Chan, KinYip (2014) Control algorithms for optimisation of engine combustion process with continuously changing fuel composition. (PhD thesis), Kingston University, .
Abstract
Current research efforts in the areas of automotive control aim at developing new control algorithms so engine can work efficiently with different fuel compositions. Progressing in this direction, this thesis pioneers and researches novel control strategies to reduce the engine emissions gasses, Carbon Dioxide (CO2), Oxygen (O2), Carbon Monoxide (CO) and Nitric Oxide (NOx), while keeping optimum performance with unknown fuel mixtures. A two-zone engine combustion model is developed and thoroughly validated against the computational data from commercial engine simulation packages. The engine model is suited for the development and testing of control systems. The simulation uses the following fuel mixtures: isooctane (C8-H18), methanol (C1-H4-O1) and ethanol (C2-H6-O1). The results obtained provide better understanding of the control parameters, including fuel-to-air ratio, ignition timing, exhaust-valve timing and intake-valve timing. Moreover, the model facilitates control design. The novel engine controller is studied on the fuel composition as the additional parameter where such parameter has not been widely considered in engine control research. Efficient methodologies to estimate the original fuel composition by using the exhaust gas composition obtained from the engine are proposed and investigated. Two novel approaches based on feed-forward neural network and Adaptive Neuro-Fuzzy Interface System (ANFIS), respectively, are proposed. The portion of mixture for Isooctane-Methanol and Isooctane-Ethanol are effectively calculated. Moreover, results suggest that the feed-forward neural network outperforms the ANFIS approach and that the performance of the fuel estimator is stable in the continuous time process. Further in this research, a Multi-Input-Multi-Output (MIMO) engine control system is developed. The methodology used is a system identification employing a state-space model. In order to reduce the complexity of the state-space model, the developed AI fuel estimator is used to facilitate on the model reduction by feeding gains to controllers for the individual components. In addition, it uses the linear quadric regulator (LQR) method to find the closed-loop gain in the development of the closed-loop control system. The above techniques have been evaluated and results show that the controller is able to identify the minimum levels of the emission gases in terms of CO2, O2, CO and NOx in a continuously changing engine speed.
Actions (Repository Editors)
 |
Item Control Page |