Research On Rule Extraction And Inference Mechanism Of Intelligent Welding Procedure Specifications System

The rail transportation manufacturing is one of the fields with core competitiveness in China’s high-end equipment manufacturing.It is gradually transforming and upgrading in Informatization and intelligent manufacturing.As a representative subsystem,the welding process intelligent system has the characteristics of knowledge structuring,information standardization,and production efficiency.However,the problems of complex knowledge acquisition and poor self-learning ability are still critical challenges to its development.In this paper,with the background of intelligent demand for welding production in rail transportation equipment manufacturing,a welding process reasoning scheme is developed based on the analysis of its domain knowledge sources and considering the complexity of welding process design and practical application situations.For the bottleneck limitation in knowledge acquisition,the strategy of an improved decision tree hybrid algorithm is used to supplement the rule base.Meanwhile,based on the multi-dimensional field classification system,an inference model based on case-based reasoning and supplemented by rule-based reasoning is used to compensate for the poor self-learning capability.To address the problem of difficult representation of tacit knowledge,which affects the construction of rule base,this paper uses neighborhood rough sets to analyze and approximate the influence features of tacit knowledge,and establishes a rule extraction model by the C4.5decision tree.The results show that the three models established in this paper have an accuracy rate of more than 80% in the development of welding specification parameters,and the usable data of the revised model accounts for more than 90%.It has good application value and expands the application of knowledge mining technology in the field of welding process intelligence system.The model provides the knowledge rule basis for the subsequent intelligent design of the welding process.The paper constructed and implemented a multi-dimensional attribute classification system and a generative-decision tree-frame knowledge rule representation system to meet the need for unified welding process design reasoning based on cases and rules.A welding process knowledge base is established using the SPO triad form.A CBR-based RBR-supported reasoning model is used to achieve the intelligent design of the welding process while having good interpretability and self-learning capability.It was verified that the model significantly shortens the welding process design time with guaranteed correctness.A welding process intelligence system based on C/S architecture was developed based on the theoretical study.The system integrates a management system,intelligent reasoning,approval process,and multi-party interaction with automation,specialization,and batching as technical means to achieve self-learning,interpretable,and three-dimensional assembly in collaboration with three-dimensional dynamic assembly methods.The system effectively improves the efficiency of preparing welding process documents,changes the traditional guidance mode in welding assembly production,solves the problems of difficult knowledge acquisition and poor self-learning ability,and provides a reference for engineering information integration.