Improvement and prediction of particles emission from diesel particulate filter based on an integrated artificial neural network

Abstract

Diesel Particulate Filter (DPF) stands out as a highly effective device for mitigating emissions in engines. To enhance DPF regeneration performance, the numerical model and the GA-BP neural network model are developed, which delves into the impacts of velocity, temperature, oxygen mass fraction, and particles size on particulate conversion. The results show that conversion rate of carbon particles can be elevated by increasing the oxygen mass fraction and inlet velocity. Specifically, the conversion rate demonstrates a remarkable improvement of 37.41% at Tin = 600 K, de = 5 μm, Vin = 10 m/s, and mO2 increased from 0.01 to 0.04. Additionally, conversion rates are increased as the size of the carbon particles gradually reduced. Besides, a GA-BP neural network is deployed to analyze and predict the numerical results of 1818 sets of DPFs under different operating conditions. From the analysis and prediction of 132 data sets, and it is discerned that a high state of contamination transformation can be achieved at Tin = 525 K, de = 5 μm, Vin = 12 m/s and mO2 = 0.04. This demonstrates the significance of judiciously selecting boundary conditions for realizing effective regenerative emission reduction.