Research Papers: Fuel Combustion

Using Machine Learning to Analyze Factors Determining Cycle-to-Cycle Variation in a Spark-Ignited Gasoline Engine

[+] Author and Article Information
Janardhan Kodavasal, Ahmed Abdul Moiz, Muhsin Ameen, Sibendu Som

Argonne National Laboratory,
Energy Systems Division,
Argonne, IL 60439

Contributed by the Internal Combustion Engine Division of ASME for publication in the JOURNAL OF ENERGY RESOURCES TECHNOLOGY. Manuscript received February 15, 2018; final manuscript received March 13, 2018; published online May 15, 2018. Editor: Hameed Metghalchi.

J. Energy Resour. Technol 140(10), 102204 (May 15, 2018) (9 pages) Paper No: JERT-18-1132; doi: 10.1115/1.4040062 History: Received February 15, 2018; Revised March 13, 2018

In this work, we have applied a machine learning (ML) technique to provide insights into the causes of cycle-to-cycle variation (CCV) in a gasoline spark-ignited (SI) engine. The analysis was performed on a set of large eddy simulation (LES) calculations of a single cylinder of a four-cylinder port-fueled SI engine. The operating condition was stoichiometric, without significant knock, at a load of 16 bar brake mean effective pressure (BMEP), at an engine speed of 2500 rpm. A total of 123 cycles was simulated. Of these, 49 were run in sequence, while 74 were run in parallel. For the parallel approach, each cycle is initialized with its own synthetic turbulent field to generate CCV, as a part of another work performed by us. In this work, we used 3D information from all 123 cycles to compute flame topology and pre-ignition flow-field metrics. We then evaluated correlations between these metrics and peak cylinder pressure (PCP) employing an ML technique called random forest. The computed metrics form the inputs to the random forest model, and PCP is the output. This model captures the effect of all inputs, as well as interactions between them owing to its decision-tree structure. The goal of this work is to demonstrate (as a first step) that ML models can implicitly learn complex relationships between the pre-ignition flow-fields, the flame shapes, and the eventual outcome of the cycle (whether a cycle will be a high or a low cycle).

Copyright © 2018 by ASME
Your Session has timed out. Please sign back in to continue.


Patterson, D. J. , 1966, “ Cylinder Pressure Variations, a Fundamental Combustion Problem,” SAE Paper No. 660129.
Ozdor, N. , Dulger, E. , and Sher, E. , 1994, “ Cyclic Variability in Spark-Ignition Engines. A Literature Survey,” SAE Paper No. 940987.
Srinivasan, K. K. , Krishnan, S. R. , and Qi, Y. , 2013, “ Cyclic Combustion Variations in Dual Fuel Partially Premized Pilot-Ignited Natural Gas Engines,” ASME J. Energy Resour. Technol., 136(1), p. 012003. [CrossRef]
Zhou, L. , Shao, A. , Hua, J. , Wei, H. , and Feng, D. , 2018, “ Effect of Retarded Injection Timing on Knock Resistance and Cycle to Cycle Variation in GDI Engine,” ASME J. Energy Resour. Technol., 140(7), p. 072202.
Scarcelli, R. , Sevik, J. , Wallner, T. , Richards, K. , Pomraning, E. , and Senecal, P. K. , 2015, “ Capturing Cyclic Variability in EGR Dilute SI Combustion Using Multi-Cycle RANS,” ASME Paper No. ICEF2015-1045.
Scarcelli, R. , Richards, K. , Pomraning, E. , Senecal, P. K. , Wallner, T. , and Sevik, J. , 2016, “ Cycle-to-Cycle Variations in Multi-Cycle Engine RANS Simulations,” SAE Paper No. 2016-01-0593.
Moureau, V. , Barton, I. , Angelberger, C. , and Poinsot, T. , 2004, “ Toward Large Eddy Simulation in Internal-Combustion Engines: Simulation of a Compressed Tumble Flow,” SAE Paper No. 2004-01-1995.
Vermorel, O. , Richard, S. , Colin, O. , Angelberger, C. , Benkenida, A. , and Veynante, D. , 2009, “ Toward the Understanding of Cyclic Variability in a Spark Ignited Engine Using Multi-Cycle LES,” Combust. Flame, 156(8), pp. 1525–1541. [CrossRef]
Fontanesi, S. , Paltrinieri, S. , Tiberi, A. , and D'Adamo, A. , 2013, “ LES Multi-Cycle Analysis of a High Performance GDI Engine,” SAE Paper No. 2013-01-1080.
Fontanesi, S. , d'Adamo, A. , and Rutland, C. J. , 2015, “ Large-Eddy Simulation Analysis of Spark Configuration Effect on Cycle-to-Cycle Variability of Combustion and Knock,” Int. J. Engine Res., 16(3), pp. 403–418. [CrossRef]
d'Adamo, A. , Breda, S. , Fontanesi, S. , and Cantore, G. , 2015, “ LES Modeling of Spark-Ignition Cycle-to-Cycle Variability on a Highly Downsized DISI Engine,” SAE Int. J. Fuels Lubr., 8(5), pp. 2029–2041.
Enaux, B. , Granet, V. , Vermorel, O. , Lacour, C. , Pera, C. , Angelberger, C. , and Poinsot, T. , 2011, “ LES Study of Cycle-to-Cycle Variations in a Spark Ignition Engine,” Proc. Combust. Inst., 33(2), pp. 3115–3122. [CrossRef]
Koch, J. , Schmitt, M. , Wright, Y. M. , Steurs, K. , and Boulouchos, K. , 2014, “ LES Multi-Cycle Analysis of the Combustion Process in a Small SI Engine,” SAE Int. J. Fuels Lubr., 7(1), pp. 269–285.
Truffin, K. , Angelberger, C. , Richard, S. , and Pera, C. , 2015, “ Using Large-Eddy Simulation and Multivariate Analysis to Understand the Sources of Combustion Cyclic Variability in a Spark-Ignition Engine,” Combust. Flame, 162(12), pp. 4371–4390. [CrossRef]
Tatschl, R. , Bogensperger, M. , Pavlovic, Z. , Priesching, P. , Schuemie, H. , Vitek, O. , and Macek, J. , 2013, “ LES Simulation of Flame Propagation in a Direct-Injection SI-Engine to Identify the Causes of Cycle-to-Cycle Combustion Variations,” SAE Paper No. 2013-01-1084.
Granet, V. , Vermorel, O. , Lacour, C. , Enaux, B. , Dugué, V. , and Poinsot, T. , 2012, “ Large-Eddy Simulation and Experimental Study of Cycle-to-Cycle Variations of Stable and Unstable Operating Points in a Spark Ignition Engine,” Combust. Flame, 159(4), pp. 1562–1575. [CrossRef]
Ameen, M. M. , Yang, X. , Kuo, T. W. , Xue, Q. , and Som, S. , 2015, “ LES for Simulating the Gas Exchange Process in a Spark Ignition Engine,” ASME Paper No. ICEF2015-1002.
Zhao, L. , Moiz, A. A. , Som, S. , Fogla, N. , Bybee, M. , Wahiduzzaman, S. , and Mirzaeian, M. , 2017, “ Examining the Role of Flame Topologies and In-Cylinder Flow Fields on Cyclic Variability in Spark-Ignited Engines Using Large Eddy Simulation,” Int. J. Engine Res., in press.
Zhao, L. , Moiz, A. A. , Som, S. , Fogla, N. , Bybee, M. , Wahiduzzaman, S. , Mirzaeian, M. , Millo, F. , and Kodavasal, J. , 2017, “ Multi-Cycle Large Eddy Simulation to Capture Cycle-to-Cycle Variation (CCV) in Spark-Ignited (SI) Engines,” Tenth U.S. National Combustion Meeting, College Park, MD, Apr. 23–26. https://iris.polito.it/handle/11583/2673451?mode=full.4423#.WvFOErlDGYk
Ameen, M. M. , Yang, X. , Kuo, T. W. , and Som, S. , 2017, “ Parallel Methodology to Capture Cyclic Variability in Motored Engines,” Int. J. Engine Res., 18(4), pp. 366–377. [CrossRef]
Ho, T. K. , 1995, “ Random Decision Forests,” Third International Conference on Document Analysis and Recognition (ICDAR), Montreal, QC, Canada, Aug. 14–15, p. 278. https://dl.acm.org/citation.cfm?id=844681
Kodavasal, J. , Pei, Y. , Harms, K. , Ciatti, S. , Wagner, A. , Senecal, P. K. , García, M. , and Som, S. , 2016, “ Global Sensitivity Analysis of a Gasoline Compression Ignition Engine Simulation With Multiple Targets on an IBM Blue Gene/Q Supercomputer,” SAE Paper No. 2016-01-0602.
Kodavasal, J. , Van Dam, N. , Pei, Y. , Harms, K. , Maheshwari, K. , Wagner, A. , García, M. , Ciatti, S. , Senecal, P. K. , and Som, S. , 2016, “ Sensitivity Analysis on Key CFD Model Inputs for Gasoline Compression Ignition on an IBM Blue Gene/Q Supercomputer,” THIESEL Conference on Thermo- and Fluid Dynamic Processes in Direct Injection Engines, Valencia, Sept. 13–16.
Kodavasal, J. , Kolodziej, C. P. , Ciatti, S. A. , and Som, S. , 2015, “ Computational Fluid Dynamics Simulation of Gasoline Compression Ignition,” ASME J. Energy Resour. Technol., 137(3), p. 032212. [CrossRef]
Kodavasal, J. , Kolodziej, C. P. , Ciatti, S. A. , and Som, S. , 2017, “ Effects of Injection Parameters, Boost, and Swirl Ratio on Gasoline Compression Ignition Operation at Idle and Low-Load Conditions,” Int. J. Engine Res., 18(8), pp. 824–836. [CrossRef]
Kolodziej, C. , Kodavasal, J. , Ciatti, S. , Som, S. , Shidore, N. , and Delhom, J. , 2015, “ Achieving Stable Engine Operation of Gasoline Compression Ignition Using 87 AKI Gasoline Down to Idle,” SAE Paper No. 2015-01-0832.
Kodavasal, J. , and Som, S. , 2018, “ Gasoline Compression Ignition—A Simulation-Based Perspective,” Advances in Internal Combustion Engine Research. Energy, Environment, and Sustainability, D. Srivastava , A. Agarwal , A. Datta , and R . Maurya , eds., Springer, Singapore.
Kodavasal, J. , Lavoie, G. A. , Assanis, D. N. , and Martz, J. B. , 2014, “ The Effects of Thermal and Compositional Stratification on the Ignition and Duration of Homogeneous Charge Compression Ignition Combustion,” Combust. Flame, 162(2), pp. 451–461. [CrossRef]
Kodavasal, J. , Lavoie, G. A. , Assanis, D. N. , and Martz, J. B. , 2015, “ The Effect of Diluent Composition on Homogeneous Charge Compression Ignition Auto-Ignition and Combustion Duration,” Proc. Combust. Inst., 35(3), pp. 3019–3026. [CrossRef]
Kodavasal, J. , Lavoie, G. A. , Assanis, D. N. , and Martz, J. B. , 2015, “ Reaction-Space Analysis of Homogeneous Charge Compression Ignition Combustion With Varying Levels of Fuel Stratification Under Positive and Negative Valve Overlap Conditions,” Int. J. Engine Res., 17(7), pp. 776–794. [CrossRef]
Kodavasal, J. , 2013, “ Effect of Charge Preparation Strategy on HCCI Combustion,” Ph.D. thesis, University of Michigan, Ann Arbor, MI. https://deepblue.lib.umich.edu/handle/2027.42/99766
Kodavasal, J. , Keum, S. H. , and Babajimopoulos, A. , 2011, “ An Extended Multi-Zone Combustion Model for PCI Simulation,” Combust. Theory Modell., 15(6), pp. 893–910. [CrossRef]
Kodavasal, J. , McNenly, M. J. , Babajimopoulos, A. , Aceves, S. M. , Assanis, D. N. , Havstad, M. A. , and Flowers, D. L. , 2013, “ An Accelerated Multi-Zone Model for Engine Cycle Simulation of Homogeneous Charge Compression Ignition Combustion,” Int. J. Engine Res., 14(5), pp. 416–433. [CrossRef]
Kodavasal, J. , Harms, K. , Srivastava, P. , Som, S. , Quan, S. , Richards, K. , and García, M. , 2016, “ Development of a Stiffness-Based Chemistry Load Balancing Scheme, and Optimization of Input/Output and Communication, to Enable Massively Parallel High-Fidelity Internal Combustion Engine Simulations,” ASME J. Energy Resour. Technol., 138(5), p. 052203. [CrossRef]
Sorgun, M. , Ozbayoglu, A. M. , and Ozbayoglu, M. E. , 2014, “ Support Vector Regression and Computational Fluid Dynamics Modeling of Newtonian and Non-Newtonian Fluids in Annulus With Pipe Rotation,” ASME J. Energy Resour. Technol., 137(3), p. 032901. [CrossRef]
Moiz, A. A. , Pal, P. , Probst, D. , Pei, Y. , Zhang, Y. , Som, S. , and Kodavasal, J. , 2018, “ A Machine Learning-Genetic Algorithm Approach for Rapid Virtual Optimization Using High-Performance Computing,” SAE Paper No. 2018-01-0190.
Rostami, H. , and Manshad, A. K. , 2014, “ A New Support Vector Machine and Artificial Neural Networks for Prediction of Stuck Pipe in Drilling of Oil Fields,” ASME J. Energy Resour. Technol., 136(2), p. 024502. [CrossRef]
El-Emam, R. S. , and Dincer, I. , 2016, “ Assessment and Evolutionary Based Multi-Objective Optimization of a Novel Renewable-Based Polygeneration Energy System,” ASME J. Energy Resour. Technol., 139(1), p. 012003. [CrossRef]
Van, S. L. , and Chon, B. H. , 2017, “ Effective Prediction and Management of a CO2 Flooding Process for Enhancing Oil Recovery Using Artificial Neural Networks,” ASME J. Energy Resour. Technol., 140(3), p. 032906. [CrossRef]
Richards, K. J. , Senecal, P. K. , and Pomraning, E. , 2017, “CONVERGE v2.3 Manual,” Convergent Science, Madison, WI.
Reitz, R. , and Diwakar, R. , 1987, “ Structure of High-Pressure Fuel Sprays,” SAE Paper No. 870598.
Senecal, P. , Richards, K. , Pomraning, E. , Yang, T. , Dai, M. Z. , McDavid, R. M. , Patterson, M. A. , Hou, S. , and Shethaji, T. , 2007, “ A New Parallel Cut-Cell Cartesian CFD Code for Rapid Grid Generation Applied to In-Cylinder Diesel Engine Simulations,” SAE Paper No. 2007-01-0159.
Pomraning, E. , and Rutland, C. J. , 2002, “ Dynamic One-Equation Nonviscosity Large-Eddy Simulation Model,” AIAA J., 40(4), pp. 689–701. [CrossRef]
Williams, F. , 1985, “ Turbulent Combustion,” The Mathematics of Combustion, J. D. Buckmaster , ed., Society for Industrial and Applied Mathematics, Philadelphia, PA, pp. 197–1318.
Peters, N. , 2000, Turbulent Combustion, Cambridge University Press, Cambridge, UK.
Pitsch, H. , 2002, “ A G-Equation Formulation for Large-Eddy Simulation of Premixed Turbulent Combustion,” Annual Research Briefs, Center for Turbulence Research, Stanford, CA.
Balaprakash, P. , Tiwari, A. , Wild, S. M. , Carrington, L. , and Hovland, P. D. , 2016, “ AutoMOMML: Automatic Multi-Objective Modeling With Machine Learning,” ISC High Performance (LNCS), Frankfurt, Germany, June 19–23, pp. 219–239.
Berthold, M. R. , Cebron, N. , Dill, F. , Gabriel, T. R. , Kötter, T. , Meinl, T. , Ohl, P. , Sieb, C. , Thiel, K. , and Wiswedel, B. , 2007, “ The Konstanz Information Miner,” Studies in Classification, Data Analysis, and Knowledge Organization, Springer, Berlin.


Grahic Jump Location
Fig. 2

Snapshot showing axes orientation with respect to the CFD model

Grahic Jump Location
Fig. 1

Experimental cycles (left), simulated consecutive cycles (middle), simulated parallel cycles (right)

Grahic Jump Location
Fig. 6

Sample decision tree showing the partitioning of the data based on a random subset of four features of the ten input features (4 = √10 rounded up) representing flame shape and pre-ignition velocities, using an if/else structure; the random forest is an ensemble of multiple decision trees

Grahic Jump Location
Fig. 5

Pearson correlation between ten metrics chosen to represent flame topologies, pre-ignition velocity flow-fields, and PCP

Grahic Jump Location
Fig. 4

A single pair of one low cycle and one high cycle at a common mass percent burned of 3% showing similar physical behavior when simulated using the parallel and consecutive approaches (similar to the behavior observed with them at the same CA)

Grahic Jump Location
Fig. 3

A single pair of one low cycle and one high cycle at 722 deg. CA showing similar physical behavior when simulated using the parallel and consecutive approaches.

Grahic Jump Location
Fig. 7

KNIME workflow for building a random forest model

Grahic Jump Location
Fig. 11

Predictions on the 15% test set considering two pre-ignition flow metrics—U5.5 mm, and W5.5 mm

Grahic Jump Location
Fig. 9

Linear fit of PCP to high-ranking inputs, with scatter around the fit shown

Grahic Jump Location
Fig. 10

Predictions on the 15% test set considering four metrics—sphericity, U5.5 mm, W5.5 mm, and COMoX

Grahic Jump Location
Fig. 8

Feature (or input) importance for predicting PCP using random forest



Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In