A production line process knowledge evolution method and system

By constructing a hierarchical representation model of process knowledge and intelligent algorithms, the problem of low update efficiency of process knowledge base is solved, realizing intelligent generation and updating of process knowledge, and improving the efficiency and quality of intelligent process planning for production lines.

CN118709767BActive Publication Date: 2024-11-15SHANDONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411197416.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-15
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

Existing process knowledge bases are inefficient in updating and maintaining, making it difficult to adapt to the rapidly evolving needs of manufacturing technologies. Furthermore, the lack of diversity and universality of process knowledge limits the efficiency of intelligent process planning for production lines.

Method used

A hierarchical representation model of process knowledge is constructed using ontology modeling technology. By combining SMOTE-NC oversampling, CTGAN model and NSGA-II algorithm, process route evolution, process parameter evolution and process case integration are implemented to realize intelligent generation and updating of process knowledge.

Benefits of technology

It has improved the diversity and universality of the process knowledge base, enhanced the intelligent process planning capabilities of the production line, and improved production quality, efficiency, and cost-effectiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118709767B_ABST
    Figure CN118709767B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of intelligent manufacturing technology, and specifically provides a production line process knowledge evolution method and system, the method comprising the following steps: acquiring process knowledge from a process knowledge source and using ontology modeling technology to represent the acquired process knowledge, and constructing a hierarchical representation model of process knowledge; taking process route data sets of different production tasks and process parameter data sets of different production processes as input, based on the process knowledge evolution goal, sequentially implementing process route evolution, process parameter evolution and process case integration, and realizing intelligent generation of process knowledge; storing the process knowledge formed in the entire evolution process into a process knowledge base. The process knowledge is hierarchically represented, and a standardized process knowledge representation model is constructed, realizing a unified modeling representation of heterogeneous process knowledge.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0002] Manufacturing is an important part of the product design to output process. It involves the integrated application of multiple processes and technologies to achieve efficient and high-quality product production. In this process, intelligent process planning plays a vital role. The soul of intelligent process planning is the process intelligent decision-making technology based on knowledge. Its main idea is to realize automatic decision-making and optimization of production processes based on existing knowledge and artificial intelligence methods. Among them, the process knowledge base, as a way to collect and manage knowledge, provides a new idea for the reuse of process knowledge.

[0003] The process knowledge base can collect and manage heterogeneous and dispersed process knowledge, and provide tools for efficient aggregation, storage and decision-making of process knowledge. It can centrally store and manage dispersed and heterogeneous process information, so that when encountering new production tasks, it can make full use of historical process knowledge for rapid retrieval and intelligent decision-making, and produce feasible process solutions. Therefore, increasing the diversity and universality of process knowledge in the process knowledge base is crucial to improving the efficiency of process planning.

[0004] At present, most of the process knowledge bases store process knowledge and experience at one time in the early stage of knowledge base system construction. Later updates and maintenance still rely on manual work, which is not only inefficient but also difficult to adapt to the rapidly developing manufacturing technology needs. In addition, the process knowledge in the actual production process is diverse in form, complex in composition, and often implicit and vague. Its collection, management and application processes require a lot of manpower and material resources, resulting in insufficient diversity and universality of process knowledge in the process knowledge base. This has limited the development and application of the knowledge base to a certain extent, and has hindered manufacturing companies from improving the efficiency of intelligent process planning for production lines. Summary of the invention

[0005] In view of the above problems existing in the process knowledge base, the present invention provides a production line process knowledge evolution method and system

[0006] In a first aspect, the technical solution of the present invention provides a production line process knowledge evolution method, comprising the following steps:

[0007] Acquire process knowledge from process knowledge sources and use ontology modeling technology to represent the acquired process knowledge, and build a hierarchical representation model of process knowledge;

[0008] Taking the process route data sets of different production tasks and the process parameter data sets of different production processes as input, based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are implemented in sequence to realize the intelligent generation of process knowledge;

[0009] The process knowledge formed during the entire evolution process is stored in the process knowledge base.

[0010] As a preferred embodiment of the technical solution of the present invention, the process knowledge source includes a process knowledge base, a process database, process documents and a production line resource document of the production line;

[0011] The hierarchical representation model of process knowledge includes production task ontology model, production resource ontology model and production process ontology model;

[0012] The production task ontology model describes product information and production requirements;

[0013] The production resource ontology model describes the basic attributes and capability attributes of production process equipment and production process tools;

[0014] The production process ontology model describes a complete and orderly description of process steps, including basic production tasks, production resources used and specific process operation procedures.

[0015] As a preferred embodiment of the technical solution of the present invention, the process route data sets of different production tasks and the process parameter data sets of different production processes are used as inputs, and based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are sequentially implemented to realize the intelligent generation of process knowledge, including the following steps:

[0016] Based on the acquired process knowledge, construct the process route data set of different production tasks; use the SMOTE-NC oversampling strategy to oversample the process route data set, and mine the process method generation rules and production resource selection rules of different production tasks based on the rough set model;

[0017] According to the acquired process knowledge, the process parameter data sets of different production processes are constructed to train the CTGAN model to generate a process parameter generation model. Based on the process parameter generation model, samples are taken from the normal distribution as input to output the process parameter generation results of different production processes.

[0018] Determine the goal of process knowledge integration, build a process case integration model based on the NSGA-II algorithm, take the results of process route evolution and process parameter evolution as input, solve based on the process case integration model, and obtain complete process solutions for multiple product levels.

[0019] As a preferred embodiment of the technical solution of the present invention, a process route data set of different production tasks is constructed based on the acquired process knowledge; the process route data set is oversampled using the SMOTE-NC oversampling strategy, and the steps of mining process method generation rules and production resource selection rules of different production tasks based on the rough set model include:

[0020] Construct process route data sets for different production tasks based on the acquired process knowledge;

[0021] The SMOTE-NC oversampling strategy is used to oversample the process route dataset, and the process attributes of the process route dataset are divided into discrete data and continuous data.

[0022] Construct the process route decision table for different part features and implement the attribute reduction algorithm based on core attributes and dependency to simplify the decision table;

[0023] The process method generation rules and production resource selection rules for different production tasks are extracted from the simplified process route decision table.

[0024] As a preferred embodiment of the technical solution of the present invention, after the step of extracting the process method generation rules and production resource selection rules for different production tasks from the simplified process route decision table, the following steps are included:

[0025] Calculate the support and confidence of the rules, and based on the preset minimum support and minimum confidence of the rules, filter out the rules that do not meet the minimum support and minimum confidence requirements to optimize the rule set;

[0026] As a preferred embodiment of the technical solution of the present invention, according to the acquired process knowledge, a process parameter data set of different production processes is constructed to train the CTGAN model to generate a process parameter generation model, and based on the process parameter generation model, sampling from a normal distribution as input, the steps of outputting the process parameter generation results of different production processes include:

[0027] According to the acquired process knowledge, construct the process parameter data sets of different production processes and pre-process the process parameter data sets;

[0028] The preprocessed process parameter data set is passed into the CTGAN model to iteratively update the model parameters of the generator and the discriminator to generate a process parameter generation model;

[0029] Data sampling is performed from the trained process parameter generation model, with different processing stages as sampling conditions, sampling input from the normal distribution to the generation model, and the process parameter generation results are output.

[0030] As a preferred embodiment of the technical solution of the present invention, the steps of transferring the preprocessed process parameter data set into the CTGAN model to iteratively update the model parameters of the generator and the discriminator to generate the process parameter generation model include:

[0031] The preprocessed process parameter data set is passed into the CTGAN model;

[0032] The generative model and the discriminative model are optimized through mutual adversarial learning. The generative model captures the distribution of samples and updates the network weights of the generative model based on the gradient feedback of the discriminative model on the generated results. After training, the process parameter generation model is obtained.

[0033] As a preferred embodiment of the technical solution of the present invention, the goal of process knowledge integration is determined, a process case integration model is constructed based on the NSGA-II algorithm, the results of process route evolution and process parameter evolution are used as input, and the process case integration model is solved to obtain a complete process solution for multiple product levels, including the following steps:

[0034] Based on the process method generation rules, multiple candidate process methods are obtained, and based on the production resource selection rules, multiple candidate process equipment and candidate process tools are generated for each candidate process method;

[0035] From the generated process parameters of the corresponding production tasks in different production processes, match the available process parameters according to the production task requirements, selected production resource information, and process methods to construct process primitives;

[0036] Generate a process primitive constraint matrix based on the priority relationship constraints between different process primitives, set the integration target of process knowledge, build and solve the process case integration model based on the NSGA-II algorithm;

[0037] Merge and sort different process primitives to obtain complete process solutions for multiple product levels.

[0038] As a preferred embodiment of the technical solution of the present invention, the process primitive constraint matrix is ​​generated according to the priority relationship constraints between different process primitives, the integration target of process knowledge is set, and the process case integration model is constructed and solved based on the NSGA-II algorithm, including the following steps:

[0039] Taking the processing rules as constraints, a processing constraint set is generated and converted into a processing primitive constraint matrix, which is used to describe the processing priority relationship between different processing primitives.

[0040] Set multiple process knowledge evolution goals, build a process case integration model based on the NSGA-II algorithm, and convert it into the algorithm fitness function in the NSGA-II multi-objective optimization algorithm model;

[0041] Taking the processing technology primitive constraint matrix as the constraint condition, the processing technology primitives are arranged and combined to construct a processing technology sequence that meets the requirements of the processing technology primitive constraint matrix as a feasible solution; the population size is set to generate an initial population containing different processing technology primitive combinations to obtain a set number of initial solutions;

[0042] The maximum number of iterations of the NSGA-II multi-objective optimization algorithm is set. In the evolution process of each generation, the fitness function value of each chromosome in the population is calculated. The chromosomes in the population are non-dominated according to the obtained fitness function value, and the crowding distances of different chromosomes in the same dominance level are calculated. Based on the calculation results, two chromosomes are selected, and the adaptive crossover and adaptive mutation strategies are implemented to generate multiple new processing technology primitive combinations. Among them, the adaptive crossover and adaptive mutation strategies dynamically update the crossover and mutation probabilities of the chromosomes according to the current number of iterations.

[0043] After completing each round of selection, crossover and mutation operations, the chromosome feasibility is evaluated, and the solutions that violate the processing technology primitive constraint matrix are adjusted and repaired;

[0044] The new population obtained in each iteration is merged with the parent population to obtain a new generation population, and the above steps are repeated to finally obtain the Pareto optimal solution set, which is decoded and output as a complete processing solution at multiple part levels.

[0045] In a second aspect, the technical solution of the present invention provides a production line process knowledge evolution system, including a knowledge evolution resource layer, a knowledge evolution method layer and a knowledge evolution application layer;

[0046] At the knowledge evolution resource layer, process knowledge is acquired from each process knowledge source, and the acquired process knowledge is provided to the knowledge evolution method layer;

[0047] At the knowledge evolution method layer, ontology modeling technology is used to represent the acquired process knowledge and construct a hierarchical representation model of process knowledge; taking the process route data sets of different production tasks and the process parameter data sets of different production procedures as input, based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are implemented in sequence to realize the intelligent generation of process knowledge; the evolution method described in the first aspect is implemented at the knowledge evolution layer.

[0048] At the knowledge evolution application layer, based on the generated process knowledge, a process knowledge base interface, a graphical interface, and an intelligent process equipment interface are provided;

[0049] Through the process knowledge base interface, the generated process knowledge is stored in the process knowledge base, and the process knowledge base is updated;

[0050] Through the graphical interface that interacts with users, it is possible to browse the generated process knowledge, collect expert evaluation feedback and on-site operator evaluation;

[0051] Through the intelligent process equipment interface, on-site feedback on the use of process knowledge and integration of intelligent process equipment based on the process knowledge base are realized.

[0052] It can be seen from the above technical scheme that the present invention has the following advantages: 1. The present application determines the data source of the evolution of production line process knowledge, extracts process knowledge from it, and uses ontology modeling technology to establish a hierarchical representation model of process knowledge from three aspects: production tasks, production resources and production processes, so as to realize the hierarchical representation and standardized organization of process knowledge; taking production quality, production efficiency and production cost as evolution goals, the evolution task of production line process knowledge is divided into three modules: process route evolution, process parameter evolution and process case integration, and a process route generation model, a process parameter generation model and a process case integration model are designed to mine and learn the generation rules of process knowledge, and on this basis realize the intelligent evolution of process knowledge, obtain complete process cases and store them in the process knowledge base, so as to realize the update of the process knowledge base.

[0053] 2. Aiming at the need to express various types of process knowledge in production line process planning, the present invention represents the process knowledge in layers, constructs a standardized process knowledge representation model, and realizes a unified modeling representation of heterogeneous process knowledge.

[0054] 3. In order to solve the problem of insufficient dynamic updating capability of the process knowledge base in the process planning of the production line, the present invention uses intelligent algorithms such as data mining and machine learning to study the evolution of the process knowledge of the production line, takes production quality, production efficiency and production cost as evolution goals, constructs a process knowledge generation model to learn the distribution law of process knowledge, generates process knowledge to accelerate the process of collecting process knowledge in the knowledge base, increases the diversity and universality of process knowledge in the process knowledge base, and enables it to better serve the actual process planning.

[0055] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] 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 or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 A block diagram of a production line process knowledge evolution method provided in Example 1 of the present invention.

[0058] Figure 2 This is a flowchart for implementing process route evolution based on a rough set model according to Embodiment 2 of the present invention.

[0059] Figure 3 This is a flowchart of the implementation of process parameter evolution based on the CTGAN model in Example 2 of the present invention.

[0060] Figure 4 This is a flowchart of the process case integration implementation based on the NSGA-II model of Example 2 of the present invention.

[0061] Figure 5 This is a schematic diagram of the structure of the production line process knowledge evolution system provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0062] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0063] In the process of developing intelligent process planning based on the process knowledge base, there are problems such as insufficient dynamic update capability of the process knowledge base, insufficient diversity and universality of process knowledge, etc. To solve the above problems, the first embodiment of the present invention provides a production line process knowledge evolution method, such as Figure 1 As shown, this method uses knowledge engineering technology, data mining algorithm and machine learning algorithm as supporting technologies to hierarchically represent the process data of the production line and implement three process knowledge evolution processes: process route evolution, process parameter evolution and process case integration.

[0064] The process instance knowledge is acquired from process knowledge sources such as process knowledge base, process database, process documents and production line resource documents, and a hierarchical representation model of process knowledge is established from three aspects: production tasks, production resources and production processes using ontology modeling technology. The heterogeneous process knowledge is described in a standardized manner by defining classes, object attributes, data attributes, cardinality restrictions, etc., to achieve hierarchical expression and standardized organization of process instance knowledge. Specifically, the production task ontology model mainly describes the production task type, production task priority, specific production requirements, etc. The production resource ontology model is mainly oriented to the various process equipment and process tools on a production line, and describes their components, performance parameters and production capabilities, which is the main execution basis for the production line to complete production tasks. The production process model links the production task model with the production resource model, and describes the process methods and process parameters used for different production tasks under certain production resource constraints.

[0065] The acquired process knowledge is input into the process knowledge evolution link. Based on the process knowledge evolution goal, the process route generation model, process parameter generation model and process case integration model are constructed in sequence to implement the process knowledge evolution method. Among them, the process knowledge evolution goals include production quality, production efficiency and production cost.

[0066] The following three models are designed and analyzed in turn, and the evolution of process knowledge is realized by learning the distribution and organization rules of historical process data; the details are as follows:

[0067] (1) A process route generation model is constructed based on a rough set model. The process route data sets of different production tasks are extracted from the process instance knowledge obtained from the process knowledge source. The process method decision table and the production resource decision table are constructed respectively to mine the process method generation rules of different production tasks and the selection rules of different production resources. The mined rules are expressed in the form of production rules to provide a visual presentation of the process rules.

[0068] (2) A process parameter generation model is constructed based on the conditional table generative adversarial network (CTGAN) model. The process parameter data sets of different production processes are extracted from the process instance knowledge obtained from the process knowledge source. The CTGAN model is trained to learn the data distribution law of the process parameter data samples and capture the distribution variation law from random noise to the real process parameters. Data is sampled from the trained process parameter generation model and the process parameter generation results of different production processes are output.

[0069] (3) A process case integration model is constructed based on the non-dominated sorting genetic algorithm (NSGA-II) model. First, for production tasks, multiple candidate process methods are obtained based on the process method generation rules output by the process route evolution stage. Multiple candidate process equipment and candidate process tools are generated for each candidate process method based on the production resource selection rules output by the process route evolution stage. Secondly, process primitives are constructed from the generated process parameters of the corresponding production tasks in different production processes according to the production task requirements, the selected production resource information, and the process method matching available process parameters. Finally, a process primitive constraint matrix is ​​generated based on the priority relationship constraints between different process primitives, the integration target of process knowledge is set, a process case integration mathematical model is constructed and solved, and different process primitives are merged and sorted to achieve the acquisition of complete process solutions for multiple product levels.

[0070] The obtained multiple complete product-level process plans are stored in the process knowledge base to realize the update of the process knowledge base.

[0071] In order to more deeply demonstrate the application of the present invention in the actual production process, the production line process knowledge evolution method provided in this embodiment 2 takes cutting processing as an example, and describes in detail the specific implementation steps and processes of each process knowledge evolution link. Cutting processing is a widely used mechanical manufacturing process, involving the precise processing of a variety of materials, and is suitable for demonstrating the actual effect and application scope of the process knowledge evolution method proposed in the present invention. In this embodiment, the production tasks mainly involve different feature processing tasks of parts, the production resources mainly involve the machine tool resources and tool resources of the production line, and the production processes mainly involve different cutting processing processes such as turning, milling, and drilling.

[0072] like Figure 2 As shown in the figure, a process route generation model is constructed based on the rough set model to realize the process route evolution for different part features, including:

[0073] Taking the production line cutting process route cases obtained from the process knowledge source as the data source, firstly, the machine tool resources and tool resources in the process route data set are numbered to realize the unique identification of different machine tools and different tools in the production line; secondly, the process route evolution data set is constructed for different part features respectively; wherein, the process route evolution data set includes feature size parameters, blank type, material type, heat treatment, machining accuracy level, machining surface roughness, feature machining method, selected tool number and selected machine tool number.

[0074] A process route generation model is constructed based on the rough set model. The process route evolution data set of different part features is used as the input of the process route generation model to realize the mining of processing method generation rules with different features and selection rules of different tools and machine tools for production lines.

[0075] Firstly, the process route data is preprocessed, and the SMOTE-NC oversampling strategy is used to oversample the process route data set; the process attributes of the process route data set are divided into discrete data and continuous data; specifically, the discrete attributes refer to the blank type, material type, heat treatment, machining accuracy level, machining surface roughness, feature machining method, selected tool number and selected machine tool number, which are represented by digital codes; the continuous attributes refer to the feature size parameters, and the trapezoidal membership function is used to design the size interval for discretization processing.

[0076] Construct process route decision tables for different part features, including feature processing method decision tables and processing resource decision tables; specifically, the feature processing method decision table uses blank type, material type, heat treatment, processing accuracy level, and processing surface roughness as conditional attributes, and the feature processing method as a decision attribute; the processing resource decision table uses blank type, material type, heat treatment, processing accuracy level, processing surface roughness, and processing method as conditional attributes, and uses selected machine tool number and selected tool number as decision attributes.

[0077] Based on the process route decision table with different characteristics, the attribute reduction algorithm based on core attributes and dependency is implemented to remove redundant conditional attributes, simplify the decision table, and determine the core process attributes that affect the selection of processing methods and the core process attributes that affect the selection of processing resources. The definition of dependency is shown in formula (1);

[0078] (1)

[0079] In the formula, C is the condition attribute; D is the decision attribute; U is the domain, which represents the set of all objects in the decision table; It is the C-positive domain of D, which represents the set of all objects in the domain U that can be accurately divided into the equivalence class of D according to the classification information.

[0080] Based on the process route decision table after attribute simplification, a heuristic attribute value simplification algorithm is implemented to delete the redundant attribute values ​​of each object, and the rules for distinguishing different decision attribute categories are extracted from the decision table and expressed in the form of IF THEN production rules; specifically, for different features, the processing method generation rules are mined based on the feature processing method decision table, and the selection rules of different machine tools and cutting tools are mined based on the processing resource decision table.

[0081] The support and confidence of the rules are used as evaluation indicators to measure and evaluate the mined decision rules and evaluate the effectiveness of the decision rules, as shown in formulas (2) and (3);

[0082] (2)

[0083] (3)

[0084] In the formula, Satisfies the decision rule The total number of samples; Indicates the antecedents that satisfy the decision rule The number of samples; Represents the total number of samples in the entire domain; Indicates the consequent that satisfies the decision rule The total number of samples.

[0085] Set the minimum support and minimum confidence of the rules, filter out the rules that do not meet the minimum support and minimum confidence requirements, and optimize the rule set.

[0086] like Figure 3 As shown in the figure, a process parameter generation model is constructed through the conditional table generative adversarial network (CTGAN) model to realize the process parameter evolution of different cutting processes, including:

[0087] Taking the production line cutting process parameter cases obtained from the process knowledge source as the data source, the process parameter evolution training data set is constructed for different processing procedures respectively; wherein the process parameter evolution training data set content is divided into part feature parameters, resource capability parameters, processing method attributes, process parameter values ​​and processing effect labels, specifically including feature size parameters, material type, material hardness, tool material, tool size, machine tool power, processing stage, back cutting amount, feed rate, cutting speed, processing surface roughness, processing time, etc.;

[0088] A process parameter generation model is built based on the conditional table generative adversarial network (CTGAN) model. The real process parameter data sets of different processing steps are used as the input of the process parameter generation model. The data distribution law of the process parameter data samples is learned to realize the intelligent generation of process parameters.

[0089] First, the process parameter data is preprocessed, and the process attributes of the process parameter data set are divided into discrete attributes and continuous attributes. Specifically, discrete attributes refer to categorical variables such as material type, tool material, machine tool power, and processing stage. Conditional fair sampling is used for preprocessing, and each categorical variable is encoded into a conditional vector, which is used as both the input of the generator and as a filtering condition for sampling from the real data distribution. Continuous attributes refer to numerical variables such as characteristic size parameters, material hardness, back cutting amount, feed rate, cutting speed, machining surface roughness, and machining time. Mode-specific Normalization is used for preprocessing, and each continuous value variable is represented as a one-hot variable representing the mode and a scalar representing the value within the mode.

[0090] The preprocessed process parameter data set is passed to the CTGAN model, and the training strategy of "binary minimax game" is introduced to iteratively update the model parameters of the generator and discriminator. Specifically, the generative model and the discriminative model are optimized through mutual adversarial learning. The generative model captures the distribution of samples and updates the network weights of the generative model based on the gradient feedback of the discriminative model to the generated results, so as to maximize the closeness between the generated distribution and the true distribution of the process parameters, as shown in formula (4).

[0091] (4)

[0092] In the formula, is a generative model, is the discriminant model; is the binary cross entropy function; Represents real data, represents random noise data, represents the distribution of real process parameter data, Represents the distribution of random noise data; Represents the data generated by the generative model under the input of given conditions y and noise z; It represents the probability that the discriminant model judges that the input belongs to the real process parameter data. It represents the probability that the discriminant model judges that the input belongs to the data generated by the generative model.

[0093] Set the maximum number of training iterations for the process parameter generation model, use the Adam optimizer to optimize the training process, calculate the adaptive learning rate for the model parameters, and realize dynamic adjustment and update of the model parameters; sample data from the trained process parameter generation model, use different processing stages as sampling conditions, sample input from the normal distribution to the generation model, and output the process parameter generation results.

[0094] like Figure 4 As shown in the figure, a process case integration model is constructed through the non-dominated sorting genetic algorithm (NSGA-II) model to obtain a complete processing solution at the part product level, including:

[0095] Acquire the feature information of the parts, and generate a set of candidate processing methods for each feature based on the feature processing method generation rules generated in the process route evolution stage; generate a set of candidate machine tools and a set of candidate tools for each processing method based on the tool and machine tool selection rules generated in the process route evolution stage, and construct multiple processing operation units.

[0096] For each processing operation unit, the capability attributes of the selected machine tool and the selected tool are obtained from the production line resource document; the available process parameter set is matched with the process parameters generated in the process parameter evolution link according to the dimensional parameters, material type, material hardness, processing surface roughness requirements, processing method, selected machine tool power, selected tool material, and selected tool size of the feature to form multiple processing process primitives, including: processing process primitive number, processing feature, processing method, selected machine tool number, selected tool number, back cutting amount, feed rate, cutting speed, and processing time.

[0097] With the processing technology rules as constraints, a processing technology constraint set is generated and converted into a processing technology primitive constraint matrix to describe the processing priority relationship between different processing technology primitives; the processing technology rules include surface first and hole later, rough first and fine later, datum first, primary first and secondary later, etc.

[0098] Set multiple process knowledge integration goals, build a mathematical model for process case integration, and convert it into an algorithm fitness function in the NSGA-II multi-objective optimization algorithm model; the process integration goals include optimal production quality, highest production efficiency, lowest production cost, etc.

[0099] The processing features, processing methods, selected machine tool number, selected tool number, back cutting depth, feed rate and cutting speed information are extracted from each processing technology primitive, and the combination of different processing features and processing methods is converted into different processing procedures, and the combination of different back cutting depths, feed rates and cutting speeds is converted into different cutting parameter schemes; a four-layer encoding strategy is designed to express the information of each processing technology primitive in the form of chromosome genes; among which, the first layer of encoding represents different processing procedures, the second layer of encoding is the machine tool number selected for each processing procedure, the third layer of encoding is the tool number selected for each processing procedure; the fourth layer of encoding is the cutting parameter scheme selected for each processing procedure.

[0100] Taking the processing primitive constraint matrix as the constraint condition, the processing primitives of all the features of the part are arranged and combined, and a processing sequence that meets the requirements of the processing primitive constraint matrix is ​​constructed as a feasible solution; the population size is set, and an initial population containing different combinations of processing primitives is generated to obtain a certain number of initial solutions.

[0101] The maximum number of iterations of the NSGA-II multi-objective optimization algorithm is set. In the evolution process of each generation, the fitness function value of each chromosome in the population is calculated. The chromosomes in the population are non-dominated according to the obtained fitness function value, and the crowding distances of different chromosomes in the same dominance level are calculated. Based on the calculation results, two chromosomes are selected by the tournament selection method, and the adaptive crossover and adaptive mutation strategies are implemented to generate multiple new processing technology primitive combinations to ensure the diversity of the solution set search space. The adaptive crossover and adaptive mutation strategies can dynamically update the crossover and mutation probabilities of the chromosomes according to the current number of iterations, as shown in equations (5) and (6).

[0102] (5)

[0103] (6)

[0104] In the formula, is the current iteration number, It is The value of the crossover operator in iterations, It is The value of the mutation operator in iterations, and is the maximum and minimum value of the crossover operator, and is the maximum and minimum value of the mutation operator, is the number of iterations of the NSGA-II multi-objective optimization algorithm.

[0105] After completing each round of selection, crossover and mutation operations, the chromosome feasibility is evaluated, and the solutions that violate the processing technology primitive constraint matrix are adjusted and repaired, so that the entire search process is always maintained in the feasible domain.

[0106] The termination condition of the NSGA-II multi-objective optimization algorithm is set to stop iteration after reaching the maximum number of iterations. The new population obtained in each iteration is merged with the parent population to obtain a new generation of population, and the non-dominated sorting, crowding distance calculation, selection, crossover, mutation and chromosome feasibility evaluation operations are repeated. Through continuous iterative optimization of the population, the Pareto optimal solution set is finally obtained, which is decoded and output as a complete machining process plan at multiple part levels, which is stored in the process knowledge base to update the process knowledge base.

[0107] Through the above steps, this embodiment not only demonstrates the evolution of process knowledge in cutting processing, but also verifies the practical application value of the method of the present invention in improving the intelligence and automation level of the production line.

[0108] Embodiment 3 of the present invention provides a production line process knowledge evolution system, such as Figure 5 As shown in the figure, the system is divided into the knowledge evolution resource layer, the knowledge evolution method layer and the knowledge evolution application layer, including:

[0109] At the knowledge evolution resource layer, process knowledge is acquired from each process knowledge source, and the acquired process knowledge is provided to the knowledge evolution method layer;

[0110] At the knowledge evolution method level, the process knowledge obtained from the process knowledge source is processed. Based on ontology modeling technology, a hierarchical representation model of process knowledge is established from three aspects: production tasks, production resources, and production processes, so as to realize the hierarchical representation and standardized organization of process knowledge. Based on data mining and machine learning related technologies, process route evolution, process parameter evolution, and process case integration are implemented to realize the intelligent generation of production line process knowledge.

[0111] At the knowledge evolution application layer, based on the generated process knowledge, an interactive interface with the process knowledge base, process planning system users, and intelligent process equipment is provided. Through the process knowledge base interface, the generated knowledge is stored in the process knowledge base, the diversity of process knowledge in the process knowledge base is increased, the process knowledge base is updated, and support is provided for intelligent process decision-making based on the process knowledge base; through the graphical interface that interacts with the process planning system users, the browsing of the generated process knowledge, expert evaluation feedback, and the collection of on-site operator evaluations are realized; through the intelligent process equipment interface, on-site feedback on the use of process knowledge and the integration of intelligent process equipment based on the process knowledge base are realized.

[0112] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0113] The disclosed embodiment provides an embodiment of a production line process knowledge evolution system. This system and the production line process knowledge evolution method of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the production line process knowledge evolution system, reference can be made to the embodiment of the above-mentioned production line process knowledge evolution method.

[0114] The production line process knowledge evolution system is a combination of the units and algorithm steps of each example described in the embodiments disclosed herein, which can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0115] Those skilled in the art will appreciate that various aspects of the production line process knowledge evolution method can be implemented as a system, method or program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which can be collectively referred to as "circuit", "module" or "system".

[0116] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A production line process knowledge evolution method, characterized in that: The steps include: Acquire process knowledge from process knowledge sources and use ontology modeling technology to represent the acquired process knowledge, and build a hierarchical representation model of process knowledge; Taking the process route data sets of different production tasks and the process parameter data sets of different production processes as input, based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are implemented in sequence to realize the intelligent generation of process knowledge; The process knowledge formed during the entire evolution process is stored in the process knowledge base; Taking the process route data sets of different production tasks and the process parameter data sets of different production processes as input, based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are implemented in sequence to realize the intelligent generation of process knowledge. The steps include: Based on the acquired process knowledge, construct the process route data set of different production tasks; use the SMOTE-NC oversampling strategy to oversample the process route data set, and mine the process method generation rules and production resource selection rules of different production tasks based on the rough set model; According to the acquired process knowledge, the process parameter data sets of different production processes are constructed to train the CTGAN model to generate a process parameter generation model. Based on the process parameter generation model, samples are taken from the normal distribution as input to output the process parameter generation results of different production processes. Determine the goal of process knowledge integration, build a process case integration model based on the NSGA-II algorithm, take the results of process route evolution and process parameter evolution as input, solve based on the process case integration model, and obtain complete process solutions for multiple product levels.

2. The production line process knowledge evolution method according to claim 1, characterized in that: The process knowledge sources include the process knowledge base, process database, process documents and production line resource documents of the production line; The hierarchical representation model of process knowledge includes production task ontology model, production resource ontology model and production process ontology model; The production task ontology model describes product information and production requirements; The production resource ontology model describes the basic attributes and capability attributes of production process equipment and production process tools; The production process ontology model describes a complete and orderly description of process steps, including basic production tasks, production resources used and specific process operation procedures.

3. The production line process knowledge evolution method according to claim 2, characterized in that: Based on the acquired process knowledge, construct process route data sets for different production tasks; The process route data set is oversampled using the SMOTE-NC oversampling strategy. The steps of mining the process method generation rules and production resource selection rules for different production tasks based on the rough set model include: Construct process route data sets for different production tasks based on the acquired process knowledge; The SMOTE-NC oversampling strategy is used to oversample the process route dataset, and the process attributes of the process route dataset are divided into discrete data and continuous data. Construct the process route decision table for different part features and implement the attribute reduction algorithm based on core attributes and dependency to simplify the decision table; The process method generation rules and production resource selection rules for different production tasks are extracted from the simplified process route decision table.

4. The production line process knowledge evolution method according to claim 3 is characterized in that: The steps of extracting the process method generation rules and production resource selection rules of different production tasks from the simplified process route decision table include: The support and confidence of the rules are calculated, and based on the preset minimum support and minimum confidence of the rules, the rules that do not meet the minimum support and minimum confidence requirements are filtered out to achieve optimization of the rule set.

5. The production line process knowledge evolution method according to claim 4, characterized in that: According to the acquired process knowledge, a process parameter data set of different production processes is constructed to train the CTGAN model to generate a process parameter generation model. Based on the process parameter generation model, sampling from a normal distribution is used as input to output the process parameter generation results of different production processes. The steps include: According to the acquired process knowledge, construct the process parameter data sets of different production processes and pre-process the process parameter data sets; The preprocessed process parameter data set is passed into the CTGAN model to iteratively update the model parameters of the generator and the discriminator to generate a process parameter generation model; Data sampling is performed from the trained process parameter generation model, with different processing stages as sampling conditions, sampling input from the normal distribution to the generation model, and the process parameter generation results are output.

6. The production line process knowledge evolution method according to claim 5, characterized in that: The steps of passing the preprocessed process parameter data set into the CTGAN model to iteratively update the model parameters of the generator and the discriminator to generate the process parameter generation model include: The preprocessed process parameter data set is passed into the CTGAN model; The generative model and the discriminative model are optimized through mutual adversarial learning. The generative model captures the distribution of samples and updates the network weights of the generative model based on the gradient feedback of the discriminative model on the generated results. After training, the process parameter generation model is obtained.

7. The production line process knowledge evolution method according to claim 6, characterized in that: Determine the goal of process knowledge integration, build a process case integration model based on the NSGA-II algorithm, take the results of process route evolution and process parameter evolution as input, solve based on the process case integration model, and obtain the complete process solutions for multiple product levels. The steps include: Based on the process method generation rules, multiple candidate process methods are obtained, and based on the production resource selection rules, multiple candidate process equipment and candidate process tools are generated for each candidate process method; From the generated process parameters of the corresponding production tasks in different production processes, match the available process parameters according to the production task requirements, selected production resource information, and process methods to construct process primitives; Generate a process primitive constraint matrix based on the priority relationship constraints between different process primitives, set the integration target of process knowledge, build and solve the process case integration model based on the NSGA-II algorithm; Merge and sort different process primitives to obtain complete process solutions for multiple product levels.

8. The production line process knowledge evolution method according to claim 7, characterized in that: The steps of generating a process primitive constraint matrix according to the priority relationship constraints between different process primitives, setting the integration target of process knowledge, and constructing and solving the process case integration model based on the NSGA-II algorithm include: Taking the processing rules as constraints, a processing constraint set is generated and converted into a processing primitive constraint matrix, which is used to describe the processing priority relationship between different processing primitives. Set multiple process knowledge evolution goals, build a process case integration model based on the NSGA-II algorithm, and convert it into the algorithm fitness function in the NSGA-II multi-objective optimization algorithm model; Taking the processing technology primitive constraint matrix as the constraint condition, the processing technology primitives are arranged and combined to construct a processing technology sequence that meets the requirements of the processing technology primitive constraint matrix as a feasible solution; the population size is set to generate an initial population containing different processing technology primitive combinations to obtain a set number of initial solutions; The maximum number of iterations of the NSGA-II multi-objective optimization algorithm is set. In the evolution process of each generation, the fitness function value of each chromosome in the population is calculated. The chromosomes in the population are non-dominated according to the obtained fitness function value, and the crowding distances of different chromosomes in the same dominance level are calculated. Based on the calculation results, two chromosomes are selected, and the adaptive crossover and adaptive mutation strategies are implemented to generate multiple new processing technology primitive combinations. Among them, the adaptive crossover and adaptive mutation strategies dynamically update the crossover and mutation probabilities of the chromosomes according to the current number of iterations. After completing each round of selection, crossover and mutation operations, the chromosome feasibility is evaluated, and the solutions that violate the processing technology primitive constraint matrix are adjusted and repaired; The new population obtained in each iteration is merged with the parent population to obtain a new generation population, and the non-dominated sorting, crowding distance calculation, selection, crossover, mutation and chromosome feasibility evaluation operations are repeated; through continuous iterative optimization of the population, the Pareto optimal solution set is finally obtained, which is decoded and output as a complete processing technology plan for multiple part levels.

9. A production line process knowledge evolution system, characterized in that: It includes knowledge evolution resource layer, knowledge evolution method layer and knowledge evolution application layer; At the knowledge evolution resource layer, process knowledge is acquired from each process knowledge source, and the acquired process knowledge is provided to the knowledge evolution method layer; At the knowledge evolution method level, ontology modeling technology is used to represent the acquired process knowledge and to construct a hierarchical representation model of process knowledge; Taking the process route data sets of different production tasks and the process parameter data sets of different production processes as input, based on the process knowledge evolution goal, the process route evolution, process parameter evolution and process case integration are implemented in sequence to realize the intelligent generation of process knowledge; The specific steps include: constructing process route data sets for different production tasks based on the acquired process knowledge; oversampling the process route data sets using the SMOTE-NC oversampling strategy, and mining the process method generation rules and production resource selection rules for different production tasks based on the rough set model; According to the acquired process knowledge, the process parameter data sets of different production processes are constructed to train the CTGAN model to generate a process parameter generation model. Based on the process parameter generation model, samples are taken from the normal distribution as input to output the process parameter generation results of different production processes. Determine the goal of process knowledge integration, build a process case integration model based on the NSGA-II algorithm, take the results of process route evolution and process parameter evolution as input, solve based on the process case integration model, and obtain complete process solutions for multiple product levels; At the knowledge evolution application layer, based on the generated process knowledge, a process knowledge base interface, a graphical interface, and an intelligent process equipment interface are provided; Through the process knowledge base interface, the generated process knowledge is stored in the process knowledge base, and the process knowledge base is updated; Through the graphical interface that interacts with users, it is possible to browse the generated process knowledge, collect expert evaluation feedback and on-site operator evaluation; Through the intelligent process equipment interface, on-site feedback on the use of process knowledge and integration of intelligent process equipment based on the process knowledge base are realized.