Fault-tolerant mining algorithm of sampling data from dynamic system

Time series data mining is an useful tool for us to design data-driven condition monitoring as well as fault diagnosis system. Aiming at monitoring abnormal changes of dynamic process, a series of mining algorithms are built up to mine signal form, model structure of process and statistical properties of noise in sampling data series, the architecture of information mining system of sampling time series is set up. These algorithms are very fault tolerant for patchy outliers in sampling data set of the complex system. Results given in this paper can be used not only in the safety analysis and fault diagnosis of complicated dynamic process but also in change detection as well as other related fields.