Application of Analytical Redundancy to the Detection of Sensor Faults on a Turbofan Engine

Abstract

In current generations of aero-engines the need to ensure the reliability of sensor measurements is met by using two or more sensors to read the same engine parameter. This technique is known as “hardware redundancy” and although reliable it does increase the cost of the sensor suite and also the weight of the engine. With the increasing computing power being fitted to new engines the opportunity has arisen of using “analytical redundancy” instead. Here the outputs of a software model of the engine are used to validate the real sensor outputs. Analytical redundancy divides into two main parts: the detection of a fault on a sensor, and the reconfiguration of the control system in response to this fault. The work conducted at DRA Pyestock has concentrated on the detection of sensor faults. The method employed uses a correlation approach to compare the shapes of the signals from the sensor and model. The sensor is declared to be faulty if the shapes become too dissimilar. Sea-level static engine trials have been conducted at Pyestock in which different faults were added to the sensor outputs of a Spey turbofan engine. The results were encouraging and indicated that such a fault detection approach could form part of a practical analytical redundancy scheme.