Elman and Jordan neural networks for prediction of transient thermal contact for engine exhaust valve

In this study, feedback neural networks namely Elman and Jordan are used for prediction of exhaust valve temperature for air cooled engines. Input-output data are extracted from an experimental setup including the valve mechanism of an air cooled engine. Inverse heat transfer problem applying the Adjoint problem is used to address the thermal flux through exhaust valve and seat. Elman and Jordan neural networks are used to predict the transient valve temperature using the experimental data. The results show that Elman and Jordan neural networks predicts well the transient exhaust valve temperature. However, Jordan neural network with training algorithm of Gradient Descent with Adaptive Learning Rate performs better with RMSE error of 16.3 for prediction of exhaust valve temperature.