Insights into the Fusion Correction Algorithm for On-Board NOx Sensor Measurement Results from Heavy-Duty Diesel Vehicles

Abstract

Over the last decade, Nitrogen Oxide (NOx) emissions have garnered significantly greater attention due to the worldwide emphasis on sustainable development strategies. In response to the issues of dynamic measurement delay and low measurement accuracy in the NOx sensors of heavy-duty diesel vehicles, a novel Multilayer Perceptron (MLP)–Random Forest Regression (RFR) fusion algorithm was proposed and explored in this research. The algorithm could help perform post-correction processing on the measurement results of diesel vehicle NOx sensors, thereby improving the reliability of the measurement results. The results show that the measurement errors of the On-board Nitrogen oxide Sensors (OBNS) were reduced significantly after the MLP-RFR fusion algorithm was corrected. Within the concentration range of 0–90 ppm, the absolute measurement error of the sensor was reduced to ±4 ppm, representing a decrease of 73.3%. Within the 91–1000 ppm concentration range, the relative measurement error was optimised from 35% to 17%, providing a reliable solution to improve the accuracy of the OBNS. The findings of this research make a substantial contribution towards enhancing the efficacy of the remote monitoring of emissions from heavy-duty diesel vehicles.

Keywords

• heavy-duty diesel vehicles
• on-board nitrogen oxide sensors (OBNS)
• fusion correction algorithm
• multilayer perceptron (MLP)–random forest regression (RFR)
• machine learning