Experimental Characterization of Motor Fault Conditions Using Vibration, Gyroscope, and Acoustic Measurements with an Arduino-Based Monitoring System

Abstract

Background: Electric motors are widely used in industrial and electromechanical systems, making reliable condition monitoring important for detecting mechanical abnormalities before performance deterioration occurs. However, individual fault conditions may produce different responses depending on operating speed, fault geometry, mounting conditions, and sensor modality. Low-cost multisensor platforms integrating vibration, angular-motion, and acoustic measurements therefore provide a practical approach for experimentally characterising distinct motor fault signatures. Aims: To develop and experimentally evaluate a multisensor platform that identifies controlled motor fault conditions using vibration, angular motion, and acoustic telemetry. Study Design: A benchtop repeated-measures experiment was conducted to test a healthy baseline and six controlled mechanical conditions across four rotational speeds. Methodology: A 12-V geared DC motor was controlled by an Arduino Mega 2560 and monitored with an Arduino Nano 33 BLE Sense Rev2, logging telemetry at 40, 80, 100, and 130 RPM. Six fault states were evaluated: four rotor mass-distribution configurations, a 15-degree shaft misalignment, and loosened motor-fixing screws, with three replicates per fault-speed combination. Trial-level features included AC resultant acceleration RMS, Dynamic-G variability, gyroscope motion, relative acoustic RMS, and packet-loss rate. Results: Vibration magnitude depended on mass distribution rather than total added mass. Two adjacent eccentric masses produced the strongest vibration, three masses approached the one-mass response, and four symmetrically distributed masses reduced vibration despite the greatest attached mass. Configuration effects were strongest at 100 RPM (one-way ANOVA, F = 66.90, P < .001) and 130 RPM (F = 32.69, P < .001). The 15-degree misalignment produced a directional, non-monotonic vibration response, while loosened motor mounting produced a comparatively flat vibration response consistent with altered structural coupling. Acoustic RMS correlated with speed rather than fault type; packet loss remained at 13-14% regardless of mechanical state. Conclusion: An embedded multisensor platform reliably identifies motor fault signatures when speed-specific and multidimensional features are considered. The results also show that a mechanical fault should not be interpreted simply as an increase in overall vibration amplitude.