Investigation of Fault Detection and Isolation Methods for Induction Motor Drives in Off-Grid Solar Power Systems in Adamawa State

Induction motor drives powered by solar photovoltaic systems are increasingly deployed for water pumping and agricultural processing in off-grid communities in Adamawa State, North East Nigeria. Electrical and mechanical faults in these drives cause system shutdowns that can have severe humanitarian consequences in remote communities, yet automated fault detection systems are rarely integrated into off-grid solar motor installations. This study investigates model-based and signal-processing fault detection and isolation methods for common induction motor drive faults, including stator winding inter-turn short circuits, rotor bar breakage, bearing defects, and inverter switch faults. A comprehensive fault simulation platform comprising a 3 kW induction motor coupled to a programmable load and driven by a three-phase VSI inverter is constructed and instrumented with current, voltage, vibration, and thermal sensors. Faults are induced sequentially under controlled conditions, and motor phase current spectra, vibration acceleration spectra, and Park's vector patterns are analysed using MCSA, wavelet transforms, and principal component analysis respectively. A model-based residual generation approach using an adaptive observer is developed and compared with signal-analysis FDI methods in terms of detection speed, sensitivity, and false positive rate. An artificial neural network classifier trained on extracted features demonstrates 94.6 percent fault classification accuracy across seven tested fault categories. The study proposes a combined observer-neural-network FDI architecture optimized for the embedded processing resources available in typical solar motor controller platforms. Keywords: fault detection, induction motor, solar power, MCSA, off-grid Nigeria