Research On Turbofan Engine Modeling And Fault Diagnosis Based On Equilibrium Manifold Theory

The fault diagnosis and fault tolerant control technology for aero engine is one of the most im-portant fields of aero engine healthy management. And a simplified model with less computationalburden and satisfying precision is the foundation of the above mentioned work. This paper focuses onthe research about the fault diagnosis and fault tolerant control technology, based on a simplified en-gine model of a turbofan-engine denoted as equilibrium manifold model.Firstly, an equilibrium manifold model, with a single scheduling variable, is built. According tothe equilibrium manifold theory for nonlinear system modeling, this model is established under theground condition, the precision of which was tested for rotor speed, total temperature, total pressureand thrust respectively.Secondly, in order to emulate the engine in the entire envelope, a whole flight envelope equili-brium manifold model with twin-scheduling variables is further developed. By introducing dualscheduling variables, the equilibrium manifold model is extended to adaptive the entire envelope. Themodeling process for this model contains two parts: One is to obtain steady-state baselines throughpolynomial fitting on the basis of aerodynamic similarity transformation theory, the other is to acquirethe dynamic coefficients mapping via neural network. The simulation results show that the equili-brium manifold model for entire flight envelope meets the dynamic and static requirements for a sim-plified on-board real-time model.Thirdly, by cooperating with the aircraft dynamics, a flight/propulsion integrated model based onthe SIMULINK is applied to study engine fault diagnosis problem not just considering engine asnormally done. The followed researches are carried out based on different baseline model for compar-ison: component level model and equilibrium manifold model, and different gas path fault diagnosismethod: Kalman filter and least squares support vector machine. The simulation results show thatwhen the baseline model is the same, the two above fault diagnosis methods are almost the same inestimating accuracy, but the latter has a priority in real time ability. Nevertheless, as using a same faultdiagnosis method, it can more accurately estimate the amount of degeneration when employing com-ponent level model as a baseline model.Finally, a research on the engine diagnostics and fault-tolerant control for sensor failure is alsoconducted, by taking the equilibrium manifold model as a reference model. Two common sensor fail-ure types including bias and drift with constant rate have been analyzed and simulated, and a sensorfault diagnosis and fault-tolerant scheme is devised. The simulation results for verifying the newscheme show that it is dramatically feasible for single or dual sensor failure happened on the non-linear component engine model.