Machinery Fault Detection Using Geodesic Distance Based on Genetic Clustering Algorithm

Abstract

Aim at the problem that there is an irregular data distribution when using multi-sensor to monitor machine conditions, a genetic clustering algorithm using geodesic distance metric (GCGD) is adopted to perform machine fault detection. In GCGD, a geodesic distance based proximity measure is employed replacing Euclidean distance that cannot correctly describe the relationship between data lying in a manifold, and GCGD determines partitioning of the feature vectors from a combinatorial optimization viewpoint. Fault detection experiments of inlet valve leakage in a two-stage reciprocating compressor reveal that GCGD achieves a better performance of fault detection than the K-means algorithm and a genetic algorithm based clustering technique.

Hybrid Genetic Clustering by Using FCM and Geodesic Distance for Complex Distributed Data

“…In the figure, the points with different color and shape represent different clusters of each data set. From the figure, we can see that the hybrid algorithm discussed in this paper also successfully acquire the desired clusters structure lying in these complex data sets, as well as the GCGD in literature [3], [4]. Consequently, the genetic clustering algorithm based on FCM and geodesic distance metric is completely feasible and availability effective on eight benchmark data sets with complex distribution.…”

“…This shortcoming makes them of limited use. In literature [3], [4], we developed a genetic algorithm based clustering using geodesic distance measure (GCGD) for complex distributed data clustering. Experimental results demonstrate the effectiveness.…”