Enhanced Symbolic Artificial Intelligence Mechanism for External Magnetic Interference Classification in Magnetostrictive Position Sensors

In order to fulfill the requirements of precision and reliability, sensors in the context of Industry 4.0 must possess self-x capabilities such as self-monitoring. Magnetostrictive Position Sensors (MPS) are used for position and velocity measurement of objects with ultra-precision. They utilize the Time-of-Flight (ToF) technique through the reception of a structure-borne sound wave generated by a magnetic field within the sensor system. However, the precision of these sensors can be compromised by unwanted External Magnetic Interference (EMI) that interacts with the generation of the structure-borne sound wave. This research contribution introduces an innovative solution which employs an enhanced symbolic Artificial Intelligence (AI) mechanism to identify and classify EMI directly from the received sound waves (signals) during the self-monitoring process. It is emphasized that contrary to conventional methods, no extra magnetometer is used for the EMI classification. In this proposed method, Fisher's discriminant analysis is employed to identify significant regions in the sensor signals for feature extraction as a first step. Subsequently, these regions are leveraged for the application of symbolic AI. The simulation results show that the proposed symbolic AI method can significantly enhance the classification accuracy by careful extraction of the relevant features. This method also provides explainability as compared to a neural network approach.