An Expert System for Fault Diagnosis in Fibre Production

Abstract

The maintenance of consistent product quality from an acrylic fibre manufacturing plant requires the rapid diagnosis of plant malfunctions. An expert system is being developed to aid process operators in this task. Various fibre quality parameters are monitored using control chart techniques. When predetermined tolerance limits are violated the system will attempt to identify the cause in terms of specific process equipment failures or mal-operations. In the proposed design, the diagnostic task is treated as a three stage process of data abstraction, hypothesis selection and solution refinement. Knowledge acquisition then involves eliciting the specific knowledge types associated with each of these tasks. Solution refinement is addressed by an hierarchical structure based on process objectives. Decomposition of the process, in terms of the fundamental tasks which it is designed to perform, aids the structuring of the knowledge base and facilitates efficient fault identification. The principal benefits of this approach are improved knowledge acquisition and knowledge base validation and modification.