Automatic qualitative trend simulation method for diagnosing faults in industrial processes

Abstract

This work proposes an automatic method of qualitative simulation for industrial processes to predict the steady-state measurement patterns arising from different faults. Due to their characteristics, qualitative simulations tend to generate multiple spurious solutions. The proposed method limits the number of such spurious solutions by automatically generating new qualitative equations from the generic quantitative model (i.e. without the need of knowing the value of its parameters) and using simple qualitative known relations or other readily available information. An algorithm to obtain such solutions from the set of qualitative equations is also presented.