The predictive ability of artificial neural networks where a large number of experimental data are available, has been studied extensively. Studies have shown that ANN models are capable of accurately predicting NOx emissions from engines under various operating conditions and different fuel types when trained well. One of the major advantages of an ANN model is its ability to re-learn when new experimental data is available, thus continuously improving its accuracy. The present work explored the potential of an ANN model to predict NOx emissions for various engine configurations outside its training envelop. This work also looked into quantifying the amount of new data required to improve the accuracy of the model when exposed to unknown conditions. The chosen ANN model was constructed using data from a high-speed direct injection diesel engine and is capable of accurate NOx emissions over a wide range of operating conditions. The optimized network utilized 14 input parameters and is using 6 neurons in a single hidden layer feed-forward neural network.

Experimental data from the various engine configurations tested, were then used to predict NOx from the existing ANN model. The results indicate that when the new data are within the original training envelop, the ANN model is capable of accurate NOx prediction even when there are substantial changes in engine configuration such as piston material. Similar results were also observed when the injector nozzle is changed. However, the model's performance drops significantly when new data, outside the original training envelop, were employed indicating that additional training is required. As such, various methods for re-training the ANN model were explored with the selected method showing the best compromise between new-data accuracy and old-data accuracy retention. Furthermore, the effect of signal noise on model accuracy was also explored. The re-trained ANN model developed was found to be an effective tool in predicting NOx emissions for different engine configurations and operating conditions.


Prediction of NOx Emissions for a Range of Engine Hardware Configurations Using Artificial Neural Networks

Division / Sponsors: 
Internal Combustion Engine Division (ICED)

Paper Proceedings: https://asmedigitalcollection.asme.org/ICEF/ICEF2020/volume/84034

ICEF2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.