Calibration of catalyst temperature in automotive engines over coldstart operation in the presence of different random noises and uncertainty: Implementation of generalized Gaussian process regression machine

Abstract

The main scope of the current study is to develop a systematic stochastic model to capture the undesired uncertainty and random noises on the key parameters affecting the catalyst temperature over the coldstart operation of automotive engine systems. In the recent years, a number of articles have been published which aim at the modeling and analysis of automotive engines’ behavior during coldstart operations by using regression modeling methods. Regarding highly nonlinear and uncertain nature of the coldstart operation, calibration of the engine system’s variables, for instance the catalyst temperature, is deemed to be an intricate task, and it is unlikely to develop an exact physics-based nonlinear model. This encourages automotive engineers to take advantage of knowledge-based modeling tools and regression approaches. However, there exist rare reports which propose an efficient tool for coping with the uncertainty associated with the collected database. Here, the authors introduce a random noise to experimentally derived data and simulate an uncertain database as a representative of the engine system’s behavior over coldstart operations. Then, by using a Gaussian process regression machine (GPRM), a reliable model is used for the sake of analysis of the engine’s behavior. The simulation results attest the efficacy of GPRM for the considered case study. The research outcomes confirm that it is possible to develop a practical calibration tool which can be reliably used for modeling the catalyst temperature.

Keywords

automotive engine / calibration / coldstart operation / Gaussian process regression machine (GPRM) / uncertainty and random noises

References

[1] Wouk V. Hybrid Electric Vehicles. Scientific American, 1997, 277(2): 70–74

[2] Vajedi M, Chehresaz M, Azad N L. Intelligent power management of plug-in hybrid electric vehicles, part I: Real-time optimum SOC trajectory builder. International Journal of Electric and Hybrid Vehicles, 2014, 6(1): 46–67

[3] Leonhard R. Reducing CO₂ emissions with optimized internal-combustion engines. 60th Automotive Press Briefing, Boxberg, 2011

[4] Sanketi P R. Coldstart modeling and optimal control design for automotive SI engines. Dissertation for the Doctoral Degree. Berkeley: University of California, 2009

[5] Azad N L, Sanketi P R, Hedrick J K. Determining model accuracy requirements for automotive engine coldstart hydrocarbon emissions control. Journal of Dynamic Systems, Measurement, and Control, 2012, 134(5): 051002.1–051002.11

[6] Mozaffari A, Azad N L. Optimally pruned extreme learning machine with ensemble of regularization techniques and negative correlation penalty applied to automotive engine coldstart hydrocarbon emission identification. Neurocomputing, 2014, 131: 143–156

[7] Zavala J C, Sanketi P R, Wilcutts M,  Simplified models of engine HC emissions, exhaust temperature and catalyst temperature for automotive coldstart. In: Proceedings of 5th IFAC Symposium on Advances in Automotive Control. Monterey Coast, 2007

[8] Zavala J C. Engine modeling and control for minimization of hydrocarbon coldstart emissions in SI engine. Dissertation for the Doctoral Degree. Berkeley: University of California, 2007

[9] Bede B. Mathematics of Fuzzy Sets and Fuzzy Logic. Berlin: Springer, 2013

[10] Franklin J. The elements of statistical learning: Data mining, inference, and prediction. The Mathematical Intelligencer, 2009, 27(2): 83–85

[11] Güvenç B A, Güvenç L, Karaman S. Robust MIMO disturbance observer analysis and design with application to active car steering. International Journal of Robust and Nonlinear Control, 2010, 20: 873–891

[12] Gibson A, Kolmonovsky I, Hrovat D. Application of disturbance observers to automotive engine idle speed control for fuel economy improvement. In: American Control Conference. Minneapolis: IEEE, 2006

[13] Turner J. Automotive Sensors. New York: Momentum Press, 2009

[14] Caponetto R, Fortuna L, Fazzino S,  Chaotic sequences to improve the performance of evolutionary algorithms. IEEE Transactions on Evolutionary Computation, 2003, 7(3): 289–304

[15] Ebden M. Gaussian Process for Regression: A Quick Introduction. Technical Report, University of Oxford, 2008

[16] Mozaffari A, Azad N L. Coupling Gaussian generalized regression neural network and mutable smart bee algorithm to analyze the characteristics of automotive engine coldstart hydrocarbon emission. Journal of Experimental & Theoretical Artificial Intelligence, 2015, 27(3): 253–272

Acknowledgment

The authors would like to thank their colleagues at University of California (UC), Berkeley, Professor Karl Hedrick and Dr. Pannag Sanketi, for their help in conducting experiments and collecting the coldstart data.

RIGHTS & PERMISSIONS

2014 Higher Education Press and Springer-Verlag Berlin Heidelberg