Development a new model based on artificial neural network to estimate torque of a conventional CI engine

Torque estimation needs intensive efforts and costly sensors. In this research, a model was proposed to estimate ITM285 tractor engine torque using some low cost sensors. Radial basis function (RBF) neural network was used for torque estimation, based on the data obtained from some inexpensive sensors including engine speed, exhaust gas opacity, fuel mass flow and exhaust gas temperature. Thirteen training algorithms were examined to train the RBF. These algorithms were compared using two statistical methods namely k-fold cross validation and completely randomized design (CRD). The Bayesian regularization (Trainbr) algorithm was the best one in regard of engine torque estimation. Based on the sensitivity analysis of the RBF, only using engine speed, fuel mass flow and exhaust gas temperature sensors are sufficient for proper engine torque estimation. R2, RMSE and EF of the RBF were 0.99, 0.50 and 0.99, respectively. It is concluded that the RBF model can be a suitable technique for estimating engine torque.