Bayesian network-based wire drawing process optimization method and device

Abstract

The invention relates to the technical field of wire drawing of metal wires, in particular to a wire drawing process optimization method and device based on a Bayesian network, and the method comprises the steps: obtaining wire drawing parameters of the metal wires, and carrying out the preprocessing to obtain an initial data set; analyzing a causal relationship among the wire drawing parameters based on existing knowledge, constructing an initial Bayesian network structure, and learning the initial Bayesian network structure based on a structure learning algorithm of score search to obtain a Bayesian network structure most suitable for the initial data set; performing parameter learning on the basis of the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model so as to perform logical reasoning on the input process parameters to obtain a prediction result of the performance parameters; and performing reverse traceability based on the constructed Bayesian network model, and dynamically adjusting process parameters. Therefore, the complex relationship among a plurality of wire drawing parameters can be comprehensively considered, and effective optimization and quality control of the process are realized.

Description

Technical Field

[0001] The present application relates to the technical field of metal wire drawing, and in particular to a wire drawing process optimization method and device based on a Bayesian network. Background Art

[0002] Metal wire drawing is a process that uses external force to force metal wire through the hole of the wire drawing die to obtain metal wire with the required cross-sectional shape and size. Through metal material drawing, the higher requirements of high-end manufacturing industry for metal wire performance can be met.

[0003] At present, the wire drawing process has the following problems: parameter adjustment mainly depends on the operator's experience, lacks a systematic parameter optimization method, and the mutual influence between parameters makes it difficult to quantify quality control issues; product quality fluctuates greatly, the frequency of wire breakage is high, the surface quality is unstable, and straightness control is difficult; frequent shutdowns for adjustments, short mold life, and serious energy waste; lack of intelligent control means, unable to achieve predictive maintenance, and difficult to optimize the system. Summary of the invention

[0004] In order to overcome the deficiencies in the prior art, the present application provides a wire drawing process optimization method and device based on a Bayesian network, which can comprehensively consider the complex relationship between multiple wire drawing parameters and achieve effective optimization and quality control of the process.

[0005] In a first aspect, the present application provides a wire drawing process optimization method based on a Bayesian network, the method comprising the following steps:

[0006] Obtaining wire drawing parameters of the metal wire and performing preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality;

[0007] Analyze the causal relationship between the drawing parameters based on existing knowledge, and construct an initial Bayesian network structure, and learn the initial Bayesian network structure based on a score search structure learning algorithm to obtain a Bayesian network structure that best suits the initial data set;

[0008] Performing parameter learning based on the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model to perform logical reasoning on the input process parameters to obtain prediction results of the performance parameters;

[0009] Reverse tracing is performed based on the constructed Bayesian network model to dynamically adjust process parameters.

[0010] In a possible implementation, preprocessing the obtained wire drawing parameters of the metal wire includes equal width discretization processing, missing data and abnormal data processing, and normalization processing.

[0011] In a possible implementation, the scoring search-based structure learning algorithm learns the initial Bayesian network structure to obtain a Bayesian network structure that best suits the initial data set, including the following steps:

[0012] Starting from the initial Bayesian network structure, a possible Bayesian network structure is searched by a greedy method;

[0013] The Bayesian information criterion is used as the scoring function, and each possible Bayesian network structure is evaluated, and the next round of search direction is carried out according to the evaluation results, until the Bayesian network structure with the smallest evaluation result is found, which is used as the Bayesian network structure most suitable for the initial data set.

[0014] In a possible implementation, parameter learning is performed based on the initial data set and the obtained Bayesian network structure, including the following steps:

[0015] Initializing the parameters of the obtained Bayesian network structure; wherein the initialized parameters include the parameters in the conditional probability table of each node in the Bayesian network structure;

[0016] Calculating the likelihood function of the initial data set;

[0017] The expectation maximization algorithm is used to update the parameters of the Bayesian network structure to maximize the likelihood function.

[0018] In a possible implementation, the constructed Bayesian network model uses a joint tree inference engine for logical reasoning.

[0019] In a possible implementation, the reverse tracing based on the constructed Bayesian network model and the dynamic adjustment of process parameters include the following steps:

[0020] The performance parameter is set as the observed data, and the observed data is input into the constructed Bayesian network model as an evidence variable;

[0021] Based on the constructed Bayesian network model, logical reasoning is performed to calculate the posterior probability of the target variable; wherein the process parameters are used as the target variables;

[0022] According to the calculated posterior probability, the process parameters are dynamically adjusted.

[0023] In a possible implementation, the initial data set is updated regularly using the wire drawing parameter data newly collected during the production process;

[0024] The local distribution function of the current Bayesian network model is updated using the updated initial data set.

[0025] In a second aspect, the present application provides a wire drawing process optimization device based on a Bayesian network, the device comprising:

[0026] An acquisition module is used to acquire the wire drawing parameters of the metal wire and perform preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, wherein the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; and the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality;

[0027] A construction module is used to analyze the causal relationship between the various drawing parameters based on existing knowledge, and to construct an initial Bayesian network structure, and to learn the initial Bayesian network structure based on a structure learning algorithm of score search to obtain a Bayesian network structure that best suits the initial data set;

[0028] A learning module, used for performing parameter learning based on the initial data set and the obtained Bayesian network structure, obtaining a constructed Bayesian network model, so as to perform logical reasoning on the input process parameters and obtain prediction results of the performance parameters;

[0029] The optimization module is used to perform reverse tracing based on the constructed Bayesian network model and dynamically adjust process parameters.

[0030] In a third aspect, the present application provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the wire drawing process optimization method based on the Bayesian network as described in any one of the first aspects are performed.

[0031] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the wire drawing process optimization method based on a Bayesian network as described in any one of the first aspects are executed.

[0032] The present embodiment provides a wire drawing process optimization method and device based on Bayesian network, which obtains the wire drawing parameters of metal wires and performs preprocessing to obtain an initial data set; wherein the wire drawing parameters of metal wires include process parameters and performance parameters, and the causal relationship between each of the wire drawing parameters is analyzed based on existing knowledge, and an initial Bayesian network structure is constructed, and the initial Bayesian network structure is learned based on a structure learning algorithm based on a scoring search to obtain a Bayesian network structure that best suits the initial data set; parameter learning is performed on the basis of the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model, so as to perform logical reasoning on the input process parameters and obtain the prediction results of the performance parameters; reverse tracing is performed based on the constructed Bayesian network model to dynamically adjust the process parameters. Thus, the wire drawing process parameters can be dynamically optimized through Bayesian network analysis, and the wire break frequency, surface quality, and wire curvature can be accurately predicted and controlled, and the mold life can be extended, downtime can be reduced, production efficiency and system stability can be improved, and dependence on operator experience can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 A flow chart of a wire drawing process optimization method based on a Bayesian network according to an embodiment of the present application is shown;

[0035] Figure 2 A schematic diagram showing an initial Bayesian network structure according to an embodiment of the present application is shown;

[0036] Figure 3 A flowchart showing the structure learning algorithm based on scoring search described in an embodiment of the present application for learning the initial Bayesian network structure is shown;

[0037] Figure 4 A flowchart of performing parameter learning based on the initial data set and the obtained Bayesian network structure according to an embodiment of the present application is shown;

[0038] Figure 5 A flowchart of reverse tracing and dynamically adjusting process parameters based on a constructed Bayesian network model according to an embodiment of the present application is shown;

[0039] Figure 6 A schematic diagram of the structure of a wire drawing process optimization device based on a Bayesian network according to an embodiment of the present application is shown;

[0040] Figure 7 A structural block diagram of an electronic device described in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0041] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0042] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0043] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0044] In view of the technical problems raised by the background technology, the present application provides a wire drawing process optimization method and device based on a Bayesian network, which can comprehensively consider the complex relationship between multiple wire drawing parameters and achieve effective optimization and quality control of the process.

[0045] In one embodiment, see the attached specification Figure 1 The present application provides a wire drawing process optimization method based on a Bayesian network, the method comprising the following steps:

[0046] S1. Obtaining wire drawing parameters of a metal wire and performing preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, wherein the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; and the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality;

[0047] S2. Analyze the causal relationship between the wire drawing parameters based on existing knowledge, and construct an initial Bayesian network structure, and learn the initial Bayesian network structure based on a score search structure learning algorithm to obtain a Bayesian network structure that best suits the initial data set;

[0048] S3, performing parameter learning based on the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model to perform logical reasoning on the input process parameters to obtain prediction results of the performance parameters;

[0049] S4. Perform reverse tracing based on the constructed Bayesian network model and dynamically adjust process parameters.

[0050] In step S1, the parameter space is defined according to the knowledge in the field of heat treatment and wire drawing, including annealing parameters (annealing temperature °C, annealing speed (m / min) / annealing time min), wire drawing speed (m / min), back tension (N), lubricant concentration (g / L), die wire drawing length (km), die quality (divided into 1 to 4, 4 levels according to the degree of wear of the die hole), wire break frequency (number of wire breakages during wire drawing, times / shift), wire surface quality (divided into 1 to 4, 4 levels according to surface quality), wire curvature (rebound height per 1m mm / m). In this application, annealing parameters, wire drawing speed, back tension, and lubricant concentration are used as process parameters, and wire break frequency, wire surface quality, wire curvature, die wire drawing length, and die quality are used as performance parameters, that is, optimization targets.

[0051] When collecting data on various wire drawing parameters, the actual temperature value during the annealing process can be obtained through thermocouples or thermistor temperature sensors installed at key positions of the annealing furnace; the wire drawing speed of the wire can be obtained through a rotary encoder installed at the traction device of the wire drawing machine; the back tension during the wire drawing process can be obtained through a tension sensor installed on the wire path before and after the wire drawing die; the lubricating fluid concentration can be obtained through a wire concentration detection sensor installed on the lubricating fluid circulation system pipeline; the wire drawing length of the die can be obtained through an encoder installed near the die installation position of the wire drawing machine; the quality grade of the die can be determined through manual regular inspection combined with image recognition technology; the wire breakage frequency can be obtained through a wire breakage detection sensor installed at key positions of the wire drawing production line; the surface quality of the wire can be evaluated by setting up surface defect detection equipment on the production line; the curl of the wire can be obtained through a curling test device set at the finished wire collection position of the production line.

[0052] Furthermore, the collected data of various wire drawing parameters are preprocessed to form a more standardized and reasonable initial data set, which is more suitable for building a Bayesian network prediction model and carrying out subsequent process parameter optimization and other work.

[0053] In one embodiment, the collected data of various drawing parameters are preprocessed including equal-width discretization, missing data and abnormal data processing, and normalization processing. Among them, equal-width discretization is to divide the continuous parameter value range into several equal-width intervals, each interval corresponds to a discrete value, the purpose of which is to facilitate the subsequent construction of the Bayesian network model and data analysis and processing. For example, for the process parameter annealing temperature, if its original value range is 200℃-600℃, it can be divided into multiple discrete intervals with 40℃ as an interval, and then different discrete values ​​(such as 1, 2, 3, etc.) are used to represent these intervals;

[0054] During the data collection process, some data may be missing due to sensor failure, transmission problems, etc. For missing data, the mean, median, etc. of the parameter are used to fill in the missing values. For example, if missing values ​​appear in some of the collected data of lubricant concentration, and the number of these missing values ​​is small, the average value of the lubricant concentration data in this time period can be used to fill in the missing values, so that the data records are complete and convenient for subsequent unified processing and analysis; Abnormal data refers to data that obviously deviates from the normal value range, which may be caused by sensor misreading, sudden equipment failure, etc. Abnormal data is identified by setting reasonable upper and lower limit thresholds. For example, for wire drawing speed, if the normal range is 10-100m / min, and 200m / min appears in the collected data, which obviously exceeds the reasonable range, it is judged as abnormal data and deleted directly;

[0055] Normalization is to convert the original data of each parameter into the range of [0,1], using the normalization formula The operation is performed, where y represents the normalized data, x identifies the original input data, and xmax and xmin represent the maximum and minimum values ​​of the input data, respectively. All parameter data are normalized in this way so that the data of different parameters have a uniform magnitude and distribution range. When constructing models such as Bayesian networks in the future, each parameter can participate in calculation and analysis at the same scale, avoiding adverse effects on model results due to large differences in data magnitude.

[0056] In step S2, first, based on the expertise in the field of heat treatment and wire drawing and past technical literature, the interaction relationship between the parameters is deeply analyzed to construct the initial Bayesian network structure. For example, the lubricant concentration directly affects the quality of the mold. The appropriate lubricant concentration can reduce the wear of the mold and extend the service life of the mold, thereby ensuring the quality of the mold; a higher lubricant concentration helps to extend the service life of the mold because it can reduce the friction loss of the mold during the working process; a high-quality mold can ensure that the wire has a good surface quality during the drawing process and reduce the generation of surface defects; the quality of the mold will affect the size of the back tension, and a good quality mold can make the back tension more stable and reasonable; as the use time of the mold increases, the wear of the mold will increase, which may cause the back tension to change and affect the stability of the drawing process; a mold used for a long time may affect the straightness of the wire, causing problems such as bending of the wire; the size of the back tension will affect the surface quality of the wire. Appropriate back tension helps to obtain better surface quality, while too large or too small back tension may cause surface defects; changes in back tension may require adjustment of annealing parameters to ensure the performance of the wire, because different back tensions may cause different internal stress states of the wire; poor wire straightness may increase the wire breakage rate during drawing, because bent or crooked wires are more susceptible to uneven stress and breakage during the drawing process; wires with poor straightness may limit the drawing speed, because wires with poor straightness are more likely to have problems during high-speed drawing; appropriate annealing parameters help to obtain good wire surface quality , because annealing can eliminate stress and defects inside the wire; too fast drawing speed may affect the surface quality of the wire, because high-speed drawing may cause scratches or other defects on the surface of the wire; annealing parameters will affect the mechanical properties of the wire, and then affect the drawing speed. Appropriate annealing parameters can make the wire have appropriate hardness and ductility, which is conducive to improving the drawing speed; too fast drawing speed may increase the wire breakage rate, because high-speed drawing may cause the wire to bear excessive stress; the drawing speed needs to be adjusted according to factors such as annealing parameters and wire straightness to ensure the smooth progress of the drawing process. In one embodiment, the initial Bayesian network structure constructed is as shown in the attached manual Figure 2 shown.

[0057] See the instruction manual Figure 3 The scoring search-based structure learning algorithm learns the initial Bayesian network structure to obtain the Bayesian network structure that best suits the initial data set, including the following steps:

[0058] S201, starting from the initial Bayesian network structure, using a greedy method to search for possible Bayesian network structures;

[0059] S202, using the Bayesian Information Criterion as a scoring function, and evaluating each possible Bayesian network structure, and conducting the next round of search direction according to the evaluation results, until the Bayesian network structure with the smallest evaluation result is found, and taking it as the Bayesian network structure that best suits the initial data set.

[0060] That is, after obtaining the initial Bayesian network structure, structural learning is required.

[0061] In step S201, starting from the constructed initial Bayesian network structure, a greedy method is used to search for possible Bayesian network structures. The greedy method is to select the option that currently appears to be the best in each search process, regardless of whether the overall optimal solution will be missed. In this scenario, starting from the initial network, each time a local modification is made to the network structure (such as adding, deleting or reversing an edge), a new possible Bayesian network structure is obtained. In step S202, the Bayesian Information Criterion (BIC) is used as a scoring function to evaluate the pros and cons of each possible structure searched. Among them, lnP(D|θ,M) is the likelihood function of data D under model M and parameter θ, d is the number of parameters of the model, and n is the number of samples. The purpose is to find a Bayesian network structure that can fit the data well and is not too complicated among many possible structures, so that the entire model has good generalization ability and interpretability. Every time a new possible structure is obtained, its score is calculated according to the BIC scoring function, and the next round of search direction is guided by the scoring result. This process continues until a structure with the smallest scoring function is found. This structure is considered to be the Bayesian network structure that best suits the current data set and problem.

[0062] See the instruction manual Figure 4 In step S3, parameter learning is performed based on the initial data set and the obtained Bayesian network structure, including the following steps:

[0063] S301, initializing the parameters of the obtained Bayesian network structure; wherein the initialized parameters include the parameters in the conditional probability table of each node in the Bayesian network structure;

[0064] S302, calculating the likelihood function of the initial data set;

[0065] S303: Using an expectation maximization algorithm to update the parameters of the Bayesian network structure to maximize the likelihood function.

[0066] Specifically, in step S301-step S303, the network parameters must first be initialized, with the focus on the parameter settings in the conditional probability table (CPT) of each node. The conditional probability table describes the probability of the current node taking different values ​​given different values ​​of its parent node. For example, for the "wire surface quality" node, its parent node may have "wire drawing speed" and "lubricant concentration", etc., then it is necessary to initialize the initial probability values ​​of "wire surface quality" at various levels (1-4 levels) when "wire drawing speed" takes different discrete values ​​and "lubricant concentration" also takes different discrete values. These initial values ​​can be given based on simple assumptions such as empirical data or uniform distribution, providing a starting point for subsequent iterative learning; the likelihood function represents the probability of the observed data appearing under the current network parameter settings. By substituting each observation record in the data set into the Bayesian network, the joint probability of the entire data set is calculated based on the conditional probability table of each node, that is, the value of the likelihood function, which reflects the degree of fit of the current parameter setting to the actual observation data and is an important basis for subsequent parameter updates; the expectation maximization (EM) algorithm is an iterative optimization algorithm used to find the maximum likelihood estimate of parameters in the presence of hidden variables (in the Bayesian network, the values ​​of some variables may not be directly observed, but they have an impact on the model, these are hidden variables). Through repeated iterations, the network parameters are continuously adjusted. When the change in the likelihood function value in several consecutive iterations is less than the set threshold, it is considered that the parameters have converged and the update process is completed. At this time, the network parameters obtained are the relatively optimal parameter estimates.

[0067] It should be noted that in this application, the constructed Bayesian network model uses a joint tree reasoning engine for logical reasoning.

[0068] First, the conditional probability distribution in the original Bayesian network is assigned to each cluster in the connection tree. The connection tree is a special data structure used for efficient reasoning calculations in the Bayesian network. Each cluster contains several variables, and its potential function is defined as the conditional probability product of the variables it contains. For example, if a cluster contains the variables "drawing speed", "reverse tension" and "wire surface quality", then the potential function of the cluster is the probability product of these three variables under the corresponding conditional probability table. In this way, the probability information in the Bayesian network is converted into the connection tree structure to prepare for subsequent message passing and reasoning; then, the message passing operation is performed in the connection tree, and each cluster sends messages to its adjacent clusters. The message passing process is actually the process of sharing and updating probability information between different clusters, so that each cluster can adjust its own probability distribution according to the received information. For example, a cluster related to "drawing speed" will integrate this information to update its estimate of the probability related to "drawing speed" after receiving messages from other clusters about variables such as "lubricant concentration". This process continues, and information is continuously propagated and integrated throughout the connection tree; the message transmission process will continue until all clusters converge, that is, the probability distribution of each cluster no longer changes significantly. Once convergence is completed, the marginal probability or conditional probability of the target variable (such as wire break frequency, wire surface quality, wire curl, mold quality, etc.) can be calculated by querying a specific cluster.

[0069] The actual parameters in production are used as evidence variables to construct the Bayesian network model. After the above-mentioned reasoning and analysis process, the key indicators such as wire breakage frequency, surface quality, wire curling and mold quality can be predicted.

[0070] In Bayesian networks, forward reasoning is usually to predict the probability of product quality related indicators with known process parameters. Reverse tracing is the other way around, starting from the observed product quality indicators and some production-related data, to infer the process parameters that may lead to these results. Figure 5 In step S4, reverse tracing is performed based on the constructed Bayesian network model to dynamically adjust the process parameters, including the following steps:

[0071] S401, setting the performance parameter as the observation data, and inputting the observation data as the evidence variable into the constructed Bayesian network model;

[0072] S402, performing logical reasoning based on the constructed Bayesian network model to calculate the posterior probability of the target variable; wherein the process parameter is used as the target variable;

[0073] S403. Dynamically adjust process parameters according to the calculated posterior probability.

[0074] Specifically, in step S401, the currently observed data such as the frequency of broken wires, the surface quality of the wires, the curvature of the wires, the quality of the molds, and the length of the molds are selected as evidence variables because they directly reflect the actual situation in the current production process and the quality status of the products, and find a suitable entry point for reverse tracing; in step S402, after the above evidence variables are input into the Bayesian network, the joint tree inference engine is used for message transmission and probability calculation. First, based on the existing conditional probability distribution of each cluster in the connection tree and the probability information after the previous transmission and convergence, combined with the value of the newly input evidence variable, the probability distribution of each cluster is readjusted. Then, by reversely transmitting messages in the connection tree (opposite to the message transmission direction during forward reasoning, transmitting messages from the cluster containing the observed data related variables to the cluster containing the target variables), the probability distribution of the target variables (annealing parameters, drawing speed, back tension, lubricant concentration) is gradually updated, and finally their posterior probabilities are calculated; in step S403, the process parameter adjustment strategy is formulated according to the calculated posterior probability of the target variables. For example, if it is found that a lower range of annealing temperature values ​​corresponds to a higher posterior probability, and the current product has a large silk thread curl and poor surface quality, and professional knowledge is used to know that appropriately increasing the annealing temperature may improve these situations, then it can be considered to adjust the annealing temperature to a higher reasonable range.

[0075] It should be noted that the adjustment of process parameters is not a one-time process, but a dynamic process. After adjusting some parameters, it is necessary to observe the various indicator data in the production process again (that is, re-obtain the data such as wire breakage frequency and wire surface quality), and input the new data into the Bayesian network for reasoning as evidence variables. According to the new reasoning results, the process parameters are further adjusted. This cycle is repeated to continuously optimize the configuration of process parameters, so as to gradually achieve the goal of optimizing wire quality and extending mold life, maintain efficient and stable operation, and produce high-quality wire products.

[0076] Further, in the present application, the initial data set is also regularly updated using the data of the wire drawing parameters newly collected during the production process; and the local distribution function of the current Bayesian network model is updated using the updated initial data set. This is because, during the production process, new data samples are constantly generated as time goes by and various internal and external factors change, and these new data contain the latest information on the current actual situation of the production process. Adding the newly collected data samples to the original data to form a new data set can enable the subsequent update of the Bayesian network to be based on more comprehensive and more practical information for the current production, ensuring that the network's characterization of the process conditions remains accurate. Among them, the local distribution function in the Bayesian network mainly involves the conditional probability distribution (CPD) of the node, which describes the probability of the current node taking different values ​​given different values ​​of its parent node, and is a key element that reflects the dependency relationship and probability transfer between variables. By continuously integrating new data into and updating the local distribution function of the Bayesian network, the dynamic update of the Bayesian network in the production process is achieved. This makes the entire wire drawing process optimization method based on Bayesian network have good adaptability and sustainability, and can continuously improve itself as production continues, always playing an important role in improving process level and product quality.

[0077] It can be seen that the wire drawing process optimization method based on Bayesian network provided in the present application can dynamically optimize the wire drawing process parameters through Bayesian network analysis, accurately predict and control the wire breakage frequency, surface quality, and wire curling, and extend the mold life, reduce downtime, improve production efficiency and system stability, and reduce dependence on operator experience.

[0078] Based on the same inventive concept, a wire drawing process optimization device based on a Bayesian network is also provided in an embodiment of the present application. Since the principle of solving the problem by the device in the embodiment of the present application is similar to the above-mentioned wire drawing process optimization method based on a Bayesian network in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0079] As the instruction manual Figure 6 As shown, an embodiment of the present application provides a wire drawing process optimization device based on a Bayesian network, the device comprising:

[0080] The acquisition module 601 is used to acquire the wire drawing parameters of the metal wire and perform preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, wherein the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality; the preprocessing of the acquired wire drawing parameters of the metal wire includes equal width discretization processing, missing data and abnormal data processing, and normalization processing;

[0081] A construction module 602 is used to analyze the causal relationship between the drawing parameters based on existing knowledge, and to construct an initial Bayesian network structure, and to learn the initial Bayesian network structure based on a structure learning algorithm of a scoring search to obtain a Bayesian network structure that best suits the initial data set;

[0082] A learning module 603 is used to perform parameter learning based on the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model to perform logical reasoning on the input process parameters to obtain prediction results of the performance parameters;

[0083] The optimization module 604 is used to perform reverse tracing based on the constructed Bayesian network model and dynamically adjust the process parameters.

[0084] In one embodiment, the construction module 602 learns the initial Bayesian network structure based on a structural learning algorithm of a scoring search to obtain a Bayesian network structure that best suits the initial data set, including: starting from the initial Bayesian network structure, using a greedy method to search for possible Bayesian network structures; using the Bayesian information criterion as a scoring function, and evaluating each possible Bayesian network structure, and conducting the next round of search direction based on the evaluation results, until the Bayesian network structure with the smallest evaluation result is found, and it is used as the Bayesian network structure that best suits the initial data set.

[0085] In one embodiment, the learning module 603 performs parameter learning based on the initial data set and the obtained Bayesian network structure, including: initializing the parameters of the obtained Bayesian network structure; wherein the initialized parameters include the parameters in the conditional probability table of each node in the Bayesian network structure; calculating the likelihood function of the initial data set; using the expectation maximization algorithm to update the parameters of the Bayesian network structure to maximize the likelihood function; wherein the constructed Bayesian network model uses a joint tree inference engine for logical reasoning.

[0086] In one embodiment, the optimization module 604 performs reverse tracing based on the constructed Bayesian network model and dynamically adjusts the process parameters, including: setting the performance parameters as observation data, and inputting the observation data as evidence variables into the constructed Bayesian network model; performing logical reasoning based on the constructed Bayesian network model to calculate the posterior probability of the target variable; wherein the process parameters are used as the target variables; and the process parameters are dynamically adjusted according to the calculated posterior probability.

[0087] In one embodiment, the device further comprises:

[0088] The updating module is used to regularly update the initial data set with the newly collected wire drawing parameter data during the production process; and to update the local distribution function of the current Bayesian network model with the updated initial data set.

[0089] The present application provides a wire drawing process optimization device based on Bayesian network, which obtains the wire drawing parameters of metal wires through an acquisition module, and performs preprocessing to obtain an initial data set; wherein, the wire drawing parameters of metal wires include process parameters and performance parameters, and the causal relationship between each of the wire drawing parameters is analyzed based on existing knowledge through a construction module, and an initial Bayesian network structure is constructed, and the initial Bayesian network structure is learned based on a structure learning algorithm based on a scoring search, and a Bayesian network structure that best suits the initial data set is obtained; a learning module performs parameter learning on the basis of the initial data set and the obtained Bayesian network structure, and a constructed Bayesian network model is obtained to perform logical reasoning on the input process parameters to obtain a prediction result of the performance parameters; an optimization module performs reverse tracing based on the constructed Bayesian network model to dynamically adjust the process parameters. Thus, the wire drawing process parameters can be dynamically optimized through Bayesian network analysis, and the wire breaking frequency, surface quality, and wire curvature can be accurately predicted and controlled, and the mold life can be extended, downtime can be reduced, production efficiency and system stability can be improved, and dependence on operator experience can be reduced.

[0090] Based on the same concept of the present invention, the specification is attached Figure 7 As shown, the structure of an electronic device 700 provided in an embodiment of the present application includes: at least one processor 701, at least one network interface 704 or other user interface 703, a memory 705, and at least one communication bus 702. The communication bus 702 is used to realize the connection and communication between these components. The electronic device 700 optionally includes a user interface 703, including a display (for example, a touch screen, LCD, CRT, holographic imaging (Holographic) or projection (Projector), etc.), a keyboard or a pointing device (for example, a mouse, a trackball (trackball), a touch pad or a touch screen, etc.).

[0091] The memory 705 may include a read-only memory and a random access memory, and provides instructions and data to the processor 701. A portion of the memory 705 may also include a non-volatile random access memory (NVRAM).

[0092] In some implementations, the memory 705 stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof:

[0093] Operating system 7051, including various system programs for implementing various basic services and processing hardware-based tasks;

[0094] The application module 7052 includes various application programs, such as a launcher, a media player, a browser, etc., which are used to implement various application services.

[0095] In an embodiment of the present application, by calling the program or instructions stored in the memory 705, the processor 701 is used to execute steps in a wire drawing process optimization method based on a Bayesian network, which can comprehensively consider the complex relationship between multiple wire drawing parameters and achieve effective optimization and quality control of the process.

[0096] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps in the wire drawing process optimization method based on a Bayesian network are executed.

[0097] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, the above-mentioned wire drawing process optimization method based on the Bayesian network can be executed.

[0098] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0099] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0100] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0101] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0102] Finally, it should be noted that the above embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A wire drawing process optimization method based on Bayesian network, characterized in that: The method comprises the following steps: Obtaining wire drawing parameters of the metal wire and performing preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality; Analyze the causal relationship between the drawing parameters based on existing knowledge, and construct an initial Bayesian network structure, and learn the initial Bayesian network structure based on a score search structure learning algorithm to obtain a Bayesian network structure that best suits the initial data set; Performing parameter learning based on the initial data set and the obtained Bayesian network structure to obtain a constructed Bayesian network model to perform logical reasoning on the input process parameters to obtain prediction results of the performance parameters; Reverse tracing is performed based on the constructed Bayesian network model to dynamically adjust process parameters.

2. A wire drawing process optimization method based on Bayesian network according to claim 1, characterized in that: in, The obtained wire drawing parameters of the metal wire are preprocessed including equal width discretization processing, missing data and abnormal data processing, and normalization processing.

3. According to the Bayesian network-based wire drawing process optimization method of claim 2, it is characterized in that: The scoring search-based structure learning algorithm learns the initial Bayesian network structure to obtain the Bayesian network structure that best suits the initial data set, including the following steps: Starting from the initial Bayesian network structure, a possible Bayesian network structure is searched by a greedy method; The Bayesian information criterion is used as the scoring function, and each possible Bayesian network structure is evaluated, and the next round of search direction is carried out according to the evaluation results, until the Bayesian network structure with the smallest evaluation result is found, which is used as the Bayesian network structure most suitable for the initial data set.

4. The wire drawing process optimization method based on Bayesian network according to claim 3 is characterized in that: Parameter learning is performed based on the initial data set and the obtained Bayesian network structure, including the following steps: Initializing the parameters of the obtained Bayesian network structure; wherein the initialized parameters include the parameters in the conditional probability table of each node in the Bayesian network structure; Calculating a likelihood function of the initial data set; The expectation maximization algorithm is used to update the parameters of the Bayesian network structure to maximize the likelihood function.

5. A wire drawing process optimization method based on Bayesian network according to claim 4, characterized in that: in, The constructed Bayesian network model uses a joint tree inference engine for logical reasoning.

6. A wire drawing process optimization method based on Bayesian network according to claim 5, characterized in that: The reverse tracing based on the constructed Bayesian network model and the dynamic adjustment of process parameters include the following steps: The performance parameter is set as the observed data, and the observed data is input into the constructed Bayesian network model as an evidence variable; Based on the constructed Bayesian network model, logical reasoning is performed to calculate the posterior probability of the target variable; wherein the process parameters are used as the target variables; According to the calculated posterior probability, the process parameters are dynamically adjusted.

7. A wire drawing process optimization method based on Bayesian network according to claim 6, characterized in that: The method further comprises the following steps: Regularly updating the initial data set with newly acquired wire drawing parameter data during the production process; The local distribution function of the current Bayesian network model is updated using the updated initial data set.

8. A wire drawing process optimization device based on Bayesian network, characterized in that: The device comprises: An acquisition module is used to acquire the wire drawing parameters of the metal wire and perform preprocessing to obtain an initial data set; wherein the wire drawing parameters of the metal wire include process parameters and performance parameters, wherein the process parameters include annealing parameters, wire drawing speed, back tension, and lubricant concentration; and the performance parameters include wire break frequency, wire surface quality, wire curvature, die drawing length, and die quality; A construction module is used to analyze the causal relationship between the various drawing parameters based on existing knowledge, and to construct an initial Bayesian network structure, and to learn the initial Bayesian network structure based on a structure learning algorithm of score search to obtain a Bayesian network structure that best suits the initial data set; A learning module, used for performing parameter learning based on the initial data set and the obtained Bayesian network structure, obtaining a constructed Bayesian network model, so as to perform logical reasoning on the input process parameters and obtain prediction results of the performance parameters; The optimization module is used to perform reverse tracing based on the constructed Bayesian network model and dynamically adjust process parameters.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the wire drawing process optimization method based on the Bayesian network are performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the wire drawing process optimization method based on the Bayesian network as described in any one of claims 1 to 7 are executed.

Images