Negative Selection Algorithm with Applications in Motor Fault Detection

Abstract

Natural immune systems are complex and enormous self-defense systems with the remarkable capabilities of learning, memory, and adaptation (Goldsby et al. 2003). Artificial Immune Systems (AIS), inspired by the natural immune systems, are an emerging kind of soft computing methods (de Castro and Timmis 2002). With the distinguishing features of pattern recognition, data analysis, and machine learning, the AIS have recently gained considerable research interest from different communities (Dasgupta 2006; Dasgupta and Attoh-Okine 1997; Garrett 2005). Being an important constituent of the AIS, Negative Selection Algorithm (NSA) is based on the principles of maturation of T cells and self/nonself discrimination in the biological immune systems. It was developed by Forrest et al. in 1994 for the real-time detection of computer viruses (Forrest et al. 1994). During the past dec ade, the NSA has been widely applied in such promising engineering areas as anomaly detection (Stibor et al. 2005), networks security (Dasgupta and Gonzalez 2002), aircraft fault diagnosis (Dasgupta 2004), and milling tool breakage detection (Dasgupta and Forrest 1995). In this paper, we first introduce the basic principle of the NSA. Two modified NSA, clonal selection algorithm-optimized and neural networks-based NSA, are next introduced. Their applications in the motor fault detection are also discussed.