Intelligent extrusion molding process parameter setting method based on scientific mold testing

Abstract

The invention discloses an extrusion molding process parameter intelligent setting method based on scientific mold testing, and belongs to the technical field of plastic extrusion molding processing. A momentum optimization algorithm is introduced in the training process of a classic BP neural network, and after the neural network is verified to be reasonable, online process parameter adjustment is performed; a particle swarm algorithm is adopted to replace a traditional manual trial and error method, automatic and global intelligent optimization of process parameters is achieved, a process parameter combination is recommended by comparing the maximum pressure of an outlet section predicted by a neural network with the actual maximum pressure of the outlet section, the simulation efficiency is improved, real-time adjustment of the process parameters is achieved, and the process parameters are optimized. According to the method, in the actual industrial production period, analysis simulation and prediction can be conducted through computer data, remote calculation and online process parameter adjustment can be achieved only by uploading product and mold data to the cloud end by an operator, and good universality and convenience are achieved in industrial production.

Description

Technical Field

[0001] This invention belongs to the field of plastic extrusion molding technology, and particularly relates to an intelligent setting method for extrusion molding process parameters based on scientific mold trials. Background Technology

[0002] Extrusion molding mainly refers to a forming method in which heated and molten polymer materials are forced through a die by a screw or plunger to form a continuous profile with a constant shape. The extrusion molding process mainly includes feeding, melt plasticizing, extrusion molding, shaping, and cooling.

[0003] In the plastic extrusion molding process, the molding process is the most important factor determining the quality of plastic products. Adjusting and optimizing process parameters has a significant impact on obtaining extruded products with stable and high quality.

[0004] Currently, the main method for setting and optimizing extruder process parameters is the traditional trial-and-error approach, which involves repeatedly modifying parameters based on existing experience. In practice, product trial production is highly dependent on the operator's experience level; repeated trials are necessary to set the extrusion molding process parameters, resulting in poor control over complex processes, low production efficiency, low yield, poor repeatability, and difficulty in guaranteeing quality. At present, the main optimization methods for extrusion molding process parameters include: (1) Numerical simulation: such as using finite element analysis (FEA) software to perform simulation, although recommended process parameters can be obtained, there are also many disadvantages: low simulation calculation efficiency, lack of information on materials, products, molds, etc. (2) Experimental design: such as orthogonal experimental design. This method can greatly reduce the number of experiments, but it can only be limited to a fixed level. Moreover, it is cumbersome to operate in the actual production process, resulting in low efficiency.

[0005] Furthermore, the process parameters set and optimized using the above methods cannot be performed simultaneously on-site during mold trials, and the relationship between process parameters and product quality caused by differences in external environment and machine specifications cannot be effectively considered. Moreover, the process parameters obtained offline deviate significantly from actual parameters, making it difficult to provide sufficient reference and application value.

[0006] The main process parameters for extrusion molding include barrel temperature, die temperature, screw feed pressure, extruder head pressure, screw speed, feed rate, die geometry design, number and structure of extruder screws, cooling parameters, and die temperature.

[0007] Among them, die temperature, extrusion speed, traction speed, and cooling temperature constitute the four core control parameters of the extrusion molding process. Their importance stems from their control over the entire process of material transformation from a molten state to the final product.

[0008] First, the die temperature controls the viscoelasticity of the melt, which in turn determines the molding stability. An optimized die temperature window is the primary condition for achieving stable extrusion and ideal surface morphology.

[0009] Second, the extrusion speed determines the economics of the production process and the thermomechanical history of the material, and needs to be optimized in conjunction with other parameters.

[0010] Third, the traction speed enables active control of product dimensional accuracy and molecular orientation. Proper control can ensure dimensional accuracy while achieving directional optimization of product mechanical properties.

[0011] Fourth, the cooling temperature solidifies the microstructure and residual stress distribution of the final product. At the same time, the uniformity of cooling directly affects the distribution of residual stress. Uneven cooling can lead to product warping and deformation and long-term dimensional instability, which is a key factor affecting product reliability.

[0012] These four parameters together constitute a sophisticated process control system, enabling complete control over the extrusion process from four dimensions: rheological regulation, process intensity, structural orientation, and thermosetting. Their synergistic effect allows producers to precisely control the entire process from material plasticization to product shaping, ensuring that the final product meets design requirements in both microstructure and macroscopic properties. Summary of the Invention

[0013] The purpose of this invention is to address the problems of low efficiency, poor accuracy, and large offline parameter deviations in the setting of extrusion molding process parameters in existing technologies, which rely on manual experience. This invention proposes an intelligent setting method for extrusion molding process parameters based on scientific trial molding, achieving automated and global intelligent optimization of process parameters, improving setting efficiency and accuracy, reducing the number of trial moldings, and ensuring stable and reliable product quality.

[0014] To achieve the above objectives, the present invention adopts the following technical solution: A method for intelligent setting of extrusion molding process parameters based on scientific trial molding includes the following steps: S1. Obtain mold and product information, and extract the fluid domain of the melt; S2. Determine the initial range of die temperature, extrusion speed, traction speed and cooling temperature through scientific die testing; S3. Randomly change the four parameters of die temperature, extrusion speed, traction speed and cooling temperature within the initial range to perform multiple simulations and obtain the maximum pressure of the flow channel outlet section; S4. Generate a dataset from the simulation data and divide it into a training set and a validation set; S5. Construct a BP neural network with die temperature, extrusion speed, traction speed, and cooling temperature as input nodes and the maximum pressure at the outlet section as the output node; S6. Train the BP neural network using the training set, introduce the momentum optimization algorithm during the training process, and verify the rationality of the neural network through the validation set after training is completed; S7. The particle swarm optimization algorithm is used to adjust the four process parameters online within the initial range. The maximum pressure at the outlet section predicted by the neural network is compared with the actual maximum pressure at the outlet section. When the relative error meets the standard, the parameter combination is the recommended process parameter combination.

[0015] As a further description of the above technical solution:

[0016] The mold and product information mentioned in S1 includes the product flow length, the inner diameter, outer diameter and length of the die, the outer diameter and length of the mandrel, the shape, quantity, position and angle of the distributor, the mold gap, the length of the shaping mold and the cooling method; the fluid domain of the melt is extracted by the geometric Boolean operation method, that is, the difference Boolean operation is performed between the mold and the solid of the outermost dimension of the mold.

[0017] As a further description of the above technical solution:

[0018] The specific sub-steps for determining the initial range of each parameter in the scientific modeling process in S2 are as follows: a. Determining the initial range of die temperature: Based on the processing temperature recommended by the material manufacturer, and using the uniformity of the velocity cloud map distribution at the exit section as a metric, determine the initial range of die temperature; b. Determination of the initial range of extrusion speed: Under the condition that the die temperature is at the lower limit of the processing range and other parameters are reasonable, the initial range of extrusion speed is determined by taking the pressure cloud diagram of the outlet section as close as possible to the actual production requirements; c. Determining the initial range of traction speed: Keep the die temperature and extrusion speed constant, and use the uniformity of the velocity and pressure cloud map when the polymer melt leaves the die head as a measure to determine the initial range of traction speed; d. Determining the initial range of cooling temperature: Non-crystalline materials are rapidly cooled in a cooling water tank, while crystalline polymers are gradually cooled down, with the water temperature decreasing step by step until room temperature is reached.

[0019] As a further description of the above technical solution:

[0020] The BP neural network described in S5 includes an input layer, a hidden layer, and an output layer. The input layer nodes are: die temperature, extrusion speed, traction speed, and cooling temperature; the output layer node is: maximum pressure at the outlet section; the hidden layer is a single layer with 4 neurons; the input layer and the hidden layer, and the hidden layer and the output layer are connected by weights.

[0021] As a further description of the above technical solution:

[0022] The weight update formula for the momentum optimization algorithm described in S6 is:

[0023]

[0024]

[0025] in, This represents the momentum at the current moment. Momentum factor For learning rate, For the current gradient, This represents the current weight.

[0026] As a further description of the above technical solution:

[0027] The velocity and position update formulas for the particle swarm algorithm described in S7 are as follows:

[0028]

[0029]

[0030] in, and They represent particles respectively Velocity and position at time t For inertial weights, and As a learning factor, and It is a random number in the range [0,1].

[0031] As a further description of the above technical solution:

[0032] The relative error standard mentioned in S7 means that the relative error is less than 5%.

[0033] As a further description of the above technical solution:

[0034] The initial range of the die temperature is 160~200℃, the initial range of the extrusion speed is 2.5mm / s~7.5mm / s, and the traction speed is expressed as the traction ratio, with an initial range of 4~8.

[0035] In summary, due to the adoption of the above technical solutions, the beneficial effects of this invention are as follows: Firstly, by obtaining the initial range of process parameters through scientific trial molding, the data range used for subsequent neural network training is narrowed. Then, online simulation is performed, using the maximum pressure at the outlet section of the flow channel as a product quality indicator. Four parameters—die temperature, extrusion speed, traction speed, and cooling temperature—are randomly changed to obtain the maximum pressure at the outlet section. A dataset is generated to train and validate the constructed neural network. To improve the convergence speed and stability of the process parameter prediction model, this technology introduces a momentum optimization algorithm into the training process of the classic BP neural network. After the neural network is validated as reasonable, online process parameter adjustments are performed. A particle swarm optimization algorithm is used to replace the traditional manual trial-and-error method, achieving automated and globally intelligent optimization of process parameters. By comparing the maximum pressure at the outlet section predicted by the neural network with the actual maximum pressure at the outlet section, process parameter combinations are recommended, improving simulation efficiency and enabling real-time adjustment of process parameters. This ensures the safety and reliability of product quality and greatly reduces the number of actual trial molding operations. This invention can use computer data for analysis, simulation and prediction during the actual industrial production cycle. Operators only need to upload product and mold data to the cloud to realize remote calculation and online process parameter adjustment, which has good universality and convenience in industrial production. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the steps of an intelligent setting method for extrusion molding process parameters based on scientific mold trial proposed in this invention.

[0037] Figure 2 This is a schematic diagram of the PB neural network for an intelligent setting method of extrusion molding process parameters based on scientific mold trial proposed in this invention.

[0038] Figure 3 This is a schematic diagram of the PB neural network training and testing process for an intelligent setting method of extrusion molding process parameters based on scientific mold trial proposed in this invention.

[0039] Figure 4 This is a schematic diagram of the error curve of the PB neural network training data for an intelligent setting method of extrusion molding process parameters based on scientific trial molding proposed in this invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] In its specific implementation, such as Figures 1-4 This invention provides a technical solution: a method for intelligent setting of extrusion molding process parameters based on scientific mold trials, comprising the following steps: S1. Ansys Polyflow mold flow analysis software is used for mold identification. Within the software, the product structure body, the inner diameter, outer diameter and length of the die, the outer diameter and length of the mandrel, the shape, number, position and angle of the distributor, the length of the shaping mold and the cooling method are determined. The fluid domain of the melt is extracted using the geometric Boolean operation method. The difference Boolean operation is performed between the mold and the solid of the outermost dimension of the mold to extract the fluid domain of the melt. S2. Scientific mold trials were conducted using Ansys Polyflow mold flow analysis software to determine the mold head temperature. Extrusion speed traction speed Cooling temperature The initial range;

[0042] a. Determination of die temperature: The LLDPE material used in the experiment showed better extrusion molding performance in the temperature range of 160~200℃. The die temperature range was selected as 160~200℃ to investigate whether the velocity contour distribution at the exit section was uniform under different temperatures.

[0043] With the extrusion speed set at 5 mm / s and the traction ratio maintained at 5, the uniformity of the velocity cloud map at the exit section is good when the die temperature is 160℃. Therefore, the lower limit of the temperature range is set to 160℃.

[0044] With the extrusion speed set at 2.5 mm / s and the draw ratio maintained at 5, the uniformity of the velocity cloud map at the exit section is good when the die temperature is 200℃. Therefore, the upper limit of the temperature range is set to 200℃.

[0045] With the extrusion speed set at 2.5 mm / s and the traction ratio maintained at 5, the uniformity of the velocity cloud map at the exit section was good when the die temperature was 180℃. The initial temperature range was finally determined to be 160~200℃.

[0046] b. Determining the extrusion speed: During the extrusion process, the extrusion speed is mainly controlled by the screw speed. Changes in the screw speed will affect the flow rate of the polymer melt in the die head and the velocity distribution at the die exit and after demolding, thus affecting the product molding quality.

[0047] When the die head temperature is set to 180℃, the extrusion speed is 2.5mm / s, and the traction ratio is kept at 5, the pressure cloud diagram of the outlet section is close to the lowest value of the actual production pressure. Therefore, the lower limit of the extrusion speed range is set to 2.5mm / s.

[0048] When the die head temperature is set to 180℃, the extrusion speed is 10mm / s, and the traction ratio is kept at 5, the pressure cloud diagram at the outlet section exceeds the maximum value of the actual production pressure. Therefore, the extrusion speed is reduced.

[0049] When the die head temperature is set to 180℃, the extrusion speed is 7.5mm / s, and the traction ratio is kept at 5, the pressure cloud diagram of the outlet section is less than the maximum value of the actual production pressure. Therefore, the upper limit of the extrusion speed range is set to 7.5mm / s.

[0050] When the die head temperature is set to 180℃, the extrusion speed is 5mm / s, and the traction ratio is kept at 5, the uniformity of the pressure cloud map at the outlet section is good. Finally, the initial extrusion speed range is determined to be 2.5mm / s~7.5mm / s.

[0051] c. Determination of traction speed: After the die temperature and extrusion speed range are determined, the traction speed should be matched with the extrusion speed. The traction ratio is used to represent the traction speed. The simulation results of the process of determining the die temperature range and extrusion speed range show that when the traction ratio is maintained between 4 and 8, the velocity and pressure cloud map distribution at the outlet section is relatively uniform. Therefore, the initial range of the traction ratio is determined to be 4 to 8.

[0052] d. Determination of cooling temperature: Since the LLDPE material used in the experiment is a crystalline polymer, the problem of cooling crystallization needs to be considered. Therefore, a gradual cooling method is adopted. The cooling water tank is divided into sections. The water temperature starts from 75℃~85℃ when the plastic extrusion layer enters the first section of the water tank, and the water temperature is gradually reduced until room temperature.

[0053] S3. After determining the initial range of each parameter, simulation is performed by randomly changing four parameters: die temperature, extrusion speed, traction speed, and cooling temperature, to obtain the maximum pressure at the outlet section of the flow channel. The die temperature ranges from 160 to 200℃; the extrusion speed ranges from 2.5 mm / s to 7.5 mm / s; the traction ratio ranges from 4 to 8; and the cooling temperature of the first section of the segmented water tank ranges from 75℃ to 85℃.

[0054] S4. Generate a dataset from the simulation data and divide it into a training set and a validation set: A total of 200 simulations were performed, obtaining 200 sets of experimental data. 180 sets were used as the training set, and the remaining 20 sets were used as the validation set, as shown in the table below:

[0055] Number of process parameter groups Die temperature (°C) Extrusion speed (mm / s) Traction speed (traction ratio) Cooling temperature (°C) Maximum pressure at the outlet section (MPa) 1 170 3.6 7 85 21.11814 2 162 7.5 7 82 17.86918 3 188 4.8 8 77 14.2928 4 190 2.7 7 76 15.24824 5 190 4.2 7 82 29.5718 6 199 4.4 6 75 12.38857 7 198 3.1 5 76 14.22962 8 171 4.8 5 77 25.47076 9 180 2.5 4 85 19.29268 10 185 7.1 8 75 16.93787 11 165 4.6 4 83 14.46397 12 197 4.6 5 80 28.09217 13 197 7.4 8 75 12.67211 14 163 7.5 8 79 28.12944 15 198 3.7 5 85 18.79796 16 179 7.1 6 83 15.99933 17 163 2.8 7 76 21.22284 18 163 3.2 4 75 11.56657 19 188 7.4 4 85 18.50372 20 178 5 5 79 17.53307

[0056] S5, such as Figure 2As shown, an input layer, an output layer, and a hidden node layer are constructed. The input layer nodes are: die temperature, extrusion speed, traction speed, and cooling temperature; the output layer node is: maximum pressure at the outlet cross-section; the hidden layer is a single layer with 4 neurons. The output of the input node layer is connected to the input of the hidden layer; the output of the hidden node layer is connected to the input of the output layer; thus, a backpropagation (BP) neural network model for predicting the maximum pressure at the outlet cross-section during the extrusion process is obtained.

[0057] Backpropagation (BP) neural networks achieve nonlinear mapping through "forward propagation - error feedback - parameter iterative optimization":

[0058] The structure contains an input layer, hidden layers, and an output layer. Neurons are connected by weights, and the input signals are weighted and summed. With activation function The results are passed layer by layer to the output layer to generate prediction results.

[0059] Loss functions such as mean squared error To measure the prediction error, the error is backpropagated using the chain rule, and the gradient of the parameter weight W with the bias b is solved.

[0060] Gradient descent method ( Similarly, the parameters are iteratively adjusted to minimize the error, ultimately enabling the network to have generalized prediction capabilities.

[0061] S6, such as Figure 3 As shown, a training set array is used to train the BP neural network. A momentum optimization algorithm is introduced into the training process of the classic BP neural network to improve the convergence speed and stability of the process parameter prediction model. The classic BP neural network relies on the gradient direction of the current batch when updating weights. When the error surface is complex, it is prone to getting trapped in local minima and oscillations in flat regions, leading to a slow and unstable training process. The momentum optimization algorithm introduces a momentum factor into the weight update term, which not only considers the current gradient but also accumulates the exponentially weighted average of historical gradients during the update process, injecting "inertia" into the optimization process. This effectively accelerates the convergence speed of the network when the gradient direction is consistent and suppresses oscillations when the gradient direction changes frequently. By applying the momentum optimization algorithm to network training, the complex nonlinear relationship between process parameters and the maximum pressure at the outlet section can be mapped more quickly and accurately. The weight update formula of the momentum optimization algorithm is:

[0062]

[0063]

[0064] in, This represents the momentum at the current moment. Momentum factor For learning rate, For the current gradient, The current weight;

[0065] After training the network using 170 training datasets, the training error was obtained, and an error curve was plotted as mean squared error versus mean squared error. Figure 4 As shown in the error curve, the training error is 0.51 after 40 training iterations.

[0066] The neural network model was tested using a validation set. The predicted maximum pressure at the outlet section was compared with the actual maximum pressure at the outlet section. The results showed that the relative error between the prediction and the actual result was small within the training set, almost all below 10%, indicating that the neural network training results were relatively accurate. See the table below:

[0067] Number of process parameter groups Die temperature (°C) Extrusion speed (mm / s) Traction speed (traction ratio) Cooling temperature (°C) Predicted maximum pressure at the outlet section (MPa) Actual maximum pressure at the outlet section (MPa) Absolute error (MPa) Relative error (%) 1 166 5.5 4 79 28.2151 26.5834 1.6317 6.1 2 193 3.9 4 85 16.7515 18.6746 1.9231 10.3 3 177 5.5 5 80 12.3259 14.0953 1.7694 12.6 4 161 6.1 4 77 20.9981 23.2048 2.2067 9.5 5 168 3.8 8 82 13.7501 12.6535 1.0966 8.7 6 170 7.4 8 84 14.1075 15.4533 1.3458 8.7 7 198 6.5 6 75 16.1486 17.3434 1.1948 6.9 8 195 6.5 6 84 12.4544 11.4098 1.0446 9.2 9 189 6.4 6 81 16.1762 18.0535 1.8773 10.4 10 197 5.6 6 82 27.7024 30.9863 3.2839 10.6 11 161 3.6 6 78 14.8496 13.8644 0.9852 7.1 12 190 3.8 8 84 28.3946 24.8534 3.5412 14.2 13 178 4.9 6 82 10.7275 12.3423 1.6148 13.1 14 170 5.2 4 81 27.6368 30.8633 3.2265 10.5 15 166 3.9 5 80 10.3160 9.4323 0.8837 9.4 16 172 7.2 4 83 13.2357 11.8308 1.4049 11.9 17 183 6.5 4 82 16.4213 15.3321 1.0892 7.1 18 169 3 8 75 28.0502 30.8533 2.8031 9.1 19 186 7.4 8 85 12.6612 11.9843 0.6769 5.6 20 189 5.1 8 81 24.3368 27.9462 3.6094 12.9 21 199 6.3 7 78 20.0545 22.9401 2.8856 12.6 22 197 6.3 5 83 17.1438 15.3426 1.8012 11.7 23 163 2.5 4 85 22.6664 20.9861 1.6803 8.0 24 181 7.2 7 83 15.3123 16.9233 1.611 9.5 25 185 3.2 5 81 24.9926 27.4092 2.4166 8.8

[0068] S7. After obtaining a high-precision prediction model, a particle swarm optimization (PSO) algorithm is used to replace the traditional manual trial-and-error method. The system adjusts the die temperature, extrusion speed, traction speed, and cooling temperature online within the initial range. During algorithm initialization, a particle swarm is randomly generated within the feasible region of the process parameters. Each particle calculates its corresponding predicted pressure based on its position information using a trained BP neural network model, and its fitness value is the reciprocal of the difference between the predicted pressure and the actual target pressure. During the iteration process, each particle updates its velocity and position by tracking its own historical optimal solution and the global optimal solution of the swarm, ultimately searching for the optimal combination of process parameters that makes the outlet section pressure closest to the target value. The velocity and position update formulas for the particle swarm optimization algorithm are:

[0069]

[0070]

[0071] in, and They represent particles respectively Velocity and position at time t For inertial weights, and As a learning factor, and A random number within the range [0,1];

[0072] Online adjustments were made within the range of die temperature 160~200℃, extrusion speed 2.5mm / s~7.5mm / s, draw ratio 4~8, and first-stage cooling temperature 75℃~85℃. Particle swarm optimization was used to find the optimal parameters. It was found that when the parameters were 185℃, 5.5mm / s, 5 draw ratio, and 80℃ for the first stage cooling of the segmented water tank, the relative error between the maximum pressure at the outlet section predicted by the neural network and the actual maximum pressure at the outlet section was less than 5%. Therefore, this parameter combination was selected as the recommended process parameter combination.

Claims

1. A method for intelligent setting of extrusion molding process parameters based on scientific trial molding, characterized in that, Includes the following steps: S1. Obtain mold and product information, and extract the fluid domain of the melt; S2. Determine the initial range of die temperature, extrusion speed, traction speed and cooling temperature through scientific die testing; S3. Randomly change the four parameters of die temperature, extrusion speed, traction speed and cooling temperature within the initial range to perform multiple simulations and obtain the maximum pressure of the flow channel outlet section; S4. Generate a dataset from the simulation data and divide it into a training set and a validation set; S5. Construct a BP neural network with die temperature, extrusion speed, traction speed, and cooling temperature as input nodes and the maximum pressure at the outlet section as the output node; S6. Train the BP neural network using the training set, introduce the momentum optimization algorithm during the training process, and verify the rationality of the neural network through the validation set after training is completed; S7. The particle swarm optimization algorithm is used to adjust the four process parameters online within the initial range. The maximum pressure at the outlet section predicted by the neural network is compared with the actual maximum pressure at the outlet section. When the relative error meets the standard, the parameter combination is the recommended process parameter combination.

2. The intelligent setting method for extrusion molding process parameters based on scientific trial molding as described in claim 1, characterized in that, The mold and product information mentioned in S1 includes the product flow length, the inner diameter, outer diameter and length of the die, the outer diameter and length of the mandrel, the shape, quantity, position and angle of the distributor, the mold gap, the length of the shaping mold and the cooling method; the fluid domain of the melt is extracted by the geometric Boolean operation method, that is, the difference Boolean operation is performed between the mold and the solid of the outermost dimension of the mold.

3. The intelligent setting method for extrusion molding process parameters based on scientific mold trial as described in claim 1, characterized in that, The specific sub-steps for determining the initial range of each parameter in the scientific modeling process in S2 are as follows: a. Determining the initial range of die temperature: Based on the processing temperature recommended by the material manufacturer, and using the uniformity of the velocity cloud distribution at the exit section as a metric, determine the initial range of die temperature; b. Determination of the initial range of extrusion speed: Under the condition that the die temperature is at the lower limit of the processing range and other parameters are reasonable, the initial range of extrusion speed is determined by taking the pressure cloud diagram of the outlet section as close as possible to the actual production requirements; c. Determining the initial range of traction speed: Keep the die temperature and extrusion speed constant, and use the uniformity of the velocity and pressure cloud map when the polymer melt leaves the die head as a measure to determine the initial range of traction speed; d. Determining the initial range of cooling temperature: Non-crystalline materials are rapidly cooled in a cooling water tank, while crystalline polymers are gradually cooled down, with the water temperature decreasing step by step until room temperature is reached.

4. The intelligent setting method for extrusion molding process parameters based on scientific mold trial as described in claim 1, characterized in that, The BP neural network described in S5 includes an input layer, a hidden layer, and an output layer. The input layer nodes are: die temperature, extrusion speed, traction speed, and cooling temperature; the output layer node is: maximum pressure at the outlet section; the hidden layer is a single layer with 4 neurons; the input layer and the hidden layer, and the hidden layer and the output layer are connected by weights.

5. The intelligent setting method for extrusion molding process parameters based on scientific trial molding as described in claim 1, characterized in that, The weight update formula for the momentum optimization algorithm described in S6 is: ； in, This represents the momentum at the current moment. Momentum factor For learning rate, For the current gradient, This represents the current weight.

6. The intelligent setting method for extrusion molding process parameters based on scientific mold trial as described in claim 1, characterized in that, The velocity and position update formulas for the particle swarm algorithm described in S7 are as follows: ； in, and They represent particles respectively Velocity and position at time t For inertial weights, and As a learning factor, and It is a random number in the range [0,1].

7. The intelligent setting method for extrusion molding process parameters based on scientific mold trial according to claim 1, characterized in that, The relative error standard mentioned in S7 means that the relative error is less than 5%.

8. The intelligent setting method for extrusion molding process parameters based on scientific mold trial according to claim 1, characterized in that, The initial range of the die temperature is 160~200℃, the initial range of the extrusion speed is 2.5mm / s~7.5mm / s, and the traction speed is expressed as the traction ratio, with an initial range of 4~8.

Images