Nonlinear Dynamic Modelling Of Automotive Engines Using Neural Networks

This paper presents some efforts on using neural networks to identify nonlinear dynamic models of the manifold pressure and the mass flow processes in automotive engines. Eternal recurrent neural networks are used for dynamic mapping. The dynamic Levenberg-Marquardt algorithm is applied to the weight-estimation. Early results indicate that the neural network based modeling of the manifold dynamics can result in a model comparable if not better than the first principles based models.