Fault Diagnosis of Multiple Components in Complex Mechanical System Using Remote Sensor

Abstract

This study proposes an approach to monitor multiple components in complex mechanical systems using a single, externally placed remote sensor. In automobiles and petrochemical plants, where numerous components (e.g., powertrain, bearing, and gear), sensor placement is often compromised by cost and installation environment constraints, resulting in sensing the components far from the regions of interest. To address this challenge, this paper proposes an Operational Transfer Path Analysis (OTPA)-based approach that derives the transfer functions between the vibration excitation source and the measurement point (i.e., receiver). The model for OTPA enables the reverse estimation of the excitation source’s signal from the receiver. Subsequently, the estimated (i.e., synthesized) source signal is fed into a diagnostic model to identify system faults. The OTPA and diagnostic models are constructed using neural network architectures, enabling better adaptation to operational conditions and system-induced nonlinearities. The proposed approach is validated from case studies using hydraulic piston pumps in construction vehicles and next-generation electric vehicles.