Simulation-Driven Machine Learning for Solving the Inverse Problem of PCB-Based Tilt-Inductive Position Sensors

PCB-based Inductive Position Sensors (IPS) are advantageous with respect to resistive, capacitive or Hall-effect alternatives since they allow for contactless sensors that are immune to stray fields and robust to temperature changes, all while being easy to manifacture by using conventional PCB fabrication techniques. In this work we present an IPS design for measuring the angle of a target conductor with respect to the sensing coils, and the distance between them. Simulations are used to build a dataset for training a neural network to solve the inverse problem of estimating tilt and airgap from the measured voltages. The trained network is tested against experimental data measured from a physical prototype.