Warpage Optimization Method for Injection Molded Optical Products Based on Reynolds Number Regulation of Cooling Channels

Abstract

The present invention discloses a warpage optimization method for optical product injection molded parts based on the regulation of the Reynolds number of cooling channels. In mold flow analysis software, a mesh is divided for the injection molded part model, finite element models of the gate runner and the cooling channels are established, and process parameters and analysis sequences are set. Multiple groups of Reynolds numbers of each cooling channel are sampled as input parameters, and the average warpage amounts of multiple points on the injection molded part are obtained as output variables according to the mold flow analysis results of each group of input parameters. An input parameter and an output variable are used to construct and train a BP neural network until the maximum number of training times is reached or the predicted value reaches the desired target. The particle swarm optimization algorithm is used to obtain the input value corresponding to the minimum objective function of the trained BP neural network as the optimal input parameter to complete the optimization. The present invention can accurately predict the minimum warpage amount and the corresponding Reynolds number of the cooling channel, improves the optimization efficiency, and is applicable to a wide range of types of injection molded parts.

Description

Technical Field

[0001] The present invention relates to the field of injection molding of optical products, and particularly to a method for optimizing the warpage of injection molded parts of optical products based on the regulation of the Reynolds number of cooling channels. Background Art

[0002] Plastic optical products have wide applications in the fields of the electronics industry, the automotive industry, aerospace, etc. due to their advantages such as light weight, high precision, and good impact resistance. Users' requirements for the precision of injection molded parts are also increasing day by day. There are many indicators to measure the quality of optical injection molded parts. Among them, warpage deformation has a greater impact on the use performance of injection molded parts. During the injection and cooling processes of injection molding, the shrinkage degree of the injection molded part is different at each position, resulting in warpage, which seriously affects the shape and dimensional accuracy of optical injection molded products.

[0003] In traditional injection mold cooling channels, due to the complexity of the cavity structure, the relative position of the cooling channel to the cavity and the flow rate of cooling water are relatively fixed, making it difficult to provide uniform cooling during the holding pressure and cooling stages of the injection molded product, resulting in warpage. At present, some scholars have proposed methods such as conformal cooling channels, by changing the diameter and position of the cooling channel to obtain injection molded products with smaller warpage amounts. However, such methods need to customize the corresponding cooling channel for a single product, which is costly and has low versatility. And currently, the research mainly controls the warpage amount of the injection molded part through the water temperature of the cooling channel, and there is little research on the optimization of the warpage of injection molded parts related to the Reynolds number of the cooling channel. Summary of the Invention

[0004] Aiming at the problems in the prior art such as uneven cooling of injection molded products, high cost and poor versatility in customizing conformal cooling channels, the present invention proposes a method for optimizing the warpage of injection molded parts of optical products based on the regulation of the Reynolds number of cooling channels. Multiple cooling channels with adjustable Reynolds numbers are embedded in the mold. Without the need to customize according to the cavity shape, while realizing the function of injection molding, uniform cooling can be provided for the polymer in the mold through algorithm optimization.

[0005] The specific technical solution is as follows:

[0006] A method for optimizing the warpage of injection molded parts of optical products based on the regulation of the Reynolds number of cooling channels, comprising the following steps:

[0007] S1: Import the injection molded part model with the warpage amount to be optimized into mold flow analysis software. After meshing it, establish the finite element models of the gate runner and the cooling channel;

[0008] S2: Sample and select multiple groups of input parameters, and each group of input parameters includes the Reynolds number of each cooling channel;

[0009] S3: Set process parameters and analysis sequences in mold flow analysis software, perform 3D mesh finite element model analysis on each set of input parameters, and obtain the mold flow analysis results of each set of input parameters;

[0010] S4. Take multiple evenly distributed points on a plane perpendicular to the axis of the gate runner, generate a path passing through all points in sequence, obtain the warpage amount of each point according to the mold flow analysis results, take the absolute value of the warpage amounts of all points and then average them as the output variable;

[0011] S5: Construct and train a BP neural network. The input samples of the BP neural network are each set of input parameters, and the objective function is the output variable corresponding to the input parameters. If the training times are reached or the predicted value reaches the expected target, the training ends, and a trained BP neural network is obtained; the expected target is that the average relative error between the predicted value of the BP neural network and the objective function is less than or equal to the set threshold;

[0012] S6: Use the particle swarm optimization algorithm to find the input value corresponding to the minimum objective function of the trained BP neural network, that is, the Reynolds number array of the cooling channel corresponding to the minimum warpage amount, as the optimal input parameter to complete the optimization.

[0013] Further, in S6, the particle swarm optimization algorithm is specifically implemented through the following sub-steps:

[0014] S6.1: Set the initial position and initial velocity of each particle in the particle swarm;

[0015] S6.2. Particle fitness calculation: Take the position of the particle swarm as the input parameter and input it into the BP neural network trained in S5, and the obtained output value is the fitness value of the particle;

[0016] S6.3: For each particle, if the fitness value of its current position is less than the fitness value of its best position reached, update the current position to the individual best position, otherwise do not update;

[0017] For the particle swarm, if the fitness value of its current position is less than the fitness value of its best position reached, update the current position to the particle swarm best position, otherwise do not update;

[0018] S6.4: Update the velocity and position of each particle, and perform particle fitness calculation using the method of S7.2 based on the updated velocity and position, and use the method of S7.3 to find the individual extreme value and the group extreme value;

[0019] According to the velocity of the i-th particle during the n-th iteration, update the velocity during the (n + 1)-th iteration. The expression is as follows:

[0020]

[0021] where ω is the inertia factor, c 1 and c 2 are learning factors; pbest i represents the best position reached by the i-th particle, and gbest i represents the best position reached by the entire particle swarm; r 1 and r 2 are different random functions, with values in the range [0, 1];

[0022] Update the position corresponding to the movement in the (n + 1)-th iteration according to the position corresponding to the movement in the n-th iteration of the i-th particle The expression is as follows:

[0023]

[0024] S7.5: Determine whether the termination condition is satisfied. If so, take the position of the particle swarm corresponding to the current fitness as the best input parameter; if not, repeat S7.4.

[0025] Further, the termination condition is specifically: the number of iterations reaches the maximum number of iterations, or in consecutive multiple iterations, the change in the objective function value of the best position of the particle swarm is less than the set threshold.

[0026] Further, in S5, divide the input of the BP neural network into a training set, a validation set, and a prediction set. This step is specifically implemented through the following sub-steps:

[0027] (1) Train the BP neural network using the training set and verify the training effect using the validation set to obtain a preliminarily trained BP neural network;

[0028] (2) Use the preliminarily trained BP neural network to predict the prediction set to obtain a predicted output; at the same time, obtain the simulation data of the prediction set, that is, the actual output variables corresponding to each input sample; (3) Calculate the average relative error between the predicted output and the simulation data. If the average relative error is less than or equal to the set threshold, the desired goal is achieved; otherwise, update the weights and execute steps (1) to (3).

[0029] (3) Calculate the average relative error between the predicted output and the simulation data. If the average relative error is less than or equal to the set threshold, the desired goal is achieved; otherwise, update the weights and execute steps (1) to (3).

[0030] Further, in S5, the training parameters of the BP neural network are as follows: the number of nodes in the input layer is 4; there are 2 hidden layers, and the number of nodes in each layer is 15. The activation functions of the hidden layers are 'tansig' and 'logsig' respectively, and the training method is 'trainlm'; the number of nodes in the output layer is 1, the learning rate is 0.02, the maximum number of iterations is 1000, and the optimization performance target is 10 -12 , and the minimum performance gradient is 10 -12 , and the maximum number of allowed failures is 6 times.

[0031] Further, the mold flow analysis software selected is Moldflow.

[0032] Further, in S2, Latin hypercube sampling is used to select input parameters, and the Reynolds number range of each cooling channel is 4000 - 12000.

[0033] Further, in S3, the analysis sequence is "filling + holding pressure + cooling + warping"; the process parameters include: mold temperature, melt temperature, injection pressure, holding pressure, the time of filling + holding pressure + cooling, clamping force, injection rate.

[0034] Further, in S4, the plane perpendicular to the axis of the gate runner is taken as a rectangle, and the multiple evenly distributed points taken include: the four vertices of the rectangle, the midpoints of the four sides, and the center point of the rectangle.

[0035] The beneficial effects of the present invention are as follows:

[0036] The present invention constructs and trains a BP neural network with the Reynolds number array of the cooling channels as the input parameter and the warpage amount as the objective function, combines the particle swarm optimization algorithm, can accurately predict the minimum value of the warpage amount and the corresponding Reynolds number of the cooling channels, realizes the optimization of the Reynolds number array of the cooling channels, so that the warpage amount of the injection molded part prepared is minimized, the optimization efficiency is high, and the types of injection molded parts applicable are wide. Description of the Drawings

[0037] Figure 1 is a flowchart of the warpage optimization method for an optical product injection molded part based on the regulation of the Reynolds number of the cooling channels proposed in the embodiment of the present invention.

[0038] Figure 2 is a schematic diagram of the injection molded part model in the embodiment of the present invention.

[0039] Figure 3 is a schematic diagram of the finite element model established in Moldflow in the embodiment of the present invention.

[0040] Figure 4 is a line graph showing the change of the warpage amount with the path distance when the number of cooling channels is four in the embodiment of the present invention.

[0041] Figure 5 It is a schematic diagram of the warpage prediction value and simulation value curves of the prediction set in the embodiment of the present invention.

[0042] Figure 6 It is a flowchart of the BP neural network - particle swarm optimization algorithm (Error back propagation training neural network - Particle Swarm Optimization, hereinafter referred to as the BPNN - PSO algorithm) in the embodiment of the present invention. Specific implementation manners

[0043] The present invention will be described in detail below according to the accompanying drawings and preferred embodiments. The purpose and effect of the present invention will become more apparent. The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] As Figure 1 shown, a warpage optimization method for an optical product injection molded part based on regulating the Reynolds number of the cooling channels specifically includes the following steps:

[0045] S1: Import the injection molded part model with the warpage to be optimized into mold flow analysis software. After meshing it, establish a finite element model of the injection molded part and the gate runner and cooling channels around the injection molded part.

[0046] In this embodiment, the mold flow analysis software selects Moldflow. The injection molded part model imported into Moldflow is as Figure 2 shown. This injection molded part model is created by Solidworks and is a cuboid thin plate of 500mm * 500mm * 5mm. A cuboid boss of 200mm * 500mm * 5mm is added in the central area of the thin plate. Since the thickness of this injection molded part model is small, the warpage deformation amount of the final product in the Z direction is mainly considered. The established finite element model is as Figure 3 shown. In this embodiment, the number of channels of the established cooling channels is 4, which are evenly distributed on both sides of the injection molded part, and the number of channels of the gate runner is 1, which is arranged at the center of the upper surface of the injection molded part.

[0047] S2: Sample and select multiple groups of input parameters. Each group of input parameters is a cooling channel Reynolds number array composed of the Reynolds numbers of each cooling channel in sequence.

[0048] In this embodiment, 89 groups of input parameters are selected by Latin hypercube sampling. The input parameters obtained by this sampling method are uniform and representative. The Reynolds number of the cooling channel can be adjusted without re-customizing the size of the cooling channel. The value range of the Reynolds number is from 4000 to 12000. This is because when the Reynolds number of the pipeline is higher than 4000, the fluid is basically in a turbulent state. However, due to reasons such as the power limitation of the pump, the Reynolds number of the cooling channel generally does not exceed 12000.

[0049] S3: Set process parameters and analysis sequences in mold flow analysis software. Take each group of input parameters as input variables and perform 3D mesh finite element model analysis to obtain the mold flow analysis results of each group of input parameters (i.e., the overall warpage amount distribution on the plane perpendicular to the axis of the gate runner).

[0050] In this embodiment, the analysis sequence of Moldflow is "filling + holding pressure + cooling + warpage", and the process parameters include: mold temperature, melt temperature, injection pressure, holding pressure, time of filling + holding pressure + cooling, clamping force, injection rate, etc., as shown in Table 1 below.

[0051] Table 1 Injection molding process parameters

[0052] Process parameters Value Mold temperature (°C) 80 Melt temperature (°C) 220 Injection pressure (MPa) 180 Packing pressure (MPa) 160 Filling + packing + cooling time (s) 30 Clamping force (ton) 35 <![CDATA[Injection rate (mm·s -1 )]]> 25

[0053] S4: Generate a path diagram in the results of mold flow analysis. The specific operation method is: take a number of evenly distributed points on the plane perpendicular to the axis of the gate runner, use the path diagram generation function of mold flow analysis software to generate a path passing through all points in sequence, and count the warpage amount of each point according to the mold flow analysis results obtained in S3. Take the average value after taking the absolute value of the warpage amounts of all points as the output variable.

[0054] In this embodiment, the upper and lower surfaces of the injection molded part are parallel and are both planes perpendicular to the axis of the gate runner. Since there is a gate runner on the upper surface, in order to facilitate statistics, points are taken on the lower surface. The lower surface is rectangular, and a total of 9 evenly distributed points including the four vertices, the midpoints of the four sides, and the center point of the rectangle are taken. The input parameters of each group and the output variables obtained are shown in Table 2 below, and the average warpage amount of each point on the path is as Figure 4 shown.

[0055] Table 2 Moldflow simulation input and output results

[0056]

[0057]

[0058]

[0059] S5: Use each group of input parameters as the input samples of the BP neural network, and use their corresponding output variables as the objective function of the BP neural network. Divide them into a training set, a validation set, and a prediction set, and construct and train the BP neural network.

[0060] In this embodiment, the training set contains 60 groups of input samples and objective functions for training the neural network; the validation set includes 20 groups of input samples and objective functions to prevent overfitting of the neural network training; the prediction set includes 9 groups of input samples for detecting the rationality of the data predicted by the BP neural network.

[0061] The training parameters of the BP neural network are as follows: the number of nodes in the input layer is 4; there are 2 hidden layers, and the number of nodes in each layer is 15. The activation functions of the hidden layers are 'tansig' and 'logsig' respectively, and the training method is 'trainlm'; the number of nodes in the output layer is 1, the learning rate is 0.02, the maximum number of iterations is 1000, and the optimization performance target is 10 -12 and the minimum performance gradient is 10 -12 and the maximum number of allowed failures is 6 times.

[0062] S6. Determine whether the BP neural network trained in S5 reaches the expected goal: Generate a comparison curve graph of the data predicted by the BP neural network validation set and the simulation data (i.e., the output variables corresponding to the prediction set), as Figure 5 shown, and calculate the average relative error value between the two as the evaluation criterion for the training effect of the BP neural network; the closer the data predicted by the BP neural network is to the simulation data, that is, the smaller the average relative error value, the better the prediction effect. If the average relative error value exceeds the set threshold, the expected goal is not reached, then update the weights and repeat the steps of training the BP neural network in S5; otherwise, continue to use the trained BP neural network for the subsequent steps. In this embodiment, the set threshold is 10%.

[0063] S7. Use the Particle Swarm Optimization (PSO) algorithm to obtain the input value corresponding to the minimum objective function of the trained BP neural network, that is, the Reynolds number array of the cooling channel corresponding to the minimum warpage amount, as the optimal input parameter to complete the optimization. Denote the combination of the BP neural network and the PSO algorithm as the BPNN - PSO algorithm.

[0064] As Figure 6 shown, the PSO algorithm is specifically implemented through the following sub - steps:

[0065] S7.1. Initialize the particle swarm: Set the initial positions and initial velocities of each particle in the particle swarm. In this embodiment, the number of particles in the particle swarm is 10, the value range of the initial position of each particle is [4000, 12000], and the value range of the initial velocity of the particle is [-40, 40].

[0066] S7.2. Calculate the particle fitness: Take the positions of each particle as input parameters, input them into the BP neural network trained in S6 to obtain the output value, and this output value is the fitness value of the particle.

[0067] S7.3. Find the individual extreme value and the global extreme value: For each particle, compare the fitness value of its current position with the fitness value of the best position pbest i it has reached. If the fitness value of the current position is less than the fitness value of the best position pbest i it has reached, then take the current position as the individual best position, that is, the individual extreme value; compare the fitness value of the current position of this particle with the fitness value of the best position gbest i the entire particle swarm has reached. If the fitness value of the current position is less than the fitness value of the best position gbest i it has reached, then take the current position as the best position of the particle swarm, that is, the global extreme value.

[0068] S7.4. Update the velocity and position of each particle, and calculate the particle fitness using the method of S7.2 based on the updated velocity and position, and find the individual extreme value and the global extreme value using the method of S7.3.

[0069] Specifically, according to the velocity when the i-th particle travels in the n-th iteration, update the velocity when traveling in the (n + 1)-th iteration, and the expression is as follows:

[0070]

[0071] In the formula, ω is the inertia factor, c 1 , c 2 are the learning factors; pbest i represents the best position the i-th particle has reached, and gbest i represents the best position the entire particle swarm has reached; r 1 , r 2 are different random functions, and the value range is [0, 1]. In this embodiment, the inertia factor ω = 1, and the learning factors c 1 = c 2 = 1.

[0072] According to the position corresponding to when the i-th particle travels in the n-th iteration Update the corresponding position after progression in the (n + 1)-th iteration The expression is as follows:

[0073]

[0074] S7.5: Determine whether the termination condition is satisfied. If so, take the particle swarm position corresponding to the current fitness as the optimal input parameter; if not, repeat S7.4. The termination condition is specifically: the number of iterations reaches 100 times, or in 10 consecutive iterations, the change in the objective function value of the best position of the particle swarm is less than 0.01.

[0075] To prove the accuracy and effectiveness of the optimal input parameter optimized by the present invention, it is verified as follows: Use the optimal input parameter (optimal cooling channel Reynolds number array) obtained in S7 to perform 3D mesh finite element model analysis in S3, and then simulate the warpage amount through the method in S4. At the same time, take this optimal input parameter as the input value of the BP neural network trained in S6 to obtain the predicted output value; calculate the average value of the relative error between the warpage amount obtained by simulation and the predicted output value of the BP neural network. If the average value of the relative error is less than the set error threshold, it indicates that the selection of this optimal input parameter meets the requirements. In this embodiment, the warpage amount obtained by simulation is 2.01 mm, the predicted output value of the BP neural network is 1.901 mm, and the relative error is 5.42%. The results show that the optimization result of the BPNN-PSO algorithm is in good agreement with the Moldflow simulation result, and the trained BP neural network model can accurately fit and predict the output variable.

[0076] Those of ordinary skill in the art can understand that the above are only preferred examples of the invention and are not used to limit the invention. Although the invention has been described in detail with reference to the foregoing examples, for those skilled in the art, they can still modify the technical solutions described in the foregoing examples, or perform equivalent substitution on some of the technical features. Any modifications, equivalent substitutions, etc. made within the spirit and principle of the invention shall be included within the protection scope of the invention.

Claims

1. A warpage optimization method for injection molded optical products based on the control of the Reynolds number of cooling channels, characterized in that: The following steps are involved: S1: Import the injection molded part model whose warpage is to be optimized into the mold flow analysis software, divide it into grids, and then establish the finite element model of the gate runner and cooling runner; S2: sampling and selecting multiple groups of input parameters, each group of input parameters includes the Reynolds number of each cooling channel; S3: setting process parameters and analysis sequence in mold flow analysis software, performing 3D mesh finite element model analysis on each set of input parameters, and obtaining mold flow analysis results for each set of input parameters; S4. Take a plurality of evenly distributed points on a plane perpendicular to the axis of the gate runner, and generate a path passing through all the points in sequence, obtain the warpage amount of each point according to the mold flow analysis result, calculate the absolute value of the warpage amount of all the points and take the average as the output variable; S5: construct and train a BP neural network, wherein the input samples of the BP neural network are each group of input parameters, and the objective function is the output variable corresponding to the input parameter. If the number of training times is reached or the predicted value reaches the expected target, the training is terminated to obtain a trained BP neural network; the expected target is that the average relative error between the predicted value of the BP neural network and the objective function is less than or equal to a set threshold; S6: Use the particle swarm optimization algorithm to obtain the input value corresponding to the minimum objective function of the trained BP neural network, that is, the cooling channel Reynolds array corresponding to the minimum warpage, as the optimal input parameter to complete the optimization.

2. The warpage optimization method for injection molded optical products based on cooling channel Reynolds number control according to claim 1, characterized in that: In S6, the particle swarm optimization algorithm is specifically implemented through the following sub-steps: S6.1: Set the initial position and initial velocity of each particle in the particle swarm; S6.2, particle fitness calculation: the position of the particle group is used as an input parameter and input into the BP neural network trained in S5, and the output value obtained is the fitness value of the particle; S6.

3. For each particle, if the fitness value of its current position is less than the fitness value of the best position it has reached, the current position is updated to the individual best position, otherwise it is not updated; For a particle swarm, if the fitness value of its current position is less than the fitness value of the best position it has reached, the current position will be updated to the best position of the particle swarm, otherwise it will not be updated; S6.4, update the speed and position of each particle, and calculate the particle fitness based on the updated speed and position using the method of S7.2, and find the individual extreme value and the group extreme value using the method of S7.3; According to the speed of the i-th particle in the n-th iteration Update the speed during the n+1th iteration The expression is as follows: In the formula, ω is the inertia factor, c1 and c2 are learning factors; pbest i represents the best position that the i-th particle has reached, gbest i Indicates the best position reached by the entire particle swarm; r1 and r2 are different random functions with a value range of [0,1]; According to the position of the i-th particle after traveling in the n-th iteration Update the corresponding position after traveling in the n+1th iteration The expression is as follows: S7.5: Determine whether the termination condition is met. If so, the particle swarm position corresponding to the current fitness is used as the optimal input parameter; if not, repeat S7.

4.

3. The warpage optimization method for injection molded optical products based on cooling channel Reynolds number control according to claim 2, characterized in that: The termination condition is specifically: the number of iterations reaches the maximum number of iterations, or the change of the objective function value of the optimal position of the particle swarm is less than a set threshold value in multiple consecutive iterations.

4. The warpage optimization method for injection molded optical products based on cooling channel Reynolds number control according to claim 1, characterized in that: In S5, the input of the BP neural network is divided into a training set, a validation set, and a prediction set. This step is specifically implemented through the following sub-steps: (1) Use the training set to train the BP neural network and use the verification set to verify the training effect, and obtain the BP neural network that has been preliminarily trained; (2) Use the preliminarily trained BP neural network to predict the prediction set and obtain the prediction output; at the same time, obtain the simulation data of the prediction set, that is, the actual output variable corresponding to each input sample; (3) Calculate the average relative error between the predicted output and the simulation data. If the average relative error is less than or equal to a set threshold, the desired goal is achieved; Otherwise, update the weights and execute steps (1) to (3).

5. The warpage optimization method for optical product injection molded parts based on cooling channel Reynolds number control according to claim 1, characterized in that: In S5, the training parameters of the BP neural network are as follows: the number of input layer nodes is 4; there are 2 hidden layers, the number of nodes in each layer is 15, the hidden layer activation functions are 'tansig' and 'logsig' respectively, and the training method is 'trainlm'; the number of output layer nodes is 1, the learning rate is 0.02, the maximum number of iterations is 1000, and the optimization performance target is 10 -12 , the minimum performance gradient is 10 -12 The maximum number of allowed failures is 6.

6. The warpage optimization method for optical product injection molded parts based on cooling channel Reynolds number control according to claim 1, characterized in that: The mold flow analysis software is Moldflow.

7. The warpage optimization method for optical product injection molded parts based on cooling channel Reynolds number control according to claim 1, characterized in that: In S2, Latin hypercube sampling is used to select input parameters, and the Reynolds number of each cooling channel ranges from 4000 to 12000.

8. The warpage optimization method for optical product injection molding parts based on cooling channel Reynolds number control according to claim 1, characterized in that: In S3, the analysis sequence is "filling+holding pressure+cooling+warping"; the process parameters include: mold temperature, melt temperature, injection pressure, holding pressure, filling+holding pressure+cooling time, clamping force, and injection rate.

9. The warpage optimization method for injection molded optical products based on cooling channel Reynolds number control according to claim 1, characterized in that: In the above S4, a plane perpendicular to the axis of the gate runner is taken as a rectangle, and the multiple evenly distributed points taken include: four vertices of the rectangle, the midpoints of the four sides, and the center point of the rectangle.

Images