Application of Parity Relation in Elevator Fault Diagnosis

A data-driven fault detection and diagnosis is a promising technology to prevent significant increase failures of elevator due to prolonged uses. But it is facing a challenge of irregular data sequence because of the randomness of elevator operation. In the paper the parity relation approach is proposed to overcome this difficulty based on artificial selection of effective time series. Firstly the left-zero matrix is obtained from the valid time series of state variables standardized by Hankle form by singular value decomposition. Then the residual space is established and the threshold is constructed by health data sets under various conditions. Finally the results are obtained by comparing the test sequence with the threshold in the residual space. Numerical simulation results validate the effectiveness of the proposed method.