Fault subspace decomposition and reconstruction theory based online fault prognosis

摘要

In this paper, a monitoring-statistic-based fault subspace decomposition and double decomposition (DD) reconstruction method is presented to deal with multivariable continuous slow-varying process fault prognosis. The new method assumed that fault is known and can be completely reconstructed. Then, the fault directions are determined by the monitoring-statistic-based fault subspace decomposition method. Finally, the DD reconstruction method and the vector autoregression (VAR) method are used to calculate and predict the magnitude degeneration process of corresponding fault directions. After that, online fault prognosis is realized. The experiments show the effectiveness of the developed method.