Silica gel formula optimization production quality control method for high-voltage silica gel line

Abstract

The invention discloses a production quality control method for optimizing a silica gel formula of a high-voltage silica gel line in the technical field of information, which comprises the following steps: determining the initial proportion of each additive in the silica gel formula according to the performance requirement of a silica gel material, and designing and optimizing the dosage of each additive through an orthogonal experiment to obtain an optimal formula with balanced performance; a support vector machine algorithm is adopted, a nonlinear mapping model between the silica gel formula and the silica gel performance is established, the comprehensive performance of the silica gel under different formulas is predicted through the model, and theoretical guidance is provided for formula optimization; constructing a knowledge base of silica gel formula-process-performance, summarizing performance change rules under different silica gel formulas and process conditions by applying a case-based reasoning technology according to historical production data, and guiding formula design and process optimization; a multi-objective optimization algorithm is adopted, the comprehensive performance of the silica gel is used as an optimization objective, formula composition and process parameters are used as optimization variables, an optimal formula and process combination is searched, and global optimization of the performance of the silica gel is achieved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to a method for optimizing the production quality control of the silicone formula of high-voltage silicone wires. Background Art

[0002] In the manufacture of high-voltage silicone wires, the silicone material formula and process control are complex technical problems. As an insulating material, the performance of silicone is crucial for product quality. However, the sensitivity of silicone to temperature and pressure poses great challenges to production.

[0003] When designing the silicone formula, the selection and proportion control of additives are particularly critical. On the one hand, there are a wide variety of additives, such as reinforcing fillers, coupling agents, flame retardants, etc. The effects of different additives on the performance of silicone are different, and even contradictory. For example, reinforcing fillers can improve the mechanical strength of silicone, but at the same time reduce its flexibility and insulation performance. Although flame retardants endow silicone with good flame retardancy, excessive use will cause the silicone to become brittle and accelerate aging. Therefore, how to balance the dosage of each additive, coordinate its influence on the performance of silicone, and achieve the optimization of performance is a technical difficulty. On the other hand, the silicone formula also needs to be adapted to the specific production process. During the silicone extrusion molding process, the control of extrusion speed, temperature, and pressure is crucial. If the extrusion speed is too fast, the silicone will flow unevenly, resulting in bubbles and defects; if the speed is too slow, the production efficiency will be affected. If the extrusion temperature is too high, the silicone will be over-vulcanized, becoming hard and brittle; if the temperature is too low, the silicone will be under-vulcanized, and the insulation performance will decline. The control of pressure also needs to be precise. Excessive pressure will deform the silicone, and too little pressure will cause the silicone to be not tightly combined with the conductor. In addition, the control of time and temperature in the silicone vulcanization process is also extremely critical. Insufficient or excessive vulcanization will seriously affect the performance of silicone. Therefore, how to optimize the extrusion and vulcanization process parameters according to the characteristics of the silicone formula, and coordinate the matching of the formula and the process is another technical problem. The high correlation between the silicone formula and the process brings many challenges to the manufacture of high-voltage silicone wires, and in-depth research and optimization are required in multiple aspects such as materials, processes, and equipment to finally achieve stable control of product performance. Summary of the Invention

[0004] The present invention provides a method for optimizing the production quality control of the silicone formula of high-voltage silicone wires, and the method includes the following steps:

[0005] Step S101, according to the performance requirements of the silicone material, determine the initial proportion of each additive in the silicone formula, and through orthogonal experimental design, optimize the dosage of each additive to obtain an optimal formula with balanced performance;

[0006] Step S102: Use the support vector machine algorithm to establish a non - linear mapping model between the silicone rubber formula and its properties. Through this model, predict the comprehensive properties of silicone rubber under different formulas to provide theoretical guidance for formula optimization;

[0007] Step S103: Obtain the rheological characteristic data of silicone rubber under different extrusion process parameters. Use data mining technology to analyze the influence rules of extrusion speed, temperature, and pressure on the uniformity and defect rate of silicone rubber, and determine the optimal process parameter combination;

[0008] Step S104: For the silicone rubber vulcanization process, establish a kinetic model to describe the influence of vulcanization time and temperature on the cross - link density of silicone rubber. If the degree of vulcanization deviates from the preset range, adjust the vulcanization process parameters in real - time to ensure the stable performance of silicone rubber;

[0009] Step S105: Install on - line detection equipment on the silicone rubber production line to monitor the performance indexes of the mechanical strength and insulation resistance of silicone rubber in real - time. If the detected value exceeds the preset threshold, adjust the formula and process parameters in time to control the product quality;

[0010] Step S105: Construct a knowledge base of silicone rubber formula - process - properties. Use case - based reasoning technology to summarize the change rules of properties under different silicone rubber formulas and process conditions according to historical production data to guide formula design and process optimization;

[0011] Step S107: Adopt a multi - objective optimization algorithm. Take the comprehensive properties of silicone rubber as the optimization goal and the formula composition and process parameters as the optimization variables to search for the optimal formula and process combination to achieve the global optimization of silicone rubber properties.

[0012] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0013] The present invention discloses a method for optimizing the production quality control of the silicone rubber formula for high - voltage silicone rubber wires. This method first optimizes the dosage of each additive in the silicone rubber formula through orthogonal experimental design, and then uses the support vector machine algorithm to establish a non - linear mapping model between the formula and the properties. During the production process, the present invention uses data mining technology to analyze the influence of extrusion process parameters on the properties of silicone rubber, establishes a kinetic model for the vulcanization process to adjust the process parameters in real - time, and monitors the key performance indexes through on - line detection equipment. In addition, the present invention constructs a formula - process - properties knowledge base and uses case - based reasoning technology to guide formula design and process optimization. Finally, a multi - objective optimization algorithm is used to search for the optimal formula and process combination. The present invention realizes the intelligent optimization of the silicone rubber formula and the quality control of the entire production process, effectively improving the performance stability and production efficiency of silicone rubber products. Brief Description of the Drawings

[0014] Figure 1It is a flowchart of the method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire of the present invention.

[0015] Figure 2 It is a schematic diagram of the method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire of the present invention.

[0016] Figure 3 It is another schematic diagram of the method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire of the present invention. Specific embodiments

[0017] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Such as Figures 1-3 , the method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire in this embodiment may specifically include:

[0019] Step S101, according to the performance requirements of the silicone material, determine the initial ratio of each additive in the silicone formula, and optimize the dosage of each additive through orthogonal experimental design to obtain an optimal formula with balanced performance.

[0020] According to the performance requirements of the silicone material, determine the key additives affecting the performance, and set the initial ratio range of each additive to construct an orthogonal experimental factor level table; use the orthogonal experimental design method to design multiple groups of experimental schemes according to the factor level table, and prepare silicone material samples according to the set additive ratio for each group; perform performance tests on each group of the silicone material samples to obtain data of various performance indicators, and establish a mathematical model between the performance indicators and the additive dosage; analyze the relationship between the dosage of each additive and the performance indicators through the mathematical model, determine the sensitivity coefficient and the optimal interval of the dosage of each additive; within the optimal interval of the dosage of each additive, use the response surface optimization or neural network algorithm to optimize the dosage ratio of each additive to obtain an optimal formula with the best comprehensive performance; verify the optimized optimal formula, and judge whether it meets the performance requirements of the silicone material by preparing samples and testing the performance; if the optimal formula meets the requirements, determine it as the final formula of the silicone material; if it does not meet the requirements, adjust the optimization range of the additive dosage and re-optimize until it meets the requirements.

[0021] Specifically, the properties of silicone materials can be adjusted by adding various additives, such as reinforcing agents, plasticizers, flame retardants, and colorants. Different additives affect different performance indicators, such as tensile strength, tear strength, hardness, heat resistance, and flame retardancy. Suppose it is necessary to improve the tensile strength and heat resistance of the silicone material. The reinforcing agent and heat-resistant agent are initially determined as the key additives. The reinforcing agent can be selected as fumed silica, and the heat-resistant agent can be selected as silicone resin. Based on experience and literature research, the addition range of fumed silica is set to 10 - 30 parts, and the addition range of silicone resin is set to 5 - 15 parts. Here, "parts" represents parts by mass, and the matrix silicone rubber is 100 parts. Construct an orthogonal test factor level table:

[0022]

[0023]

[0024] According to the 9 groups of test schemes designed by the orthogonal table, prepare silicone material samples respectively. For example, in the first group of samples, add 10 parts of fumed silica and 5 parts of silicone resin to 100 parts of the matrix silicone rubber, mix evenly and then carry out vulcanization molding. Test the tensile strength and heat resistance of each group of samples. The tensile strength test can be carried out according to the GB / T528 - 2009 standard, and the heat resistance test can use thermogravimetric analysis (TGA) to measure the weight loss of the sample at a certain temperature. The less the weight loss, the better the heat resistance.

[0025] Record the test data in the following table,

[0026] Test No. Tensile strength (MPa) Heat resistance (weight loss rate, %) 1 5 15 2 6 12 3 7 10 4 7 13 5 8 9 6 9 7 7 8 11 8 9 8 9 10 6

[0027] Establish a mathematical model between the performance indicators and the additive dosage using the data in the above table. The multiple linear regression method can be used to establish the model. For example: Tensile strength = a + b × dosage of silica + c × dosage of silicone resin, where a, b, and c are regression coefficients. By analyzing the mathematical model, the sensitivity coefficients of the dosages of each additive can be determined. For example, if the value of b is large, it indicates that the dosage of silica has a greater impact on the tensile strength. The optimal ranges of the dosages of each additive can also be determined. For example, the optimal range of the dosage of silica may be 20 - 30 parts, and the optimal range of the dosage of silicone resin may be 10 - 15 parts. Within the optimal ranges, the response surface method is used to optimize the additive dosage ratio. The response surface method can establish a three-dimensional response surface to visually display the relationship between the additive dosage and the performance indicators. By finding the highest point of the response surface, the formulation with the optimal comprehensive performance can be obtained. The neural network algorithm can also be selected for optimization. The neural network algorithm can handle more complex non-linear relationships but requires a large amount of training data. According to the optimization results of the response surface method or the neural network algorithm, an optimal formulation is determined, such as 25 parts of silica and 12 parts of silicone resin. Prepare a silicone material sample according to the optimal formulation and conduct performance tests. If the test results meet the performance requirements, it is determined as the final formulation. For example, if the required tensile strength is not less than 9 MPa and the heat resistance weight loss rate is not higher than 8%, and the test results are a tensile strength of 9.5 MPa and a weight loss rate of 7.5%, then the requirements are met. If the test results do not meet the performance requirements, the optimization range of the additive dosage needs to be adjusted and the optimization is carried out again. For example, if the heat resistance does not meet the requirements, the dosage range of the heat-resistant agent can be increased, such as 10 - 20 parts, and then the orthogonal experiment, modeling, and optimization are carried out again. This is to continuously approach the target performance and finally obtain a formulation that meets all performance requirements.

[0028] Step S102, use the support vector machine algorithm to establish a non-linear mapping model between the silicone formulation and the silicone performance, and predict the comprehensive performance of silicone under different formulations through this model to provide theoretical guidance for formulation optimization.

[0029] For the preprocessed dataset, use the support vector machine regression algorithm to construct the non-linear mapping model between the silicone formulation and the silicone performance, optimize the hyperparameters of the non-linear mapping model through cross-validation and grid search; according to the non-linear mapping model, input the silicone formulation parameters, predict the comprehensive performance indicators of silicone under this formulation, compare the predicted performance indicators with the actual performance indicators, if the error between the two is greater than the preset threshold, obtain more formulation-performance data, and retrain the non-linear mapping model; use the trained non-linear mapping model to predict the comprehensive performance of silicone under different formulations and obtain the combination of silicone formulation parameters with the optimal performance.

[0030] Specifically, data preprocessing is the key to building an accurate model. Suppose 100 sets of silicone formula data are collected, which include the dosages of three additives A, B, and C and the corresponding performance indicators of tensile strength, hardness, and heat resistance. First, the data needs to be cleaned, for example, removing outliers or missing values. For instance, if the tensile strength in a certain set of data far exceeds other data, it may be caused by measurement errors and needs to be removed. Then, data normalization is carried out to convert data with different dimensions to the same range, avoiding some indicators having too much influence on the model due to large numerical values. For example, the dosage range of additive A from 0 - 100g is normalized to 0 - 1, and the tensile strength from 10 - 100MPa is normalized to 0 - 1. The support vector machine regression algorithm can establish a non - linear mapping relationship between the silicone formula and performance. Taking tensile strength as an example, the dosages of the three additives A, B, and C can be used as input features, and the tensile strength as the output target to train a support vector machine regression model. The model will learn the complex relationship between the formula and tensile strength, even if this relationship is non - linear. Through cross - validation and grid search methods, the optimal hyperparameters of the model can be found, such as the type of kernel function, regularization parameter. For example, using 5 - fold cross - validation, the dataset is divided into 5 parts. Each time, 4 parts are used to train the model and 1 part is used to validate the model. Finally, the combination of hyperparameters with the best average performance on the validation set is selected. Using the trained support vector machine model, the performance of silicone under different formulas can be predicted. For example, when inputting 50g of additive A, 30g of additive B, and 20g of additive C, the model can predict that the tensile strength of the silicone under this formula is 60MPa, the hardness is 70HA, and the heat resistance is 150°C. Comparing the prediction results with the actual performance can evaluate the prediction accuracy of the model. If there is a large deviation between the predicted value and the actual value, for example, the predicted tensile strength is 60MPa while the actual value is only 40MPa, it indicates that the prediction accuracy of the model is not high enough and needs to be improved. The key to improving the model's prediction accuracy lies in the data. If it is found that the prediction error is large, more formula - performance data pairs need to be collected to retrain the model. For example, near the formula with a large prediction error, some new experimental data are supplemented and the model is retrained, which can improve the prediction accuracy of the model in this area. Continuous data accumulation and model training can continuously optimize the prediction performance of the model. Through the support vector machine model, the comprehensive performance of silicone under a large number of different formulas can be predicted, thus quickly screening out the formula with the best performance. For example, the goal can be set as the maximum tensile strength, moderate hardness, and relatively high heat resistance. Through model prediction, the best combination of formula parameters that meet these requirements can be found, such as 60g of additive A, 25g of additive B, and 15g of additive C. It is crucial to apply the optimal formula parameters predicted by the model to actual production for verification. For example, the predicted optimal formula parameters are input into the silicone production process flow to produce silicone samples and test their performance.If the test results are basically consistent with the model prediction results, it indicates that the formulated recipe predicted by the model is feasible and stable. If there are significant differences between the test results and the prediction results, it is necessary to analyze the reasons. It may be that the prediction accuracy of the model is insufficient, or there are some factors not considered in the production process. The optimization of the silicone recipe is an iterative improvement process. Through actual production verification, more recipe-performance data can be accumulated and used to improve the support vector machine model. For example, if it is found that the effects of certain recipes in actual production are not as expected, the reasons can be analyzed and this information can be fed back into the model to retrain the model and improve the prediction accuracy of the model. Through the closed-loop optimization process of recipe design-performance prediction-recipe optimization, the performance of silicone products can be continuously improved and the R & D cycle can be shortened.

[0031] Step S103: Obtain the rheological characteristic data of silicone under different extrusion process parameters, and use data mining techniques to analyze the influence laws of extrusion speed, temperature, and pressure on the uniformity and defect rate of silicone, and determine the optimal process parameter combination.

[0032] Obtain a silicone extrusion experiment dataset including extrusion speed, temperature, pressure, and corresponding uniformity index values and defect rates. Train a machine learning model based on the dataset, and determine the optimal combination of the extrusion speed, the temperature, and the pressure that maximizes the uniformity index value and minimizes the defect rate through a parameter search algorithm. Apply the optimal parameter combination to the silicone extrusion production process to obtain product quality data.

[0033] Specifically, the optimization of silicone extrusion process parameters is a complex problem, which involves the interaction between multiple parameters. The best parameter combination can be effectively found through machine learning methods, thereby improving product quality and production efficiency. Data is the basis of machine learning. First, it is necessary to collect silicone extrusion experimental data under different extrusion speeds, temperatures, and pressures. For example, different extrusion speeds (such as 50mm / s, 100mm / s, 150mm / s), different temperatures (such as 180℃, 190℃, 200℃), and different pressures (such as 1MPa, 2MPa, 3MPa) can be set, and the uniformity index value of silicone under each set of parameter combinations (for example, the standard deviation can be used to measure the uniformity of the cross-sectional diameter of the extrudate, and the smaller the standard deviation, the better the uniformity) and defect rate (for example, the percentage of bubbles and crack defects) are recorded. Assume that 27 experiments are conducted and a data set containing five columns of extrusion speed, temperature, pressure, uniformity index value, and defect rate is obtained. After obtaining the data, the data needs to be preprocessed. This includes processing missing values ​​and outliers. For example, if data is missing in an experiment due to equipment failure, it can be filled with the average value of adjacent experiments. If the data of an experiment deviates significantly from the normal range, it may be due to measurement errors or other abnormal conditions. It needs to be identified as an outlier and processed, such as being eliminated or replaced with a reasonable value. In addition, data standardization is also required. Due to the different dimensions of speed, temperature, and pressure, their numerical values ​​vary greatly, which will affect the training of the model. In order to eliminate the influence of the dimension, the data needs to be converted to the same dimension. The commonly used method is standardization, that is, converting the mean of each feature to 0 and the standard deviation to 1. For example, assuming that the original data range of the extrusion speed is 50 to 150, after standardization, the data range will be close to -1 to 1. Next, feature engineering is performed to extract features from the preprocessed data. In addition to the original speed, temperature, and pressure, some statistical features such as mean, variance, maximum, and minimum can also be extracted. Combined features can also be constructed, such as the product of speed and temperature and the ratio of speed to pressure. These features can better reflect the interaction between parameters and improve the prediction ability of the model. For example, the product of speed and temperature can be calculated. This feature may reflect the energy input during the extrusion process and have a certain impact on product quality. Then select a suitable machine learning model. Support vector regression, random forest regression, and gradient boosting regression are all commonly used regression models that can be used to predict the uniformity and defect rate of silicone. These models can capture nonlinear relationships and are suitable for dealing with complex process parameter optimization problems. For example, you can choose a support vector regression model, which has good generalization ability and can effectively avoid overfitting. After selecting the model, the model needs to be trained and evaluated. Input the prepared feature vector and the corresponding uniformity and defect rate label data into the model for training. The cross-validation method is used to evaluate the performance of the model.For example, the dataset can be divided into five parts. Each time, four parts of the data are used to train the model, and one part of the data is used for testing. This is repeated five times to finally obtain the average performance of the model. Select the model and parameters with the best performance. After the model is trained, the process parameters can be optimized. Using the trained model, through a parameter search algorithm, find the best combination of extrusion speed, temperature, and pressure that maximizes uniformity and minimizes the defect rate. For example, grid search can be used to try different parameter combinations within a certain range and select the best parameter combination according to the prediction results of the model. Assume that the model predicts that when the extrusion speed is 100 mm / s, the temperature is 190 °C, and the pressure is 2 MPa, the uniformity is the largest and the defect rate is the smallest. Finally, the obtained best parameter combination needs to be applied to the actual silicone extrusion production to verify the accuracy of the model prediction and the effectiveness of the best parameter combination. Collect product quality data and compare the actual results with the model prediction results. If the deviation between the prediction result and the actual result is large, it is necessary to return to the data preprocessing step, readjust the data preprocessing method or select other machine learning models. For example, if it is found in actual production that the defect rate is still high, it may be due to some unconsidered factors affecting the product quality, and it is necessary to collect data again and improve the model. Through the above seven steps, the best silicone extrusion process parameters can be effectively obtained, the product quality can be improved, and the production cost can be reduced.

[0034] Step S104: For the silicone vulcanization process, establish a kinetic model to describe the influence of vulcanization time and temperature on the crosslinking density of silicone. If the degree of vulcanization deviates from the preset range, adjust the vulcanization process parameters in real time to ensure the stable performance of silicone.

[0035] Obtain historical data of the silicone vulcanization process, where the historical data includes vulcanization temperature, vulcanization time, and crosslinking density; use a machine learning algorithm to train a vulcanization kinetic model based on the historical data to obtain the vulcanization kinetic model; obtain the vulcanization temperature and vulcanization time of the current silicone vulcanization process in real time; input the vulcanization temperature and vulcanization time into the vulcanization kinetic model to predict the crosslinking density of silicone under the current vulcanization conditions to obtain the predicted crosslinking density; obtain the preset target range of crosslinking density; determine whether the predicted crosslinking density is within the target range; if the predicted crosslinking density is not within the target range, trigger a vulcanization process parameter adjustment mechanism; use an optimization algorithm to determine the adjustment amplitude of the vulcanization temperature and time according to the deviation between the predicted crosslinking density and the target range to obtain the adjusted vulcanization temperature and the adjusted vulcanization time; apply the adjusted vulcanization temperature and the adjusted vulcanization time to the control system of the silicone vulcanization equipment; continuously monitor the adjusted vulcanization process to obtain real-time vulcanization temperature, vulcanization time, and crosslinking density data to obtain real-time vulcanization data; use the real-time vulcanization data to update the vulcanization kinetic model; regularly adjust the target range according to changes in production requirements.

[0036] Specifically, the silicone vulcanization process is a complex chemical reaction, and its ultimate goal is to achieve the desired crosslinking density by controlling the vulcanization temperature and time, thereby obtaining the required material properties. Machine learning can be used to build a vulcanization kinetics model to achieve precise control of the vulcanization process. First, a large amount of historical vulcanization data needs to be collected. This data should include the vulcanization temperature, vulcanization time, and the corresponding crosslinking density. For example, different vulcanization temperatures (such as 150°C, 160°C, 170°C) and vulcanization times (such as 30 minutes, 45 minutes, 60 minutes) can be set, and the final crosslinking density can be measured under each condition. Suppose 27 experiments are carried out to obtain a dataset containing three columns: vulcanization temperature, vulcanization time, and crosslinking density. After obtaining the data, a suitable machine learning algorithm can be selected to establish a vulcanization kinetics model. Support vector regression, random forest regression, and artificial neural networks are all commonly used regression models, which can capture non-linear relationships and are suitable for dealing with complex vulcanization kinetics problems. For example, a support vector regression model can be selected and the collected data can be used to train the model. The goal of training is to enable the model to accurately predict the crosslinking density at a given vulcanization temperature and time. After the model training is completed, it can be used in the actual vulcanization process for real-time prediction and control. For example, assume that the current vulcanization temperature is 165°C and the time is 40 minutes. These two parameters are input into the trained model, and the model will output a predicted crosslinking density value, such as 0.85. Then, the predicted crosslinking density needs to be compared with the preset target range. Suppose the target range is 0.8 to 0.9, then the current predicted value of 0.85 is within the target range. If the predicted value deviates from the target range, for example, the predicted value is 0.75, a dynamic adjustment mechanism needs to be activated. The core of the dynamic adjustment mechanism is to determine the adjustment amplitude of the vulcanization temperature and time through an optimization algorithm. For example, the gradient descent algorithm can be used to find the optimal adjustment amplitude so that the adjusted parameters can control the predicted crosslinking density within the target range. Suppose the result calculated by the algorithm is to increase the vulcanization temperature by 5°C and extend the time by 10 minutes. The optimized parameters (170°C, 50 minutes) are applied to the control system of the vulcanization equipment to achieve real-time adjustment. After adjustment, the vulcanization process needs to be continuously monitored, and new data, such as new vulcanization temperature, time, and crosslinking density, need to be collected. These data can be used to update and optimize the existing vulcanization kinetics model, thereby improving the prediction accuracy of the model. For example, the newly collected data can be added to the original dataset, and then the model can be retrained. Finally, the preset crosslinking density target range needs to be regularly evaluated and adjusted according to the changes in production requirements. For example, if the product requires higher hardness, the target range of crosslinking density needs to be increased. Through such a closed-loop control, it can be ensured that the silicone performance always meets the product quality requirements.This is like an experienced master who, through continuous observation, adjustment, and learning, finally masters the skill of controlling the firing temperature and can produce perfect ceramic works.

[0037] Step S105: Install an on-line detection device on the silicone production line to monitor the key performance indicators of the mechanical strength and insulation resistance of the silicone in real time. If the detected value exceeds the preset threshold, the formula and process parameters shall be adjusted in time to control the product quality.

[0038] Obtain the real-time data collected by the on-line detection device installed on the silicone production line, and monitor and analyze the key performance indicators of the mechanical strength and insulation resistance of the silicone; compare the detected values of the silicone performance indicators obtained from the monitoring and analysis with the preset threshold to determine whether they exceed the preset threshold range; if the detected value exceeds the preset threshold, trigger an early warning signal, analyze the historical production data using a machine learning algorithm, and obtain optimization adjustment suggestions for the silicone formula and process parameters; according to the optimization adjustment suggestions, automatically adjust the relevant parameter settings of the formula system and process control system of the silicone production line to achieve dynamic optimization of the formula and process parameters; continuously conduct real-time monitoring and dynamic optimization adjustment during the silicone production process, continuously improve the key performance indicators of the silicone products through closed-loop feedback control, conduct modeling analysis on the collected silicone performance indicator data, predict the change of performance trend, and give early warning of the quality problems that occur; according to the prediction results of the machine learning algorithm, dynamically adjust the silicone formula and process parameters, prevent quality problems in advance, optimize the production efficiency and cost at the same time, and realize the intelligent management and control of the silicone production line.

[0039] Specifically, the intelligent control of the silicone production line relies on online detection equipment and machine learning algorithms. The online detection equipment collects real-time data on various performance indicators of silicone, such as mechanical strength and insulation resistance, and transmits this data to the control system. Taking mechanical strength as an example, assume that the set tensile strength threshold for silicone is 5 MPa. The online detection equipment measures the tensile strength of the silicone sample in real time and transmits the measured value to the system. The system compares the detected values of the performance indicators collected with the preset threshold. If the measured tensile strength is 4.8 MPa, which is lower than the threshold of 5 MPa, the system determines that the mechanical strength of this batch of silicone is unqualified and triggers a warning signal. At the same time, the system records the production data that led to this result, such as the current formulation composition (e.g., 10% component A, 20% component B, 70% component C), as well as process parameters such as vulcanization temperature (160 °C) and vulcanization time (30 minutes). When the warning signal is triggered, the machine learning algorithm comes into play. The system inputs this production data and a large amount of historical production data into a pre-trained machine learning model. Assume a model constructed using the support vector machine algorithm, which has learned the complex relationships between formulations, process parameters, and final performance indicators in historical production data. By analyzing the similarities and differences between the current data and historical data, the model can give specific optimization and adjustment suggestions. For example, the model may suggest increasing the proportion of component A to 12%, decreasing the proportion of component B to 18%, and keeping the proportion of component C unchanged; at the same time, increasing the vulcanization temperature to 165 °C and extending the vulcanization time to 35 minutes. The system automatically adjusts the formulation system and process control system of the production line according to the suggestions of the machine learning model. The formulation system modifies the ratio of each component according to the suggestions, and the control system adjusts the process parameters of vulcanization temperature and time. After the adjustment is completed, the production line continues to produce silicone, and the online detection equipment continues to monitor the performance indicators of silicone in real time. The system continuously conducts real-time monitoring and dynamic optimization and adjustment, forming a closed-loop feedback control. The online detection equipment continuously collects new data and feeds the data back to the machine learning model. The model continuously updates and optimizes itself based on the new data, thereby improving the prediction accuracy and the effectiveness of the adjustment suggestions. Through such a closed-loop control, the key performance indicators of silicone products can be continuously improved and stabilized within the preset threshold range. In addition to real-time monitoring and adjustment, machine learning algorithms can also be used to predict changes in performance trends. For example, by analyzing the performance indicator data over a period of time, a neural network model can predict the change trend of the mechanical strength of silicone in the future. If it is predicted that the mechanical strength will drop below the threshold, the system will issue a warning in advance and, based on the analysis results of the model, adjust the formulation and process parameters in advance, thereby nipping potential quality problems in the bud. For example, if it is predicted that the mechanical strength will drop to 4.5 MPa after 1 hour, the system will suggest adjusting the formulation or process parameters in advance to avoid the mechanical strength falling below the threshold of 5 MPa.This machine learning-based prediction and prevention mechanism can not only improve product quality, but also optimize production efficiency and costs. Through early warning and adjustment, the generation of waste products can be reduced, the production cycle can be shortened, thereby reducing production costs. At the same time, through the analysis and learning of historical data, the formula and process parameters can be continuously optimized to find the best production plan, thereby improving production efficiency. This realizes the intelligent control of the silicone production line, making the production process more efficient, stable and controllable. Just like an experienced old master, who can predict and adjust the production process through observation and experience to ensure product quality.

[0040] Step S106, construct a knowledge base of silicone formula-process-performance, and use case-based reasoning technology to summarize the change laws of performance under different silicone formulas and process conditions according to historical production data, so as to guide formula design and process optimization.

[0041] Adopt the association rule mining algorithm to analyze the correlation between the silicone formula, the process conditions and the product performance, and obtain the association rule base of formula-process-performance; according to the strong association rules in the association rule base, summarize the influence rules of the performance change of the silicone product under different formula and process conditions; use the association rule base and the influence rules as knowledge to construct the silicone formula-process-performance knowledge base, and realize the structured representation and storage of the knowledge.

[0042] Specifically, the production process of silicone products involves numerous factors, such as raw material ratios, process parameters (e.g., temperature, pressure, time), and final product properties (e.g., mechanical strength, insulation resistance, hardness). To optimize the production process, improve product quality and efficiency, data mining and knowledge engineering techniques can be utilized to construct a silicone formula-process-property knowledge base. First, a large amount of historical production data needs to be obtained. This data includes the raw material ratios of each batch of products, such as the dosage of silicone rubber base material, the dosage of crosslinking agent, and the dosages of various additives; the process parameter settings, such as vulcanization temperature, vulcanization time, mixing speed, and extrusion speed; and the corresponding product performance test results, such as tensile strength, tear strength, hardness, and insulation resistance. Suppose the historical production data of a silicone product A includes: 100 parts of silicone rubber base material, 2 parts of crosslinking agent, 1 part of additive A, 0.5 part of additive B; vulcanization temperature of 170 °C, vulcanization time of 30 minutes; tensile strength of 10 MPa, hardness of 60 HA. Next, preprocess the original data. Since there may be data missing and recording errors in the actual production process, these noisy data need to be removed. For example, if the hardness test result is missing for a certain batch of products, the data of that batch needs to be removed. At the same time, the data ranges of different indicators may vary greatly. For example, the unit of tensile strength is MPa, and the unit of hardness is HA. To facilitate subsequent analysis, data standardization processing is required to convert the data of different indicators to the same range. For example, the min-max normalization method can be used to convert the data of all indicators to between 0 and 1. This can avoid certain indicators having too much influence on the analysis results due to large numerical ranges. Then, use the association rule mining algorithm to analyze the correlation between the formula, process conditions, and product performance. The goal of association rule mining is to discover the hidden association relationships in the data. For example, a rule like "When the dosage of silicone rubber base material is 100 parts, the dosage of crosslinking agent is 2 parts, and the vulcanization temperature is 170 °C, the tensile strength of the product is usually greater than 9 MPa, and the hardness is between 58 and 62 HA" can be discovered. The strength of association rules can be measured by support and confidence. Support represents the frequency of the rule appearing in the dataset, and confidence represents the reliability of the rule. Strong association rules with relatively high support and confidence need to be selected and added to the knowledge base. Based on the strong association rules, the influence laws of the performance changes of silicone products under different formulas and process conditions can be summarized. For example, a law like "Increasing the dosage of the crosslinking agent can increase the hardness of the product but decrease the tensile strength" can be summarized. These laws can help understand the influence mechanism of the formula and process parameters on the product performance. Organize the association rules and influence laws into a structured form to construct a silicone formula-process-property knowledge base. The knowledge base can be organized in the form of a tree structure or a graph structure. For example, the silicone rubber base material and the crosslinking agent can be used as nodes of the knowledge base, and the relationships between them and the influence on product performance can be used as the edges of the knowledge base.When a new formulation needs to be designed or the process optimized, the design or optimization requirements can be input into the knowledge base system. For example, assume that a silicone product with a tensile strength greater than 12 MPa and a hardness between 65 - 70 HA needs to be designed. The system will search the knowledge base for historical cases that are most similar to the requirements using case - based reasoning technology. For example, the system may find a case with a tensile strength of 11.5 MPa and a hardness of 68 HA, and its formulation and process conditions are as follows: 98 parts of silicone rubber base, 2.5 parts of cross - linker, 1.2 parts of additive A, a vulcanization temperature of 175°C, and a vulcanization time of 35 minutes. Based on the matching results, the system will output a formulation design plan or process parameter setting that is most similar to the requirements. For example, the system may suggest increasing the amount of cross - linker to 2.8 parts and raising the vulcanization temperature to 180°C. These suggestions can guide formulation design and process optimization, thereby improving the performance and quality of silicone products, shortening the R & D cycle, and reducing R & D costs.

[0043] Step S107, using a multi - objective optimization algorithm, with the comprehensive performance of the silicone as the optimization objective and the formulation composition and process parameters as the optimization variables, search for the optimal formulation and process combination to achieve the global optimization of the silicone performance.

[0044] Obtain the silicone formulation database and the corresponding process parameter and performance index data, pre - process the data to obtain pre - processed data; according to the pre - processed data, construct a formulation feature vector and a process feature vector; use a multi - objective optimization algorithm, with the comprehensive performance index of the silicone as the optimization objective and the formulation feature vector and the process feature vector as inputs, train a multi - objective optimization model; if there are conflicts between the comprehensive performance indexes, use the concept of Pareto optimal solution set to handle and obtain a set of non - dominated solutions; use an optimization algorithm to optimize the parameters of the multi - objective optimization model to obtain the best parameter combination; according to the best parameter combination, use the multi - objective optimization model to predict the comprehensive performance index of the silicone to obtain a prediction result; according to the prediction result, determine the best formulation and process parameter combination; conduct experimental verification according to the best formulation and process parameter combination to obtain the actual performance index of the silicone; compare the actual performance index with the prediction result to evaluate the actual application effect of the multi - objective optimization model.

[0045] Specifically, data preprocessing is the first step in constructing a knowledge base for silicone rubber formulations - processes - properties. Historical production data needs to be retrieved from the database, which typically includes raw material ratios (e.g., the type and dosage of silicone resins, the type and dosage of fillers, the dosage of crosslinking agents and catalysts), process parameter settings (e.g., mixing temperature, mixing time, vulcanization temperature, vulcanization time, molding pressure), and product performance test results (e.g., tensile strength, hardness, tear strength, heat resistance, aging resistance). Suppose 1000 pieces of historical production data are obtained. In the data cleaning stage, it is found that 50 pieces of data have missing values. For example, the hardness data of some samples is not recorded, then these 50 pieces of data need to be removed. In addition, it is also found that the tensile strength values of 30 pieces of data are abnormally high, which may be due to test errors, and these abnormal data also need to be removed. The remaining 920 pieces of data will be standardized, for example, scaling the values of all performance indicators to between 0 and 1 for subsequent model training and analysis. The purpose of feature engineering is to transform silicone rubber formulations and process parameters into numerical features that can be recognized by machine learning models. For silicone rubber formulations, multiple features can be extracted, such as the content of different types of silicone resins, the type and content of fillers, the type and dosage of crosslinking agents, and the type and dosage of catalysts. Suppose three types of silicone resins are used, then the content of each type of silicone resin can be used as a feature. For process parameters, the mixing temperature, mixing time, vulcanization temperature, vulcanization time, and molding pressure parameters can be used as features. Suppose a silicone rubber formulation contains 5 raw materials and the process contains 4 parameters, then a 9-dimensional feature vector can be constructed to represent this formulation and process. In the model training stage, the NSGA-II algorithm can be selected as the multi-objective optimization algorithm. Suppose two performance indicators of tensile strength and hardness of silicone rubber are concerned, then these two indicators need to be maximized simultaneously. Since there may be a conflict between tensile strength and hardness, that is, increasing the tensile strength may reduce the hardness and vice versa, a set of Pareto optimal solutions need to be found, and these solutions represent the best performance under different combinations of tensile strength and hardness. For example, one solution has a tensile strength of 10 MPa and a hardness of 70, and another solution has a tensile strength of 12 MPa and a hardness of 65. These two solutions are both Pareto optimal solutions because it is impossible to increase both indicators simultaneously without reducing the other indicator. In the model evaluation stage, 5-fold cross-validation can be used to evaluate the performance of the model. The dataset is divided into 5 parts. Each time, 4 parts are used to train the model and 1 part is used to test the model, and this is repeated 5 times to obtain the performance indicators of 5 models, such as the root mean square error (RMSE) of the predicted values. Finally, the average of the 5 RMSEs is used as the final performance indicator of the model. If the RMSE is small, it indicates that the prediction accuracy of the model is high. In the parameter optimization stage, the genetic algorithm can be used to optimize the parameters of the NSGA-II algorithm. The NSGA-II algorithm has some key parameters, such as population size, crossover probability, and mutation probability.Through the genetic algorithm, the optimal combination of these parameters can be found to optimize the performance of the model. For example, through the genetic algorithm, it is determined that when the population size is 100, the crossover probability is 0.8, and the mutation probability is 0.1, the performance of the model is the best. In the prediction and analysis phase, assume that a new silicone formula is designed and the corresponding process parameters are set. After converting the new formula and process parameters into feature vectors and inputting them into the trained multi-objective optimization model, the predicted values of the tensile strength and hardness of the silicone can be obtained. For example, the model predicts that the tensile strength of the silicone is 11 MPa and the hardness is 68. By comparing with the Pareto optimal solution set, it can be judged whether the new formula and process parameters can meet the expected performance requirements. Finally, the results are verified. According to the best formula and process parameters predicted by the model, actual production is carried out, and the actual performance indicators of the silicone are tested. Comparing the actual performance indicators with the predicted values of the model, for example, the actually measured tensile strength is 10.8 MPa and the hardness is 67.5, which is very close to the predicted values, indicating that the prediction results of the model are reliable and can be used to guide actual production.

[0046] Inspired by the above embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of the present invention. The technical scope of the present invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for optimizing the production quality control of silicone formula for high-voltage silicone wire, characterized in that: The method comprises the following steps: Step S101, according to the performance requirements of the silicone material, determine the initial proportion of each additive in the silicone formula, optimize the dosage of each additive through orthogonal test design, and obtain the optimal formula with balanced performance; Step S102, using a support vector machine algorithm to establish a nonlinear mapping model between silicone formula and silicone performance, and using the model to predict the comprehensive performance of silicone under different formulas to provide theoretical guidance for formula optimization; Step S103, obtaining rheological property data of silicone under different extrusion process parameters, using data mining technology to analyze the influence of extrusion speed, temperature and pressure on silicone uniformity and defect rate, and determining the best process parameter combination; Step S104, establishing a kinetic model for the silicone vulcanization process to describe the effects of vulcanization time and temperature on the silicone crosslinking density. If the degree of vulcanization deviates from the preset range, the vulcanization process parameters are adjusted in real time to ensure the stability of the silicone performance. Step S105, installing online detection equipment on the silicone production line to monitor the performance indicators of the mechanical strength and insulation resistance of the silicone in real time. If the detection value exceeds the preset threshold, the formula and process parameters are adjusted in time to control the product quality; Step S106, constructing a knowledge base of silicone formula-process-performance, using case-based reasoning technology, and summarizing the changing rules of performance under different silicone formulas and process conditions based on historical production data to guide formula design and process optimization; Step S107, using a multi-objective optimization algorithm, taking the comprehensive performance of silica gel as the optimization target, taking the formulation composition and process parameters as the optimization variables, searching for the optimal formulation and process combination, and realizing global optimization of silica gel performance.

2. The method for optimizing the production quality of high-voltage silicone rubber wire according to claim 1 is characterized in that: The step S101 also includes: According to the performance requirements of silicone materials, determine the key additives that affect the performance, set the initial proportion range of each additive, and construct an orthogonal test factor level table; Adopting the orthogonal experimental design method, multiple groups of experimental schemes are designed according to the factor level table, and each group of schemes prepares silica gel material samples according to the set additive ratio; Performing performance tests on each group of silica gel material samples, obtaining data on various performance indicators, and establishing a mathematical model between the performance indicators and the amount of additives used; Analyzing the relationship between the dosage of each additive and the performance index through the mathematical model, and determining the sensitivity coefficient and optimal range of the dosage of each additive; In the optimal range of the dosage of each additive, the response surface optimization or neural network algorithm is used to optimize the dosage ratio of each additive to obtain the formula with the best comprehensive performance; Verify the optimal formula obtained through optimization, and determine whether it meets the performance requirements of silicone materials by preparing samples and testing their performance; If the optimal formula meets the requirements, it is determined as the final formula of the silicone material; If the requirements are not met, adjust the optimization range of additive dosage and re-optimize until the requirements are met.

3. The method for optimizing the production quality of high-voltage silicone rubber wire according to claim 1 is characterized in that: The step S102 also includes: For the preprocessed data set, a support vector machine regression algorithm is used to construct a nonlinear mapping model between the silicone formula and the silicone performance, and the hyperparameters of the nonlinear mapping model are optimized through cross-validation and grid search; According to the nonlinear mapping model, input the silicone formula parameters, predict the comprehensive performance index of the silicone under the formula, compare the predicted performance index with the actual performance index, and if the error between the two is greater than a preset threshold, obtain more formula-performance data and retrain the nonlinear mapping model; The trained nonlinear mapping model is used to predict the comprehensive performance of silicone under different formulations and obtain the silicone formulation parameter combination with the best performance.

4. The method for optimizing the production quality of high-voltage silicone rubber wire according to claim 1 is characterized in that: The step S103 also includes: Obtain a silicone extrusion experimental data set including extrusion speed, temperature, pressure, and corresponding uniformity index values ​​and defect rates; Training a machine learning model based on the data set and determining the optimal combination of the extrusion speed, the temperature, and the pressure that maximizes the uniformity index value and minimizes the defect rate through a parameter search algorithm; The optimal parameter combination is applied to the silicone extrusion production process to obtain product quality data.

5. The method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire according to claim 1 is characterized in that: The step S104 also includes: Acquire historical data of the silicone vulcanization process, wherein the historical data includes vulcanization temperature, vulcanization time and cross-linking density; Using a machine learning algorithm to train a vulcanization kinetics model based on the historical data to obtain the vulcanization kinetics model; Real-time acquisition of the curing temperature and curing time of the current silicone curing process; Inputting the vulcanization temperature and the vulcanization time into the vulcanization kinetics model to predict the crosslinking density of the silicone under the current vulcanization conditions to obtain a predicted crosslinking density; Obtaining a preset cross-linking density target range; Determining whether the predicted crosslink density is within the target range; If the predicted crosslink density is not within the target range, a vulcanization process parameter adjustment mechanism is triggered; Using an optimization algorithm, according to the deviation between the predicted crosslink density and the target range, determining the adjustment range of the vulcanization temperature and time, and obtaining an adjusted vulcanization temperature and an adjusted vulcanization time; Applying the adjusted vulcanization temperature and the adjusted vulcanization time to a control system of a silicone vulcanization device; Continuously monitor the adjusted vulcanization process, obtain real-time vulcanization temperature, vulcanization time and cross-link density data, and obtain real-time vulcanization data; Using the real-time vulcanization data to update the vulcanization kinetics model; Adjust the target range regularly based on changes in production requirements.

6. The method for optimizing the production quality control of the silicone formula of the high-voltage silicone wire according to any one of claims 1 to 5, characterized in that: The step S105 also includes: Obtain real-time data collected by online testing equipment installed on the silicone production line to monitor and analyze key performance indicators of silicone mechanical strength and insulation resistance; Compare the silica gel performance index detection value obtained by the monitoring analysis with the preset threshold value to determine whether it exceeds the preset threshold range; If the detection value exceeds the preset threshold, an early warning signal is triggered, and a machine learning algorithm is used to analyze historical production data to obtain optimization adjustment suggestions for silicone formula and process parameters; According to the optimization and adjustment suggestions, the relevant parameter settings of the formula system and process control system of the silicone production line are automatically adjusted to achieve dynamic optimization of the formula and process parameters; Continuously conduct real-time monitoring and dynamic optimization and adjustment during the silicone production process, continuously improve the key performance indicators of silicone products through closed-loop feedback control, conduct modeling and analysis on the collected silicone performance indicator data, predict performance trend changes, and provide early warning of quality problems; According to the prediction results of the machine learning algorithm, the silicone formula and process parameters are dynamically adjusted to prevent quality problems in advance, while optimizing production efficiency and cost to achieve intelligent management and control of the silicone production line.

7. The method for optimizing the production quality of high-voltage silicone rubber wire according to any one of claims 1 to 5, characterized in that: The step S107 also includes: Acquire a silica gel formula database and corresponding process parameter and performance index data, and preprocess the data to obtain preprocessed data; Constructing a recipe feature vector and a process feature vector according to the preprocessed data; A multi-objective optimization algorithm is adopted, the comprehensive performance index of silica gel is used as the optimization target, the formula feature vector and the process feature vector are used as input, and a multi-objective optimization model is trained; If there is a conflict between the comprehensive performance indicators, the concept of Pareto optimal solution set is used to deal with it and obtain a set of non-inferior solutions; Using an optimization algorithm to optimize the parameters of the multi-objective optimization model to obtain the best parameter combination; According to the optimal parameter combination, the multi-objective optimization model is used to predict the comprehensive performance index of silica gel to obtain a prediction result; Determine the best formula and process parameter combination according to the prediction results; Experimental verification is performed based on the optimal formula and process parameter combination to obtain actual performance indicators of silica gel; The actual performance index is compared with the predicted result to evaluate the actual application effect of the multi-objective optimization model.

Images