A method and system for intelligent analysis of production data of carbon-carbon composite material deposition furnace

By using a clustering algorithm based on learning automata to intelligently analyze the production data of carbon-carbon composite material deposition furnaces, the set of high-quality and qualified production factors is determined, which solves the problems of long production cycles and high costs, and improves product qualification rate and production efficiency.

CN114065615BActive Publication Date: 2024-11-19TAIYUAN UNIVERSITY OF TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111283179.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-01
Publication Date
2024-11-19
Estimated Expiration
2041-11-01

AI Technical Summary

Technical Problem

In the existing carbon-carbon composite material deposition furnace production process, the production cycle is long, the cost is high, and the product qualification rate is low. Traditional analysis methods cannot determine the specific set values ​​of production factors, resulting in unqualified products.

Method used

A clustering algorithm based on learning automata is used to cluster production data. Combined with the fuzzy c-means algorithm, the set of high-quality, qualified and poor-quality production factors is determined through fuzzy membership calculation and reinforcement learning, thereby improving the product unit weight gain qualification rate.

Benefits of technology

It significantly improved the pass rate of carbon-carbon composite material deposition furnace products. By optimizing the setting of production factors through intelligent analysis methods, it improved production efficiency and product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114065615B_ABST
    Figure CN114065615B_ABST
Patent Text Reader

Abstract

The present invention relates to a method and system for intelligent analysis of production data of a carbon-carbon composite material deposition furnace. A plurality of production data are first obtained, including deposition furnace raw material data and product unit weight gain. The deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace. Then, a clustering algorithm based on a learning automaton is used to cluster all production data to obtain a plurality of clusters. Finally, all clusters are graded according to the unit weight gain of the product to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set. Then, a machine learning method is used to realize intelligent analysis of the production data of the carbon-carbon composite material deposition furnace to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set, so as to provide a basis for the deposition furnace production process of carbon-carbon composite material products and significantly improve the qualified rate of carbon-carbon composite material deposition furnace products.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method and system for intelligently analyzing production data of a carbon-carbon composite material deposition furnace. Background Art

[0002] Carbon-carbon composite products are mainly used in aerospace and military fields, which have high requirements for product material quality. In the production process of carbon-carbon composite products, the current main method is to use chemical vapor deposition furnace technology through carbon-carbon composite rapid deposition furnace, but due to its lengthy product production cycle and high cost, its wide application is limited.

[0003] At present, the production data in the production process of carbon-carbon composite material deposition furnace products at home and abroad are mainly set based on experience on the one hand, and analyzed by correlation in traditional statistical analysis methods on the other hand. There will be the problem of uncertain set values ​​of various production factors, resulting in a part of the final products failing to meet the product indicators and having a low pass rate. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for intelligent analysis of production data of a carbon-carbon composite material deposition furnace, which adopts a machine learning method to realize intelligent analysis of production data of a carbon-carbon composite material deposition furnace, so as to determine a set of high-quality production factors and improve the qualified rate of carbon-carbon composite material deposition furnace products.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for intelligent analysis of production data of a carbon-carbon composite material deposition furnace, the method comprising:

[0007] Acquire multiple production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace;

[0008] Clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters;

[0009] All the clusters are graded according to the unit weight gain of the products to determine a high-quality production factor set, a qualified production factor set and a low-quality production factor set.

[0010] A carbon-carbon composite material deposition furnace production data intelligent analysis system, the intelligent analysis system comprising:

[0011] An acquisition module, used to acquire a plurality of production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace;

[0012] A clustering module, used for clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters;

[0013] The analysis module is used to classify all the clusters according to the unit weight gain of the products, and determine a high-quality production factor set, a qualified production factor set and a low-quality production factor set.

[0014] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0015] The present invention is used to provide a method and system for intelligent analysis of production data of a carbon-carbon composite material deposition furnace. First, a plurality of production data are obtained, including deposition furnace raw material data and product unit weight gain. The deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace. Then, a clustering algorithm based on a learning automaton is used to cluster all production data to obtain a plurality of clusters. Finally, all clusters are graded according to the unit weight gain of the product to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set. Then, a machine learning clustering analysis method is used to realize intelligent analysis of the production data of the carbon-carbon composite material deposition furnace to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set, so as to provide a basis for the production process of the carbon-carbon composite material deposition furnace product and significantly improve the qualified rate of the carbon-carbon composite material deposition furnace product. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A method flow chart of the intelligent analysis method provided by Embodiment 1 of the present invention;

[0018] Figure 2 A flow chart of the clustering method provided in Example 1 of the present invention;

[0019] Figure 3 A correlation diagram between the production data provided in Example 1 of the present invention and the initial Q table;

[0020] Figure 4This is a system block diagram of the intelligent analysis system provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Since each production factor has a great influence on the final unit weight gain of the product during the actual production process of the carbon-carbon composite material deposition furnace, if the value of each production factor is not set properly, the final product will be unqualified. In order to obtain better economic benefits and better product quality, the purpose of the present invention is to provide a carbon-carbon composite material deposition furnace production data intelligent analysis method and system, by clustering the production factors using a clustering algorithm based on learning automata, so that the value of each production factor is reasonably set on the premise that the unit weight gain of the product is qualified, so as to improve the qualified rate of the product and better control the product quality during the production process.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Embodiment 1:

[0025] Carbon-carbon composite material deposition furnace products are reinforced by graphitization on the basis of carbon fiber. The reinforced products have the characteristics of high temperature resistance, long service life, and can be used repeatedly. They have obvious cost-effectiveness and a wide range of applications. When using chemical vapor deposition furnace technology to produce carbon-carbon composite materials, there are defects such as long product production cycle and high cost. Based on this defect, how to improve the qualification rate of carbon-carbon composite products is a very important issue.

[0026] The unit weight gain value is mainly used as the product quality evaluation standard for carbon-carbon composite materials, and carbon-carbon composite products will be affected by many factors during the production process. Therefore, how to set the value of each production factor to maximize the final unit weight gain is a very important issue. At present, the production process of carbon-carbon composite deposition furnace is mainly set based on experience, or based on factor analysis, a statistical analysis method is used in combination with a simulation experiment process for analysis. The statistical analysis method is only based on the existing experimental data, using the correlation between product quality and each production factor for analysis. The simulation analysis method generally uses simulation to design the heating system to simulate the production process of the carbon-carbon composite deposition furnace, and then continuously adjusts the value of each production factor to obtain the value range of the production factor in the product production process, and finally obtains a qualified product. However, these methods cannot provide a theoretical basis and setting method for determining the value of a specific production factor to improve the qualified rate. The method in this embodiment can analyze the detailed setting values ​​of each production factor in the actual production process under the premise of ensuring the qualified product from the existing production data. This embodiment is mainly based on the factor setting values ​​in the production process and the final unit weight gain, and uses an improved clustering algorithm to automatically analyze the production factors, and find reasonable factor values ​​in the product production process under the premise of qualified unit weight gain to improve the qualified rate of product production.

[0027] The purpose of this embodiment is to use a machine learning method to realize the intelligent analysis of the production data of a carbon-carbon composite material deposition furnace. As an unsupervised machine learning method, the clustering algorithm can realize the purpose of automatic clustering analysis of the data without labeling the data, so as to divide the input production data into multiple different clusters according to the specified rules, and realize the purpose of automatic analysis of the production data. The general implementation process of using the clustering algorithm for production data analysis is: first, preprocessing the production data, which mainly includes removing the data with missing information and performing numerical operations on symbolic data. Second, selecting a suitable method for calculating the distance and a similarity metric. Third, selecting a suitable clustering algorithm according to the applicable scenario of the data and the characteristics of the data, clustering the data, so that data with similar characteristics can be divided into the same cluster as much as possible. Through cluster analysis, the optimal factor batching value is found and timely feedback is given, and at the same time, a better raw material combination is found according to the product indicators to avoid a poor raw material combination, and finally the optimal ratio scheme is obtained according to the cluster analysis.

[0028] This embodiment is used to provide a method for intelligent analysis of carbon-carbon composite material production data. Figure 1 As shown, the intelligent analysis method includes:

[0029] S1: Acquire multiple production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace;

[0030] The production data includes all deposition furnace raw material data used in the production process of carbon-carbon composite products and the unit weight gain of the products after the carbon-carbon composite products are completed. The deposition furnace raw material data refers to the values ​​of various production factors in the production process of carbon-carbon composite products, where the production factors include product position, temperature, methane flow rate and deposition time, which are all factors that have a great influence on the unit weight gain of the product. The product position refers to the placement position of the carbon-carbon composite material in the chemical vapor deposition furnace. The unit weight gain of the product is used as an evaluation indicator to evaluate the values ​​of various production factors in the production process. According to the different unit weight gains of the products, the production factors are divided into different levels.

[0031] After obtaining a plurality of production data, the intelligent analysis method of this embodiment also includes: preprocessing the deposition furnace raw material data to obtain preprocessed raw material data, and using the preprocessed raw material data and the unit weight gain of the product as new production data to execute S2. The preprocessing mainly includes eliminating missing data, digitizing and normalizing symbolic data. Among them, eliminating missing data means that if a part of the value is missing in a production data, the production data is deleted. The value of the production factor of product position is symbolic data, and its digitization method is to give different numbering information corresponding to each placement position. Specifically, it can be converted into digits according to the corresponding relationship of position 1-1 corresponding to 1, position 1-2 corresponding to 2, position 1-3 corresponding to 3, position 1-outside corresponding to 4, position 2-1 corresponding to 5, position 2-2 corresponding to 6, position 2-3 corresponding to 7, and position 2-outside corresponding to 8.

[0032] Table 1 is an example of multiple original production data, and Table 2 is the production data obtained after preprocessing.

[0033] Table 1

[0034] Process Name Sedimentation time h Methane flow rate L temperature Product Location Unit weight gain kg / h 1 80 80 1120 1-1 0.29 1 80 80 1120 1-2 0.33 1 80 80 1120 1-Outside 0.20 1 80 80 1120 2-1 0.08 1 80 80 1120 2-2 0.18 1 80 2-Outside 0.12 1 80 80 1130 1-1 0.36 1 80 80 1130 1-2 0.23

[0035] Table 2

[0036] Process Name Sedimentation time h Methane flow rate L temperature Furnace loading position Unit weight gain kg / h 1 80 80 1120 1 0.29 1 80 80 1120 2 0.33 1 80 80 1120 4 0.20 1 80 80 1120 5 0.08 1 80 80 1120 6 0.18 1 80 80 1130 1 0.36 1 80 80 1130 2 0.23

[0037] After eliminating missing data and converting symbolic data into numerical data, the preprocessing of this embodiment may also include normalizing the processed production factors. Normalization mainly maps the data to between 0 and 1, which can make data processing more convenient and quick.

[0038] The formula used for normalization is as follows:

[0039]

[0040] In formula (1), X norm is the normalized production factor data; X 原 is the original production factor data; X max , X min They are respectively the maximum and minimum values ​​of all data of this type of production factors.

[0041] It should be noted that normalization is performed on each production factor separately.

[0042] S2: clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters;

[0043] The clustering algorithm based on learning automata in this embodiment is a new clustering algorithm proposed by combining the fuzzy c-means clustering algorithm and the learning automata. The fuzzy c-means clustering algorithm (FCM) obtains the membership of each sample point to all class centers by optimizing the objective function, uses the Euclidean distance to measure the similarity, assigns similar samples to the same cluster, and assigns different samples to different clusters. Learning automata (LA) are a type of algorithm in reinforcement learning that runs in a probability space and learns the optimal value by continuously interacting with an unknown environment. The learning automata adjust the probability distribution of each action being selected according to the environmental feedback (reward or punishment), and make the probability value eventually converge to the optimal action.

[0044] like Figure 2 As shown, S2 may include:

[0045] S21: Initialize the Q table of the learning automaton to obtain an initial Q table;

[0046] Take each production data as a sample, assign the probability of each sample belonging to each cluster with equal probability, and construct a Q table with J rows and K columns to form the initial Q table.

[0047] like Figure 3 As shown, it is a correlation diagram between production data and the initial Q table. Figure 3 In this example, X is the set of all deposition furnace raw material data, X = [x1, x2, ..., x j , ..., x J ] T , which is a J×L matrix; x j is the jth deposition furnace raw material data; j=1, 2, ..., J; J is the total number of deposition furnace raw material data, that is, the total number of production data obtained in S1; x j =[x j1 , xj2 , ..., x jl , ..., x jL ]; x jl is the value of the lth production factor of the jth deposition furnace raw material data; l = 1, 2, ..., L; L is the total number of production factors, and in this embodiment, L is taken as 4. Y is the set of all product unit weight gains, Y = [y1, y2, ..., y j , ..., y J ] T , which is a J×1 matrix.

[0048] In this embodiment, the action set of the learning automaton is A = [a1, a2, ..., a k , ..., a K ] T , which is a K×1 matrix, the action set includes K actions, k=1, 2, ...K; establish a mapping relationship between it and the number of clusters, and the K actions correspond to the K clusters of the clustering results. The element q in the Q table jk represents the membership of the j-th deposition furnace raw material data to the k-th cluster; q jk include and Indicates that the jth deposition furnace raw material data selects the kth action, is the probability of selecting the kth action for the jth deposition furnace raw material data corresponding to the learning automaton, q jk In That is, the probability that the raw material data of the jth deposition furnace belongs to the kth cluster. The greater the probability, the greater the possibility that the raw material data of the jth deposition furnace belongs to the kth cluster. jk The value of The values ​​are equal.

[0049] S22: Calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data;

[0050] Specifically, S22 may include:

[0051] 1) According to the initial Q table and deposition furnace raw material data, a plurality of cluster centers are obtained using a cluster center calculation formula, and an objective function value is obtained using an objective function;

[0052] The cluster center calculation formula is as follows:

[0053]

[0054] In formula (2), c k is the kth cluster center; m is the fuzzy coefficient, m>1.

[0055] The objective function is as follows:

[0056]

[0057] In formula (3), J(Q, c1, c2, ... c k , ...c K ) is the objective function value; d jk is the Euclidean distance between the jth sample and the kth cluster center; c kl is the value of the lth production factor of the kth cluster center.

[0058] 2) Divide all deposition furnace raw material data into multiple clusters according to the initial Q table;

[0059] Specifically, for each deposition furnace raw material data, the cluster corresponding to the maximum probability of the deposition furnace raw material data is determined according to the initial Q table, and then the deposition furnace raw material data is divided into the cluster. For example, for the jth deposition furnace raw material data, if q jk is the maximum probability value, which refers to all the probabilities corresponding to the raw material data of the jth deposition furnace, that is, q j1 ,q j2 ,...q jK The maximum value in , that is, the maximum value in the j-th row element of the initial Q table, then the k-th cluster is selected as the cluster to which the j-th original data belongs, and then all the deposition furnace raw material data are divided into K clusters.

[0060] Since the initial Q table is obtained by equal probability distribution in the first iteration, all probability values ​​in the initial Q table are equal at this time, so all deposition furnace raw material data are randomly divided into K clusters in the first iteration.

[0061] 3) According to the cluster center, the average distance within each cluster is calculated using the average distance within the cluster calculation formula.

[0062] The calculation formula for the average distance within a class is as follows:

[0063]

[0064] In formula (4), Adis(c k ) is the average intra-class distance between the raw material data of each deposition furnace in the kth cluster; |c k | is the number of deposition furnace raw material data included in the kth cluster; i = 1, 2, ..., |c k |;x i is the raw material data of the i-th deposition furnace belonging to the k-th cluster; il is the value of the lth production factor of the i-th deposition furnace raw material data belonging to the k-th cluster.

[0065] S23: Compare the objective function value and the average distance within the class obtained in the current iteration with the objective function value and the average distance within the class obtained in the previous iteration to obtain a comparison result; and update the initial Q table using a learning automaton according to the comparison result to obtain an updated Q table;

[0066] Specifically, S23 may include:

[0067] 1) Determine whether the objective function value obtained in the current iteration is less than the objective function value obtained in the previous iteration, and obtain a first determination result;

[0068] 2) For each cluster, determine whether the average intra-class distance obtained in the current iteration is less than the average intra-class distance obtained in the previous iteration, and obtain a second determination result;

[0069] For the first iteration, the objective function value and the average distance within the class obtained in the previous iteration are both the initial setting values. Specifically, the algorithm randomly selects the cluster center and randomly clusters, and then uses the above formula (3) and the above formula (4) to calculate the initial objective function value and the initial average distance within the class.

[0070] 3) If the first judgment result is yes and the second judgment result is yes, the learning automaton updates the partial initial Q table corresponding to the cluster according to the reward signal calculation formula; otherwise, the learning automaton updates the partial initial Q table corresponding to the cluster according to the penalty signal calculation formula; after updating the partial initial Q tables corresponding to all clusters, an updated Q table is obtained.

[0071] The learning automaton rewards or penalizes the probability value according to the judgment result. The initial Q table corresponding to a cluster is the row corresponding to all the deposition furnace raw material data included in the cluster in the initial Q table. If the new iteration reduces the objective function value and the average distance within the class at the same time, it means that this iteration makes the clustering effect better, and a reward signal is given at this time, otherwise a penalty signal is given.

[0072] The reward signal calculation formula is as follows:

[0073]

[0074] In formula (5), p j (n) is the maximum probability of the j-th deposition furnace raw material data in the n-th iteration; p j (n+1) is p j (n) is the updated probability value of the position; a is the reward coefficient; if in the previous iteration, the j-th deposition furnace raw material data belongs to the k-th cluster, and in this iteration, the j-th deposition furnace raw material data still belongs to the k-th cluster, then j=k in formula (5), otherwise, j≠k.

[0075] The penalty signal calculation formula is as follows:

[0076]

[0077] In formula (6), b is the penalty coefficient; r is the total number of actions, that is, the K value at the time of initialization; the value range of the reward coefficient and the penalty coefficient is [0, 1], and is generally 0.5.

[0078] S24: Determine whether a preset termination condition is reached;

[0079] The preset termination condition is that the current iteration number reaches the maximum iteration number or the objective function value obtained by the current iteration is less than the preset threshold.

[0080] S25: If yes, then the iteration ends, and all the production data are divided into multiple clusters according to the updated Q table;

[0081] That is, if the current iteration number reaches the maximum iteration number or the objective function value obtained by the current iteration is less than the preset threshold, the iteration ends. For each production data, the cluster corresponding to the maximum probability of the production data is determined according to the updated Q table, and the cluster is used as the cluster to which the production data belongs. That is, the corresponding column of the maximum probability of the corresponding row of each production data in the updated Q table is selected as the cluster to which the production data belongs, and then all production data are divided into multiple clusters to obtain clustering results.

[0082] S26: If not, continue iterating, use the updated Q table as the initial Q table in the next iteration, and return to step S22 of "calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data".

[0083] This embodiment uses a reward or penalty method according to a learning automaton to update the probability value of each sample in the Q table, and divides all production data into multiple clusters according to the maximum probability after the iteration, thereby completing clustering.

[0084] S3: Classifying all the clusters according to the unit weight gain of the products to determine a high-quality production factor set, a qualified production factor set and a low-quality production factor set.

[0085] Specifically, the average unit weight gain of the products in all the production data included in each cluster is calculated to obtain the average unit weight gain corresponding to each cluster. For each cluster, it is determined whether the average unit weight gain is greater than the first weight gain threshold to obtain a third judgment result. If the third judgment result is yes, all the deposition furnace raw material data included in the cluster are high-quality production factor sets, and all the deposition furnace raw material data in the cluster are stored in the high-quality production factor database. If the third judgment result is no, it is determined whether the average unit weight gain is less than the second weight gain threshold. If so, all the deposition furnace raw material data included in the cluster are low-quality production factor sets, and all the deposition furnace raw material data in the cluster are stored in the low-quality production factor database, providing a basis for early warning of subsequent product production. Otherwise, that is, when the average unit weight gain is greater than or equal to the second weight gain threshold, but less than or equal to the first weight gain threshold, all the deposition furnace raw material data included in the cluster are qualified production factor sets, and all the deposition furnace raw material data in the cluster are stored in the qualified production factor database. Then, all the raw material data are graded based on the average unit weight gain, the first weight gain threshold, and the second weight gain threshold corresponding to each cluster. The first weight gain threshold is greater than the second weight gain threshold, and both are greater than zero.

[0086] This embodiment combines the clustering algorithm in machine learning with the learning automaton in reinforcement learning, improves the fuzzy clustering algorithm (FCM), and innovatively proposes a method for intelligent analysis of production data of carbon-carbon composite material deposition furnace using unsupervised machine learning clustering algorithm, uses fuzzy clustering idea to perform intelligent analysis of production data, and uses learning automaton to calculate fuzzy membership. Then, the improved fuzzy clustering algorithm is used to process the factor values ​​in the production process of carbon-carbon composite material deposition furnace products, and the clustering method is innovatively used to screen the optimal production factor values ​​in the production process of carbon-carbon composite material deposition furnace products. According to the clustering results, the high-quality, qualified and poor-quality factor setting values ​​of the production products corresponding to different positions in the production process are determined on the premise that the unit weight gain of the product is qualified, and the values ​​are stored in the corresponding database for reference in the production process, so as to achieve the purpose of improving the product qualification rate and improving production efficiency.

[0087] The advantages of the method of this embodiment include: 1. In order to determine the optimal value of each factor value in the production process, the production information of products with qualified unit weight gain values ​​is separated in combination with the qualified unit weight gain value of the product, and the specific factor value of each qualified product production process is provided. The product production information is clustered by a clustering algorithm, and each cluster is judged as a high-quality product, a qualified product or a low-quality product according to the cluster center of each cluster combined with the qualified unit weight gain value, and the high-quality product factor setting value, the qualified product factor setting value and the low-quality product factor setting value are determined accordingly. The low-quality product factor setting value is stored in the designated low-quality product factor database. In the subsequent production process, as a comparison database information, the new production data matches the low-quality product factor database information, and an alarm message is issued to correct the product factor setting value. 2. The fuzzy membership degree is calculated by using a learning automaton, which can fully consider the influence of each factor value of each qualified product on the final clustering result, and can accurately give the optimal value of each factor in the production product under the premise of qualified unit weight gain while improving the clustering effect, thereby obtaining the best carbon-carbon composite material deposition furnace production data setting value.

[0088] Embodiment 2:

[0089] This embodiment is used to provide a carbon-carbon composite material deposition furnace production data intelligent analysis system. Figure 4 As shown, the intelligent analysis system includes:

[0090] The acquisition module M1 is used to acquire a plurality of production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace;

[0091] A clustering module M2, used for clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters;

[0092] The analysis module M3 is used to classify all the clusters according to the unit weight gain of the products, and determine a high-quality production factor set, a qualified production factor set and a low-quality production factor set.

[0093] The intelligent analysis system of this embodiment further includes a storage module for classifying and storing the production data corresponding to the high-quality production factor set, the qualified production factor set, and the low-quality production factor set.

[0094] This embodiment first quantifies the production factors, and then analyzes the production factors according to actual production needs using a clustering algorithm based on a learning automaton, so that the production data of the carbon-carbon composite material deposition furnace is divided into different categories according to the quality of the unit weight gain of the products. The production factors that have a greater impact on the unit weight gain of the products are obtained, thereby finding the production factor value with the best ingredient ratio to maximize the unit weight gain and significantly improve the pass rate.

[0095] Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0096] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for intelligent analysis of production data of a carbon-carbon composite material deposition furnace, characterized in that: The intelligent analysis method comprises: Acquire multiple production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace; Clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters; Classifying all the clusters according to the unit weight gain of the products to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set; The production factors include product location, temperature, methane flow rate and deposition time; The clustering algorithm based on learning automata is used to cluster all the production data to obtain multiple clusters, specifically including: Initialize the Q table of the learning automaton to obtain an initial Q table; Calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data; Compare the objective function value and the average distance within the class obtained in the current iteration with the objective function value and the average distance within the class obtained in the previous iteration to obtain a comparison result; based on the comparison result, update the initial Q table using a learning automaton to obtain an updated Q table; Determine whether the preset termination condition is met; If yes, then the iteration ends, and all the production data are divided into a plurality of clusters according to the updated Q table; If not, continue iterating, use the updated Q table as the initial Q table in the next iteration, and return to the step of "calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data".

2. The intelligent analysis method according to claim 1, characterized in that: After obtaining a plurality of production data, the intelligent analysis method further includes: The deposition furnace raw material data is preprocessed to obtain preprocessed raw material data, and the preprocessed raw material data and the unit weight gain of the product are used as new production data; the preprocessing includes eliminating missing data and digitizing and normalizing symbolic data.

3. The intelligent analysis method according to claim 1, characterized in that: The calculating of the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data specifically includes: According to the initial Q table and the deposition furnace raw material data, a plurality of cluster centers are obtained using a cluster center calculation formula, and an objective function value is obtained using an objective function; Divide all the deposition furnace raw material data into a plurality of clusters according to the initial Q table; According to the cluster center, the intra-class average distance of each cluster is obtained by using the intra-class average distance calculation formula.

4. The intelligent analysis method according to claim 3, characterized in that: The objective function value and the average distance within the class obtained by the current iteration are compared with the objective function value and the average distance within the class obtained by the previous iteration to obtain a comparison result; According to the comparison result, the initial Q table is updated by using a learning automaton to obtain an updated Q table, which specifically includes: Determine whether the objective function value obtained in the current iteration is less than the objective function value obtained in the previous iteration, and obtain a first determination result; For each of the clusters, determining whether the average intra-class distance obtained in the current iteration is smaller than the average intra-class distance obtained in the previous iteration, and obtaining a second determination result; If the first judgment result is yes and the second judgment result is yes, the learning automaton updates the initial Q table corresponding to the cluster according to the reward signal calculation formula; Otherwise, the learning automaton updates the initial Q table corresponding to the cluster according to the penalty signal calculation formula; After updating the parts of the initial Q-tables corresponding to all the clusters, an updated Q-table is obtained.

5. The intelligent analysis method according to claim 1, characterized in that: The preset termination condition is that the current iteration number reaches the maximum iteration number or the objective function value obtained by the current iteration is less than a preset threshold.

6. The intelligent analysis method according to claim 1, characterized in that: The dividing all the production data into a plurality of clusters according to the updated Q table specifically includes: For each of the production data, a cluster corresponding to the maximum probability of the production data is determined according to the updated Q table, and the cluster is used as the cluster to which the production data belongs.

7. The intelligent analysis method according to claim 1, characterized in that: The step of classifying all the clusters according to the unit weight gain of the product to determine a high-quality production factor set, a qualified production factor set, and a low-quality production factor set specifically includes: Calculating the average unit weight gain of all the products included in the cluster to obtain the average unit weight gain corresponding to each cluster; Determine whether the average unit weight gain is greater than a first weight gain threshold, and obtain a third determination result; If the third judgment result is yes, then all the deposition furnace raw material data included in the cluster are high-quality production factor sets; If the third judgment result is no, determining whether the average unit weight gain is less than a second weight gain threshold; If yes, then all the deposition furnace raw material data included in the cluster are a set of inferior production factors; Otherwise, all the deposition furnace raw material data included in the cluster are qualified production factor sets.

8. An intelligent analysis system for production data of a carbon-carbon composite material deposition furnace, characterized in that: The intelligent analysis system comprises: An acquisition module, used to acquire a plurality of production data; the production data includes deposition furnace raw material data and product unit weight gain; the deposition furnace raw material data is the value of each production factor in the production process of the carbon-carbon composite material deposition furnace; A clustering module, used for clustering all the production data using a clustering algorithm based on a learning automaton to obtain a plurality of clusters; An analysis module, used for classifying all the clusters according to the unit weight gain of the products, and determining a high-quality production factor set, a qualified production factor set and a low-quality production factor set; The production factors include product location, temperature, methane flow rate and deposition time; The clustering algorithm based on learning automata is used to cluster all the production data to obtain multiple clusters, specifically including: Initialize the Q table of the learning automaton to obtain an initial Q table; Calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data; Compare the objective function value and the average distance within the class obtained in the current iteration with the objective function value and the average distance within the class obtained in the previous iteration to obtain a comparison result; based on the comparison result, update the initial Q table using a learning automaton to obtain an updated Q table; Determine whether the preset termination condition is met; If yes, then the iteration ends, and all the production data are divided into a plurality of clusters according to the updated Q table; If not, continue iterating, use the updated Q table as the initial Q table in the next iteration, and return to the step of "calculating the objective function value and the average distance within the class according to the initial Q table and the deposition furnace raw material data".