A Non-Intrusive Deep Learning Based Diagnosis System for Elevators

Abstract

With the ever-growing number of elevators coupled with the aging workforce, diminishing new installations and limited use of maintenance technology, it is increasingly challenging for the owners and responsible parties to maintain the safe and reliable operation of the lift systems. To address this issue, a non-intrusive artificial intelligence (AI) based diagnosis system, aiming at providing fault detection and potential fault prediction for multi-brand lifts without intervening the existing circuitry of the lift installations, is proposed in this paper. The proposed system employs the multivariate long short term memory fully convolutional network (MLSTM-FCN) to learn and analyze the measured signals from the non-intrusive detection system of the elevators. It is capable of (i) giving advance and clear warnings of corrective actions to prevent major equipment breakdowns, and (ii) indicating just-in-time maintenance for enhancing the lift reliability at a low cost. The implementation of the non-intrusive detection system is provided. The design of the diagnostic algorithm is elaborated. Both the simulations and experiments of a commercial elevator have been conducted to verify the effectiveness of the proposed system.