A control optimization method for glass fiber process

Abstract

The present application provides a control optimization method for a glass fiber process, which relates to the field of material processing technology, including: obtaining multiple continuous process nodes, establishing a production quality prediction model, receiving multiple real-time process parameters, inputting a global quality analysis network, generating a global quality index, and if the global quality index does not meet the preset quality index, inputting multiple real-time process parameters into multiple node quality prediction networks respectively, generating multiple node quality indexes, and performing iterative optimization of node process parameters until the optimized global quality index meets the preset quality index, generating optimized process parameters, and processing and controlling the glass fiber. The present application can solve the technical problems in the prior art that the traditional glass fiber production process relies on a simple feedback mechanism, resulting in low production efficiency and unstable product quality. Through global quality analysis, each process node can be accurately controlled, thereby improving the overall production efficiency and product quality.

Description

Technical Field

[0001] The present application relates to the technical field of material processing, and in particular to a control optimization method for a glass fiber process. Background Art

[0002] With the development of science and technology, glass fiber, as a lightweight and high-strength material, has been widely used in many fields. The production process of glass fiber is relatively complex and involves multiple key process steps, such as raw material preparation, melting, drawing, cooling, oiling and winding. Each step has an important impact on the quality and production efficiency of the final product. However, in the existing glass fiber production process, the control of process parameters mostly relies on experience or a simple feedback mechanism, resulting in low production efficiency, unstable product quality, and difficulty in real-time adjustment to cope with process changes. At present, the control methods used in the glass fiber production process are usually based on manual experience and simple feedback control systems, which cannot monitor and adjust process parameters in real time, resulting in product quality fluctuations and low production efficiency. Lack of comprehensive analysis of the entire production process makes it difficult to accurately predict and control product quality.

[0003] In summary, the prior art has technical problems such as low production efficiency and unstable product quality due to the traditional glass fiber production process relying on a simple feedback mechanism. Summary of the invention

[0004] The purpose of this application is to provide a control optimization method for a glass fiber process, so as to solve the technical problems in the prior art that the traditional glass fiber production process relies on a simple feedback mechanism, resulting in low production efficiency and unstable product quality.

[0005] In view of the above problems, the present application provides a control optimization method for a glass fiber process, wherein the control optimization method for a glass fiber process includes: obtaining multiple continuous process nodes of the glass fiber, and establishing a production quality prediction model, wherein the production quality prediction model includes a global quality analysis network and multiple node quality prediction networks corresponding to the multiple continuous process nodes; receiving multiple real-time process parameters of the multiple continuous process nodes, inputting the parameters into the global quality analysis network, and generating a global quality index; judging whether the global quality index meets the preset quality index, and if not, respectively inputting the multiple real-time process parameters into the multiple node quality prediction networks to generate multiple node quality indexes; iteratively optimizing the node process parameters according to the multiple node quality indexes until the optimized global quality index meets the preset quality index, and generating optimized process parameters; and controlling the processing of the glass fiber with the optimized process parameters.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By acquiring multiple continuous process nodes of glass fiber, a production quality prediction model is established, wherein the production quality prediction model includes a global quality analysis network and multiple node quality prediction networks corresponding to the multiple continuous process nodes; multiple real-time process parameters of the multiple continuous process nodes are received, input into the global quality analysis network, and a global quality index is generated; it is determined whether the global quality index meets the preset quality index, if not, the multiple real-time process parameters are respectively input into the multiple node quality prediction networks to generate multiple node quality indexes; according to the multiple node quality indexes, the node process parameters are iteratively optimized until the optimized global quality index meets the preset quality index, and the optimized process parameters are generated; the glass fiber is processed and controlled with the optimized process parameters. In other words, each process node is accurately controlled through global quality analysis, which improves the overall production efficiency and product quality.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0010] Figure 1 A schematic flow chart of a control optimization method for a glass fiber process according to the present application.

[0011] Figure 2 A schematic diagram of a flow chart for generating optimized process parameters in a control optimization method for a glass fiber process according to the present application. DETAILED DESCRIPTION

[0012] This application provides a control optimization method for glass fiber process, which solves the technical problem in the prior art that the traditional glass fiber production process relies on a simple feedback mechanism, resulting in low production efficiency and unstable product quality. Through global quality analysis, each process node is precisely controlled, thereby improving the overall production efficiency and product quality.

[0013] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0014] For examples, please see the attached Figure 1 The present application provides a control optimization method for a glass fiber process, wherein the control optimization method for a glass fiber process specifically comprises the following steps:

[0015] Step 1: Acquire multiple continuous process nodes of glass fiber and establish a production quality prediction model, wherein the production quality prediction model includes a global quality analysis network and multiple node quality prediction networks corresponding to the multiple continuous process nodes.

[0016] Specifically, multiple continuous process nodes in the glass fiber production process are obtained, including key steps such as raw material preparation, melting, drawing, and coating. Sensors and data acquisition systems are installed at each process node to collect process parameters such as temperature, humidity, speed, tension, flow, pressure, etc. in real time. Data preprocessing is performed on the collected data to extract and transform the data and generate features that are helpful for model learning. Machine learning models suitable for global analysis are selected, such as deep learning networks, random forests, support vector machines, etc. The historical process parameter sets of multiple continuous process nodes are obtained, and the production quality is marked to generate a quality indicator set. The node historical process parameter set and quality indicator set are used as training data to train the global quality analysis network so that it can predict the production quality as a whole, comprehensively consider the data of all process nodes, and predict the quality of the final product.

[0017] For each process node, a node segmentation parameter set is generated for each process node, and the process parameters of each node are used as variables, while the parameters of other nodes are used as quantitative data under the qualified mark. The node segmentation parameter set and the quality indicator set are used to train the corresponding prediction network so that it can accurately predict the production quality of the node. The global quality analysis network and multiple node quality prediction networks are integrated into one system, and the cross-validation technology is used to evaluate the accuracy and generalization ability of the model. The model parameters are adjusted according to the verification results to improve the prediction accuracy. The trained model is deployed to the production line to monitor the process parameters in real time. By establishing such a prediction model, the quality can be monitored and predicted in real time during the production process, so as to discover potential problems in advance.

[0018] Step 2: receiving a plurality of real-time process parameters of the plurality of continuous process nodes, inputting the parameters into the global quality analysis network, and generating a global quality index.

[0019] Specifically, ensuring that each process node is equipped with appropriate sensors and monitoring equipment to collect process parameters in real time, including temperature, pressure, speed, etc., is a key factor affecting product quality. Configure a data transmission system to transmit the real-time process parameters of each node to the central control system, preprocess the data, remove obvious errors or outliers, and then normalize them to ensure that they are on the same scale. Integrate the preprocessed real-time process parameters into a data set and input them into the global quality analysis network. The global quality analysis network is a model based on big data analysis and machine learning technology that can comprehensively consider the data of all process nodes. The global quality analysis network processes the input real-time process parameters, analyzes the data through its internal algorithms and structures, and outputs one or more global quality indicators, which represent the overall quality status of the current production process. By receiving real-time process parameters and inputting them into the global quality analysis network, the system can monitor the quality status of the entire production process in real time.

[0020] Step three: determine whether the global quality index meets the preset quality index; if not, input the multiple real-time process parameters into the multiple node quality prediction networks respectively to generate multiple node quality indexes.

[0021] Specifically, a quality indicator is pre-set according to the production objectives and product quality standards, and the obtained global quality indicator is compared with the pre-set quality standard. If the global quality indicator meets or exceeds the preset quality indicator, the production process can continue; if the global quality indicator does not meet the preset quality indicator, it indicates that there are quality problems in the production process and further detailed analysis is required. Multiple real-time process parameters are input into the corresponding multiple node quality prediction networks respectively. For each node, its specific quality prediction network is used to process the real-time process parameters. Each node quality prediction network will output one or more node quality indicators, reflecting the production quality status of the node. By judging whether the global quality indicator meets the preset standard, quality problems in the production process can be quickly identified, which helps to achieve more accurate and efficient quality control and reduce the production of unqualified products.

[0022] Step 4: According to the multiple node quality indicators, iteratively optimize the node process parameters until the optimized global quality indicator meets the preset quality indicator, and generate optimized process parameters.

[0023] Specifically, according to multiple node quality indicators, the process parameters of the corresponding nodes are iteratively optimized, and for each node, the process parameters of each node are improved to improve its production quality. After each optimization, the global quality indicator is re-evaluated to see if it meets the preset quality indicator, and the effect of the parameter adjustment is evaluated. The iterative process continues until the global quality indicator meets the preset quality indicator and the quality indicators of all process nodes meet the predetermined standards. Once the preset quality indicator is met, the iterative optimization process ends, and the process parameters at this time are the final optimized process parameters. These parameters are based on the entire optimization process, combined with global and local quality analysis, and the results of iterative optimization, which can ensure the efficiency of the production process and the stability of product quality. By iteratively optimizing according to the node quality indicators, the quality performance of each process node is improved in a targeted manner, and the final optimized process parameters are generated to ensure the efficiency of the production process and the stability of product quality.

[0024] Step 5: Controlling the processing of the glass fiber using the optimized process parameters.

[0025] Specifically, the generated optimized process parameters are applied to the glass fiber processing process, and the process settings of the glass fiber production line are adjusted and controlled according to these parameters, including but not limited to the raw material mixing ratio, melting temperature, drawing speed, cooling efficiency and oiling uniformity, etc., to achieve the best processing effect. By accurately controlling these optimized process parameters, the processing quality and efficiency of glass fiber are improved, ensuring that each link of the production process can operate in the best state. By applying the optimized process parameters, the generation of unqualified products is reduced, thereby improving the overall production efficiency and product quality.

[0026] Furthermore, step one of this application includes:

[0027] Connect to the production log storage platform on the glass fiber production line, retrieve the node historical process parameter sets of the multiple continuous process nodes, perform production quality marking, and generate a quality indicator set; use the node historical process parameter sets and the quality indicator set as training data to construct the global quality analysis network; use the production log storage platform to establish multiple node segmentation parameter sets, wherein the process parameters of the corresponding nodes in any node segmentation parameter set are variables, and other process node parameters are quantitative data under qualified marking; use the multiple node segmentation parameter sets and the quality indicator set as training data to construct the multiple node quality prediction networks.

[0028] Specifically, the production log storage platform connected to the glass fiber production line is responsible for recording and storing key data in the production process, such as process parameters, equipment status, etc. The historical process parameter sets of multiple continuous process nodes are retrieved, including the process parameter records of each process node over a period of time, such as temperature, pressure, speed, etc. These historical process parameters are marked with production quality, and the actual process parameters are compared with the quality results of the final product to establish the relationship between parameters and quality and determine the quality status of each data point. Based on the quality mark, a quality indicator set is generated, which is the target variable for model training.

[0029] The node historical process parameter sets and the generated quality indicator sets obtained from the production log storage platform are used as training data to construct a global quality analysis network. The global quality analysis network is a model based on machine learning algorithms, such as neural networks, support vector machines, or random forests, which is used to analyze and predict the quality of the entire production process. By analyzing historical data, the global quality analysis network learns and understands the relationship between process parameters and product quality, so that it can predict the quality of new process parameters. Based on the production log storage platform, multiple node segmentation parameter sets are established. In each node segmentation parameter set, the process parameters of the corresponding node are set as variables and serve as the main research object in subsequent analysis. The parameters of other process nodes are regarded as quantitative data under the qualified mark and remain unchanged in the analysis as a reference benchmark. Using multiple node segmentation parameter sets and quality indicator sets as training data, multiple node quality prediction networks are constructed. The node quality prediction network is a model based on machine learning algorithms that is used to analyze and predict the quality of a specific process node. Each network is specifically targeted at a process node and is used to predict the production quality of that node. By constructing a global quality analysis network and multiple node quality prediction networks, a comprehensive production quality prediction model is established to analyze the quality status of the entire production process from a global perspective.

[0030] Furthermore, the present application also includes the following steps:

[0031] Through the production log storage platform, multiple sets of production parameter record sets are collected, and the parameters under any node are marked as qualified to generate multiple sets of node qualified label sets; the first continuous process node is extracted from the multiple continuous process nodes; the first continuous process node is used as a variable, combined with the multiple sets of node qualified label sets, the production parameter records of the nodes other than the first continuous process node that are qualified are extracted from the multiple sets of production parameter record sets to generate a first node segmentation parameter set; the first node segmentation parameter set is added to the multiple node segmentation parameter sets.

[0032] Specifically, through the production log storage platform, a production parameter record set of multiple continuous process nodes is collected, including the process parameter data of each process node in the production process. Each set of production parameter record sets is marked as qualified under any node, usually based on a pre-set quality standard or process specification. For each process node, its process parameters are judged to be qualified according to the product quality standard. According to the parameter qualified mark, a corresponding node qualified label set is generated for each set of production parameter records, and multiple sets of node qualified label sets are generated, including the qualified conditions of the parameters of each process node. The first continuous process node is extracted from the multiple continuous process nodes, and it is used as a variable, combined with the multiple sets of node qualified label sets generated previously. The production parameter records whose parameters of all other nodes except the first continuous process node are qualified are screened out from the multiple sets of production parameter record sets, and only those records where all other nodes meet the qualified standards are selected. Based on the extracted records, a first node segmentation parameter set is generated for training the quality prediction network for the first process node. The generated first node segmentation parameter set is added to the existing multiple node segmentation parameter sets, and the multiple node segmentation parameter sets are data sets for different process nodes, each of which contains the process parameter variables of a specific node and quantitative data of other nodes. Each node will have a corresponding segmentation parameter set to train its quality prediction network. Accurately analyze and optimize the parameters of each process node to ensure that each link in the production process can operate in the best state and predict and adjust the process parameters more effectively.

[0033] Further, as attached Figure 2 As shown, step 4 of this application includes:

[0034] The multiple node quality indicators are serialized in order from small to large to generate a node quality indicator sequence; a first optimization node is extracted from the node quality indicator sequence, the process parameters of the first optimization node are iteratively optimized, and a quality assessment is performed through a first node quality prediction network corresponding to the first optimization node to generate first node optimization parameters that meet the preset quality indicators; the multiple real-time process parameters are updated with the first node optimization parameters and then input into the global quality analysis network for quality analysis to output a first updated global quality indicator; it is determined whether the first updated global quality indicator also meets the preset quality indicator, and if not, the process parameter optimization of the second optimization node is continued to generate the optimized process parameters.

[0035] Specifically, multiple node quality indicators are sorted in order from small to large to generate a node quality indicator sequence, which reflects the relative size and order of the quality indicators of each process node. The first node (i.e., corresponding to the smallest indicator) is extracted from the node quality indicator sequence as the first optimization node. This node is usually the node with the lowest quality indicator in the sequence, indicating that its quality performance needs to be improved the most. Starting from the node corresponding to the lowest indicator, not optimizing all nodes, the efficiency is improved and the optimization complexity is reduced. The process parameters of the first optimization node are iteratively optimized, including adjusting process parameters such as temperature, pressure, speed, etc. to improve the quality performance of the node. After each iteration, the first node quality prediction network corresponding to the first optimization node is used for quality assessment until the generated parameters enable the quality indicators of the first optimization node to meet the preset quality standards. Multiple real-time process parameters are updated with the first node optimization parameters, and the updated real-time process parameters are input into the global quality analysis network for quality analysis. The global quality analysis network is a model based on big data analysis and machine learning technology that can comprehensively consider the data of all process nodes.

[0036] Using the updated process parameters, the global quality analysis network performs quality analysis and outputs updated global quality indicators, which reflect the impact of the updated process parameters on the quality of the entire production process. The first updated global quality indicator is also compared with the preset quality indicator. If the preset quality indicator is met, it means that the adjustment of the process parameters is successful and the quality of the production process has been improved, then the optimization process ends; otherwise, it indicates that the adjustment of the process parameters may not be sufficient, and the process parameter optimization of the next optimization node is continued. Find the adjacent optimization node, that is, the second optimization node, iteratively optimize the process parameters of the second optimization node, and perform quality evaluation until the preset quality indicator is met, then generate the optimized process parameters. The entire optimization process forms a closed-loop system, which automatically performs iterative optimization until the quality requirements are met. Through this iterative and serialized optimization process, the production quality is effectively improved, ensuring that each process node operates in the best state, and improving production efficiency and product quality.

[0037] Furthermore, the present application also includes the following steps:

[0038] The second optimization node is a node adjacent to the first optimization node in the node quality indicator sequence; the process parameters of the second optimization node are iteratively optimized, and quality assessment is performed through a second node quality prediction network corresponding to the second optimization node to generate second node optimization parameters that meet the preset quality indicators; the multiple real-time process parameters are updated with the first node optimization parameters and the second node optimization parameters, and then input into the global quality analysis network for quality analysis, and a second updated global quality indicator is output; it is determined whether the second updated global quality indicator meets the preset quality indicator, and if so, the optimized process parameters are generated with the updated process parameters.

[0039] Specifically, the node adjacent to the first optimization node, i.e., the second optimization node, is found in the node quality index sequence, which ensures the continuity and systematicness of the optimization process. The process parameters of the second optimization node are iteratively optimized, similar to the optimization of the first optimization node. After each optimization, the second node quality prediction network is used to perform quality assessment to determine whether the preset quality index has been reached. After multiple rounds of optimization, the second node optimization parameters that meet the preset quality index are generated. The first node optimization parameters and the second node optimization parameters are used to update multiple real-time process parameters together. The updated real-time process parameters are input into the global quality analysis network, and the second updated global quality index is output to evaluate whether the overall production quality has been improved. The second updated global quality index is compared with the preset quality index to determine whether the preset quality index is met. If it is met, the updated process parameters are used as the optimized process parameters, and the optimization process ends. If not, the optimization process may need to continue, considering similar iterative optimization of the process parameters of other nodes until the global quality index reaches the preset standard. Through this continuous and iterative method, the parameters of each key process node are systematically optimized, thereby improving the overall production quality.

[0040] Furthermore, the present application also includes the following steps:

[0041] Based on the training data of the first node quality prediction network, a change relationship between process parameters and quality indicators is established; based on the change relationship, multiple rounds of optimization are performed on the process parameters of the first optimization node according to a preset optimization step size to generate first node optimization parameters that meet the preset quality indicators.

[0042] Specifically, based on the training data of the first node quality prediction network, the relationship between process parameters and quality indicators is analyzed to understand which changes in process parameters will significantly affect the quality indicators of the first optimization node. Using the established process parameter-quality indicator change relationship, the process parameters of the first optimization node are optimized for multiple rounds according to the preset optimization step size. Select a suitable optimization algorithm, such as gradient descent, simulated annealing, genetic algorithm, etc., to effectively find the optimal or approximately optimal process parameters. According to the characteristics of the process parameters and production experience, a reasonable optimization step size is preset to determine the amplitude of the adjustment of the process parameters in each optimization iteration. During the optimization process, the optimization step size can be adjusted according to the actual situation to balance the convergence speed and stability. After each optimization, the first node quality prediction network is used for quality assessment to determine whether the preset quality indicators are achieved. Multiple rounds of optimization are performed until the first node optimization parameters that meet the preset quality indicators are generated. By establishing the change relationship between process parameters and quality indicators, the impact of process parameter adjustment on quality is understood and predicted, and the process parameters of the first optimization node are optimized until the generated quality indicators meet the preset quality standards, thereby improving the overall performance of the production line and product quality.

[0043] Furthermore, the present application also includes the following steps:

[0044] If the first updated global quality index meets the preset quality index, the updated process parameters are received and input into the multiple node quality prediction networks to generate multiple updated node quality indexes; quality uniformity optimization is performed according to the updated node quality index to generate multiple uniformly optimized nodes and multiple optimized quality ranges that meet the preset uniform constraints; the node quality prediction networks corresponding to the multiple uniformly optimized nodes are called to optimize the node process parameters again with the multiple optimized quality ranges as the target to generate the optimized process parameters.

[0045] Furthermore, the present application also includes the following steps:

[0046] The preset uniform constraint is that the multiple updated node quality indicators are all greater than the preset node quality indicators, and the deviation between the quality indicators of any two nodes is less than the preset deviation.

[0047] Specifically, if the first updated global quality index meets the preset quality index, the updated process parameters are received and input into multiple node quality prediction networks to predict and evaluate the quality impact of each process node. Through multiple node quality prediction networks, multiple updated node quality indicators are obtained, reflecting the quality status of each process node after using the updated process parameters. Quality uniformity optimization is performed according to the updated node quality indicators so that the quality indicators of all nodes meet the preset uniform constraints. Quality uniformity optimization is an optimization method that aims to ensure that the quality indicators of all process nodes meet the preset uniform constraints. The preset uniform constraints are set to require that all updated node quality indicators are greater than the preset node quality indicators, and the deviation between the quality indicators of any two nodes is less than the preset deviation. The preset node quality indicator refers to the quality standard that each process node should achieve, which is pre-set according to the product quality standard and production requirements, and is used to compare and evaluate the optimized node quality indicator. The preset deviation is the upper limit of the difference between a quality indicator set during the optimization process, which is used to ensure that the difference between the optimized node quality indicators is within an acceptable range. This deviation value is used to limit the difference between the quality indicators of any two process nodes to avoid excessive fluctuations in the quality indicators, thereby ensuring the stability and consistency of product quality.

[0048] According to the updated node quality index, quality uniform optimization is performed to generate multiple uniform optimization nodes and multiple optimized quality ranges that meet the preset uniform constraints. Multiple uniform optimization nodes are process parameter settings that meet the preset uniform constraints and are used to guide the production process; multiple optimized quality ranges are the changes in the quality indicators of each node during the optimization process, reflecting the amplitude of parameter adjustment. For each uniform optimization node, the node quality prediction network corresponding to multiple uniform optimization nodes is called. With the optimization quality amplitude as the goal, the node process parameters are optimized again, including further adjustment of process parameters such as temperature, pressure, speed, etc., to achieve more refined quality control. After re-optimization, the final optimized process parameters are generated. These parameters are based on the entire optimization process, combined with global and local quality analysis, and the results of iterative optimization, which can ensure the efficiency of the production process and the stability of product quality. The uniformity of the process parameters is optimized to ensure that the difference in quality indicators of each node is minimized, and to ensure that the quality indicators of each node in the production process are not only within the preset range, but also the differences between them are kept at an acceptable level, so as to achieve balanced optimization of the entire production line.

[0049] In summary, the control optimization method of a glass fiber process provided by the present application has the following technical effects:

[0050] By acquiring multiple continuous process nodes of glass fiber, a production quality prediction model is established, wherein the production quality prediction model includes a global quality analysis network and multiple node quality prediction networks corresponding to the multiple continuous process nodes; multiple real-time process parameters of the multiple continuous process nodes are received, input into the global quality analysis network, and a global quality index is generated; it is determined whether the global quality index meets the preset quality index, if not, the multiple real-time process parameters are respectively input into the multiple node quality prediction networks to generate multiple node quality indexes; according to the multiple node quality indexes, the node process parameters are iteratively optimized until the optimized global quality index meets the preset quality index, and the optimized process parameters are generated; the glass fiber is processed and controlled with the optimized process parameters. In other words, each process node is accurately controlled through global quality analysis, which improves the overall production efficiency and product quality.

[0051] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0052] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A control optimization method for a glass fiber process, characterized in that: include: Acquire multiple continuous process nodes of the glass fiber and establish a production quality prediction model, wherein the production quality prediction model includes a global quality analysis network and multiple node quality prediction networks corresponding to the multiple continuous process nodes; Receiving a plurality of real-time process parameters of the plurality of continuous process nodes, inputting the parameters into the global quality analysis network, and generating a global quality index; Determine whether the global quality index meets the preset quality index, and if not, input the multiple real-time process parameters into the multiple node quality prediction networks respectively to generate multiple node quality indexes; Iteratively optimize node process parameters according to the multiple node quality indicators until the optimized global quality indicator meets the preset quality indicator, thereby generating optimized process parameters; Controlling the processing of the glass fiber using the optimized process parameters; According to the multiple node quality indicators, iterative optimization of node process parameters is performed until the optimized global quality indicator meets the preset quality indicator, and the optimized process parameters are generated, including: Serializing the plurality of node quality indicators in order from small to large to generate a node quality indicator sequence; Extracting a first optimization node from the node quality indicator sequence, iteratively optimizing process parameters of the first optimization node, and performing quality assessment through a first node quality prediction network corresponding to the first optimization node to generate first node optimization parameters that meet the preset quality indicators; The plurality of real-time process parameters are updated with the first node optimization parameters and then input into the global quality analysis network for quality analysis, and a first updated global quality index is output; Determine whether the first updated global quality index also meets the preset quality index, and if not, continue to optimize the process parameters of the second optimization node to generate the optimized process parameters; The process parameter optimization of the second optimization node is continued to generate the optimized process parameters, including: The second optimized node is a node adjacent to the first optimized node in the node quality indicator sequence; Iteratively optimize the process parameters of the second optimization node, and perform quality assessment through a second node quality prediction network corresponding to the second optimization node to generate second node optimization parameters that meet the preset quality index; The plurality of real-time process parameters are updated with the first node optimization parameter and the second node optimization parameter, and then input into the global quality analysis network for quality analysis, and a second updated global quality index is output; Determining whether the second updated global quality index meets the preset quality index, and if so, generating the optimized process parameters with the updated process parameters; The step of generating the first node optimization parameter that meets the preset quality index includes: Based on the training data of the first node quality prediction network, a relationship between process parameters and quality indicators is established; Based on the change relationship, performing multiple rounds of optimization on the process parameters of the first optimization node according to a preset optimization step length to generate first node optimization parameters that meet the preset quality index; Among them, the production quality prediction model is established, including: Connecting to the production log storage platform on the glass fiber production line, retrieving the node historical process parameter set of the plurality of continuous process nodes, and marking the production quality to generate a quality indicator set; Using the node historical process parameter set and the quality indicator set as training data, constructing the global quality analysis network; Using the production log storage platform, multiple node segmentation parameter sets are established, wherein the process parameters of the corresponding node in any node segmentation parameter set are variables, and other process node parameters are quantitative data under qualified marks; Using the plurality of node segmentation parameter sets and the plurality of quality indicator sets as training data, constructing the plurality of node quality prediction networks; Wherein, using the production log storage platform, multiple node segmentation parameter sets are established, including: Through the production log storage platform, multiple sets of production parameter record sets are collected, and parameter qualification marking is performed under any node to generate multiple sets of node qualification label sets; extracting a first continuous process node from the plurality of continuous process nodes; Taking the first continuous process node as a variable, in combination with the multiple groups of node qualified label sets, extracting the production parameter records of the nodes other than the first continuous process node whose parameters are qualified from the multiple groups of production parameter record sets to generate a first node segmentation parameter set; The first node segmentation parameter set is added to the plurality of node segmentation parameter sets.

2. A control optimization method for a glass fiber process as claimed in claim 1, characterized in that: According to the multiple node quality indicators, iterative optimization of node process parameters is performed, further comprising: If the first updated global quality index satisfies the preset quality index, receiving the updated process parameters, inputting them into the multiple node quality prediction networks, and generating multiple updated node quality indexes; Performing uniform quality optimization according to the updated node quality index to generate multiple uniform optimization nodes and multiple optimized quality ranges that meet preset uniform constraints; The node quality prediction network corresponding to the plurality of uniformly optimized nodes is called, and the node process parameters are optimized again with the plurality of optimized quality ranges as targets to generate the optimized process parameters.

3. A control optimization method for a glass fiber process as claimed in claim 2, characterized in that: The preset uniform constraint is that the multiple updated node quality indicators are all greater than the preset node quality indicators, and the deviation between the quality indicators of any two nodes is less than the preset deviation.

Images