Development of A Diagnostic Expert System (FDD-Expert) for Woven Fabric Defects

Abstract

A variety of computer aidedtechniques have been proposed to help solving diagnostic problems in differentdisciplines.  Commonly known as expert systems, approaches,  undertaking distinct formalisms for modellingdomain knowledge, include: rule-based and frame-based systems or semantic andneural networks.  Explanatory mechanismsare more easily implemented with rule-based systems and graphical models makingthese preferred solutions to the diagnostic problem. Since the pioneering workby Feigenbaum (DENDRAL, 1971) theory and utility of expert systems haveprogressively evolved, and  accompaniedby the appearance of various expert systems development tools. In textile industry, expert systems aregenerally used to increase production, improve quality and reduce costs.This paper describes FabricDefect Diagnosing System (FDD-EXPERT), an expert system in the domain of  fabric defects.  Fabric defect is known as an undesirablefault in the fabric which deteriorates the quality fabric and makes itinferior.   FDD-EXPERT contains 173 different defects  based on TS471 ISO8498 Woven Fabric Description of defects with additional 47defects.  Defects depicted in this domainare classified based upon their similarity of appearance,  multiple attributes relevant to effect uponvisual quality and sources of defects. Possible reasons  for each defect and remedial solutions arealso included. A rule-based knowledge domain and interference engine isgenerated in PROLOG.  User friendlyinterface of the system is  developedthrough Visual Basic (VB.Net).  It is acombination of logic and declarative programming. This diagnostic system may beutilised as problem solving  and trainingtool by novice operators  or  as a supplementary knowledge for expertoperators.