Non Linear Predictive Modelling for IC Engine Using Artificial Neural Network

Abstract

Artificial neural networks are powerful data computational models which have the capability of representation of complex input-output relationships of physical systems. Further they could perform “intelligent” tasks that performed by the human brain. In this work a predictive nonlinear model of an internal combustion engine is simulated using Elman recurrent neural work, Cascade Forward Neural Network and a Feed Forward Neural Network to predict the operational parameters engine torque and the nitrous oxide emissions. The parameters fuel rate and speed of the engine serve as input. A standard bench mark dataset is used for training the Elman neural network. The simulations results confirm that the Neural Network models can map the nonlinear input -output relationships in an effective manner. All the three different neural networks could map the input-output relationship and the test results confirm that Elman Neural Network has the best performance.