CN-122020552-A - Multi-source information fusion-based process parameter decision system and method

CN122020552ACN 122020552 ACN122020552 ACN 122020552ACN-122020552-A

Abstract

The invention discloses a process parameter decision system and a process parameter decision method based on multisource information fusion, and relates to the technical field of conflict detection; the method comprises the steps of establishing a process knowledge graph, calculating distances and conflict entropy between different evidences, calculating evidence credibility weights between the evidences, correcting each evidence by the aid of the credibility weights, carrying out evidence fusion by using evidence theory to obtain a probability distribution function of comprehensive evidence, determining a current process state and confidence coefficient according to the fused evidences, extracting different process parameter adjustment schemes related to the current process state from the knowledge graph, selecting an optimal process parameter adjustment scheme by means of a neural network algorithm, executing the optimal process parameter adjustment scheme, collecting multi-source heterogeneous data of the next production period, and judging whether the optimal process parameter adjustment scheme is effective.

Inventors

LIAO ZHENDONG
MO YONG
LI XUESONG

Assignees

三众智能精密机械(江苏)有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. A process parameter decision method based on multi-source information fusion is characterized by comprising the following steps: s100, acquiring multi-source heterogeneous data in the production process of the process, and preprocessing the multi-source heterogeneous data; S200, carrying out feature extraction and uncertainty quantization on the preprocessed multi-source heterogeneous data; S300, constructing a process knowledge graph, mapping the extracted features on knowledge graph nodes, and performing conflict detection by using the knowledge graph; S400, constructing a basic probability distribution function aiming at a process state, taking the characteristics of different data sources as evidences, calculating distances and conflict entropy between different evidences, calculating evidence credibility weights among the evidences, and correcting each evidence by utilizing the credibility weights; S500, determining the current process state and the confidence coefficient according to the fused evidence, extracting different process parameter adjustment schemes related to the current process state from the knowledge graph, and selecting an optimal process parameter adjustment scheme by using a neural network algorithm; And S600, executing an optimal process parameter adjustment scheme, collecting multi-source heterogeneous data of the next production period, and judging whether the optimal process parameter adjustment scheme is effective.
2. The method for determining process parameters based on multi-source information fusion according to claim 1, wherein the uncertainty quantization in S200 is specifically: uncertainty quantization is performed on different multi-source heterogeneous data, and a data quality score is calculated.
3. The method for determining process parameters based on multi-source information fusion according to claim 1, wherein the step S300 of performing conflict detection by using a knowledge graph is specifically as follows: Extracting process objects and process elements in process production as entities, extracting different production relations in process production as edges, taking the entities as knowledge graph nodes, and constructing a knowledge graph by taking the production relations as the knowledge graph edges; Mapping the characteristics of the multi-source heterogeneous data in the corresponding knowledge graph nodes, and supplementing knowledge graph edges according to causal relations among different multi-source heterogeneous data characteristics in production records in history; Calculating the confidence coefficient of each judgment path respectively and calculating the conflict measurement when the multi-source heterogeneous data characteristics of different data sources are different according to the judgment of the edges of the knowledge graph aiming at the same technological state in the knowledge graph; a conflict threshold T c is preset, and when the conflict measure is larger than the conflict threshold, the existence of obvious conflict is judged.
4. The method for determining process parameters based on multi-source information fusion as set forth in claim 1, wherein the constructing a basic probability distribution function for the process state in S400 is specifically: The method comprises the steps of defining an identification frame aiming at the same process state identification problem, wherein the identification frame represents a set constructed by all mutually exclusive process states, constructing all subsets, empty sets and self of the identification frame as power sets, constructing a basic probability distribution function m aiming at each data source according to multi-source heterogeneous data characteristic values and knowledge patterns, constraining the probability of the basic probability distribution function as the empty set to be 0, constraining the sum of the probabilities of all elements in the power set to be 1, and obtaining the basic probability distribution function by multiplying fuzzy membership and data quality scores, wherein the fuzzy membership is calculated by using a Gaussian membership function.
5. The method for determining process parameters based on multi-source information fusion according to claim 2, wherein the probability distribution function for obtaining comprehensive evidence by evidence fusion using evidence theory in S400 is specifically as follows: The method comprises the steps of taking characteristic values of multi-source heterogeneous data of each data source as evidence, calculating network distance and conflict entropy aiming at basic probability distribution functions of two evidence, calculating credibility weight by using network distance and data quality scores among different evidence, correcting each evidence by using the credibility weight, and fusing all evidences with obvious conflict by using a Dempster evidence theory combination rule to obtain a probability distribution function mfinal of comprehensive evidence.
6. The method for determining process parameters based on multi-source information fusion as set forth in claim 1, wherein the selecting an optimal process parameter adjustment scheme by using a neural network algorithm in S500 is specifically: Extracting a process parameter adjustment scheme with a relation side with the process current state from the knowledge graph to form a candidate set, and calculating the influence and the confidence coefficient of the process parameter adjustment scheme in the candidate set on multiple targets by using a neural network algorithm; Constructing a multi-target space, wherein each process parameter adjustment scheme represents a point in the multi-dimensional space, the point coordinates comprise the influence of the process parameter adjustment scheme on the multi-target, and the optimal process parameter adjustment scheme is obtained by solving the multi-target space by utilizing a pareto optimal algorithm.
7. The method for determining process parameters based on multi-source information fusion as set forth in claim 1, wherein the step S600 of determining whether the optimal process parameter adjustment scheme is valid is specifically: executing the optimal process parameter adjustment scheme, collecting new multi-source heterogeneous data in the next production period, calculating the data quality score of the new multi-source heterogeneous data, judging that the scheme is effective when the data quality score is higher than that before executing the optimal process parameter adjustment scheme, otherwise judging that the scheme is ineffective, discarding the adjustment when the scheme is ineffective, and returning to the step S400 for recalculation.
8. The technological parameter decision system based on multi-source information fusion is characterized by comprising a data acquisition module, a feature extraction module, a conflict detection unit, a conflict elimination module, an optimal scheme screening module and a technological parameter adjustment module; The data acquisition module is used for acquiring multi-source heterogeneous data in the process of production and preprocessing the multi-source heterogeneous data; The characteristic extraction module is used for carrying out characteristic extraction and uncertainty quantization on the preprocessed multi-source heterogeneous data; the conflict detection unit is used for constructing a process knowledge graph, mapping the extracted features on knowledge graph nodes and carrying out conflict detection by using the knowledge graph; the conflict elimination module is used for constructing a basic probability distribution function aiming at the process state, taking the characteristics of different data sources as evidences, calculating the distances and conflict entropy between different evidences, calculating the credibility weight of the evidences by using the distance between the evidences, and correcting each evidence by using the credibility weight; the optimal scheme screening module is used for determining the current process state and the confidence coefficient according to the fused evidence, extracting different process parameter adjustment schemes related to the current process state from the knowledge graph, and selecting the process parameter adjustment schemes by using a neural network algorithm; the process parameter adjusting module is used for executing an optimal process parameter adjusting scheme, collecting multi-source heterogeneous data of the next production period and judging whether the optimal process parameter adjusting scheme is effective.
9. The system for determining process parameters based on multi-source information fusion according to claim 8, wherein the feature extraction module comprises a feature extraction unit and a data quality scoring unit; the characteristic extraction unit is used for extracting characteristics by adopting different methods aiming at different multi-source heterogeneous data; The data quality scoring unit is used for carrying out uncertainty quantification on different multi-source heterogeneous data and calculating data quality scores.
10. The system for determining process parameters based on multi-source information fusion according to claim 8, wherein the conflict elimination module comprises a basic probability distribution function unit, an evidence correction unit and an evidence fusion unit; The basic probability distribution function unit is used for defining an identification frame aiming at the same process state identification problem, and constructing all subsets, empty sets and self of the identification frame into power sets; the evidence correction unit is used for calculating credibility weights by utilizing network distances and data quality scores among different evidences, and correcting each evidence by utilizing the credibility weights; the evidence fusion unit is used for fusing all evidences with obvious conflict by using the Dempster evidence theory combination rule, and the corrected evidences to obtain a probability distribution function mfinal of the comprehensive evidence.

Description

Multi-source information fusion-based process parameter decision system and method Technical Field The invention relates to the technical field of conflict detection, in particular to a process parameter decision system and method based on multi-source information fusion. Background With the deep advancement of intelligent manufacturing, the manufacturing industry changes to flexible, customized and green, and the customized production mode of multiple varieties and small batches becomes the mainstream, and a process parameter decision system is required to rapidly respond to the process requirements of different products, so that the parameter configuration of one-key switching is realized. For example, in the manufacture of 3C electronic products, the same production line needs to be compatible with the assembly process of different products such as mobile phones and tablets, and the conventional fixed parameter mode cannot meet the requirement of fast switching. The requirements of consumers on the quality of products are increasingly severe, and the industrial field makes mandatory demands on the whole life cycle of the products, so that a process parameter decision system is required to be capable of correlating with the production of whole-flow data, and the accurate positioning of quality problems and the reverse optimization of process parameters are realized. Along with the development of industrial Internet and big data technology, the manufacturing industry is pushed to enter an informatization and industrialization deep fusion stage, and the multi-source information fusion technology is widely applied to process parameter decision-making; however, the traditional process parameter decision often depends on a single data source (such as sensor data or an empirical formula), or adopts a simple data superposition and fusion mode, so that the information utilization is unilateral, the conflict information processing capability is weak, and complex working conditions are difficult to deal with. Disclosure of Invention The invention aims to provide a process parameter decision system and a process parameter decision method based on multi-source information fusion, which are used for solving the problems in the prior art. In order to achieve the above purpose, the present invention provides the following technical solutions: a process parameter decision method based on multi-source information fusion is characterized by comprising the following steps: s100, acquiring multi-source heterogeneous data in the production process of the process, and preprocessing the multi-source heterogeneous data; Further, the specific steps of preprocessing the multi-source heterogeneous data are as follows: S101, acquiring multi-source heterogeneous data with different sources, different formats and different acquisition frequencies in the process of production, wherein the multi-source heterogeneous data comprises physical sensing data, quality detection data, production management data and expert experience knowledge; The physical sensing data are exemplified by time sequence data such as temperature, pressure, vibration and the like, the quality detection data are exemplified by discrete data such as product size, surface defects, performance indexes and the like, the production management data are exemplified by structural data such as material batch, equipment ID, maintenance record and the like, and the expert experience knowledge is exemplified by generation rules or language description; S102, preprocessing comprises format standardization, time alignment and space alignment; The format standardization means that all data are mapped to a preset standardized data model in a unified mode, the time alignment means that time synchronization is carried out on time sequence data by interpolation or resampling based on unified time stamps, and the space alignment means that quality detection data are accurately associated with processing positions and equipment units. The method has the advantages that four major types of data including cover sensing, quality detection, production management and expert experience are covered, the limitation of a single data source is avoided, the integrity of decision basis is ensured, and the problems of single data source and one-sided information in traditional technological parameter decision are solved. S200, carrying out feature extraction and uncertainty quantization on the preprocessed multi-source heterogeneous data; further, the specific steps of extracting the characteristics and quantifying the uncertainty of the preprocessed multi-source heterogeneous data are as follows: s201, performing time-frequency domain analysis on physical sensing data to extract time domain features, frequency domain features and time-frequency domain features, wherein the time domain features comprise mean values, standard deviations, peak values, root mean square skewness and the like, the freque