The Application Of Data Mining On Machine Condition Monitoring

This paper starts with the concept and application of machine condition monitoring, and introduces the research background of the technology. Based on these work, the paper proposes a two-step strategy to fulfill the monitoring:LISDC and SBR. More specifically, firstly clustering the large history data of the machine running under good condition by the clustering algorithm LISDC, then giving the realtime prediction values of the machine condition by the regression algorithm SBR. The first step-clustering-is the foundation of the whole process. This paper is organized as follows:Chapter 1 is the introduction, it introduces the background of machine condition monitoring, followed by the main work of this paper.Chapter 2 focuses on the three approaches of the implementation of the machine condition monitoring. It gives many algorithms related to the technology. Especially it gives the details of kernel regression and some useful clustering algorithms. They have been used in the application of machine condition monitoring.Chapter 3 introduces the two-step strategy:LISDC and SBR, gives the theory of the algorithms.Chapter 4 gives an implementation of the proposed algorithms, using the data from a power station. In the end, it's the performance evaluation. Chapter 5 discusses the three directions of the potential improvements of the pro-posed strategy, they are the data inputting sequences, the rule of the choosing of the similar state and the initial values of the arguments.The last part concludes the paper, describes the current and future situation of the machine condition monitoring.