Fault Diagnosis of Sticking of Fuel Metering Valve in Aero-Engine Based on 1D CNNs

In aero-engine, the opening of the fuel metering valve determines the amount of fuel supplied. During engine operation, the movement of the fuel metering valve will be stuck, failing to supply fuel as expected because of the existence of unfiltered impurities or fuel viscosity. This fault is the sticking of the fuel metering valve (SFMV). In this paper, a method based on one-dimensional Convolutional Neural Networks (1D CNNs) is proposed to diagnose SFMV, this method can diagnose the SFMV of different levels under multiple operating conditions of engine. It directly processes the original fault data, integrates feature extraction and classification, thus saving calculation costs and speeding up detection. The effectiveness of the proposed method is verified by numerical simulation.