Development, characterization, and modeling of an electronic particulate matter sensor for internal combustion engines

U.S. Federal regulations requiring on-board diagnostics of diesel particulate filters have created a demand for compact, inexpensive, fast, and accurate sensors for measuring the particulate matter (PM) content of diesel exhaust. An electronic sensor capable of measuring the carbonaceous fraction (soot) of PM has been developed at The University of Texas at Austin. The behavior and performance of this sensor was characterized in both an older style non-emission controlled diesel engine and a modern heavy-duty diesel certified in 2008 to meet current federal emissions standards. The ability of the sensor to detect particulates at the regulated level of 15 mg/bhp-hr downstream of a leaking particulate filter was demonstrated. Under optimal conditions, the sensor was shown to have a resolution of 0.003 mg/bhp-hr, or 0.005 mg/m3. The sensor operated by measuring the flux of charged particles, ions, and electrons to an electrode immersed in an exhaust gas flow. Two distinct modes of operation were demonstrated. In the first, the sensor detected particles carrying residual charge from the combustion process. In this mode, the sensor was shown to be relatively insensitive to particle morphology and to be sensitive to exhaust gas velocity. In the second, charge carriers (particles, electrons, and ions) were created in the strong electric field produced by a second electrode at high voltage. In this mode, the sensor was found to be relatively insensitive to exhaust gas velocity, but quite sensitive to the orientation of the sensor in the exhaust flow. The size and number density of the particles was found to have a strong influence on the sensor sensitivity: as number density increased with increasing load or decreasing EGR rate, so did sensor sensitivity. Thus, as changes in engine operating condition affect particle morphology, the behavior of the sensor changes. A numerical model of the discharge mechanism in the form of an atmospheric pressure glow discharge was implemented to model the charge creation and transport. The model accurately predicted the nanoamp-level electrode currents produced in a real sensor to within a half order of magnitude with no empirical fits. The model tended to over-predict the sensitivity of sensor output to applied voltage but matched the observed sensitivity within an order of magnitude. Due to the lack of modeling flow field effects it predicted a 250% increase in sensitivity for a gap width reduced by 50% where a comparison of real sensors showed a decrease in sensitivity of 25% with a 50% reduction in gap width.