This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Hybrid plant modelling of diesel engine and After treatment systems using Artificial Neural Networks
ISSN: 0148-7191, e-ISSN: 2688-3627
Published December 19, 2019 by SAE International in United States
Annotation ability available
For Euro VI & JOBD-II emission compliance, emission control software and fault monitors are complex. In order to test such complex functionalities on a Hardware-In-Loop (HIL) environment, a realistic plant model is necessary. A realistic plant model can replicate real life scenarios accurately and help create scenarios difficult to test on a vehicle. A realistic plant model can increase the scope of emission software controls and OBD fault monitor testing on a HIL system.
- (b)Problem statement:
Emission control software interacts with emission control devices based on complex chemical and physical interactions. Although physical and empirical approaches of modeling the complex emission plant models have been explored earlier, there is a tradeoff between plant model complexity and real time performance on HIL system, also there is a large effort and equipment infrastructure spent on parametrization of the complex physical and empirical models using techniques of Design of Experiments (DOE) and data analysis.
One promising avenue is to introduce machine learning and sophisticated control techniques inspired by artificial intelligence into plant modelling.
Within machine learning and artificial intelligence, neural networks are particularly well-suited to modelling, control, and diagnostic analysis of complex, nonlinear, and time-varying systems.
Consequently, the use of neural network-based modeling and control techniques could be of significant benefit to diesel engine and after treatment system modelling.
The purpose of this paper is to introduce a hybrid plant modeling technique which combines neural networks controls with a traditional physical plant model and to report on work in this hybrid modelling methodology used on engine and emission controls plant models.
We describe the challenges of inculcating neural control with traditional physical modelling approach and highlight the adaptations required to the neural model to deploy the hybrid into a real time HIL system. Further, we present the initial experimental results computed using neural networks trained with diesel engine and after treatment controls data.
In this paper, we conclude with presenting the output of the plant models, developed using the hybrid modelling technique, once deployed on a real time HIL system.
CitationKumar, S., Gope, S., Vijapur, A., and Nakayama, S., "Hybrid plant modelling of diesel engine and After treatment systems using Artificial Neural Networks," SAE Technical Paper 2019-01-2292, 2019, https://doi.org/10.4271/2019-01-2292.
- R. Noorossana , Sam Davanloo Tajbakhsh , A. Saghaei An artificial neural network approach to multiple-response optimization Int J Adv Manuf Technol 2009
- S. Halbach , P. Sharer , S. Pagerit , C. Folkerts , A. Rousseau Model Architecture, Methods, and Interfaces for Efficient Math-Based Design and Simulation of Automotive Control Systems Argonne National Laboratory 2010
- Prof. P. J. Shayler , Dr. M. S. Goodman , Dr. T. Ma APPLICATIONS OF NEURAL NETWORKS IN AUTOMOTIVE ENGINE MANAGEMENT SYSTEMS
- P Brijesh , S Sreedhara Exhaust emissions and its control methods in compression ignition engines: A review International Journal of Automotive Technology 2012
- Zongyan Li , Matt Best Optimization of the Input Layer Structure for Feed-Forward Narx Neural Networks International Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering 2015
- Yazeed A. Al-Sbou1 ,* and Khaled M. Alawasa2 Nonlinear Autoregressive Recurrent Neural Network Model for Solar Radiation Prediction International Journal of Applied Engineering Research 2017
- Design Time Series NARX Feedback Neural Networks https://de.mathworks.com/help/deeplearning/ug/design time series narx feedback neural networks.html