Research On Reasoning Mechanism And Knowledge Base Of Jacket Welding Process

With the rapid development of the domestic wind power industry,jackets have become an important support and frame structure for wind turbines,and welding is a core process that affects the quality of jacket production.Its intelligent programming has become the focus of current research.The jacket welding process has characteristics such as high current,thick plate,multi-layer,and multi-channel,which make it time-consuming and labor-intensive.Its manufacturing market is developing rapidly,requiring technicians to quickly write new welding process documents.Therefore,the establishment of an intelligent system is needed to solve the current problems.This project aims to establish a welding process reasoning system based on the characteristics of the jacket welding process,which will promote the digital and intelligent development of welding technology.The system will have great research value for the rapid generation of effective jacket welding process documents.The project consists of the following completed content:(1)Construction of jacket welding process knowledge module: Based on the characteristics of the jacket welding process,the data types of characters and numbers are separated,and then defined in a unified knowledge model.Rule models,synonym models,and unit models are built around the knowledge.After the jacket information is converted into knowledge or rules,it is added to the knowledge base.Finally,access rights to jacket welding knowledge are obtained through different retrieval methods.(2)Research on reasoning mechanism of jacket welding process knowledge:Knowledge reasoning establishes character reasoning and number reasoning on the basis of character and number welding process knowledge.Character reasoning loads the initial and default welding knowledge or rules of the jacket,selects forward or reverse reasoning,determines the upper limit of the confidence of the rules according to the quality requirements,and finally obtains the target welding knowledge with uncertainty production reasoning.After testing,this method can quickly generate applicable welding process files based on historical files.Digital reasoning uses the jacket welding process as a sample to establish multiple sets of models.The mean square error is used as the cost function,the initial welding parameters are used as the input,and the core parameters are used as the output to establish a multiple linear regression model.The analysis of the training results shows that the welding current is used as the model fitting effect,and the cross entropy is used as the cost function,the core parameter is the input,and the initial parameter is the output.Combined with softmax,the Logistic regression model is established.The analysis of the training results shows that the classification of other initial parameters is accurate except for the thickness of the base material.(3)Establishment of jacket welding process reasoning system: Python is used as the programming language,combined with Qt to complete the interface design,and SQLite is used as the data interaction object to establish a jacket welding process reasoning system.This system consists of a user module and functional module.The functional module includes information retrieval,information analysis,information reasoning,and help design.The user module provides access rights.Information retrieval involves screening pipeline welding information and knowledge.Information reasoning is based on character inference to generate the welding process file.Information analysis uses digital reasoning to add pipe rack welding knowledge and help craftsmen obtain answers for the welding process.