scenarios.

[0230] In summary, the present invention has significant technical value and application prospects in improving modeling efficiency in the injection molding field, enhancing model interpretability, optimizing injection molding process control and improving product quality.

[0231] In addition, the beneficial effects of the product embodiments provided in the following embodiments of the present application are the same as the beneficial effects of the process parameter optimization method in the injection molding process provided in the above embodiments, and the other technical features in the product embodiments are the same as the features disclosed in the above embodiment methods, which will not be repeated here.

[0232] Reference below Figure 10 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application can include but is not limited to a mobile terminal and / or a fixed terminal. Figure 10 The device shown is only an example and should not limit the functions and scope of use of the embodiments of the present application.

[0233] like Figure 10 As shown, the electronic device may include a processing device 1001, such as a central processing unit and / or a graphics processing unit, which can perform various appropriate actions and processes according to programs stored in ROM 1002 or loaded from storage device 1003 into RAM 1004. RAM 1004 also stores various programs and data required for the operation of the electronic device. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to bus 1005. Typically, the following systems can be connected to I / O interface 1006: input devices 1007, such as a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, and / or gyroscope; output devices 1008, such as an LCD (Liquid Crystal Display), speaker, and / or vibrator; storage device 1003, such as a magnetic tape and / or hard disk; and communication device 1009. Communication device 1009 enables the electronic device to communicate with other devices wirelessly or wired to exchange data. Although the figures show a model building device with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems can be implemented or have instead.

[0234] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0235] The present application provides a computer-readable storage medium having computer-readable program instructions stored thereon, namely the above-mentioned computer program, which is used to execute the process parameter optimization method in the injection molding process in the above-mentioned embodiment.

[0236] The computer-readable storage medium carries one or more programs that, when executed by the model building device, can be written in one or more programming languages ​​or a combination thereof to perform the operations of the present application. The programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, or C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a LAN (Local Area Network) or a WAN (Wide Area Network), or can be connected to an external computer, for example, via the Internet using an Internet service provider.

[0237] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0238] The modules described in the embodiments of the present application can be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0239] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0240] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included in the protection scope of the present application.

Claims

1. A method for optimizing process parameters in an injection molding process, characterized in that: include: Collecting production data of an injection molding process, wherein the production data includes a production sample and corresponding multiple process parameters; Using principal component analysis (PCA) to extract principal components corresponding to the production samples, projecting the production data onto the principal components to perform dimensionality reduction, and obtaining a principal component matrix; Using the principal component matrix, calculating a comprehensive contribution score of each of the multiple process parameters on the principal component, and screening the multiple process parameters according to the comprehensive contribution score to obtain a contribution process parameter matrix; Input the contribution process parameter matrix into the quality prediction model built based on XGBoost, and predict the corresponding predicted quality index; optimizing the process parameters using the predicted quality indicators; The step of inputting the contribution process parameter matrix into the quality prediction model constructed based on XGBoost and predicting and outputting the corresponding predicted quality index includes: Use XGBoost to build a quality prediction model; wherein the quality prediction model includes a base learner and an objective function, and the base learner is a regression tree; Constructing a regularization term of the quality prediction model using the penalty coefficient of the regression tree corresponding to the tree structure complexity, the regularization term coefficient, and the prediction weight of the leaf node; Using the regularization term in combination with the loss function, the objective function is constructed; Construct a training set using historical data and corresponding preset quality labels; Processing the historical data using the principal component analysis method PCA to obtain a contribution process parameter matrix corresponding to the historical data; Training and validating the quality prediction model using the contribution process parameter matrix corresponding to the historical data and the preset quality labels; Optimizing parameters of the quality prediction model during training and validation until the objective function converges; When the objective function converges, the contribution process parameter matrix corresponding to the production data is input into the quality prediction model, and the predicted quality index is obtained as output.

2. The method according to claim 1, wherein The step of collecting production data of the injection molding process includes: Collecting production data in the injection molding process, wherein the production data includes a production sample and corresponding multiple process parameters; constructing a production data matrix using the production samples and corresponding multiple process parameters, wherein the production data matrix includes all collected production data; All production data in the production data matrix are standardized to obtain a standardized data matrix corresponding to the production data matrix.

3. The method according to claim 2, wherein The step of extracting the principal components corresponding to the production samples using principal component analysis (PCA), projecting the production data onto the principal components for dimensionality reduction, and obtaining a principal component matrix includes: Constructing a covariance matrix corresponding to the standardized data matrix; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​corresponding to the process parameters in the principal component direction and eigenvectors corresponding to the eigenvalues; Sort the eigenvalues ​​from large to small, according to the relationship between the cumulative equation contribution rate and the variance contribution rate threshold: The first k eigenvalues ​​that satisfy the size relationship are selected, and the eigenvectors corresponding to the first k eigenvalues ​​are used as the principal components corresponding to the production sample; wherein, represents the eigenvalue, Indicates the equation contribution rate threshold; The first k eigenvalues ​​are used to form a feature matrix, and the standardized data matrix is ​​combined with the feature matrix to construct a principal component matrix, so as to project the production data onto the principal component for dimensionality reduction.

4. The method according to claim 1 or 3, wherein: The step of using the principal component matrix to calculate the comprehensive contribution score of each of the multiple process parameters on the principal component, and screening the multiple process parameters according to the comprehensive contribution score to obtain a contribution process parameter matrix includes: Using the principal component matrix, calculating the load of each of the plurality of process parameters on the principal component; Calculate a comprehensive contribution score of each of the multiple process parameters using the load combined with the cumulative equation contribution rate; Sort by the comprehensive contribution scores from high to low, and select a predetermined number of process parameters as contributing process parameters; The contributing process parameters are used to construct the contributing process parameter matrix.

5. The method according to claim 1, wherein The step of optimizing the process parameters using the predicted quality indicators comprises: After using the quality prediction model to predict and output a predicted quality indicator corresponding to the contributing process parameter, use the predicted quality indicator to estimate the quality status corresponding to the contributing process parameter, and provide a fluctuation warning corresponding to the predicted quality indicator; as well as, Using different contributing process parameters to form process parameter combinations, and using the quality prediction model to output predicted quality indicators corresponding to the different process parameter combinations; The injection molding process is reverse optimized according to the predicted quality indicators corresponding to the different process parameter combinations.

6. The method according to claim 5, wherein The step of reversely optimizing the injection molding process according to the predicted quality indicators corresponding to the different process parameter combinations includes: Set target quality indicators; Using an optimization algorithm in combination with the predicted quality index, inferring an approximately optimal process parameter combination from the different process parameter combinations; Using the target quality indicator and the predicted quality indicator, according to the indicators satisfying the conditions: Filter out the optimal parameter combination from the approximately optimal parameter combination; wherein, is the predicted quality index corresponding to the process parameter combination, is the target quality indicator.

7. The method according to claim 1, wherein After the step of optimizing the process parameters using the predicted quality indicators, the method further comprises: In combination with the process setting constraints of the injection molding process, a recommended process parameter solution is screened out according to the predicted quality indicators output by the quality prediction model; According to the recommended process parameter scheme, the process parameters of the injection molding site are adjusted using a relevant control system, and the production status information of the injection molding site is read in real time, wherein the production status information includes the production data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for optimizing process parameters in an injection molding process as claimed in any one of claims 1 to 7 is implemented.

9. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for optimizing process parameters in an injection molding process according to any one of claims 1 to 7 is implemented.

Images