Fault Analysis For Aero-Engine Temperature Sensor

The temperature sensor of aero-engine is an important part of the control system,so the faults of the temperature sensor affects the safety of the control system.The faults analysis of aero-engine temperature sensor and the establishment of Faults Detection,Isolation and Accommodation(FDIA)system can ensure the normal work of the control system.A knowledge-based fault analysis method is applied to build up the FDIA system to detect,isolate and accommodation the faults through the residual error between the predicted value and the output value of the temperature sensor.And the predicted value will be obtained by neural networkDifferent types of temperature sensors are introduced in this paper,especially the resistance temperature sensor and the thermocouple temperature sensor.The failure modes,phenomena and causes of these two temperature sensors are analyzed.Simulink software is used to simulate two kinds of temperature sensors and fault signals.Matlab software is used to analyze the advantages and disadvantages of Back Propagation(BP)neural network prediction.BP neural network is optimized by Genetic Algorithm(GA),and GA-BP neural network is established which can control the prediction error within 3%.The two factors that affect the performance of GA-BP neural network are analyzed: the number of hidden layer nodes,the interval between training samples and prediction samples.At the same time,the performance of two kinds of neural networks is compared.It is verified that the GA-BP neural network has the advantage of higher accuracy,higher stability and less iteration steps than the BP neural network.The faults detection based on the residual between the predicted output of the GA-BP neural network and the actual output of the temperature sensor is carried out.The simulation platform of aero-engine temperature sensor faults detection,isolation and accommodation is set up by using Simulink software.Through this platform,the faults detection,isolation and signal accommodation are successfully realized.Finally,the influence of the hard faults amplitude and the soft faults change rate on the faults detection are analyzed.The bigger the amplitude of the faults and the faster the change rate of the faults,the better the detection andreconstruction effect is.