Vibration-Based Multicomponent Condition Monitoring and Diagnostics Approach for Elevator Door System

Abstract

Condition monitoring of automatic elevator doors is essential for ensuring operational reliability, passenger safety, and optimized maintenance scheduling. Although scientific literature extensively covers fault diagnostics for some types of machinery (e.g., rotating machinery), complex electromechanical door systems also deserve close attention. This paper proposes an intuitive and easily deployable monitoring methodology designed for broad industrial implementation. Feature extraction is performed using Autoregressive (AR) models on time-series signals. A condition monitoring index is formulated by combining Mahalanobis Distance with Box–Cox transformation, thereby normalizing the baseline data and enabling straightforward and reliable threshold setting. The methodology can be applied to both individual components and the whole system, within the same theoretical framework. Fault classification is effectively achieved using Linear Discriminant Analysis (LDA). A diagnostic interpretation tool, termed Pairwise Inter-Class Mahalanobis Distance Analysis (PICMDA), is introduced to facilitate the visual assessment of class separation. Validation on a full-scale elevator door system demonstrates high sensitivity to incipient faults and excellent diagnostic capabilities. The framework provides an effective tool for condition monitoring and fault diagnosis, with a flexible architecture that is not strictly restricted to elevator doors but can be readily extended to various other mechanical systems.