ARTIFICIAL INTELLIGENT MODELING OF THE BI-FUEL ENGINE

In this paper, a new method for modeling of bi-fuel (Gasoline and liquid natural gas (LNG)) SI (spark ignition) engine is studied and introduced; using feed forward (FF) artificial neural network (ANN). The engine (for each fuel) has 3 inputs including the engine speed, ignition spark timing (IGT), and air fuel ratio (AFR), and 4 outputs including, brake power (BP), brake torque (BT), brake mean effective pressure (BMEP) and brake specific fuel consumption (BSFC). For improving in this model, eight parallel ANN’s have been used, each has three of the mentioned inputs and one output. Experimental data obtained from testing on a real engine is used for training and evaluation of ANN. Moreover, the data for training and evaluation are divided into two methods; Group and Points and one for training of ANN’s both standard back propagation and its modified method are used. ANN’s training is done with 70% of experimental data and evaluated with the remaining 30%. Model validation results with comparison of experimental data show that modified back propagation with classification of Points method, significantly improves the engine ordinary ANN models performance for prediction.