A method for optimizing molding quality of glass fiber injection molding products

Abstract

The present invention discloses a method for optimizing the molding quality of glass fiber injection molded products. The warpage deformation and volume shrinkage of glass fiber injection molded products are used as quality optimization targets. First, the test scheme results are obtained based on CCF test design and injection molding simulation. Then, the injection molding quality optimization model is constructed using ILSO‑RF. Then, multi-objective optimization is performed using MSINSWOA, and the optimal combination of process parameters is determined in combination with the Critic method. The ILSO‑RF model of the glass fiber injection molded product molding quality optimization method controls the prediction error of the warpage deformation and volume shrinkage within 5%, and the model prediction accuracy is good. The warpage deformation and volume shrinkage of the glass fiber injection molded products after optimization are significantly reduced compared with the simulation results before optimization, respectively, which can provide a theoretical basis and data support for obtaining the optimal combination of optimal process parameters for GFRP injection molding.

Description

Technical Field

[0001] The present invention relates to a method for optimizing the molding quality of injection molded products, in particular to a method for optimizing the molding quality of glass fiber injection molded products based on an ILSO-RF model and MSINSWOA, and belongs to the technical field of injection molding. Background Art

[0002] Glass fiber reinforced plastic (GFRP) is a composite material made by pultruding and winding resin and then thermally curing glass fiber (GF). GFRP is widely used in aerospace, chemical engineering, and other fields due to its advantages such as high strength, corrosion resistance, high temperature resistance, high thermal insulation, light weight, strong impact resistance, and good flame retardancy. It is mostly used as a structural engineering material in the structural parts of products. However, the addition of GF not only increases the melt viscosity, deteriorates the fluidity, and increases the water absorption, but also causes the melt to orient along the flow direction during the injection molding process, thereby causing the mechanical properties and shrinkage rate to increase in the orientation direction, making the injection molded products prone to quality defects such as warping, deformation, and volume shrinkage, thereby reducing the molding quality of glass fiber injection molded products. Therefore, compared with ordinary plastic injection molding, GFRP injection molding is more difficult to improve product quality. Warping and volume shrinkage are two common quality defects of GFRP injection molded parts.

[0003] The molding quality of injection molded products depends largely on process parameters. Traditional process parameter optimization methods require repeated mold trials, resulting in a waste of time and cost. In recent years, researchers have begun using mathematical surrogate models to construct relationships between quality objectives and process parameters. These models then use search-based optimization algorithms to identify optimal combinations of process parameters, significantly reducing research time and improving production efficiency. For example, Cao Y et al. constructed an adaptive network fuzzy inference system (ANFIS) between warpage and process parameters and used a genetic algorithm (GA) to identify optimal process parameter solutions. Sun Zheng et al. constructed mathematical surrogate models between warpage, volumetric shrinkage, and process parameters based on the gradient-enhanced Kriging (GEK) model. They then used a multi-objective differential evolution (MODE) algorithm to search for Pareto solutions and determined the optimal process parameter solution based on weighted coefficients. Liu X et al. constructed mathematical surrogate models between warpage, volumetric shrinkage, and process parameters using a GA-optimized extreme learning machine (ELM) model. They then applied a multi-objective firefly algorithm combined with the Grammar Analysis-Topsis (GRA-TOPSIS) multi-objective decision-making method to determine the optimal process parameter solution.

[0004] Since plastic parts usually have more than one defect, multi-objective optimization is required. Current research on multi-objective optimization usually transforms multi-objective problems into single-objective problems through the idea of ​​weighted combination. Commonly used methods include grey relational analysis (GRA), topology of nearest ideal solutions (TOPSIS), entropy weight, and fuzzy comprehensive evaluation (FCE). However, from the perspective of multiple objectives, all objectives are usually mutually constrained, and improving one objective often comes at the expense of other objectives. Therefore, for a multi-objective optimization problem, a large number of non-dominated solutions (Pareto solutions) are usually generated. Currently, in the study of multi-objective optimization of injection molding process parameters, researchers often determine the final Pareto optimal solution based on engineering experience and repeated experiments. Therefore, for the optimization of injection molding process parameters of GFRP plastic parts, how to determine the optimal trade-off solution to achieve the best comprehensive goal is still a difficult problem in the industry. Summary of the Invention

[0005] In response to the above problems, the present invention provides a method for optimizing the molding quality of glass fiber injection molded products, which can effectively reduce the warping deformation and volume shrinkage of GFRP injection-molded parts, thereby obtaining high-quality GFRP plastic products. It can provide a theoretical basis and data support for obtaining the optimal combination of process parameters for GFRP injection molding.

[0006] To achieve the above objectives, the present method for optimizing the molding quality of glass fiber injection molded products specifically comprises the following steps:

[0007] Step 1: Take warpage and volume shrinkage as optimization targets, select process parameters that affect these two targets as optimization design variables, determine the value range of process parameters based on the recommended process parameter values ​​of GFRP materials, and use the value range of each process parameter as the experimental design space;

[0008] Step 2: Design a CCF test based on the value range of each process parameter, and perform injection molding simulation based on Moldflow to obtain the quality optimization target test results;

[0009] Step 3. Based on the simulation test results, the LSO is improved by using the Tent-Logistic-Cosine chaotic map initialization method to replace the random generation initialization method in the LSO to generate the ILSO. The ILSO is then used to optimize the three parameters in the RF model: the number of decision trees, the minimum number of samples at leaf nodes, and the number of randomly selected features at nodes, to construct the ILSO-RF prediction model.

[0010] Step 4: Improve NSWOA by using four strategies: good point set initialization, nonlinear inertia weight, adaptive spiral update based on Levy flight, and improved convergence factor to generate MSINSWOA. Apply MSINSWOA to the ILSO-RF model to obtain the Pareto solution set of the quality optimization objective.

[0011] Step 5: Determine the optimal combination of process parameters from the Pareto solution set through the Critic method.

[0012] Furthermore, in Step 3, the mathematical definition of the Tent-Logistic-Cosine chaotic map is as follows:

[0013]

[0014] Where: X i is the current solution; r is the chaos mapping parameter, which is 0.7.

[0015] Furthermore, in Step 4, when the nonlinear inertia weight strategy is used to improve NSWOA, the expression of the nonlinear inertia weight is as follows:

[0016]

[0017] Where: w is the nonlinear inertia weight; t is the current iteration number; T max is the maximum number of iterations; k is the adjustment coefficient;

[0018] The position update formula of NSWOA is:

[0019]

[0020] Where: is the position vector of the current optimal solution in the whale individual; w is the nonlinear inertia weight; is the position vector of the current solution in the whale; is the distance between the whale and its prey; is the coefficient vector; is the distance between the current solution and the current optimal solution; b is the spiral shape parameter, which is used to limit the logarithmic spiral shape; l is the spiral development parameter, which is a random number in [-1,1];

[0021] The definition is as follows:

[0022]

[0023] Where: is a random number between [0,1], is the convergence factor;

[0024] The definition is as follows:

[0025]

[0026] Where: t is the current iteration number, T max is the maximum number of iterations.

[0027] Furthermore, the value of the adjustment coefficient k is 2.2.

[0028] Furthermore, in Step 4, when the adaptive spiral update strategy based on Levy flight is used to improve NSWOA, the spiral shape parameter b of NSWOA is set to a dynamic value that changes with the number of iterations. The improved spiral shape parameter b' is as follows:

[0029]

[0030] Where: b' is the improved spiral shape parameter; λ is the spiral update coefficient; t is the current iteration number; T max is the maximum number of iterations; b is the original spiral shape parameter;

[0031] The specific calculation of Levy flight is as follows:

[0032]

[0033] σ v =1

[0034] Where: L v is the Levy flight step length; u is a random number that satisfies the normal distribution v is a random number that satisfies the normal distribution β is the probability coefficient, which is a number between [0,2];

[0035] Use Levy flight step length L v Instead of the spiral development parameter l of NSWOA, b' and L are introduced v Finally, the spiral update position formula of NSWOA is as follows:

[0036]

[0037] Where: w is the nonlinear inertia weight; b' is the improved spiral shape parameter; L v Levy flight step length.

[0038] Furthermore, the value of β is 1.5.

[0039] Furthermore, in Step 4, when the improved convergence factor strategy is used to improve NSWOA, the improved convergence factor a is as follows:

[0040]

[0041] Where: t is the current iteration number, T max is the maximum number of iterations.

[0042] Compared with the existing technology, the present glass fiber injection molding product molding quality optimization method adopts the Tent-Logistic-Cosine chaotic map initialization method instead of the random initialization method in LSO to improve the LSO to generate ILSO, which can make the LSO produce better fitness. Using ILSO to scientifically solve the three parameters of the RF model, namely the number of decision trees, the minimum number of samples at leaf nodes, and the number of randomly selected features at nodes, the prediction accuracy of the RF model can be effectively guaranteed, and the constructed ILSO-RF prediction model has better prediction ability. The NSWOA is improved by adopting four strategies: optimal point set initialization, nonlinear inertia weight, adaptive spiral update based on Levy flight, and improved convergence factor, which can produce a Pareto solution set with better uniformity. The critic method is used to determine the optimal combination of injection molding process parameters. Simulation verification shows that the ILSO-RF model has a good prediction accuracy for the prediction error of warpage deformation and volume shrinkage within 5%. In addition, the warpage deformation and volume shrinkage of the glass fiber injection molding products after optimization are significantly reduced compared with the simulation results before optimization. This provides a theoretical basis and data support for obtaining the optimal combination of process parameters for GFRP injection molding. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is an injection molding simulation model of the bracket plastic parts in the electric control cabinet;

[0044] Figure 2 This is a comparison diagram of the fitness curves of the LSO-RF model and the ILSO-RF model;

[0045] Figure 3 This is a comparison diagram of the initialization of the population generated by initialization of the good point set and the population generated by random initialization;

[0046] Figure 4 is a graph showing the nonlinear inertia weight curve;

[0047] Figure 5 is a graph showing the change of the improved spiral shape parameter b' with the number of iterations when λ takes different values;

[0048] Figure 6 is a graph showing the convergence factor curve;

[0049] Figure 7 This is a comparison chart of the warpage deformation and volume shrinkage of the plastic parts of the bracket in the electric control cabinet before and after optimization. DETAILED DESCRIPTION

[0050] This glass fiber injection molding product molding quality optimization method takes the warpage deformation and volume shrinkage of glass fiber injection molding products as the quality optimization targets. First, the test scheme results are obtained based on CCF experimental design and injection molding simulation. Then, the ILSO-RF method is used to construct the injection molding quality optimization model. Then, MSINSWOA is used for multi-objective optimization, and the critic method is combined to determine the optimal combination of process parameters.

[0051] The present invention is described in detail below by taking a GFRP injection-molded bracket plastic part of an electric control cabinet as an example.

[0052] The geometric model of the plastic part of the bracket in the electric control cabinet is as follows: Figure 1 As shown in (a), the plastic part has an overall size of 135mm×74mm×38mm, an average wall thickness of 5mm, and is made of PA6+GF30%. The 3D model of the plastic part is imported into Moldflow software for dual-domain meshing. After meshing is completed, the following is established: Figure 1 (b) The mold flow analysis model including the pouring system and cooling system.

[0053] Step 1: Determine the quality optimization goal, select the injection molding process parameters and their value ranges, and use the value ranges of each process parameter as the experimental design space.

[0054] Warpage (y1; mm) and volume shrinkage (y2; %) were selected as optimization targets. Six process parameters influencing these two targets were selected as design variables for optimization: melt temperature (A; °C), mold temperature (B; °C), holding pressure (C; MPa), holding time (D; s), cooling time (E; s), and injection time (F; s). Based on the recommended process parameter values ​​for PA6+GF30% products, the experimental factor ranges and levels are shown in Table 1.

[0055] Table 1 Experimental factor variables and levels

[0056]

[0057] Step 2: Design a CCF test based on the value range of each process parameter, and perform injection molding simulation based on Moldflow to obtain the quality optimization target test results.

[0058] As a type of CCD, the CCF method has only three levels for each experimental variable. It is the simplest and fastest central composite experimental design method. It is also not easily invalidated by error sources. Therefore, the CCF method is used for experimental design.

[0059] Using Minitab software, 90 CCF test plans were designed. Each plan was then simulated in Moldflow software for injection molding, yielding the corresponding warpage y1 and volumetric shrinkage y2. The test plans and simulation results are shown in Table 2.

[0060] Table 2 Experimental scheme and simulation results

[0061]

[0062]

[0063] Step 3: Construct the ILSO-RF prediction model based on the simulation test results.

[0064] Random Forest (RF) is an ensemble learning method that constructs and combines multiple decision trees to perform prediction or classification tasks. Practical applications have shown that random forests are widely used in many fields due to their excellent performance in handling complex nonlinear relationships between prediction values ​​and response values. Therefore, this paper uses the RF model as a foundation to establish a mathematical proxy model for injection molding quality optimization.

[0065] The Light Spectral Optimization (LSO) algorithm is a physics-based meta-heuristic optimization algorithm proposed in 2022 by Abdel-Basset M et al., inspired by the physical phenomenon of sunlight dispersion when passing through raindrops. The LSO algorithm works as follows:

[0066] ①Initialization.

[0067] ② Produce three types of light: internal refraction light, internal reflection light and external refraction light.

[0068] ③Generate new colored light based on the three types of light.

[0069] ④ Scatter the light in three directions: the current best solution, the current solution, and a random solution from the current group as candidate solutions.

[0070] ⑤Evaluate and update the candidate solutions.

[0071] ⑥When the termination condition is reached, output the optimal solution.

[0072] In order to enable LSO to further generate a better initial solution, the present invention adopts the improved spectral optimization algorithm (ILSO) and uses the Tent-Logistic-Cosine chaotic map initialization method to replace the random generation initialization method in LSO. The mathematical definition of the Tent-Logistic-Cosine chaotic map is as follows:

[0073]

[0074] Where: X i is the current solution; r is the chaos mapping parameter, which is 0.7.

[0075] Based on the RF model, injection molding quality optimization models were established for the relationship between process parameters and warpage and volumetric shrinkage. To verify the effectiveness of ILSO for achieving optimal prediction accuracy, LSO and ILSO were used to optimize the number of decision trees, the minimum number of leaf node samples, and the number of randomly selected features at nodes in the RF model. The optimal parameter ranges are shown in Table 3.

[0076] Table 3 Optimization parameter range

[0077]

[0078] 72 sets of data from the 90 experimental schemes were used as training sets to train the model, and 18 sets of data were used as test sets to test the model accuracy. The relative error between the prediction value of the LSO-RF and ILSO-RF models and the test set was used as the fitness function of LSO and ILSO. The initial population size was set to 100, and the maximum number of iterations was set to 500. The optimization results of the LSO-RF and ILSO-RF models are shown in Table 4. The fitness curve comparison results are shown in Figure 2 As shown. Figure 2 As can be seen from (a), when the iteration is in the middle and late stages, the fitness generated by LSO is 0.0128, and the fitness generated by ILSO is 0.0078; Figure 2 As can be seen in (b), in the middle and late stages of the iteration, the fitness generated by LSO is 0.0127, and the fitness generated by ILSO is 0.0114. Therefore, ILSO can produce better fitness than LSO.

[0079] Table 4 Parameter optimization results

[0080]

[0081] To compare the prediction capabilities of the RF, LSO-RF, and ILSO-RF models, the root mean square error (RMSE), mean square error (MSE), mean relative error (MRE), and mean absolute error (MAE) were selected as evaluation metrics for the prediction accuracy of the RF, LSO-RF, and ILSO-RF models. The prediction accuracy errors for warpage and volumetric shrinkage of the RF, LSO-RF, and ILSO models are shown in Tables 5 and 6. As can be seen from Tables 5 and 6, the ILSO-RF model's prediction accuracy errors for both warpage and volumetric shrinkage are smaller than those of the RF and LSO-RF models, indicating that the ILSO-RF model has superior prediction capabilities.

[0082] Table 5 Prediction accuracy error of warpage

[0083]

[0084]

[0085] Table 6 Prediction accuracy error of volume shrinkage

[0086]

[0087] Step 4. In order to make the distribution of Pareto solution sets more uniform, multiple strategies are used to improve NSWOA to generate MSINSWOA. MSINSWOA is applied to the ILSO-RF model to obtain the Pareto solution set of the quality optimization objective.

[0088] The Whale Algorithm (WOA) is a swarm intelligence algorithm proposed by Mirjalili et al. in 2016, inspired by the hunting behavior of humpback whales. WOA searches for the optimal solution through three methods: surrounding prey, attacking with a bubble net, and searching for prey. The details are as follows:

[0089] ① Surround the prey

[0090] In the initial stages of hunting, since the optimal whale's position in the search space is unknown, the target prey is assumed to be the current optimal solution. The whales then encircle and contract the prey and update their group position. The mathematical model is as follows.

[0091]

[0092] Where: is the distance between the whale and its prey; is the position vector of the current optimal solution among individual whales; is the position vector of the current solution in the whale; is a coefficient vector, defined as follows:

[0093]

[0094] Where: is a random number between [0,1]; is the convergence factor, which decreases linearly from 2 to 0 and is defined as follows:

[0095]

[0096] Where: t is the current iteration number, T max is the maximum number of iterations.

[0097] ②Bubble Net Attack

[0098] In order to simulate the bubble net attack behavior of humpback whales using mathematical models, Mirjalili et al. designed two mechanisms: shrinking encirclement and spiral position update.

[0099] Shrinkage and encirclement mechanism: This behavior is achieved through the convergence factor a.

[0100] Spiral position update mechanism: The spiral equation is used to simulate the hunting behavior of humpback whales spitting out bubble nets along a spiral line. The mathematical model is as follows.

[0101]

[0102] Where: is the distance between the current solution and the current optimal solution; b is the spiral shape parameter, which is used to limit the logarithmic spiral shape; l is the spiral development parameter, which is a random number in [-1,1].

[0103] When hunting, humpback whales not only approach their prey in a spiral manner but also perform a contraction-encirclement behavior. Therefore, to achieve these two operations, it is necessary to select a probability coefficient p for the contraction-encirclement mechanism and the spiral position update. The mathematical model is as follows.

[0104]

[0105] Where: p is the probability coefficient.

[0106] ③Searching for prey

[0107] During the prey search phase, individual whales no longer update their positions based on the current optimal individual, but instead randomly select an individual from the current population as the target for position update. The mathematical model is as follows.

[0108]

[0109] Where: is the distance between the whale and its prey; is a whale individual randomly selected from the current whale group.

[0110] Because the WOA algorithm cannot handle multi-objective optimization problems, Pradeep Jangir and Narottam Jangir introduced multi-objective optimization theory and proposed the non-dominated sorting multi-objective whale optimization algorithm (NSWOA) in 2017, using the WOA framework. NSWOA uses non-dominated rankings to sort and select whales, adding non-dominated sorting to the WOA framework. Through crowding calculation and an elite retention strategy, it selects outstanding non-dominated individuals to guide the population and evolve it.

[0111] In order to make the distribution of Pareto solutions more uniform, the present invention adopts multiple strategies to improve NSWOA. The specific strategies are as follows:

[0112] ① Initialization of good point set

[0113] NSWOA uses random generation to initialize the population of whale individuals, which can easily lead to repeated positions of whale individuals during the distribution process, and the distance between individuals is too far or too close. The uneven distribution of the population will reduce the quality of the global optimal solution. Compared with the random production of whale individuals, good point set initialization is an effective uniform point selection method, which helps the initialized population generated by NSWOA to have better uniformity. In order to verify the uniformity of good point set initialization, the initial number is set to 100, and the population is generated in two-dimensional space using good point set initialization and random initialization respectively. The comparison results are shown in the figure below. Figure 3 As shown in the figure, it can be observed that the population generated by initializing with the good point set is more evenly distributed, which is more conducive to optimizing the convergence speed of NSWOA and preventing the algorithm from falling into local optimality.

[0114] ② Nonlinear inertia weight

[0115] Inertia weight is a key parameter in metaheuristic algorithms. In NSWOA, using a constant inertia weight of 1 results in inefficient algorithms and, to a certain extent, limits their search capabilities. Because a larger inertia weight favors global search, while a smaller inertia weight favors local search, this paper introduces a nonlinear inertia weight, as shown in the following equation, to balance NSWOA's global and local search capabilities.

[0116]

[0117] Where: w is the nonlinear inertia weight; t is the current iteration number; T max is the maximum number of iterations; k is the adjustment coefficient.

[0118] like Figure 4 As shown in FIG, when k takes different values, the nonlinear inertia weight w changes accordingly. It can be seen from the figure that when k is larger, the nonlinear inertia weight w decreases faster, and the nonlinear inertia weight is more inclined to guide the algorithm to perform local search development. Therefore, the present invention selects k = 2.2.

[0119] After introducing the nonlinear inertia weight, the position update formula of NSWOA is:

[0120]

[0121] ③ Adaptive spiral update based on Levy flight

[0122] In NSWOA, the spiral shape parameter b is a constant, forcing individual whales to follow a fixed spiral during the spiral search phase, which can lead to premature convergence. This method sets the spiral shape parameter b to a dynamic value that changes with the number of iterations. This allows whales to dynamically adjust the spiral shape during the spiral search phase, thereby enhancing the algorithm's global search capabilities and improving its convergence accuracy. The improved spiral shape parameter b' is shown in the following formula.

[0123]

[0124] Where: b' is the improved spiral shape parameter; t is the current iteration number; T max is the maximum number of iterations; λ is the spiral update coefficient; b is the original spiral shape parameter.

[0125] like Figure 5 As shown in the figure, when λ takes different values, the improved spiral shape parameter b' changes with the number of iterations. It can be seen that the larger λ is, the steeper the slope of the curve is, the larger the range of the spiral shape parameter is, and the stronger the global search ability of the algorithm is.

[0126] In NSWOA, the spiral development parameter l is a random number between [-1,1], and the development range is small, resulting in a weak global search capability of the algorithm. To improve this situation, the present invention uses the Levy flight step length L v Instead of the spiral development parameter l, Levy flight has a chance of large steps during the random walk, which can help the algorithm escape from the local optimum when it is trapped in the local optimum. The specific calculation principle of Levy flight is shown in the following formula.

[0127]

[0128] σ v =1

[0129] Where: L v is the Levy flight step length; u is a random number that satisfies the normal distribution v is a random number that satisfies the normal distribution β is a probability coefficient, which is a number between [0, 2] and is set to 1.5 in the present invention.

[0130] Introducing b', L v Finally, the spiral update position formula of NSWOA is shown as follows.

[0131]

[0132] ④ Improved convergence factor a

[0133] Variations in the convergence factor a affect NSWOA's global exploration capabilities and local exploitation. Using a strategy where a decreases linearly with the number of iterations can lead to slow convergence and trapping in local optima. The improved convergence factor a is shown in the following formula.

[0134]

[0135] like Figure 6 As shown in the figure, in the early stage of iteration, the value range of a is wider, and the range that can be used for exploration is also wider; in the later stage of iteration, a is smaller, which helps the algorithm to perform local development and speed up its convergence.

[0136] In order to compare the uniformity of the Pareto solution set distribution obtained by NSWOA and MSINSWOA, the spatial distribution index Δ is used for evaluation. Δ can be used to measure the uniformity of the solution set distribution. The smaller the value, the better the uniformity of the solution set distribution. The calculation formula is as follows:

[0137]

[0138] Where: S is the Pareto solution set obtained by the algorithm; d i is the Euclidean distance between two consecutive vectors on the non-dominated boundary of the solution set; d i Mean.

[0139] NSWOA and MSINSWOA were used to perform multi-objective optimization on the ILSO-RF model for warpage and volumetric shrinkage, respectively. During the optimization search, the population size and maximum number of iterations were set to 100, and all other parameters were kept at their default values. The spatial distribution index Δ obtained by NSWOA and MSINSWOA was 3.7204 and 3.5054, respectively. Compared to NSWOA, MSINSWOA achieved a smaller Δ index. Therefore, MSINSWOA was able to generate Pareto solutions with better uniformity than NSWOA.

[0140] Step 5: Determine the optimal combination of process parameters from the Pareto solution set through the Critic method.

[0141] The Critic method was used to comprehensively evaluate the Pareto solution set obtained by MSINSWOA. The objective weights for warpage and volumetric shrinkage were calculated to be 0.6086 and 0.3914, respectively. The analysis results are shown in the table below. As can be seen, the second set of solutions has the lowest comprehensive evaluation score, corresponding to a warpage of 0.3375 mm and a volumetric shrinkage of 7.266%.

[0142] Table 7 Analysis results of the Critic method

[0143]

[0144]

[0145] The optimized process parameter combination was substituted into Moldflow software for injection molding simulation verification, and compared with the analysis results of the process parameters (melt temperature: 230℃; mold temperature: 70℃; holding pressure: 40MPa; holding time: 20s; cooling time: 20s; injection time: 1s) set based on the actual production requirements of glass fiber injection molding products and the recommendations of Moldflow software. The comparison results are as follows: Figure 7 The warpage deformation value after optimization is 0.3468 mm, with an error of 2.68% from the model prediction value, which is a decrease of 33.55% compared with the value before optimization. The volume shrinkage rate after optimization is 6.979%, with an error of 4.11% from the model prediction value, which is a decrease of 17.28% compared with the value before optimization.

[0146] According to the comparison results, it can be seen that the optimal combination of process parameters obtained by the ILSO-RF-MSINSWOA method (melt temperature: 230°C; mold temperature: 91.89°C; holding pressure: 50MPa; holding time: 34.4s; cooling time: 20s; injection time: 4.72s) can achieve a simultaneous reduction in warpage deformation and volume shrinkage, and the optimization effect is good; at the same time, the model prediction errors of warpage deformation and volume shrinkage are both less than 5%, and the model prediction accuracy is also good.

Claims

1. A method for optimizing the molding quality of glass fiber injection molded products, characterized in that: The specific steps include: Step 1: Take the warpage deformation and volume shrinkage as optimization targets, select the process parameters that affect these two targets as optimization design variables, determine the value range of the process parameters according to the recommended values ​​of the process parameters of GFRP materials, and use the value range of each process parameter as the experimental design space; Step 2: Design a CCF test based on the value range of each process parameter, and perform injection molding simulation based on Moldflow to obtain the quality optimization target test results; Step 3. Based on the simulation test results, LSO is improved by using the Tent-Logistic-Cosine chaotic mapping initialization method to replace the random generation initialization method in LSO to generate ILSO. ILSO is used to optimize the three parameters of the number of decision trees, the minimum number of samples at leaf nodes, and the number of randomly selected features at nodes in the RF model to construct the ILSO-RF prediction model. The mathematical definition of Tent-Logistic-Cosine chaotic mapping is as follows: Where: X i is the current solution; r is the chaos mapping parameter, which is 0.7; Step 4: Use the four strategies of good point set initialization, nonlinear inertia weight, adaptive spiral update based on Levy flight, and improved convergence factor to improve NSWOA to generate MSINSWOA, and apply MSINSWOA on the ILSO-RF model to obtain the Pareto solution set of the quality optimization objective; Step 5: Determine the optimal combination of process parameters from the Pareto solution set through the Critic method.

2. The method for optimizing the molding quality of glass fiber injection molded products according to claim 1, characterized in that: In Step 4, when the nonlinear inertia weight strategy is used to improve NSWOA, the expression of the nonlinear inertia weight is as follows: Where: w is the nonlinear inertia weight; t is the current iteration number; T max is the maximum number of iterations; k is the adjustment coefficient; The position update formula of NSWOA is: Where: is the position vector of the current optimal solution in the individual whale; w is the nonlinear inertia weight; is the position vector of the current solution in the whale; is the distance between the whale and its prey; is the coefficient vector; is the distance between the current solution and the current optimal solution; b is the spiral shape parameter, which is used to limit the logarithmic spiral shape; l is the spiral development parameter, which is a random number in [-1,1]; The definition is as follows: Where: is a random number between [0,1], is the convergence factor; The definition is as follows: Where: t is the current iteration number, T max is the maximum number of iterations.

3. The method for optimizing the molding quality of glass fiber injection molded products according to claim 2, characterized in that: The value of the adjustment coefficient k is 2.

2.

4. The method for optimizing the molding quality of glass fiber injection molded products according to claim 2, characterized in that: In Step 4, when the adaptive spiral update strategy based on Levy flight is used to improve NSWOA, the spiral shape parameter b of NSWOA is set to a dynamic value that changes with the number of iterations. The improved spiral shape parameter b' is as follows: Where: b' is the improved spiral shape parameter; λ is the spiral update coefficient; t is the current iteration number; T max is the maximum number of iterations; b is the original spiral shape parameter; The specific calculation of Levy flight is as follows: s v =1 Where: L v is the Levy flight step length; u is a random number that satisfies the normal distribution v is a random number that satisfies a normal distribution β is the probability coefficient, which is a number between [0,2]; Use Levy flight step length L v Instead of the spiral development parameter l of NSWOA, b' and L are introduced v After that, the spiral update position formula of NSWOA is as follows: Where: w is the nonlinear inertia weight; b' is the improved spiral shape parameter; L v The length of the Levy flight step.

5. The method for optimizing the molding quality of glass fiber injection molded products according to claim 4, characterized in that: The value of β is 1.

5.

6. The method for optimizing the molding quality of glass fiber injection molded products according to claim 2, characterized in that: In Step 4, when the improved convergence factor strategy is used to improve NSWOA, the improved convergence factor a is as follows: Where: t is the current iteration number, T max is the maximum number of iterations.

Images