Modelling and prediction of air path behaviour in a heavy-duty engine using artificial neural networks

The correct management of air delivery to the combustion chamber is vital to the economic and clean operation of modern internal combustion engines. However, estimation of air mass trapped in the cylinders prior to combustion in these engines proved to be challenging and yet is fundamental to the engine control process.If such an engine is boosted and equipped with an exhaust after-treatment device, the result is many degrees of control freedom compounded with highly nonlinear behaviour. Control solutions require embedded models and on-line optimisation in order to manage the often conflicting objectives of fuel economy and low exhaust emissions. The work reported in this thesis addresses the particular issue of trapped air mass estimation in a heavy-duty engine using artificial neural networks (ANN).