Injection molding process multivariable control method and system

By using partial least squares regression model to screen target process parameters and multi-objective optimization algorithm, combined with experimental deviation verification, the problem of insufficient model accuracy in multi-index optimization in injection molding process was solved, and the reliability and precise control of injection molding process parameters were achieved.

CN123103990APending Publication Date: 2026-11-17HAITIAN PLASTICS MACHINERY GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611073723.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing injection molding processes, the multiple linear regression method is prone to overfitting when dealing with multiple conflicting quality indicators, resulting in insufficient model prediction accuracy. Furthermore, it fails to effectively coordinate and optimize indicators such as product quality, warpage, and production cycle, and the reliability of the model prediction results is insufficient.

Method used

Partial least squares regression model is used to screen out the target process parameter categories that meet the conditions for the comprehensive influence of multiple quality indicators. The optimization is carried out by combining multi-objective optimization algorithm. The final parameter combination is verified by measured deviation. Partial least squares regression model is constructed and iteratively updated to improve model accuracy.

Benefits of technology

It achieves synergistic optimization of multiple quality indicators under small sample conditions, improves the reliability and prediction accuracy of injection molding process parameters, and ensures the consistency between the measured and predicted results of the injection molding process parameter combination.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123103990A_ABST
    Figure CN123103990A_ABST
Patent Text Reader

Abstract

This invention provides a multivariate control method and system for injection molding processes, belonging to the field of injection molding process optimization technology. The method includes: acquiring a partial least squares regression model constructed based on an initial sample set, and target process parameter categories screened based on variable projection importance values; using the function constructed based on the partial least squares regression model as the fitness function of a multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter categories, obtaining candidate parameter combinations and corresponding predicted quality indices; acquiring the actual quality indices generated by performing injection molding operations according to the candidate parameter combinations, and calculating the deviation between the actual quality indices and the predicted quality indices; when the deviation meets a preset threshold condition, determining the candidate parameter combination as the target injection molding parameter. This method eliminates process parameter categories with insignificant contributions, achieves collaborative optimization of conflicting quality indices, and ensures the reliability of the target injection molding parameters.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of injection molding process optimization technology, and in particular to a method and system for multivariate control of injection molding processes. Background Technology

[0002] Injection molding is a core manufacturing process in the field of polymer material processing, widely used in industries such as automotive, electronics, and medical. With the continuous increase in product precision requirements, how to simultaneously ensure product quality, control warpage, and shorten production cycles during injection molding has become a crucial need in the field of injection molding process optimization.

[0003] In existing technologies, ordinary multiple linear regression is typically used to model and analyze injection molding processes. This method uses process parameters such as injection speed, transfer position, and holding pressure as independent variables, and a single quality indicator such as product quality, warpage, or production cycle as the dependent variable. It establishes a linear mapping relationship between process parameters and quality indicators through least squares fitting, and then adjusts the process parameters based on the output of the regression model.

[0004] However, multicollinearity is common among injection molding process parameters, and the sample size of injection molding experiments is usually limited. Under these conditions, ordinary multiple linear regression is prone to overfitting, resulting in insufficient model prediction accuracy. Furthermore, this method only constructs a regression model for a single quality indicator, failing to simultaneously handle multiple conflicting quality indicators such as product quality, warpage, and production cycle. This leads to insufficient synergistic assurance of the final process parameter combination for each quality indicator, and the resulting process parameter combination has not been verified against the deviation between measured and predicted results, thus lacking reliability. Summary of the Invention

[0005] This invention provides a multivariate control method and system for injection molding processes to address the shortcomings of existing technologies, achieve synergistic optimization of conflicting quality indicators, and ensure the reliability of target injection molding parameters.

[0006] This invention provides a multivariate control method for injection molding processes, comprising the following steps: The target process parameter category and prediction model are obtained. The prediction model is a partial least squares regression model constructed based on an initial sample set. The initial sample set contains multiple sets of process parameters and quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from the multiple process parameter categories based on variable projection importance values. The variable projection importance values ​​are used to characterize the comprehensive influence of a single process parameter category on all the quality indicators. The function constructed based on the prediction model is used as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, thereby obtaining candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations. Obtain the actual quality indicators generated by performing injection molding operations according to the candidate parameter combinations; Calculate the deviation between the actual quality index and the predicted quality index; When the deviation is determined to meet the preset threshold condition, the candidate parameter combination is determined as the target injection molding parameter.

[0007] According to a multivariate control method for injection molding process provided by the present invention, after calculating the deviation between the actual quality index and the predicted quality index, the method further includes: When it is determined that the deviation does not meet the preset threshold condition, the candidate parameter combination and its corresponding actual quality index are used as new samples and merged into the initial sample set to form an updated sample set. Based on the updated sample set, latent variables are re-extracted to update the partial least squares regression model, and the process returns to the step of using the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category until the deviation meets the preset threshold condition.

[0008] According to the multivariate control method for injection molding process provided by the present invention, the step of obtaining the target process parameter category and prediction model includes: The process parameters and quality indicators in the initial sample set are decentralized to obtain a centered process parameter matrix and a quality indicator matrix, and the process parameter matrix and the quality indicator matrix are used as the input residual matrix and the output residual matrix, respectively. Based on the input residual matrix and the output residual matrix, latent variables are extracted, and the partial least squares regression model containing the regression coefficient matrix is ​​constructed, wherein the regression coefficient matrix is ​​used to characterize the independent influence of a single process parameter category on a single quality index. Using the parameters of the latent variables extracted by the partial least squares regression model, the projected importance value of the variables corresponding to each category of process parameters is calculated; The process parameter categories whose variable projection importance values ​​are less than a preset importance threshold are removed, and the remaining process parameter categories are used as the target process parameter categories.

[0009] According to the multivariate control method for injection molding process provided by the present invention, the step of extracting latent variables based on the input residual matrix and the output residual matrix, and constructing the partial least squares regression model containing the regression coefficient matrix, includes: The following processes are performed iteratively until a predetermined number of latent variables are extracted: Calculate the weight vector between the input residual matrix and the output residual matrix; multiply the input residual matrix and the weight vector to obtain the factor score; calculate the factor loadings corresponding to the input residual matrix and the response loadings corresponding to the output residual matrix based on the factor scores; update the input residual matrix and the output residual matrix by matrix subtraction based on the factor scores, the factor loadings, and the response loadings, as input data for the next iteration round. The weight vectors, factor loads and response loads extracted in each iteration are combined column-wise to construct the weight matrix, factor load matrix and response load matrix in sequence. The regression coefficient matrix is ​​calculated based on the product of the weight matrix, the factor loading matrix, and the response loading matrix.

[0010] According to a multivariate control method for injection molding process provided by the present invention, the step of calculating the regression coefficient matrix based on the product of the weight matrix, the factor loading matrix, and the response loading matrix includes: Calculate the first product matrix of the transpose of the factor loading matrix and the weight matrix; Invert the first product matrix, multiply the weight matrix on the right by the inverse of the first product matrix, and then multiply it on the right by the transpose of the response load matrix to obtain the regression coefficient matrix.

[0011] According to a multivariate control method for injection molding process provided by the present invention, the step of calculating the projected importance value of the variables corresponding to each category of process parameters includes: The importance value of the variable projection is calculated using the following formula: ; ; in, For the first j The projected importance value of the variable corresponding to each of the aforementioned process parameter categories. The total number of the process parameter categories. H The total number of the latent variables. h Let be the iteration round number of the latent variable. For the first h The weight vector extracted in the nth iteration corresponds to the weight vector in the nth iteration. j Weight components for each process parameter category, For the first h The square of the magnitude of the weight vector extracted in each iteration round; For the first hThe sum of squares explaining the quality indicators in the initial sample set by the latent variables. Indicates the first h The transpose matrix of factor scores for each iteration round; Indicates the first h The output residual matrix used before each iteration round update; Indicates the first h The response load for each iteration round.

[0012] According to the multivariate control method for injection molding process provided by the present invention, the decentralization process of multiple sets of process parameters and quality indicators in the initial sample set includes: Calculate the first mean vector corresponding to each column of process parameters in the initial sample set, and the second mean vector corresponding to each column of quality indicators. Subtract the first mean matrix of the corresponding dimension, which is composed of the first mean vector, from the process parameter matrix in the initial sample set to obtain the input residual matrix; The output residual matrix is ​​obtained by subtracting the second mean matrix of the corresponding dimension, which is composed of the second mean vector, from the quality index matrix in the initial sample set.

[0013] According to a multivariate control method for injection molding process provided by the present invention, the step of using a function constructed based on the prediction model as the fitness function of a multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter categories, and obtaining candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations, includes: In the search space of the multi-objective optimization algorithm, a population containing multiple particles is initialized, where each particle represents a set of parameter values ​​for the target process parameter category within the feasible region. Multiply the input row vector formed by each set of parameter values ​​by the regression coefficient matrix in the prediction model to obtain the predicted values ​​of the corresponding particles in multiple quality index dimensions. According to the preset optimization direction corresponding to each of the quality index dimensions, the predicted values ​​are converted into the multi-target fitness values ​​of each particle. The velocity and position of each particle are updated based on the multi-objective fitness value, and the Pareto optimal solution set is obtained through iterative optimization. A set of solutions that meet the preset priority order is selected from the Pareto optimal solution set as the candidate parameter combination, and the prediction value corresponding to the set of solutions is used as the prediction quality index.

[0014] According to a multivariate control method for injection molding process provided by the present invention, the step of converting each predicted value into a multi-objective fitness value of each particle according to a preset optimization direction corresponding to each of the quality index dimensions includes: If the preset optimization direction of a certain quality indicator is to maximize it, then the negative number corresponding to the predicted value is taken as the minimum fitness value of that quality indicator dimension. If the preset optimization direction of a certain quality indicator is minimization, then the corresponding predicted value will be used as the minimum fitness value of that quality indicator dimension.

[0015] The present invention also provides a multivariable control system for injection molding process, comprising the following modules: The first acquisition module is used to acquire the target process parameter category and the prediction model. The prediction model is a partial least squares regression model constructed based on an initial sample set. The initial sample set contains multiple sets of process parameters and quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from the multiple process parameter categories based on the variable projection importance value. The variable projection importance value is used to characterize the comprehensive influence of a single process parameter category on all the quality indicators. The parameter optimization module is used to use the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, and obtain candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations. The second acquisition module is used to acquire the actual quality indicators generated by performing the injection molding operation according to the candidate parameter combination; The deviation calculation module is used to calculate the deviation between the actual quality indicator and the predicted quality indicator. The target parameter determination module is used to determine the candidate parameter combination as the target injection molding parameter when the deviation is determined to meet the preset threshold condition.

[0016] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multivariate control method for injection molding process as described above.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multivariate control method for injection molding process as described above.

[0018] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multivariate control method for injection molding process as described above.

[0019] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By constructing a partial least squares regression model based on an initial sample set and calculating the projected importance values ​​of variables, target process parameter categories that satisfy the condition of comprehensive influence on all quality indicators are selected from multiple process parameter categories. This eliminates process parameter categories that do not contribute significantly to quality indicators, reducing the decision space dimension of subsequent multi-objective optimization and improving optimization efficiency. By using the prediction function constructed based on the partial least squares regression model as the fitness function of the multi-objective optimization algorithm, the parameter values ​​of the target process parameter categories are globally optimized within the process feasible region. This achieves synergistic optimization among multiple conflicting quality indicators such as product quality, warpage, and production cycle, obtaining candidate parameter combinations that take into account multiple optimization objectives and corresponding predicted quality indicators. By obtaining the actual quality indicators generated by performing injection molding operations according to the candidate parameter combinations and calculating the deviation between the actual and predicted quality indicators, the prediction results of the partial least squares regression model are quantitatively compared with the actual production results of the injection molding machine, establishing an objective evaluation benchmark for the model's prediction accuracy. By determining the candidate parameter combination as the target injection molding parameter when the judgment deviation meets the preset threshold condition, the injection molding process parameter combination that is finally put into production has been effectively verified by the actual test results. This avoids the problem of insufficient parameter reliability caused by relying solely on model prediction results to directly guide production. Overall, it achieves precise control of multivariable parameters in the injection molding process and stable assurance of quality indicators. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is one of the flowcharts of the multivariable control method for injection molding process provided by the present invention.

[0022] Figure 2 This is the second flowchart of the multivariable control method for injection molding provided by the present invention.

[0023] Figure 3 This is the third flowchart of the multivariable control method for injection molding provided by the present invention.

[0024] Figure 4 This is the fourth flowchart of the multivariable control method for injection molding provided by the present invention.

[0025] Figure 5 This is the fifth flowchart of the multivariable control method for injection molding process provided by the present invention.

[0026] Figure 6 This is the sixth flowchart of the multivariable control method for injection molding provided by the present invention.

[0027] Figure 7 This is a schematic diagram of the multivariable control system for injection molding process provided by the present invention.

[0028] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0030] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships according to the accompanying drawings, are only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0031] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0032] The following is combined with Figures 1 to 8This invention describes the multivariable control method, system, electronic device, storage medium, and computer program product for injection molding processes provided by this invention.

[0033] The multivariable control method for injection molding processes provided in this application can be executed by a controller, processor, injection molding process optimization system, injection molding process multivariable control system, or other electronic devices with data processing capabilities. For ease of explanation, the following embodiments will uniformly use the injection molding process multivariable control system (hereinafter referred to as the system) as the execution subject for description.

[0034] Reference Figure 1 , Figure 1 This is one of the flowcharts illustrating the multivariate control method for injection molding processes provided by this invention. The multivariate control method for injection molding processes provided in this application includes steps 110 to 150: Step 110: Obtain the target process parameter category and the prediction model. The prediction model is a partial least squares regression model built based on the initial sample set. The initial sample set contains multiple sets of process parameters and the quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from multiple process parameter categories based on the variable projection importance value. The variable projection importance value is used to characterize the comprehensive influence of a single process parameter category on all quality indicators.

[0035] Specifically, the system first obtains an initial sample set through experimental design and data acquisition. During the experiment, the operator controls the injection speed. Transfer location Holding pressure These process parameters are categorized with different parameter values, and n sets of experiments are conducted, with product quality data collected for each set of experiments. Warpage Production cycle Quality indicators and values.

[0036] The system constructs the collected process parameters into an n×3-dimensional parameter matrix X and the collected quality indicators into an n×3-dimensional response matrix Y. Each row of the parameter matrix X corresponds to the specific values ​​of injection speed, transfer position, and holding pressure in a set of experiments, and each row of the response matrix Y corresponds to the product quality, warpage, and production cycle values ​​produced by the same set of experiments.

[0037] Next, the system constructs a partial least squares regression model using the initial sample set. The partial least squares regression model extracts latent variables by linearly combining the original process parameters; for example, it constructs latent variables by multiplying the injection rate by 0.5, the transfer position by 0.3, and the holding pressure by 0.2. This eliminates collinearity among process parameters and improves prediction stability.

[0038] After the prediction model is built, the system further filters multiple process parameter categories such as injection speed, transfer position, and holding pressure based on the variable projection importance value, and retains the process parameter categories whose comprehensive influence on all quality indicators meets the conditions as the target process parameter categories.

[0039] Step 120: Use the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, and obtain the candidate parameter combination and the prediction quality index corresponding to the candidate parameter combination.

[0040] Specifically, the system constructs the partial least squares regression model as the prediction function. , where x is the input row vector consisting of parameter values ​​of the target process parameter category, and B is the regression coefficient matrix. The output row vector consists of the predicted product quality, predicted warpage, and predicted production cycle.

[0041] The system uses the prediction function as the fitness function of the multi-objective particle swarm optimization algorithm to optimize the parameter values ​​of the target process parameter categories within the feasible region defined by the process range. Through iterative search using the multi-objective particle swarm optimization algorithm, the system obtains a set of Pareto optimal solutions and selects a set of solutions from the Pareto optimal solutions as candidate parameter combinations based on the priority order of actual production needs. Simultaneously, the system records the prediction quality index corresponding to the candidate parameter combinations.

[0042] Step 130: Obtain the actual quality indicators generated by performing the injection molding operation according to the candidate parameter combination.

[0043] Specifically, the system sends candidate parameter combinations to the injection molding machine, which then performs the injection molding operation according to the parameter values ​​such as injection speed, transfer position, and holding pressure in the candidate parameter combinations to produce the actual product. The system collects actual product quality indicators such as product quality, warpage, and production cycle through sensors and detection equipment.

[0044] Step 140: Calculate the deviation between the actual quality indicators and the predicted quality indicators.

[0045] Specifically, the system calculates the difference between the actual quality indicators (product quality, warpage, and production cycle) and the corresponding dimensions of the predicted quality indicators to obtain the deviation in each dimension of the quality indicators. Alternatively, it comprehensively characterizes the deviation between the actual and predicted quality indicators by calculating the mean square error, relative error, and other methods.

[0046] Step 150: When the judgment deviation meets the preset threshold condition, the candidate parameter combination is determined as the target injection molding parameter.

[0047] Specifically, the system compares the deviation with a preset threshold condition. When the deviation falls within the preset threshold range, it is determined that the partial least squares regression model meets the requirements for predicting the current process conditions. The candidate parameter combination is then determined as the final target injection molding parameters for production, and the injection molding machine performs mass production accordingly.

[0048] It should be noted that the preset threshold conditions can be set according to the actual production situation, and no further restrictions are imposed in this application.

[0049] Through steps 110 to 150, this implementation combines the partial least squares regression model with a multi-objective optimization algorithm to achieve synergistic optimization of multiple quality indicators in the injection molding process scenario with small sample size and multivariate coupling. The reliability of the candidate parameter combination is ensured through actual measurement deviation verification, thereby significantly improving the prediction accuracy and practical usability of the injection molding process optimization results.

[0050] In one possible implementation, refer to Figure 2 , Figure 2 This is the second flowchart of the multivariate control method for injection molding process provided by the present invention. After calculating the deviation between the actual quality index and the predicted quality index in step 140, the multivariate control method for injection molding process further includes the following steps: Step 160: When the judgment deviation does not meet the preset threshold condition, the candidate parameter combination and the actual quality index corresponding to the candidate parameter combination are added as new samples and merged into the initial sample set to form an updated sample set; based on the updated sample set, latent variables are re-extracted to update the partial least squares regression model, and the step of using the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category is returned until the deviation meets the preset threshold condition.

[0051] Specifically, when the system determines that the deviation between the actual quality index and the predicted quality index exceeds a preset threshold range, it considers that the partial least squares regression model's ability to fit the current process condition is insufficient, requiring supplementary data and retraining. The system constructs new samples by using candidate parameter combinations as a new set of process parameter samples and actual quality indices as corresponding quality index samples, and then appends and merges these new samples into the initial sample set to form an updated sample set. Based on the updated sample set, the system re-executes processes such as decentralization, latent variable extraction, and regression coefficient matrix solving to obtain an updated partial least squares regression model. It then returns to execute the multi-objective optimization process defined in step 120 to obtain new candidate parameter combinations and compare their deviation with the actual quality indices again until the deviation falls within the preset threshold range.

[0052] Through step 160, this embodiment establishes a closed-loop feedback and model iterative update mechanism based on measured deviation, enabling the partial least squares regression model to continuously self-correct based on new samples during the production process, effectively responding to dynamic changes such as mold wear and material batch differences, thereby significantly improving the robustness and adaptability of the injection molding process multivariate control method in long-term use.

[0053] In one possible implementation, refer to Figure 3 , Figure 3 This is the third flowchart of the multivariate control method for injection molding provided by the present invention. Step 110, which obtains the target process parameter category and prediction model, further includes steps 111 to 114: Step 111: Decentralize the multiple sets of process parameters and quality indicators in the initial sample set to obtain the centered process parameter matrix and quality indicator matrix, and use the process parameter matrix and quality indicator matrix as the input residual matrix and output residual matrix, respectively.

[0054] Specifically, the system decentralizes the parameter matrix X and the response matrix Y according to the following formula: ; in, Let X be the column mean vector of the parameter matrix. The column mean vector of the response matrix Y, where 1 represents an n×1 vector of all 1s. After decentralization, the system will use the centralized process parameter matrix. As the initial input residual matrix The centralized quality index matrix As the initial output residual matrix ,Right now , After decentralization, the covariance can directly reflect the linear correlation strength between process parameters and between process parameters and quality indicators, so that the partial least squares regression model no longer needs an additional intercept term and eliminates the interference of the benchmark value on the interpretation of the coefficients.

[0055] In one possible implementation, refer to Figure 4 , Figure 4 This is the fourth flowchart of the multivariate control method for injection molding provided by the present invention. Step 111 decentralizes multiple sets of process parameters and quality indicators in the initial sample set, and further includes the following steps: Step 1111: Calculate the first mean vector corresponding to each column of process parameters in the initial sample set, and the second mean vector corresponding to each column of quality indicators.

[0056] Specifically, the system calculates the column mean for each column of the parameter matrix X to form the first mean vector. First mean vector The elements in the matrix Y correspond to the column mean of injection speed, the column mean of holding position, and the column mean of holding pressure, respectively. Simultaneously, the column mean of each column in the response matrix Y is calculated to form a second mean vector. Second mean vector The elements in the table correspond to the column mean of product quality, the column mean of warpage, and the column mean of production cycle, respectively.

[0057] Step 1112: Subtract the first mean matrix of the corresponding dimension, which is composed of the first mean vector, from the process parameter matrix in the initial sample set to obtain the input residual matrix.

[0058] Specifically, the system utilizes an n×1 all-one vector 1 and the first mean vector Performing an outer product operation on the transpose of X yields a first mean matrix with the same dimensions as the parameter matrix X. Then follow the formula Subtracting the parameter matrix X from the first mean matrix yields the centered process parameter matrix. and process parameter matrix As the initial input residual matrix .

[0059] Step 1113: Subtract the second mean matrix of the corresponding dimension, which is composed of the second mean vector, from the quality index matrix in the initial sample set to obtain the output residual matrix.

[0060] Specifically, the system utilizes an n×1 all-one vector 1 and the second mean vector The outer product operation is performed on the transpose of the matrix to obtain the second mean matrix with the same dimension as the response matrix Y. Then follow the formula Subtracting the response matrix Y from the second mean matrix yields the centered quality index matrix. and the quality indicator matrix As the initial output residual matrix .

[0061] Through steps 1111 to 1113, this embodiment eliminates the interference of the benchmark values ​​of process parameters and quality indicators by calculating the column mean and performing matrix subtraction, so that the covariance can directly reflect the linear correlation strength between variables and eliminate the dependence of the regression process on the intercept term, thereby providing the necessary preconditions for the partial least squares regression model to correctly extract latent variables.

[0062] Step 112: Extract latent variables based on the input residual matrix and the output residual matrix, and construct a partial least squares regression model containing a regression coefficient matrix, where the regression coefficient matrix is ​​used to characterize the independent influence of a single process parameter category on a single quality index.

[0063] Specifically, the system extracts latent variables from the input and output residual matrices through iterative extraction, ultimately obtaining a weight matrix W, a factor loading matrix P, and a response loading matrix Q, each composed of column-wise combinations of the weight vectors, factor loadings, and response loadings from each round. The system then solves for the regression coefficient matrix B based on the weight matrix W, factor loading matrix P, and response loading matrix Q. The regression coefficient matrix B is a 3x3 matrix, where the element in the i-th row and j-th column of matrix B... This represents the magnitude and direction of the independent influence of the i-th process parameter category on the j-th quality index, with a positive sign indicating a positive influence and a negative sign indicating a negative influence. For example, the rows of the regression coefficient matrix B correspond sequentially to the injection speed. Transfer location Holding pressure The columns correspond to product quality in order. Warpage Production cycle .

[0064] In one possible implementation, refer to Figure 5 , Figure 5 This is the fifth flowchart of the multivariate control method for injection molding process provided by the present invention. Step 112, which involves extracting latent variables based on the input residual matrix and the output residual matrix and constructing a partial least squares regression model containing a regression coefficient matrix, further includes the following steps: Step 1121: Iteratively execute the following processes until a preset number of latent variables are extracted: calculate the weight vector between the input residual matrix and the output residual matrix; multiply the input residual matrix by the weight vector to obtain the factor score; calculate the factor loadings corresponding to the input residual matrix and the response loadings corresponding to the output residual matrix based on the factor scores; update the input residual matrix and the output residual matrix by matrix subtraction based on the factor scores, factor loadings, and response loadings, as input data for the next iteration round.

[0065] Specifically, the system sets the preset number of latent variables to... H , H The value of depends on how many composite factors are used to interpret the data. For example, when H When the value is 3, the iterative process extracts 3 latent variables, and the constructed weight matrix W has 3 columns. The factor loading matrix P and the response loading matrix Q also both have 3 columns. H When the value is 5, the iteration process extracts 5 latent variables, and the weight matrix W has 5 columns.

[0066] System Initialization , Afterwards, The following iterative steps are performed repeatedly: Step 1, calculate the following formula: h Weight vector for each iteration round : ; in, Let h be the input residual matrix for the (h-1)th iteration. This is the output residual matrix for the (h-1)th iteration. Weight vector. The input process parameters such as injection speed, transfer position, and holding pressure are combined into a comprehensive index, so that the comprehensive index has the strongest combination relationship with output quality indicators such as product quality, warpage, and production cycle.

[0067] Step two, calculate the following formula: h Factor scores for each iteration round : ; Factor scores The input values ​​of each sample are arranged according to the weight vector. The samples are combined to calculate the overall score for each sample.

[0068] Step 3, calculate the following formula: h Factor loadings in each iteration round : ; Factor Loading Indicates factor score The average degree of change in injection speed, transfer position, and holding pressure for each unit change; Furthermore, the first step is calculated according to the following formula. h Response load in each iteration round : ; Response Load Indicates factor score The average degree of change in product quality, warpage, and production cycle for each unit change.

[0069] Step 4: Update the input residual matrix and output residual matrix respectively by matrix subtraction according to the following formula: ; And the updated , Using this data as input for the next iteration, the calculation process for the weight vector, factor scores, factor loadings, and response loadings is repeated to find the weight vector for the next iteration. Factor scores for the next iteration round and the corresponding factor loadings for the next iteration round Response load in the next iteration round Until the iteration round number h Reaching the preset quantity H .

[0070] Step 1122: Combine the weight vectors, factor loads and response loads extracted from each iteration into columns to construct the weight matrix, factor load matrix and response load matrix in sequence.

[0071] Specifically, the system concatenates all weight vectors obtained during the H-round iterations column-wise to form a weight matrix. All factor loads are concatenated column-wise to form a factor load matrix. All response loads are concatenated column-wise to form a response load matrix. .

[0072] Step 1123: Calculate the regression coefficient matrix based on the product of the weight matrix, factor loading matrix, and response loading matrix.

[0073] Specifically, the partial least squares regression model can be expressed as: The form is given by , where E is the residual, and it has an approximate relationship. as well as .Will Multiplying both sides by the weight matrix W, we get Thus, the solution is obtained. Substitute T into ,get Based on this, the system obtains the regression coefficient matrix B by multiplying the weight matrix W, the factor loading matrix P, and the response loading matrix Q.

[0074] Through steps 1121 to 1123, this implementation constructs the weight matrix, factor loading matrix, and response loading matrix of the partial least squares regression model using an iterative extraction method, and derives the regression coefficient matrix through explicit matrix multiplication. This approach not only preserves the ability to maximize the discovery of the correlation structure between latent variables and process parameters and quality indicators, but also ensures the stability and interpretability of the partial least squares regression model in small sample and multivariate coupling scenarios.

[0075] In one possible implementation, step 1123, which calculates the regression coefficient matrix based on the product of the weight matrix, factor loading matrix, and response loading matrix, further includes the following steps: Step 11231: Calculate the first product matrix of the transpose of the factor loading matrix and the weight matrix.

[0076] Specifically, the system performs a transpose operation on the factor loading matrix P to obtain the transpose matrix of the factor loading matrix. Then transpose the factor loading matrix Multiplying it by the weight matrix W yields the first product matrix. .

[0077] Step 11232: Invert the first product matrix. Multiply the weight matrix on the right by the inverse of the first product matrix, and then multiply it on the right by the transpose of the response load matrix to obtain the regression coefficient matrix.

[0078] Specifically, the system performs a multiplication on the first product matrix. Performing the inverse operation yields the inverse matrix. Then, the weight matrix W is right-multiplied by the inverse matrix. Obtain the intermediate matrix Then multiply the intermediate matrix by the transpose of the response load matrix. Finally, the regression coefficient matrix B is obtained, that is: ; The regression coefficient matrix B is a 3x3 matrix, and the elements of the regression coefficient matrix B are... Let represent the magnitude and direction of the influence of the i-th process parameter category on the j-th quality index. Then matrix B is: ; In this matrix, the rows correspond to the input factors: (Injection speed) (Transfer location) (Pressure holding pressure); Columns of matrix B correspond to the following responses: (Product quality) (warpage) (Production cycle). A positive sign indicates a positive impact, and a negative sign indicates a negative impact.

[0079] Step 113: Calculate the projected importance values ​​of the variables corresponding to each process parameter category using the parameters of the latent variables extracted by the partial least squares regression model.

[0080] Specifically, during the construction of the partial least squares regression model, the system extracts latent variables that characterize the linear combination relationship of the original process parameters, and subsequently obtains the parameters of these latent variables. The parameters of the latent variables include a weight vector, factor scores, factor loadings, and response loadings. The weight vector represents the weight allocation of process parameter categories such as injection speed, transfer position, and holding pressure when combined into a comprehensive index, ensuring the strongest combination relationship between the comprehensive index and each quality index. The factor score represents the comprehensive score obtained by combining the process parameter values ​​of each sample in the initial sample set according to the weight vector. The factor loading represents the average change in injection speed, transfer position, and holding pressure when the factor score changes by a single unit. The response loading represents the average change in product quality, warpage, and production cycle when the factor score changes by a single unit.

[0081] Furthermore, based on the parameters of the aforementioned latent variables such as weight vectors, factor scores, and response loads, the system calculates the variable projection importance value for each process parameter category to characterize the comprehensive influence of individual process parameter categories such as injection speed, transfer position, and holding pressure on all quality indicators.

[0082] In one possible implementation, the variable projection importance value corresponding to each process parameter category is further calculated according to the following formula: ; ; in, For the first j Projected importance values ​​of variables corresponding to each category of process parameter p This represents the total number of process parameter categories. H The total number of latent variables. h The iteration round number of the latent variable. For the first h The weight vector extracted in the nth iteration corresponds to the weight vector in the nth iteration. j Weight components for each process parameter category, For the first hThe square of the magnitude of the weight vector extracted in each iteration round; For the first h The sum of squares explaining the quality indicators in the initial sample set by the latent variables. Indicates the first h The transpose matrix of factor scores for each iteration round; Indicates the first h The output residual matrix used before each iteration round update; Indicates the first h The response load for each iteration round.

[0083] Specifically, in implementation scenarios where injection speed, transfer position, and holding pressure are used as process parameter categories, the total number of process parameter categories is... p The value is set to 3. The system calculates the corresponding projected importance value for each process parameter category based on the above formula. The calculated projected importance value comprehensively reflects the explanatory power of each latent variable on the set of quality indicators, and uses this as a basis to assess the overall contribution of each process parameter category to all quality indicators. The larger the projected importance value, the more significant the comprehensive impact of the corresponding process parameter category on multiple quality indicators.

[0084] The above formula describes the specific calculation method of the variable projection importance value, enabling those skilled in the art to accurately and reproducibly achieve a quantitative assessment of the comprehensive influence of process parameter categories, and providing an objective and quantitative basis for subsequent process parameter category screening and process adjustment.

[0085] Step 114: Eliminate process parameter categories whose variable projection importance values ​​are less than the preset importance threshold, and use the remaining process parameter categories as the target process parameter categories.

[0086] Specifically, the system compares the projected importance value of the variable corresponding to each process parameter category with a preset importance threshold. In one implementation, the preset importance threshold is set to 0.8. When the projected importance value of a variable for a process parameter category is greater than 1, it is considered very important to the explanatory quality index; when the projected importance value of a variable for a process parameter category is between 0.8 and 1, it is considered moderately important; and when the projected importance value of a variable for a process parameter category is less than 0.8, it is considered unimportant and is removed. The system uses the process parameter categories retained after the removal process as the target process parameter categories for subsequent multi-objective optimization.

[0087] Through steps 111 to 114, this implementation not only uses the regression coefficient matrix to achieve a quantitative analysis of the influence relationship between a single process parameter category and a single quality index, but also achieves an objective screening of process parameter categories through variable projection importance values. This reduces the dimensionality of the optimization space while retaining key variables, thereby improving the solution efficiency of the subsequent multi-objective optimization process.

[0088] In one possible implementation, refer to Figure 6 , Figure 6 This is the sixth flowchart of the multivariate control method for injection molding provided by the present invention. In step 120, the function constructed based on the prediction model is used as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category. This further includes steps 121 to 124: Step 121: Initialize a population containing multiple particles within the search space of the multi-objective optimization algorithm. Each particle represents a set of parameter values ​​for the target process parameter category within the feasible region.

[0089] Specifically, the system sets the decision variables as follows: The feasible domain of the decision variables is determined by the process range. For example, the feasible domain for injection speed is 30-120 mm / s, the feasible domain for holding position is 10-50 mm, and the feasible domain for holding pressure is 20-100 bar. The system randomly generates multiple particles within the feasible domain, and the position coordinates of each particle correspond to a set of parameter values ​​for the target process parameter category.

[0090] Step 122: Multiply the input row vector consisting of each set of parameter values ​​with the regression coefficient matrix in the prediction model to obtain the predicted values ​​of the corresponding particles in multiple quality index dimensions.

[0091] Specifically, the system, for each particle, according to the prediction function... The input row vector corresponding to the particle position Perform matrix multiplication with the regression coefficient matrix B, where the input row vectors are... The product is a 1×3 row vector, and the regression coefficient matrix B is a 3×3 matrix. It is a 1×3 row vector, which corresponds to the predicted value of the particle in three quality index dimensions: product quality, warpage, and production cycle.

[0092] Step 123: According to the preset optimization direction corresponding to each quality index dimension, convert each predicted value into the multi-objective fitness value of each particle.

[0093] Specifically, the system pre-sets optimization directions for quality indicators such as product quality, warpage, and production cycle. When the optimization objectives are to maximize product quality, minimize warpage, and minimize production cycle, the system unifies the directions of each predicted value to obtain three multi-objective fitness values ​​that converge in the same minimization direction, so that the multi-objective particle swarm optimization algorithm can complete the optimization within a unified minimization framework.

[0094] In one possible implementation, step 123, which converts each predicted value into the multi-objective fitness value of each particle according to the preset optimization direction corresponding to each quality index dimension, further includes the following steps: If the preset optimization direction of a certain quality indicator is to maximize it, then the negative of the corresponding predicted value is taken as the minimum fitness value of the corresponding quality indicator dimension. If the preset optimization direction of a certain quality indicator is minimization, then the corresponding predicted value will be used as the minimum fitness value of the corresponding quality indicator dimension.

[0095] Specifically, in implementation scenarios where product quality is maximized, the system targets the predicted product quality value. Take the negative number, As the minimum fitness value of the product quality dimension, the goal of maximizing product quality is thus uniformly transformed into the goal of minimizing.

[0096] In implementation scenarios where both warpage and production cycle minimization are pursued, the system directly outputs the predicted warpage value. As and production cycle forecast values As , respectively serving as the minimum fitness value for the warpage dimension and the minimum fitness value for the production cycle dimension.

[0097] This implementation provides a unified direction transformation processing method for maximizing and minimizing objectives, enabling quality indicators with different optimization directions, such as product quality, warpage, and production cycle, to be uniformly solved by the multi-objective particle swarm optimization algorithm under the same minimization framework.

[0098] Step 124: Update the velocity and position of each particle based on the multi-objective fitness value, iteratively optimize to obtain the Pareto optimal solution set, select a set of solutions that satisfy the preset priority order from the Pareto optimal solution set as candidate parameter combinations, and use the prediction value corresponding to this set of solutions as the prediction quality index.

[0099] Specifically, in each iteration, the system updates the individual optimal position and global optimal position of each particle based on its multi-objective fitness value, and updates the velocity and position vectors of each particle according to the velocity and position update rules of the multi-objective particle swarm optimization algorithm. When the number of iterations meets the convergence condition, the system obtains a set of Pareto optimal solutions that are mutually independent in the three dimensions of product quality, warpage, and production cycle. Based on the priority order set according to the current actual production needs, such as product quality first, warpage second, and production cycle third, the system selects a set of solutions that meet the priority order from the Pareto optimal solution set as candidate parameter combinations, and uses the predicted values ​​corresponding to the selected set of solutions as the predicted quality index.

[0100] Through steps 121 to 124, this embodiment deeply couples the partial least squares regression model with the multi-objective particle swarm optimization algorithm, enabling the process parameter optimization process to collaboratively optimize multiple conflicting quality indicators within the feasible domain of process constraints. Furthermore, by using the Pareto optimal solution set, it provides flexible process parameter selection for actual needs of different production priorities, thereby significantly improving the flexibility and engineering applicability of the multivariate control method for injection molding processes.

[0101] Reference Figure 7 , Figure 7 This is a schematic diagram of the multivariable control system for injection molding process provided by the present invention. The system includes: The first acquisition module is used to acquire the target process parameter category and the prediction model. The prediction model is a partial least squares regression model built based on the initial sample set. The initial sample set contains multiple sets of process parameters and the quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from multiple process parameter categories based on the variable projection importance value. The variable projection importance value is used to characterize the comprehensive influence of a single process parameter category on all quality indicators. The parameter optimization module is used to use the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, and obtain candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations. The second acquisition module is used to acquire the actual quality indicators generated by performing injection molding operations according to the candidate parameter combinations; The deviation calculation module is used to calculate the deviation between the actual quality indicators and the predicted quality indicators. The target parameter determination module is used to determine the candidate parameter combination as the target injection molding parameter when the judgment deviation meets the preset threshold condition.

[0102] It should be noted that the injection molding process multivariate control system provided by the present invention can execute the injection molding process multivariate control method of any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0103] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 8 As shown, the electronic device may include a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute the multivariate control method for injection molding processes provided in the above embodiments.

[0104] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0105] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the injection molding process multivariate control method provided in the above embodiments.

[0106] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the injection molding process multivariate control method provided in the above embodiments.

[0107] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multivariate control method for injection molding process, characterized in that, include: The target process parameter category and prediction model are obtained. The prediction model is a partial least squares regression model constructed based on an initial sample set. The initial sample set contains multiple sets of process parameters and quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from the multiple process parameter categories based on variable projection importance values. The variable projection importance values ​​are used to characterize the comprehensive influence of a single process parameter category on all the quality indicators. The function constructed based on the prediction model is used as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, thereby obtaining candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations. Obtain the actual quality indicators generated by performing injection molding operations according to the candidate parameter combinations; Calculate the deviation between the actual quality index and the predicted quality index; When the deviation is determined to meet the preset threshold condition, the candidate parameter combination is determined as the target injection molding parameter.

2. The multivariate control method for injection molding process according to claim 1, characterized in that, After calculating the deviation between the actual quality indicator and the predicted quality indicator, the method further includes: When it is determined that the deviation does not meet the preset threshold condition, the candidate parameter combination and its corresponding actual quality index are used as new samples and merged into the initial sample set to form an updated sample set. Based on the updated sample set, latent variables are re-extracted to update the partial least squares regression model, and the process returns to the step of using the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category until the deviation meets the preset threshold condition.

3. The multivariate control method for injection molding process according to claim 1, characterized in that, The acquisition of the target process parameter category and prediction model includes: The process parameters and quality indicators in the initial sample set are decentralized to obtain a centered process parameter matrix and a quality indicator matrix, and the process parameter matrix and the quality indicator matrix are used as the input residual matrix and the output residual matrix, respectively. Based on the input residual matrix and the output residual matrix, latent variables are extracted, and the partial least squares regression model containing the regression coefficient matrix is ​​constructed, wherein the regression coefficient matrix is ​​used to characterize the independent influence of a single process parameter category on a single quality index. Using the parameters of the latent variables extracted by the partial least squares regression model, the projected importance value of the variables corresponding to each category of process parameters is calculated; The process parameter categories whose variable projection importance values ​​are less than a preset importance threshold are removed, and the remaining process parameter categories are used as the target process parameter categories.

4. The multivariate control method for injection molding process according to claim 3, characterized in that, The step of extracting latent variables based on the input residual matrix and the output residual matrix, and constructing the partial least squares regression model containing the regression coefficient matrix, includes: The following processes are performed iteratively until a predetermined number of latent variables are extracted: Calculate the weight vector between the input residual matrix and the output residual matrix; multiply the input residual matrix and the weight vector to obtain the factor score; calculate the factor loadings corresponding to the input residual matrix and the response loadings corresponding to the output residual matrix based on the factor scores; update the input residual matrix and the output residual matrix by matrix subtraction based on the factor scores, the factor loadings, and the response loadings, as input data for the next iteration round. The weight vectors, factor loads and response loads extracted in each iteration are combined column-wise to construct the weight matrix, factor load matrix and response load matrix in sequence. The regression coefficient matrix is ​​calculated based on the product of the weight matrix, the factor loading matrix, and the response loading matrix.

5. The multivariate control method for injection molding process according to claim 4, characterized in that, The calculation of the regression coefficient matrix based on the product of the weight matrix, the factor loading matrix, and the response loading matrix includes: Calculate the first product matrix of the transpose of the factor loading matrix and the weight matrix; Invert the first product matrix, multiply the weight matrix on the right by the inverse of the first product matrix, and then multiply it on the right by the transpose of the response load matrix to obtain the regression coefficient matrix.

6. The multivariate control method for injection molding process according to claim 4, characterized in that, The calculation of the variable projection importance value corresponding to each of the process parameter categories includes: The importance value of the variable projection is calculated using the following formula: ; ; in, For the first j The projected importance value of the variable corresponding to each of the aforementioned process parameter categories. The total number of the process parameter categories. H The total number of the latent variables. h Let be the iteration round number of the latent variable. For the first h The weight vector extracted in the nth iteration corresponds to the weight vector in the nth iteration. j Weight components for each process parameter category, For the first h The square of the magnitude of the weight vector extracted in each iteration round; For the first h The sum of squares explaining the quality indicators in the initial sample set by each latent variable. Indicates the first h The transpose matrix of factor scores for each iteration round; Indicates the first h The output residual matrix used before each iteration round update; Indicates the first h The response load for each iteration round.

7. The multivariate control method for injection molding process according to claim 3, characterized in that, The process of decentralizing multiple sets of process parameters and quality indicators in the initial sample set includes: Calculate the first mean vector corresponding to each column of process parameters in the initial sample set, and the second mean vector corresponding to each column of quality indicators. Subtract the first mean matrix of the corresponding dimension, which is composed of the first mean vector, from the process parameter matrix in the initial sample set to obtain the input residual matrix; The output residual matrix is ​​obtained by subtracting the second mean matrix of the corresponding dimension, which is composed of the second mean vector, from the quality index matrix in the initial sample set.

8. The multivariate control method for injection molding process according to claim 1, characterized in that, The step of using the function constructed based on the prediction model as the fitness function of a multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, thereby obtaining candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations, includes: In the search space of the multi-objective optimization algorithm, a population containing multiple particles is initialized, where each particle represents a set of parameter values ​​for the target process parameter category within the feasible region. Multiply the input row vector formed by each set of parameter values ​​by the regression coefficient matrix in the prediction model to obtain the predicted values ​​of the corresponding particles in multiple quality index dimensions. According to the preset optimization direction corresponding to each of the quality index dimensions, the predicted values ​​are converted into the multi-target fitness values ​​of each particle. The velocity and position of each particle are updated based on the multi-objective fitness value, and the Pareto optimal solution set is obtained through iterative optimization. A set of solutions that meet the preset priority order is selected from the Pareto optimal solution set as the candidate parameter combination, and the prediction value corresponding to the set of solutions is used as the prediction quality index.

9. The multivariate control method for injection molding process according to claim 8, characterized in that, The step of converting each predicted value into a multi-objective fitness value for each particle according to a preset optimization direction corresponding to each of the quality index dimensions includes: If the preset optimization direction of a certain quality indicator is to maximize it, then the negative number corresponding to the predicted value is taken as the minimum fitness value of that quality indicator dimension. If the preset optimization direction of a certain quality indicator is minimization, then the corresponding predicted value will be used as the minimum fitness value of that quality indicator dimension.

10. A multivariable control system for injection molding process, characterized in that, include: The first acquisition module is used to acquire the target process parameter category and the prediction model. The prediction model is a partial least squares regression model constructed based on an initial sample set. The initial sample set contains multiple sets of process parameters and quality indicators corresponding to each set of process parameters. Each set of process parameters includes parameter values ​​under multiple process parameter categories. The target process parameter category is selected from the multiple process parameter categories based on the variable projection importance value. The variable projection importance value is used to characterize the comprehensive influence of a single process parameter category on all the quality indicators. The parameter optimization module is used to use the function constructed based on the prediction model as the fitness function of the multi-objective optimization algorithm to optimize the parameter values ​​of the target process parameter category, and obtain candidate parameter combinations and the prediction quality index corresponding to the candidate parameter combinations. The second acquisition module is used to acquire the actual quality indicators generated by performing the injection molding operation according to the candidate parameter combination; The deviation calculation module is used to calculate the deviation between the actual quality indicator and the predicted quality indicator. The target parameter determination module is used to determine the candidate parameter combination as the target injection molding parameter when the deviation is determined to meet the preset threshold condition.