Plant Fault Diagnosis System using Negative Selection Algorithm

Early fault detection and correct diagnosis are required for chemical plants. Therefore, the existing fault detection systems using upper/lower thresholds have difficulties to detect faults when the correlation among process variables breaks without excess of any thresholds. In our previous study, an artificial immune system—especially, negative selection algorithm— had been adopted to fault detection system. Negative selection algorithm is one of methods of artificial immune systems which imitate the vital immune system. We have built up a multiagent based fault detection system using negative selection algorithm. In our system, a set of detectors is generated in each two-dimensional variable space consists of two process variables. In this study, we extend the system to plant fault diagnosis based on the fault detection result using negative selection algorithm. In this paper, we will illustrate our fault detection and diagnosis system using negative selection algorithm. And we will show the detection and diagnosis results when a malfunction occurs in a dynamic plant simulator of a boiler plant.