Neural Networks Models Usage in Methods for Combustion Process Information Extraction in IC Engines

Abstract

The In-cylinder pressure profile contains valuable information on the combustion process and its availability is greatly desirable in closed loop IC Engine control systems. The low lifetime and high costs of the currently available sensors are still preventing high-volume production of in-cylinder based engine control system. This paper deals with the potentials of Artificial Neural Networks (ANN) and their application in combustion features extraction, based solely on the crankshaft angular speed measurements. The focus of this paper is put on two concepts of ANN, based on a radial basis function (RBF) and a local linear Neuro-fuzzy models (LLNFM) and their applicability in virtual sensing of crucial combustion process parameters. Training and validation of the suggested ANN models is based on comprehensive engine test bed data set.