Abstract:Wiebe formula as a semi-empirical formula that calculates heat release in the zero-dimensional combustion model of diesel engine was widely used in the simulation of working process of diesel engine, the accuracy of simulation was mainly dependent on choice of parameters in the formula, but there were some blindness and poor universality in the traditional Wiebe parameters selection method. So in view of shortcoming of the selection method, the D4114B type electricitygenerating diesel engine was taken as an example, a diesel engine Wiebe formula parameters prediction method was proposed based on neural network. By experimental measurement of cylinder pressure curve, the combustion heat release rate was backward deducted, the heat release rate curve was numerically fitted, and the neural network was trained, a neural network which can be used to predict the parameters of Wiebe formula was established. Through the comparison of predicted results and experimental data, and evaluation of prediction accuracy, the accuracy and feasibility of prediction method were verified. On this basis, D4114B diesel engine dynamic simulation model was set up by using Modelica language, and the transient performance of diesel engine was simulated and researched. Finally, the simulation results of neural network input parameters were compared with experimental value, which proved the Wiebe formula parameters prediction method can be applied to the diesel engine dynamic simulation study.