Neuro-fuzzy approach for performance optimisation of variable nozzle turbofan engine

Abstract

AbstractAn algorithm employing adaptive neuro-fuzzy online identification and sequential quadratic programming optimisation techniques is developed to enhance aircraft engine performance. This algorithm is implemented and tested using digital simulation for two spool, mixed exhaust, variable geometry turbofan engine. Parametric study is conducted to select the proper measurable parameter that can represent the actual thrust during online optimisation. Subtractive clustering technique is applied to generate the minimum number of fuzzy rules that can model the engine performance at various input parameters and flight conditions. The resulting neuro-fuzzy system is then optimised through training algorithm to accurately represent the engine. This system can address engine variations by relearning the network using online measurements from the actual engine. Generating the optimum schedules and comparing them with those obtained from the complete non-linear engine model verify the algorithm. Benefits from this algorithm include fuel consumption savings, reductions in turbine inlet temperature, and preventing limit exceeding.