Research On Knowledge Modeling And Inference Methods In Ship Welding Intelligence Systems

Welding intelligentization is a tendency of ship welding technology.Domain knowledge acquisition is a bottleneck problem in building anintelligence system. Knowledge modeling based on the Rough Set theory(RS) provides a new approach to the development of welding designintelligence system for shipbuilding. Arc welding is a classical complexprocess. As a new method applied in welding engineering area, modelingbased on RS has been proved applicable to welding process. However, interms of handling uncertain information, there are still some issues in thetraditional RS to be addressed. Thus, it is becoming an important researchtopic to extend the existing theories and methods of rough set to deal withfuzzy-valued or continuous-valued data.An appropriate reasoning strategy plays an important role inimproving ability to solve problems of intelligence system. As twoinfluential inference methods in approximate reasoning field, theCompositional Rule of Inference (CRI) and the Similarity-basedApproximate Reasoning (SAR) have successful applications in somefields, but there are still some deficiencies that need to be addressed.Further improving inference mechanism, as well as applying theextensional fuzzy set theory to approximate reasoning field, has become aresearch focus of fuzzy inference in recent years.In this dissertation, we first summarize the application status ofexpert system and intelligent modeling in welding field. Then, throughanalyzing the problems on knowledge acquisition and knowledge inference in intelligence system, we propose the related improvement andextension methods, which are applicable to welding production design ofshipbuilding, welding deformation prediction of shiphull, weldingoperational parameters design and weld formation prediction, etc. Themajor research results achieved in this dissertation are listed as follows：(1) RS modeling and inference with application to weldingproduction planning for shipbuilding.The Rough Set theory is introduced into knowledge modeling ofwelding production design. The procedures on obtaining knowledgemodel of welding design based on the RS method are given, and aninference algorithm based on attribute significance was provided in thisdissertation,(2) Process modeling of welding deformation of shiphull based onthe vague rough set theory.Combining rough set with vague set, we present an approach toobtaining knowledge model of complex process based on the vague roughset theory. Using a study case on welding deformation prediction ofmarine high tensile steel, we introduce the application of modelingmethod based on vague rough set in knowledge modeling of weldingprocess.(3) Approximate reasoning based on vague set with application towelding field.In view of problems in the existing SAR algorithms, we propose anovel inference method based on similarity degree of vague sets. A studycase on welding deformation prediction of marine high tensile steel isused to illustrate application of the proposed method in weldingdeformation prediction field. Compositional Rule of Inference is one ofthe popular algorithms in approximate reasoning field. We give anextensional CRI algorithm based on vague set. A study case concerningwelding procedure parameters design for carbon-dioxide arc welding isused to illustrate application of the proposed algorithm to welding processdesign.(4) Knowledge modeling in weld formation prediction system.The systematic parameters in weld formation process are largelybased on continuous-valued attributes. Extending the RS to handleContinuous-valued Attributes Decision System (CADS) is a current research focus in the RS research. We present a CADS modeling methodin this dissertation. A study case concerning weld formation prediction forcarbon-dioxide welding is used to illustrate application of the proposedmethod in welding process modeling.