This content is not included in your SAE MOBILUS subscription, or you are not logged in.

Development of a Neuro Genetic Algorithm Based Virtual Sensing Platform for the Simultaneous Prediction of NOx, Opacity and BSFC in a Diesel Engine Operated in Dual Fuel Mode with Hydrogen under Varying EGR Conditions

Journal Article
2011-01-2472
ISSN: 1946-3936, e-ISSN: 1946-3944
Published April 01, 2012 by SAE International in United States
Development of a Neuro Genetic Algorithm Based Virtual Sensing Platform for the Simultaneous Prediction of NOx, Opacity and BSFC in a Diesel Engine Operated in Dual Fuel Mode with Hydrogen under Varying EGR Conditions
Sector:
Citation: Banerjee, R. and Bose, P., "Development of a Neuro Genetic Algorithm Based Virtual Sensing Platform for the Simultaneous Prediction of NOx, Opacity and BSFC in a Diesel Engine Operated in Dual Fuel Mode with Hydrogen under Varying EGR Conditions," SAE Int. J. Engines 5(2):119-140, 2012, https://doi.org/10.4271/2011-01-2472.
Language: English

Abstract:

The present study is dedicated to the development of a Coupled Multi Objective Neuro Genetic Algorithm-based model to emulate the NOx, opacity and BSFC (diesel) of an existing single-cylinder four-stroke CI engine operated in various dual-fuel modes with hydrogen operated under EGR of varying thermal signatures. The associated MIMO problem has been transformed into an equivalent MISO model through unique coupling and decoupling algorithms. GA has been used as the search method to optimize the MISO topology. The MISO model substantially reduced the network weight in contrast to its MIMO counterpart thereby ensuring faster convergence and better accuracy with a consequently lower computational footprint and a promising potential to be used as a virtual sensing platform in real-time optimization strategies in engines. The MISO model successfully captured the thermal effects of EGR on the desired outputs. The performance of the proposed MISO model has been benchmarked against a conventional equivalent MIMO platform through comprehensive absolute, relative and correlation metrics. Exemplary (R2) scores of 0.9999 with the experimental values in predicting opacity, NOx and BSFC simultaneously, together with extremely low Standard Error of Estimate of R2, GMRAE and Thiel U2 uncertainty coefficient, consistently, in comparison to its competing MIMO counterpart and was also complimented with an ability to predict each of the desired output with a certainty of 97.92% within a 99.5% confidence interval. The developed model was furthermore capable of mapping the NOx-Opacity-BSFC trade-off potential of the dual-fuel modes of engine operation, even under EGR, for all cases of actual observations with remarkable accuracy, thereby establishing its inherent proficiency in providing a holistic and robust predictive platform for virtual sensing in real-time optimization strategies for such dual-fuel operation.