Modeling of internal combustion engine emissions by LOLIMOT algorithm

In this work, a neural network model of internal combustion engine is presented. First, we collect data of a 1.6L gasoline engine which was used for deriving an engine neural model. The engine was coupled to a hydraulic dynamometer to provide load. 3 inputs (engine speed, injection angle and the amount of injected fuel) were excited into a specific value range and then emissions were measured. The data of these variables are collected by a real time system. A local linear radial basis function network (LOLIMOT) was used in addition to a training algorithm for online adaptation of neural network parameters, which has a reduced convergence time. The results show the effectiveness of the proposed approach in modeling the studied gasoline engine.