Abstract

Acoustic diagnostics, traditionally associated with mechanical fault modes, can potentially solve a wider range of monitoring applications. Typically, fault modes are induced purposefully by the researcher through physical component damage whilst the system is shutdown. This paper presents low-cost real-time fault diagnostics of a previously unreported acute electrical origin fault that manifests sporadically during system operation with no triggering intervention. A suitability study into acoustic measurements from readily available consumer-grade sensors for low-cost real-time diagnostics of audible faults, and a brief overview of the theory and configuration of the wavelet packet transform (including optimal wavelet selection methods) and empirical mode decomposition processing algorithms is also included. The example electrical origin fault studied here is an unpredictable current instability arising with the PWM-controller of a BrushLess DC motor. Experimental trials positively detect 99.9 % of the 1160 resultant high-bandwidth torque transients using acoustic measurements from a USB microphone and a smartphone. While the use of acoustic techniques for detecting emerging electrical origin faults remains largely unexplored, the techniques demonstrated here can be readily adopted for the prevention of catastrophic failure of drive and power electronic components.