• Home
• About Us
  • Welcome Message
  • Mission
  • Facts/History
  • Diversity
  • Honors and Awards
  • Recent Graduates
  • Statistics Donations
  • Travel Information
  • Contact Us
• People
  • Faculty
  • Staff
  • Graduate Students
  • Certificate Students
  • Student Organizations
  • Alums
• Academic Programs
  • Courses
  • Undergraduate
  • Graduate
  • Certificate and Non-Degree
• Research
  • Specialty Areas
  • Centers and Groups
  • Technical Reports
  • Dissertations
  • Research Profiles
  • Research Videos
• News
  • Newsletters
  • Faculty Positions
  • Announcements
  • Departmental News
• Calendar
  • Seminars and Events
  • Academic Calendar
• External Relations
  • K-12 Outreach
  • StatCom
  • StatCom P-12 Outreach
• Resources
  • Business Office
  • Computing Facilities
  • Jobs
  • Statistics Graduate Student Resumes
  • Tutor List
  • Grader Job Application
  • Internal (Stat Central)
• Statistical Consulting Service

Research Profile

Kristofer Jennings - Modeling exhaust emissions in diesel engines

Written by: Andrea Rau, Ph.D. candidate in Statistics

In recent years, as concerns about global warming have intensified and gasoline prices have escalated, automobile manufacturers and owners have paid increasing attention to the fuel economy and gas emissions of their vehicles.  Although diesel engines can be more efficient, and have greater fuel economy and power output, than gasoline engines, they also tend to produce more exhaust emissions. As a result, in 2007 the U.S. Environmental Protection Agency (EPA) developed a new set of restrictions for emissions from diesel engines in order to make them more environmentally friendly.

Because failure to respect these regulations can result in costly fines, it is particularly important for diesel engine manufacturers to be able to accurately monitor the gas emissions of their engines.  To this end, Professor Kristofer Jennings, in collaboration with colleagues Drs. Peter Meckl and Galen King in the School of Mechanical Engineering and The Ray W. Herrick Labs with support from Cummins Engines, is working to develop a protocol for a model to chart emissions in diesel engines.

Generally, emissions can be roughly split into two groups: particulates and gases.  For both types, it is both expensive and difficult to directly obtain accurate measurements, particularly as they may change with how hard or fast an engine is being run. Using about two to three dozen sensors set in different parts of the engine, researchers must thus be able to indirectly decide whether or not it is running in a faulty state, which subsequently may account for excess emissions.  To do so, the researchers often start with an engine in a healthy state, and induce faults – such as blockages or simulating sticky valves – and then monitor the sensors to attempt to distinguish the fault.

Most groups working to accurately model engine emissions concentrate on fixed engine emissions, where torque and speed remain constant; the work being done by Professor Jennings and his colleagues is unique in its focus on the dynamic modeling of engines running through various speeds and load states.  Models for engines running under different conditions can be very different, and for this reason, Professor Jennings is working to develop a model-building protocol.  These models generally include several incidental variables, like temperature, pressure, and electric current, as well as some ambient information, and two non-stochastic variables: speed and torque, or how much load the engine is bearing. Based on this information, the researchers can better understand whether a particular engine will produce excess emissions and why.

Professor Jennings has been involved with this project over the past five years, and he cited "environmental stewardship" as being one of his favorite parts of the project.  "I definitely like the environmental aspect of it," said Jennings.  In coming years, the group hopes to simulate more complicated engine faults and to improve modeling techniques by using information theoretic indices to select important variables and by using singular spectrum analysis.  In addition, they hope to implement bootstrap methods in complex systems to put bounds on the distribution of signals, to better understand where the health of the engines is expected to be.

Professor Jennings obtained his B.A. in Mathematics from Oberlin College in 1996 before moving on to Stanford University to earn his M.S. and Ph.D. in Statistics.  He joined the faculty of the Purdue Statistics department in 2002, and is currently an Assistant Professor. He has taught graduate courses in applied regression analysis, sampling and survey techniques, and experimental design.  His research interests include bootstrap and resampling methods, applied statistics, and astrostatistics.  For more information about Professor Jennings, please visit his homepage.

References:

A. A. Joshi, S. M. James, P. H. Meckl, G. B. King, and K. Jennings. (2007) "Information-Theoretic Feature Selection for Classification," Proceedings of the 2007 American Control Conference 2000-2005. [pdf]

A. A. Joshi, P. B. Deignan, P. H. Meckl, G. B. King, and K. Jennings. (2005) "Information theoretic fault detection." Proceedings of the American Control Conference 1642-1647. [pdf]

A. A. Joshi, P. H. Meckl, G. B. King, and K. Jennings. (2008) "Data-dimensionality reduction using information-theoretic stepwise feature selector." To appear in ASME Journal of Dynamic Systems, Measurement, and Control. [pdf]

A. A. Joshi, S. M. James, P. H. Meckl, G. B. King, and K. Jennings. (2008) "Assessment of charge-air cooler health in diesel engines using nonlinear time series analysis of intake manifold temperature." To appear in ASME Journal of Dynamic Systems, Measurement, and Control. [pdf]

October 2008