A LTSA and artificial immune based early fault detection for reciprocating compressor

Aim at the problem that it is difficult to detect reciprocating compressor early fault data with complex shape clusters, a novel fault detection algorithm is put forward based on antibody clonal selection and immune memory principle. Firstly, high dimension space of raw feature signals is constructed by multivariate statistical analysis, and then the local tangent space alignment (LTSA) algorithm is employed for extracting one dimension principle manifold. Consequently incorporating artificial immune system, a dynamic memory unit evolvement operator is designed for fine-tune search when the antibody population carry out global search to improve the hyper mutation rate. Finally fault detection is performed via memory cells. Experimental results on early fault detection of reciprocating compressor´s valve leakage show that the new algorithm is efficient and effective as a detection technique. Compared with general clonal selection algorithm, the new algorithm not only has faster convergence speed, but also can achieve higher detection accuracy, and can be adopted for detecting machine fault with complex data distribution.