Literature review indicates that a large amount of failures are caused by abnormal wear of the diesel engine components. It is therefore essential to monitor the engine condition using the tribological information. Although the wear debris analysis has been proven to be effective for condition monitoring and fault diagnosis (CMFD) of diesel engines, limited work has been done to address the remote on-line CMFD system in practice. To extend the oil monitoring technology into industrial application level, a new remote on-line fault diagnosis system for marine diesel engines has been proposed in this paper. The new system consists of an on-line tribological signal acquisition model in the ship, a remote feature extraction model and a fault diagnosis model in the laboratory center. Nine wear characteristics were extracted to detect the engine faults, including the surface roughness of wear particles, oil moisture and viscosity, and index of particle covered area (IPCA), etc. In order to select a relative best feature for the on-line fault detection, the interaction information based feature selection method was employed to determine the suitable indicator. This study has found that the IPCA is the best feature among other eight features to on-line respond the engine condition changes. The diagnosis results show that the new system offers satisfactory on-line fault diagnosis ability and is effective for the diesel engine fault diagnosis in practice.
