Neurofuzzy-based learning algorithm for fault detection & simulation

Early fault detection is critical for safe and optimum plant operation and maintenance in any chemical plant. Quick corrective action can help in minimizing quality and productivity offsets and can assist in averting hazardous consequences in abnormal situations. In this paper, fault diagnosis based on trends analysis is considered where integrated equipment behaviors and operation trajectory are analyzed using a trend-matching approach. A qualitative representation of these trends using IF-THEN rules based on neuro-fuzzy approach is used to find root causes and possible and consequences for any detected abnormal situation. Experimental plant is constructed to provide real time fault simulation data for fault detection method verification.