Permanent magnet synchronous motors (PMSM) are an emerging high-power energy system that is prevalent for industrial manufacturing applications. However, there is a scarcity of experimental data, due to the novelty of the system and the cost of experimental testing. Nonetheless, it is imperative to accurately and efficiently monitor the health of these systems. Unexpected breakdowns can lead to catastrophic failures, from extreme revenue loss to even human life. In recent literature, physics-informed machine learning has shown success for fault detection within various engineering applications. These include but are not limited to electric vehicles, propulsion aircrafts, ultra-high-speed elevators, additive manufacturing, and many other impactful concentrations. This study aims to develop a fault detection framework for PMSMs, which will enable efficient health monitoring and fault detection. In particular, the proposed method utilizes generative machine learning techniques to simultaneously determine the optimal placement of sensors while training a classifier of faults. In addition, the case where a sensor fails is considered, ensuring one level of resilience for the chosen design. Predicting these faults will enable appropriate maintenance plans, which ensures that manufacturing will safely meet the expected demands. Various search algorithms are implemented to solve the generally applicable mathematical formulation, which utilizes predictor accuracy as the fitness function. Overall, this proposed method converges to a design that has high accuracy for detection of faults, and also satisfies a N-1 redundancy criterion.