An Adaptive Neuro-Fuzzy Inference System (ANFIS) model for prediction of thermal contact conductance between exhaust valve and its seat

Exhaust valve and its seat in an internal combustion engine reach high temperatures due to the hot gases exit through the engine exhaust port. This high temperature must decrease in order to avoid damaging of the combustion chamber and the engine itself. This high amount of heat can be taken away from the exhaust valve to its seat when they are in contact, then it will be transferred to the coolant. Therefore, study of heat transfer between the exhaust valve and its seat is necessary. In this paper, the capabilities of a neuro-fuzzy approach, namely ANFIS (Adaptive Network based Fuzzy Inference System) have been studied for estimating the rate of the heat transfer between the exhaust valve and its seat. The ANFIS model is formed by means of input-output data set taken from the experimental study and the inverse solution using the Conjugate Gradient Method (CGM) with adjoint problem. It is shown that the ANFIS architecture can estimate the heat transfer rate between the exhaust valve and its seat very accurately by means of input-output pairs obtained either from the actual system or the inverse method. It is also shown that based on the error analysis between the different algorithms, gaussmf membership function provided the best model for estimating the thermal contact conductance between exhaust valve and its seat. The calculated Root Mean Square Error (RMSE) of the ANFIS architecture with the gaussmf membership function when compared against the experiment is 0.00017439.