Abnormal Derivative Frequency for Sensor and Wiring Prognostics

This paper presents a new method for the detection and prognosis of sensor and wiring failures in rotorcraft and aircraft. Our approach, called the Abnormal Derivative Frequency (ADF), is able to detect sensor and wiring failures in the challenging setting where sensor readings do not exceed pre-determined thresholds and thus do not trigger a rotorcraft's onboard diagnostics algorithms to issue warnings, such as fault messages. The ADF is straightforward to compute, which makes it amenable to implementation in onboard mission processor software. We demonstrate the effectiveness of our method on field data collected from hundreds of Apache rotorcraft. We show that the ADF is able to reliably detect the onset of sensor and wiring problems well before they are caught by onboard diagnostic algorithms and maintenance personnel or crew members. Our approach has the added benefit of helping to distinguish between sensor/wiring problems and actual component failures, thus reducing trouble-shooting time and maintenance costs.