Design and Implementation of an Intelligent Adaptive Controller for Air-fuel Ratio Control of an Automotive Engine System

Abstract

Air-fuel ratio (AFR) control is essential for maintaining the best engine performance. Parctical technique for AFR control is the propeortional-integral-derivative control in which the process in deriving the best controller parameters is very tedious and even a well-tuned controller still cannot guarantee long-term control performance. In the literature, various strategies have been developed for AFR control. These include sliding mode control, fuzzy logic control and model predictive control based on intelligent techniques. However, all of these aforesaid methods require prior expert knowledge of the engine before the controller construction. If no prior knowledge is available or the available knowledge is not sufficient, it is parctically impossible to construct a reliable controller. To address this issue, this project aims to design an intelligent adaptive controller for AFR control, in which no prior knowledge, no pre-trained model and no optimizer are required. An intelligent method called fully online sequential extreme learning machine (FOS-ELM) is used to construct the controller. Simulations have been conducted to verify the designed controller. Moreover, most of studies in the literature only perform simulations for controller verification. In this project, to evaluate the effectiveness of the designed controller, experimemnts are further set up on a real test engine. The signals of the engine sensors were studied and analyzed so that the controller could be successfully implemented on the test engine. Both simulation and experimental results demostrated that the designed intelligent adaptive AFR controller is effective for maintaining the AFR to a desired level. In addition, the designed controller has been compared with the engine built-in AFR controller. The comparison shows that the designed controller can achieve better tracking performance than the built-in one, indicating that the designed controller in this project is feasible and promising.