Detection Method for Abnormal Conditions in EMAs

This technical report presents a method for the early detection of abnormal conditions in Electro-Mechanical Actuators (EMAs) using Physics-Informed Long Short-Term Memory networks (PILSTMs). The objective is to ensure the safe and reliable operation of aircraft by identifying potential failures in the mechanical components of EMAs. The proposed method utilizes a signal reconstruction model that estimates the motor position under normal operating conditions and detects anomalies by analyzing the difference between the actual motor position and the reconstructed position. The architecture of the PILSTM combines a physics-informed layer that solves the governing differential equations and a data-driven LSTM layer that reconstructs the expected signals. The method is validated using data simulated from a high-fidelity model of EMAs, demonstrating its effectiveness in providing accurate estimates and reducing missed and false detection alarms compared to existing methods. The report discusses the challenges and advantages of integrating physics-based and data-driven approaches for fault detection.