Sensor Validation and Signal Recovery for Helicopter Engine Torque Sensors

Errors in helicopter engine torque sensors are a major cause of costly maintenance actions, dangerous flight situations, and the underutilization of engine power. Torque sensors which read at the low-end but in range can result in an aircraft being taken offline, even requiring removal or partial disassembly of the engine, to investigate problems such as the engine torque splits that have plagued the CH-47 community for years. Furthermore, while an engine that is delivering more power than expected can be destructive to the transmission and airframe, it is also problematic if an engine is being underutilized - i.e. the sensor underestimates the amount of power that can be delivered. In this respect, more accurate torque measurement assessment can not only reduce unscheduled maintenance actions with increased situational awareness, it can also extend engine and powertrain life by more effectively using the power that an engine already possesses. The authors are developing innovative methods for validating and synthesizing torque sensor readings for the aircraft turboshaft engines with the goal of prototyping a complete torque sensor validation and signal recovery module. The solution consists of several submodules: a preprocessing module for signal denoising and operational mode detection; a diagnostic feature extraction module for extracting condition indicators based on the correlations between the operating parameters (i.e., torque and engine parameters) or electrical signals (i.e., voltages and currents); a torque signal generation module based on an empirical datadriven modal (i.e. SignalPro™) to synthesize a torque signal using the engine parameters and electrical signals; and a fault detection and classification module using pattern recognition and data mining techniques for detecting and classifying the faults on a torque system. The resulting outputs from the validation process will be failure modes of torque sensor itself and peripheral devices as well as a synthesized torque signal. Validation of the sensitivity of algorithms and techniques was accomplished though careful analysis of baseline and seeded fault test data generated by a Honeywell helicopter engine test cell as well as a scaled torque sensor system test stand utilizing a magnetostrictive torque sensor. The developed algorithmic approach promises to reduce operational cost, increase availability, and improve troubleshooting effectiveness as a retrofit to existing platforms, and as a future enhancement to development engines.