A statistical approach to reduce failure facilities based on predictive maintenance

The aim of this study is to reduce the number of building facilities getting into failure efficiently based on predictive maintenance. These facilities diagnose their health states by time series sensing data and transmit event data to Monitoring Center (MC) as needed. In this paper, a statistical approach to determine when to dispatch maintenance workers is proposed. Firstly, the characteristic index is extracted from cumulative failure probability distribution using the event data. Secondly, the priority of the maintenance site visit for each facility is determined. Finally, the departure time of maintenance workers is calculated. Experimental results to evaluate whether the facilities can be maintained before failure occurs prove that this approach can reduce the number of failure facilities and mitigate the burden of maintenance workers.