Prediction of diesel particulate filter regeneration conditions and diesel engine performance under regeneration mode using AMSO-BPNN and combined with XGBoost

This research proposed a dependable model for predicting Diesel Particulate Filter (DPF) regeneration conditions and emission performance in diesel engines. The proposed model aims to aid engineers in optimizing diesel engine DPF regeneration efficiency, reducing emissions, and minimizing fuel consumption, thereby ensuring a highly efficient and safe DPF regeneration process. In this paper, A novel multi-model fusion prediction model, combining Improved Seagull Optimization Algorithm (AMSO)-Backpropagation neural network (BPNN) with the extreme Gradient Boosting (XGBoost) model was proposed. A test design method for DPF regeneration conditions was proposed, and the test data were obtained from the engine bench. An AMSO was proposed to improve the prediction capability of the BPNN. To enhance O2 and smoke prediction, the XGBoost model was introduced in conjunction with the AMSO-BP. Finally, the prediction model's response was experimentally validated. The results show that: After AMSO optimization, the BPNN model's predictions for T4, T5, NOx, brake specific fuel consumption (BSFC), and exhaust temperature was significantly improved, which were 0.97, 0.99, 0.99, and 0.98, respectively. The improved model Mean absolute percentage error (MAPE) is reduced by 0.01 %, 0.38 %, 0.08 %, 0.36 %, and 0.17 %. However, in the prediction of O2 and smoke density, the model's fitting coefficient (R2) remains relatively low, at 0.95 and 0.96, respectively. After adding XGBoost, the prediction accuracy of O2 and smoke is significantly improved, R2 is increased to 0.97 and 0.98, and MAPE is reduced by 1.76 % and 14.93 %, respectively. The results of the AMSO-BP-XGBoost model are consistent with the experimental, providing a foundational model for optimizing the performance of DPF.