Research On Equipment Performance Degradation Based On SVDD And Information Fusion

Intelligent Maintenance for large equipment has been becoming a new hot spot in fault diagnosis research field. The equipment performance degradation assessment is one of important parts of the Intelligent Maintenance, which is also the foundation of the equipment running states prediction. Generally speaking, the process of the equipment performance degradation can be divided into many different performance degradation states. It will be easier for engineer to organize production and maintain equipment if the equipment state and its degradation level could be identified when it running, which will prevent the fault happening effectively.With the support of Key Project supported by National High-tech R&D Program (863 Program) of China (2006AA04Z175) and project supported by National Natural Science Foundation of China (50675140), a novel method for performance degradation assessment is proposed, which bases on two techniques: Support Vector Data Description (SVDD) and Information fusion technology.SVDD is a new data domain description method, which proposed in the statistical Learning Theory. This data description should cover the class of objects represent by the training set, and ideally should reject all other possible objects in the object space, which can be widely applied to fault diagnosis. The new method based on SVDD is proposed in our research. The algorithm was used to assess the parts performance according to the data from sensor. Simultaneously, considering the computational complexity brought by state recognition, a feature reduction algorithm based on genetic algorithm is adopted to extract the feature information which makes the assessment result more accurately and quickly.As for the need of comprehensiveness equipment performance assessment, multi-sensor data acquisition was adopted, which means that the information fusion become necessary in assessment process. Dempeter-Shafer theory is adopted in our research, which can deal with the evidence redundancy, uncertainty, even contradiction effectively.As shown in experience, the assessment result using the process proposed in this paper accords with the practical situation, which means this assess method could be used in future. An equipment performance degradation assessment system based on SOA was proposed, which lays a very solid foundation for application practical application of this assessment method.