Research On LEAP Engine Variable Stator Vane’s Control Logic And Condition Monitoring Based On QAR Data

Variable stator vane(VSV)is an important mechanism of LEAP engine airflow control system,which can enlarge the aerodynamic stable margin of LEAP engine compressor and effectively improve the aerodynamic performance of compressor.The working environment of civil aero engine is quite complex.In order to avoid the occurrence of engine surge,the variable stator vane continuously adjust the attach angle frequently,so the direction of the air flow into the high-pressure compressor is always keeping in a reasonable range.The failure of the variable stator vane will lead to the engine surge,therefore bring great threaten to the aircraft safety,so research of variable stator vane control logic and condition monitoring is very essential.Surely the method of mathematical modeling can deeply analyze the control logic of the LEAP engine variable stator vane from the principle level,but this method requires profound technical information of LEAP engine.At the same time,LEAP engine is designed and produced by foreign manufacturers,that makes mathematical modeling method become difficult and almost impossible in the research of LEAP engine variable stator vane control logic.Different from the mathematical modeling method,the data driven method does not rely on the technical information of LEAP engine too much,just rely on massive LEAP engine operation data,while massive LEAP engine operation data has been accumulated with the operation of LEAP engine in Chinese airlines.The variable stator vane control logic is different under different operation conditions and operation status,but due to the confidential limitation of LEAP engine,there is a dilemma to distinguish variable stator vane’s operation data under the different operation conditions and operation status.Therefore,based on analysis of massive and real operation data of LEAP engine,this thesis first selects the steady-state extraction method to standardize the data division of different operation conditions and operation status.Then based on the massive and real LEAP engine operation data recording in the aircraft rapid access recorder(Quick Access Recorder,QAR),this thesis adapts 1DCNN-Resnet-LSTM and 1DCNN-Bi LSTM-Attention to simulate the full state control logic and divided state control logic of variable stator vane.Validation results verify that both models have great performance no matter in the divided state or full state.At the end of this thesis,comparing full state control logic and divided state control logic based on 1DCNN-Resnet-LSTM and 1DCNN-Bi LSTM-Attention,full state control logic based1DCNN-Bi LSTM-Attention model is applied to establish a monitoring system of the variable stator vane.Validated with real variable stator vane fault data,this system can detect the fault of variable stator vane,also can locate the specific starting and ending time of the fault,and obtain the offset angle of the variable stator vane.