Expert system for water treatment plant operation

Abstract

An expert system based on a combination of forward chaining, backward chaining, and metarules has been developed for water treatment plant operations. The expert system was developed by interviewing a number of plant operators and chemists. Decision trees and fault trees were used to speed the development of the expert system. The combined strategy of using both backward and forward chaining within the expert system allows the number of questions asked of an operator to be minimized to prevent the asking of redundant questions. The methodology developed is applied to a case study water treatment plant. Two of the modules of the expert system are described in this paper. In the first, turbidity treatment module assists an operator in deciding a change in alum dosing in response to a change in treated water quality. In the second, a fault diagnosis module to detect the cause of equipment failure in the plant is presented.