Bilateral contour descriptor and manifold learning based fault diagnosis for non-periodic vibrating equipments on ships

In this paper, the bilateral contour descriptor and the nonlinear manifold learning were presented for fault diagnosis of non-periodic vibrating equipments on ships. The fault diagnosis for periodic or rotating equipments was researched widely. There were also many non-periodic and non-rotating equipments on ships. Due to the unreasonable description and the redundant signal, it was difficult to diagnose fault of non-periodic and non-rotating equipments. The bilateral contour descriptor was presented to describe the reasonable feature of the non-periodic and non-rotating equipments. The bilateral contour descriptor fused the spatial position information, the feature weight and the difference resulted by the sensor distribution. Then the original high-dimensional feature data was handled by the nonlinear manifold learning. The lower-dimensional description of the original vibration feature showed the intrinsic characteristic of the non-periodic and non-rotating equipments. In the end, the accurate results of fault diagnosis was obtained for the non-periodic and non-rotating equipments.