An intelligent adjustment method for injection molding process parameters based on hybrid model

Abstract

The present invention discloses a method for intelligently adjusting injection molding process parameters based on a hybrid model. The initial data of the Kriging proxy model is obtained by mapping simulation between injection molded product defects and process parameters, and injection molding machine time measurement data and Kriging fitting data are used as subsequent data. At the same time, a minimum signal-to-noise ratio function is defined; an optimal process parameter combination fitness function is defined; a process parameter contribution rate is defined; the optimal initial process parameters are determined; a GA-BP model with high prediction accuracy is constructed; a global optimization Kriging proxy model based on EGO is constructed; and the Kriging model and the GA-BP model are used to dynamically adjust the process parameters. The present invention can solve the problem that the Kriging model has high calculation accuracy but is sensitive to noise data, and combines the prediction model with machine vision detection means to realize dynamic adjustment of injection molding process parameters, effectively improving the adjustment efficiency and automation level in the injection molding process.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning and industrial automation, and specifically relates to an intelligent adjustment method for injection molding process parameters based on a hybrid model. Technical Background

[0002] Injection molding is a complex process characterized by multiple variables, distributed parameters, intermittent operation, large hysteresis, strong coupling, nonlinearity, and high dispersion. Numerous factors influence the quality of molded parts, broadly categorized as injection molding machine parameters, injection molding material parameters, injection molding process parameters, and disturbances. Process parameters are crucial for adjustment during the molding process and are a key determinant of molded product quality.

[0003] Due to the multi-parameter coupling of injection molding, the adjustment of the process parameters of injection molding machines currently relies mostly on the experience and professional knowledge accumulated by skilled workers, resulting in large randomness in the production quality of injection molded products and low adjustment efficiency.

[0004] Therefore, under the premise that the product mold and machine have been selected, how to effectively replace the physical model of the injection molding process with multiple independent variables and multiple uncontrollable factors with a data-driven proxy model, and at the same time set a dynamic adjustment strategy to establish a dynamic adjustment framework, conduct real-time monitoring and feedback on the product quality status, and dynamically adjust the process parameters in real time through the injection molding process parameter adjustment system, is the key to improving the quality and efficiency of the injection molding process. Summary of the Invention

[0005] In order to understand the problems in the background technology, the present invention proposes an intelligent adjustment method for injection molding process parameters based on a hybrid model.

[0006] The technical solution of the present invention comprises the following steps:

[0007] Step 1: Use Latin hypercube sampling to randomly sample values ​​within the empirical range of process parameters required for the injection molding process. Use the Taguchi orthogonal method to combine the randomly sampled values ​​to obtain multiple experimental schemes. Use Moldflow to simulate the experimental schemes to obtain target values. Perform variance analysis on the simulation results to determine the contribution rate of all process parameters, thereby determining the main process parameters that affect the quality of injection molded products.

[0008] Step 2: Construct a BP-GA neural network model using the main process parameters of step 1;

[0009] Step 3: Construct a global optimization Kriging surrogate model based on EGO;

[0010] Step 4: Start injection molding and produce products. Perform periodic inspections on the injection molded products during the production process to obtain multiple sets of process parameter combination data and their corresponding target values.

[0011] Step 5: Select valid data from the multiple sets of process parameter combination data in step 4 to calculate the signal-to-noise ratio, and update the GA-BP neural network model. At the same time, Latin hypercube sampling and EGO global optimization method are used to perform targeted point addition, supplement and update the training data of the Kriging model, and retrain the Kriging prediction model.

[0012] Step 6: Visually monitor the target values ​​of the injection molded products in real time during the production process and check whether they exceed the set threshold. Correct the process parameters corresponding to the target values ​​that exceed the threshold. Input the corrected process parameters into the GA-BP neural network model to predict the signal-to-noise ratio value and determine whether the signal-to-noise ratio meets the set range. If it exceeds the threshold, continue to correct the process parameters until the signal-to-noise ratio meets the required range.

[0013] Step 7: Input the process parameter combination that meets the signal-to-noise ratio requirements obtained in step 6 into the Kriging model to predict the target value of the injection molded product and determine whether it meets the set threshold:

[0014] If not satisfied, return to step 6 to continue to modify the process parameters;

[0015] If so, a process parameter combination that meets both the signal-to-noise ratio requirement and the quality standard is obtained;

[0016] Step 8: Use the process parameter combination obtained in step 7 to adjust the process parameters of the injection molding machine. Compare the actual defect detection results of the adjusted injection molding machine with the target value predicted in step 7, and feed the error back to the Kriging model for correction.

[0017] Through steps 6 to 8, a dynamic adjustment system for injection molding process parameters is formed.

[0018] In step 1), the process parameter contribution rate θ is calculated by the following formula:

[0019]

[0020]

[0021]

[0022]

[0023]

[0024] Where, SS t is the total sum of squares, SS i is the sum of squares of process parameters, SS e is the sum of squared errors, y is the response value, is the sum of all experimental indicators, T iis the sum of the experimental indicators of the same level of each process parameter, N is the total number of experiments, and n is the number of factors; DF i =a i -1, DF T =N-1,a i is the number of levels of the i-th factor; F is the statistic, MSB is the between-group variance, and MSE is the within-group variance;

[0025] Contribution rate θ = SeqSS i / ∑SeqSS i ;

[0026] Where,

[0027] Where s j represents the normalized mean value of defects, n is the number of process parameters, and i represents the i-th process parameter;

[0028] The main process parameters in step 1 are process parameters with a contribution rate greater than 5%.

[0029] The target values ​​include defects, glossiness and other main quality judgment criteria of injection molded products.

[0030] The step 2 is specifically as follows:

[0031] 2.1) Select the main process parameters (melt temperature x1, mold temperature x2, injection rate x3, holding time x4, holding pressure x5) as the input of the BP-GA neural network, and use the corresponding signal-to-noise ratio as the output to train the BP-GA neural network model;

[0032] 2.2) Increase the number of samples: Use the Latin hypercube sampling method to randomly sample within the process parameter range to obtain process parameter combinations as additional samples, and use simulation or actual experiments to obtain the corresponding target values;

[0033] 2.3) Optimize the BP-GA neural network structure: Simplify the model structure by reducing the number of hidden layer nodes. Specifically, the number of hidden layer nodes is determined by the following formula:

[0034]

[0035] h=log2m

[0036]

[0037] Where h is the number of hidden layer nodes; m is the number of input layer nodes; l is the number of output layer nodes; α is a constant, which is an integer between 1 and 10.

[0038] 2.4) Use the Bayesian method to optimize the weights and thresholds of the model to ensure accurate model fitting so that the prediction accuracy meets the requirements, thereby completing the construction of the BP-GA neural network model.

[0039] The step 3 is specifically as follows:

[0040] 3.1) Construct Kriging model using initial samples:

[0041] The initial sample includes the main process parameters determined in step 1 and the process parameters added in step 2 through the Latin hypercube sampling method, as well as the target values ​​corresponding to all process parameters;

[0042] 3.2) For sample data X * Perform normalization processing;

[0043] 3.3) Use Latin hypercube sampling and EGO global optimization to obtain new sample points:

[0044] Perform Latin hypercube sampling within the process parameter range to obtain M new sample points;

[0045] New sample points are determined by minimizing the response surface and maximizing the expected improvement function. The formula is as follows:

[0046]

[0047]

[0048] Where x (i) Represents the i-th new sample point, which contains the characteristics of process parameter combination and target value, i∈{1, 2,…M}, x (i) ∈X={x (1) ,...,x (i) ,...x (M)};y(x (i) ) is the sample x (i) Target value obtained through experiment or simulation;

[0049] The improvement is defined as I = max(y * -y(x (i) ), 0), y is the target value of the sample

[0050] Expected improvement function EI(x (i) )for:

[0051]

[0052] Where CDF and PDF are the cumulative distribution function and probability density function; y(μ(x)) is the sample data in step 3.2) plus the sample point x (i)The process parameters are obtained using the kriging model to obtain the target value mean, σ(x) is the sample data of step 3.2) plus the sample point x (i) The variance of the process parameters;

[0053] Select the x corresponding to the maximum EI(x) (i) As a new sample point x * , change x * Join X * , get a new set X * , using the updated sample dataset X * Refit the Kriging surrogate model to complete the update;

[0054] 3.4) Forecast error term of the updated Kriging model:

[0055] The optimization goal of the Kriging model is to minimize the prediction error. The objective function is as follows:

[0056]

[0057]

[0058] Where, is the objective function, y is the target value of the sample, is the predicted value of the sample target value; g j (x) is the objective function Constraint function of are the process parameters x in sample x i The upper and lower limits, x∈X * ; N c is the number of constraints;

[0059] 3.5) When the objective function If it is not less than the set target value, repeat 3.3) to 3.4);

[0060] When the objective function If the value is less than the set target value, the update of the Kriging model is stopped, thereby completing the construction of the global optimization Kriging model based on EGO.

[0061] In step 2 and step 5, the signal-to-noise ratio corresponding to the process parameters is obtained by the following signal-to-noise ratio function:

[0062]

[0063] Where SNR is the signal-to-noise ratio; is the average value of all sample target values; is the sample point x (i)The target value under the mth repeated simulation experiment; t is the target value; S 2 is the variance;

[0064]

[0065] Where N represents the number of experimental repetitions.

[0066] The effective data in step 5 is the process parameters corresponding to the time period when the injection molding machine enters stable production, that is, the sample data X′.

[0067] In step 5:

[0068] The method for updating the GA-BP neural network model is steps 2.3) to 2.4);

[0069] The method for supplementing and updating the training data of the Kriging model is steps 3.3) to 3.5).

[0070] In step 6, the process parameters are modified according to empirical rules, dynamic rules and summarized qualitative rules. The specific rules are as follows:

[0071] Set each process parameter x i The conservative threshold Eligibility threshold and change threshold Calculate the time series of process parameters within the detection period T1 The rate of change series and the overall rate of change Set the adjustment coefficient θ, θ is set according to experience, θ∈(-1,1);

[0072] a) When the target value to be corrected does not exceed the conservative threshold, it is directly determined that no process parameter adjustment is performed;

[0073] b) When the target value to be corrected is greater than the conservative threshold and less than or equal to the qualified threshold, the process parameter variation is judged: if the variation is less than the variation threshold, no adjustment is made to the injection molding process parameters; otherwise, the process parameters are slightly adjusted. The adjustment basis is:

[0074] Will The final time length is T1, which is multiplied by θ to obtain and Based on the interval t of the time series, N intervals are considered as a cycle, and the process parameters corresponding to each interval are used as the adjustment target, and each process parameter is gradually adjusted;

[0075] c) When the target value to be corrected is greater than the qualified threshold, the degree of change judgment condition is skipped and the process parameters are adjusted. The basis for the adjustment is: multi-objective optimization of the points on the kriging prediction model to obtain the process parameter combination that optimizes the target value to be corrected.

[0076] Beneficial effects of the present invention:

[0077] The method of the present invention can upgrade the manually driven injection molding process parameter adjustment to data-driven injection molding process parameter adjustment, obtain product quality data through real-time monitoring, realize dynamic process parameter adjustment of injection molded products, and effectively improve the quality and efficiency of the injection molding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 Determine the flow chart for the initial process parameters of injection molding;

[0079] Figure 2 This is a flow chart for dynamic adjustment of injection molding process parameters. DETAILED DESCRIPTION

[0080] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0081] like Figure 1 As shown:

[0082] Step 1: Set the signal-to-noise ratio function:

[0083]

[0084] Where SNR is the signal-to-noise ratio, Refers to the target value under the mth repeated simulation experiment (the defect value of the injection molded product corresponding to the process parameters), and t is the target value;

[0085]

[0086] Where N represents the number of repetitions;

[0087] Step 2: Define the optimal process parameter combination fitness function:

[0088]

[0089] st

[0090] LS j ≤x j ≤US j ; j = 1, 2, ... n

[0091] Where T(X) is the objective function; s represents the total number of defects; P sni is the signal-to-noise ratio prediction value of the i-th defect; SN iis the highest signal-to-noise ratio value of the i-th defect value; x j is the process parameter of injection molding process; LS j , US j are the lower and upper limits of the process parameters respectively; n is the total number of process parameters;

[0092] Step 3: Define the contribution of process parameters:

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] Where, SS t is the total sum of squares, SS i is the sum of squares of process parameters, SS e is the sum of squared errors, y is the response value, T is the sum of all experimental indicators, T i is the sum of the experimental indicators of the same level of each process parameter, N is the total number of experiments, and n is the number of factors; DF i =a i -1, DF T =N-1,a i is the number of levels of the i-th factor; F is the statistic, MSB is the between-group variance, and MSE is the within-group variance;

[0099]

[0100] Contribution rate θ = Seq SS i / ∑Seq SS i ;s j represents the normalized mean value of defects.

[0101] Step 4: Determine the optimal initial process parameters based on Taguchi orthogonal method and variance analysis;

[0102] 4.1: All process parameters are combined using the Taguchi orthogonal method to obtain multiple experimental schemes. The experimental schemes are simulated using Moldflow. The simulation results are analyzed for variance to determine the contribution rate of all process parameters and the main process parameters affecting product quality.

[0103] The contribution rate of the main process parameters is greater than 5%, and the process parameters with a contribution rate of less than 5% are excluded;

[0104] 4.2: The orthogonal experimental process parameter combination and the corresponding signal-to-noise ratio (S / N) data are put into the BP-GA neural network for training to establish a prediction model, as shown in the following table;

[0105] Process parameter calculation table

[0106]

[0107]

[0108] 4.3: Calculate the contribution rate of each process parameter:

[0109]

[0110]

[0111] ANOVA results table

[0112]

[0113] BP-GA was used to predict the minimum signal-to-noise ratio and obtain the corresponding process parameter combination, which was used as the initial process parameter combination for injection molding.

[0114] Step 5: Build a GA-BP model and improve its prediction accuracy;

[0115] The GA-BP network construction in step 5 specifically includes the following steps:

[0116] 5.1: Select the main process parameters (melt temperature x1, mold temperature x2, injection rate x3, holding time x4, holding pressure x5) as the BP neural network input, and use the signal-to-noise ratio value as the output to train the BP-GA neural network model;

[0117] 5.2: Use the Latin hypercube sampling method to randomly sample within the process parameter range to obtain process parameter combinations and increase the number of samples;

[0118] 5.3: Optimize the neural network structure, reduce the number of hidden layer connections, and simplify the model structure;

[0119] Determine the number of hidden layer nodes:

[0120]

[0121] h=log2 m

[0122]

[0123] Where: h is the number of hidden layer nodes, n is the number of input layer nodes, s is the number of output layer nodes, α is a constant, an integer from 1 to 10;

[0124] 5.4: Use Bayesian methods to optimize the weights and thresholds of the model to ensure accurate model fitting so that the prediction accuracy meets the requirements

[0125] Step 6: Construct a global optimization Kriging surrogate model based on EGO, which includes the following steps:

[0126] 6.1: Constructing a Kriging model using initial discrete data samples;

[0127] The initial samples are all process parameter combinations of steps 4 and 5 and their corresponding defect values;

[0128] 6.2: Normalize the sample data X*:

[0129] 6.3: Use Latin hypercube sampling and EGO global optimization to obtain new sample points:

[0130] New sample points are obtained by minimizing the response surface and maximizing the expected improvement function (EI), as follows:

[0131]

[0132]

[0133] Where: x (i) Represents the i-th new sample point, i∈{1, 2, …M}, x (i) ∈X={x (1) ,...,x (i) ,...x (M)};y(x (i) ) is the sample x (i) Target value obtained through experiment or simulation;

[0134] The improvement is defined as I = max(y * -y(x (i) ), 0), y is the target value of the sample

[0135] The expected improvement function is:

[0136]

[0137] Where CDF and PDF refer to the cumulative distribution function and probability density function;

[0138] Select the x corresponding to the maximum EI(x) (i) As a new sample point x * , change x * Join X * ;

[0139] 6.4: Prediction error term of the updated Kriging model:

[0140] The optimization goal of the Kriging model is to minimize the prediction error. The objective function is as follows:

[0141]

[0142]

[0143] Where, is the objective function; g i (x) is the constraint function; x u , x l are the upper and lower limits of the process parameter x; N c is the number of constraints;

[0144] 6.5: Repeat 6.3 to 6.4 until the objective function If the value is less than the set target value, the update of the Kriging model is stopped.

[0145] Step 7: Use Kriging model and GA-BP model to dynamically adjust process parameters.

[0146] like Figure 2 As shown, the dynamic adjustment method in step 7 specifically includes the following steps:

[0147] 7.1 Select an injection molded product and conduct an experimental design using Taguchi orthogonal experiments with multiple process parameters as variables and multiple product defect types as indicators. Use Moldflow to obtain simulation data to determine the main process parameters for multiple defects and the significance of the impact of process parameters on various defects in the product. Based on the results, determine the combination of significant influencing factors for multiple defect types.

[0148] 7.2 Carry out actual tests to obtain real data and start injection molding production;

[0149] 7.3 Periodically inspect injection molded products during production to obtain multiple sets of process parameter combination data and their corresponding defect values. Select historical data corresponding to the period when the injection molding machine entered stable production as valid data. Calculate the signal-to-noise ratio using this valid data and periodically update the GA-BP neural network. Simultaneously, perform targeted additions using Latin hypercube sampling and EGO optimization methods. Periodically supplement and update the Kriging model training data and retrain the Kriging prediction model.

[0150] 7.4 Conduct real-time visual monitoring of defects in injection molded products. If one or more defects exceed the set threshold, obtain the current process parameters and defect value data, and modify the process parameters based on empirical rules, dynamic rules, and summarized qualitative laws. Input the modified process parameters into the GA-BP signal-to-noise ratio model to predict the signal-to-noise ratio value and determine whether the signal-to-noise ratio meets the set range. If it exceeds the threshold, continue to modify the process parameters until the signal-to-noise ratio meets the requirements.

[0151] 7.5 The obtained process parameter combination that meets the signal-to-noise ratio requirements is introduced into the Kriging model to predict the defect value of the injection molded product and determine whether it meets the set threshold. This cycle is repeated to eventually obtain a process parameter combination that meets both the signal-to-noise ratio requirements and the quality standards.

[0152] Adjust the process parameters of the injection molding machine, compare the adjusted visual inspection results with the predicted results, feed the error back to the prediction model, and correct the prediction model;

[0153] 7.6 Combine steps 7.4 and 7.5 to form a dynamic adjustment system for injection molding process parameters.

[0154] While the present invention has been disclosed above with reference to preferred embodiments, this is not intended to limit the present invention. Persons skilled in the art will readily appreciate that various modifications and variations can be made without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A method for intelligently adjusting injection molding process parameters based on a hybrid model, characterized in that: The following steps are involved: Step 1: Use Latin hypercube sampling to randomly sample values ​​within the empirical range of process parameters required for the injection molding process. Use the Taguchi orthogonal method to combine the randomly sampled values ​​to obtain multiple experimental schemes. Use Moldflow to simulate the experimental schemes to obtain target values. Perform variance analysis on the simulation results to determine the contribution rate of all process parameters, thereby determining the main process parameters that affect the quality of injection molded products. Step 2: Construct a BP-GA neural network model using the main process parameters of step 1; Step 3: Construct a global optimization Kriging surrogate model based on EGO; Step 4: Start injection molding and produce products. Perform periodic inspections on the injection molded products during the production process to obtain multiple sets of process parameter combination data and their corresponding target values. Step 5: Select valid data from the multiple sets of process parameter combination data in step 4 to calculate the signal-to-noise ratio, and update the GA-BP neural network model. At the same time, Latin hypercube sampling and EGO global optimization method are used to perform targeted point addition, supplement and update the training data of the Kriging model, and retrain the Kriging prediction model. Step 6: Monitor the target value of the injection molded product in real time during the production process and check whether it exceeds the set threshold. Modify the process parameters corresponding to the target value that exceeds the threshold. Input the modified process parameters into the GA-BP neural network model to predict the signal-to-noise ratio value and determine whether the signal-to-noise ratio meets the set range. If it exceeds the threshold, continue to modify the process parameters until the signal-to-noise ratio meets the requirement. Step 7: Input the process parameter combination that meets the signal-to-noise ratio requirements obtained in step 6 into the Kriging model to predict the target value of the injection molded product and determine whether it meets the set threshold: If not satisfied, return to step 6 to continue to modify the process parameters; If so, a process parameter combination that meets both the signal-to-noise ratio requirement and the quality standard is obtained; Step 8: Use the process parameter combination obtained in step 7 to adjust the process parameters of the injection molding machine. Compare the actual defect detection results of the adjusted injection molding machine with the target value predicted in step 7, and feed the error back to the Kriging model for correction. A dynamic adjustment system for injection molding process parameters is formed through steps 6 to 8; In step 1), the process parameter contribution rate θ is calculated by the following formula: Contribution rate θ = Seq SS i / ∑Seq SS i ; Where, Where s j represents the normalized mean value of defects, n is the number of process parameters, and i represents the i-th process parameter; The main process parameters in step 1 are process parameters with a contribution rate greater than 5%; The step 3 is specifically as follows: 3.1) Construct the Kriging model using the initial sample: The initial sample includes the main process parameters determined in step 1 and the process parameters added in step 2 through the Latin hypercube sampling method, as well as the target values ​​corresponding to all process parameters; 3.2) For sample data X * Perform normalization processing; 3.3) Use Latin hypercube sampling and EGO global optimization to obtain new sample points: Perform Latin hypercube sampling within the process parameter range to obtain M new sample points; New sample points are determined by minimizing the response surface and maximizing the expected improvement function. The formula is as follows: Where x (i) Represents the i-th new sample point, which contains the characteristics of process parameter combination and target value, i∈{1,2,…M}, x (i) ∈X={x (1) ,…,x (i) ,…x (M) };y(x (i) ) is the sample x (i) Target value obtained through experiment or simulation; Expected improvement function EI(x (i) )for: Where CDF and PDF are the cumulative distribution function and probability density function; y(μ(x)) is the sample data in step 3.2) plus the sample point x (i) The process parameters are obtained using the kriging model to obtain the target value mean, σ(x) is the sample data of step 3.2) plus the sample point x (i) The variance of the process parameters; Select the x corresponding to the maximum EI(x) (i) As a new sample point x * , change x * Join X * , get a new set X * , using the updated sample dataset X * Refit the Kriging surrogate model to complete the update; 3.4) Forecast error term of the updated Kriging model: The optimization goal of the Kriging model is to minimize the prediction error. The objective function is as follows: Where, is the objective function, y is the target value of the sample, is the predicted value of the sample target value; g j (x) is the objective function Constraint function of are the process parameters x in sample x i The upper and lower limits, x∈X * ; N c is the number of constraints; 3.5) When the objective function If it is not less than the set target value, repeat 3.3) to 3.4); When the objective function If the value is less than the set target value, the update of the Kriging model is stopped, thereby completing the construction of the global optimization Kriging model based on EGO.

2. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 1, characterized in that: The target values ​​include defects and glossiness of the injection molded product.

3. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 1, characterized in that: The step 2 is specifically as follows: 2.1) Select the main process parameters as the input of the BP-GA neural network and use the corresponding signal-to-noise ratio as the output to train the BP-GA neural network model; 2.2) Increase the number of samples: Use the Latin hypercube sampling method to randomly sample within the process parameter range to obtain process parameter combinations as additional samples, and use simulation or actual experiments to obtain the corresponding target values; 2.3) Optimize the BP-GA neural network structure: Simplify the model structure by reducing the number of hidden layer nodes. Specifically, the number of hidden layer nodes is determined by the following formula: h=log2m Where h is the number of hidden layer nodes; m is the number of input layer nodes; l is the number of output layer nodes; α is a constant, which is an integer between 1 and 10. 2.4) Use the Bayesian method to optimize the weights and thresholds of the model to ensure accurate model fitting so that the prediction accuracy meets the requirements, thereby completing the construction of the BP-GA neural network model.

4. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 3, characterized in that: The signal-to-noise ratio corresponding to the process parameters is obtained through the following signal-to-noise ratio function: Where SNR is the signal-to-noise ratio; is the average value of all sample target values; is the target value of the sample point in the mth repeated simulation experiment; t is the target value; S 2 is the variance; Where N represents the number of experimental repetitions.

5. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 1, characterized in that: The effective data in step 5 is the process parameters corresponding to the time period when the injection molding machine enters stable production, that is, the sample data X′.

6. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 3, characterized in that: In step 5: The method for updating the GA-BP neural network model is steps 2.3) to 2.4); The method for supplementing and updating the training data of the Kriging model is steps 3.3) to 3.5).

7. The method for intelligently adjusting injection molding process parameters based on a hybrid model according to claim 1, characterized in that: In step 6, the process parameters are modified according to empirical rules, dynamic rules and summarized qualitative rules. The specific rules are as follows: Set each process parameter x i The conservative threshold Eligibility threshold and change threshold Calculate the time series of process parameters within the detection period T1 The rate of change series and the overall rate of change Set the adjustment coefficient θ, θ is set according to experience, θ∈(-1,1); a) When the target value to be corrected does not exceed the conservative threshold, it is directly determined that no process parameter adjustment is performed; b) When the target value to be corrected is greater than the conservative threshold and less than or equal to the qualified threshold, the process parameter variation is judged: if the variation is less than the variation threshold, no adjustment is made to the injection molding process parameters; otherwise, the process parameters are slightly adjusted. The adjustment basis is: Will The final time length is T1, which is multiplied by θ to obtain and Based on the interval t of the time series, N intervals are considered as a cycle, and the process parameters corresponding to each interval are used as the adjustment target, and each process parameter is gradually adjusted; c) When the target value to be corrected is greater than the qualified threshold, the degree of change judgment condition is skipped and the process parameters are adjusted. The basis for the adjustment is: multi-objective optimization of the points on the kriging prediction model to obtain the process parameter combination that optimizes the target value to be corrected.

Images