Classification of Faults in DAMADICS Benchmark Process Control System Using Self Organizing Maps

This paper presents a new approach for classification of faults in a process control system with complex overlapping fault classes. It is based on the application of Self Organising Maps that possess the capability of efficient unsupervised learning. Using the SOM training process, the proposed approach derives a set of neurons by considering process monitoring dataset comprising of multiple measured attributes. This set of neurons constitutes the multilayered SOM, in which each neuron corresponds to a class of faults. The neurons with similar attribute values are spatially arranged in adjoining localities, to set up an exploratory linkage between the SOM and the fault dataset. The performance of the proposed method is found to be satisfactory for fault diagnosis in the DAMADICS Benchmark Process Control System, even for the overlapping fault classes that pose considerable difficulty to other classification approaches applied by researchers.