Application of Hybrid Engine Modeling Method Based on Neural Network Group and PSO with Adaptive Inertia Factor in Engine Calibration

The traditional hybrid engine calibration method has low efficiency and high cost of manpower and material resources, which can not meet the calibration requirements of complex engine electronic control system. The calibration method based on a mathematical model can greatly reduce the test workload and improve efficiency. Therefore, the black-box model of the engine is constructed by using the results of Spearman correlation analysis, and nine variables are selected as input, at the same time five variables are used as outputs. The improved RSR-BPNNG neural network group method is used to construct the hybrid engine economy and emission model. The model prediction results show that the R² value of fuel consumption prediction reaches 0.9975, and the R² value of NOx emission prediction reaches 0.9933, which achieves high precision modeling. On this basis, the performance of the engine under the WHSC cycle is simulated and optimized by the improved adaptive PSO algorithm. The optimization results show that the NOx emission of the engine is reduced by 8.16%, and the fuel consumption is reduced by 4.55%.