| The National VI emission regulation has imposed stricter controls on the particulate matter emissions of mobile machinery with diesel engines as the prime mover.Equipping with a particulate filter(DPF)on diesel engine to trap particulates and burning them at an appropriate time(when carbon load reaches a certain threshold)to achieve regeneration is an important technical approach to reduce diesel particulate emissions.However,the operating conditions of actual vehicle diesel engines are very complicated.Traditional soot loading estimation from diesel particulate filter(DPF)based on experimental calibration and pressure drop model is greatly different from the experimental results,which seriously affects the determination accuracy of DPF regeneration timing.This paper applies data-driven machine learning algorithms to predict the DPF soot loading of automotive diesel engines.Firstly,emission collection and test system based on diesel engine was built,and the road load conditions of actual vehicle diesel engine under NRTC transient cycle conditions could be simulated,then the emission characteristics,pollutants(PM,NOx,HC)characteristics and the impact of post-processing devices on pollutant emissions were analyzed in detail.After that,various engine sensor tag data related to particulate emissions were collected such as speed,torque,exhaust temperature,DPF oxygen concentration,to establish a soot loading dataset(SLDDE)of more than150,000 data.After detailed data analysis and data processing of this dataset,traditional machine learning algorithm and neural network algorithm were applied to construct and analyze its particulate emission prediction model respectively.Combining the advantages of the two types of models in terms of accuracy and stability,self-learning strategy was further applied to optimize the accuracy and generalization of the model.Through analyzing the combustion of soot inside the DPF under the NRTC cycle,a detailed regeneration mathematical model was also built to couple the optimized particulate emission model.Finally,based on the particle swarm optimization algorithm,the parameters of coupled DPF soot loading model were revised,and accuracy of the model was verified by experimental data furtherly.The research in this paper provides a new idea for accurately predicting DPF soot loading of diesel engine.In the same time,it also provides a basis for the determination of DPF regeneration timing and the development of diesel engine aftertreatment system,especially the control strategy. |
