Advanced Design and Anomaly Detection Techniques for Inductive Position Sensors

In recent years, Inductive Position Sensors (IPS) have emerged as a robust and low cost technology for position measurement in harsh industrial and automotive environments. However, these sensors still face two major challenges: the coil design of IPS and their inherent sensitivity to misalignment. This thesis addresses both challenges by introducing advanced techniques for IPS design. Beyond their conventional use as position sensors, integrating anomaly detection functionality directly into the IPS offers significant practical advantages. For this reason, this thesis also proposes an innovative method that exploits undesired signal variations produced under misalignment for anomaly detection in rotating machinery. A fast electromagnetic simulation framework based on integral methods is employed as the core basis for this work, enabling rapid IPS performance evaluation, up to three orders of magnitude faster than commercial Finite Element Method tools. First, advanced design methodologies are introduced to accelerate and improve IPS development. A machine learning based Design Support System is developed to guide designers through large and highly nonlinear design spaces. Furthermore, a novel coil optimization strategy based on geometric harmonics is presented, providing a fast and controlled method for reducing the linearity error of IPS. Second, it is investigated how IPS sensitivity to rotor misalignment, typically regarded as a limitation, can instead be exploited for anomaly detection in rotating machinery. A detailed analysis of the characteristic distortion patterns generated by different mechanical deviations is performed, enabling a real-time detection method capable of identifying deviations from nominal behavior. The research results are finally evaluated experimentally using two different test benches. Based on the results from these tests, IPS can detect mechanical anomalies and that the proposed method is inherently independent of rotational speed, as it relies on positional information rather than vibration frequency.