ANFIS-based comparative exhaust gases emissions prediction model of a military aircraft engine

Abstract

In this paper, comparison of estimation methods for exhaust gaseous emissions developed for a military aircraft engine via adaptive neuro-fuzzy inference system (ANFIS) structure is introduced. For system identification process, combustion efficiency, engine shaft RPM and air-fuel ratio are preferred to be system inputs to obtain emission indexes of carbon monoxide, carbon dioxide, nitrogen oxides and unburned hydrocarbon as system outputs. While comparing the estimation methodologies, two clustering methods in adaptive neuro-fuzzy inference system structure, grid partitioning and subtractive clustering, are benefited to define membership functions. Hybrid optimisation is preferred in training parts. As a conclusion remark of the present study, estimation error values of both clustering methods are found for different number of membership functions with the common training method. Nonetheless, training time saving is the advantage of subtractive clustering method in our study.