Negative Selection Inspired Machine Learning Approach for Damage Detection

ABSTRACT The fault detection of dynamics systems is of great importance when it comes to ensuring safety and increasing their performance. The present paper uses the concept of the natural immune systems and takes the advantage of its major characteristics like evolution and trainability for fault detection purposes. In the presented approach, first, both negative and clonal selection methods are used in order to improve the rate of convergence. Second, the generated population is compared to the members of memory cells (as the set of best samples of overall population) that results in a more efficient local search and faster convergence. Finally, the obtained results from the artificial immune system (AIS) are compared to the genetic algorithm (GA). The simulation results indicate that the proposed approach shows a higher accuracy in terms of identifying the location and intensity of the fault in the system under study.