Relationship between Carbon Monoxide and Particulate Matter Levels across a Range of Engine Technologies 2012-01-1346

Relationships between diesel particulate matter (PM) mass and gaseous emissions mass produced by engines have been explored to determine whether any gaseous species may be used as surrogates to infer PM quantitatively. It was recognized that sulfur content of fuel might independently influence PM mass, since PM historically is composed of elemental carbon, organic carbon, sulfuric acid, ash and wear particles. Previous research has suggested that PM may be correlated with carbon monoxide (CO) for an engine that is exercised through a variety of speed and load cycles, but that the correlation does not extend to a group of engines. Large databases from the E-55/59 and Gasoline/Diesel PM Split programs were employed, along with the IBIS bus emissions database and several additional data sets for on- and off-road engines to examine possible relationships. Regressions using the E-55/59 database confirmed that Oxides of Nitrogen, Hydrocarbons, and Carbon Dioxide did not correlate satisfactorily with PM. Linear regression analysis of the E-55/59 diesel truck data, on all cycles yielded a weak correlation using a best fit line constrained to pass through the origin for PM (g/mile) versus CO (g/mile), with a resulting coefficient of determination (R₂) = 0.276. However, the relationship was lost (R₂ = -0.122) when the data were limited to trucks with pre-1994 engines at 56,000 lbs over one cycle, the Urban Dynamometer Driving Schedule. The Gasoline/Diesel PM Split study vehicles yielded a value of R₂ = 0.254, and the slope of the regression line differed substantially from the E-55/59 line. If the PM Split study data were used to predict PM emissions from CO for the E-55/59 trucks, the median under-prediction of PM would be 44%. Zero-intercept constrained linear regression of distance specific PM on CO for the IBIS buses yielded a value of R₂ = 0.313, but for the two-stroke Detroit Diesel-powered buses, it was found that the R₂ value was negative. Data for a fleet of buses in Florida showed good correlation between PM and CO (R₂ = 0.727), but this was because the fleet consisted of buses with and without diesel particulate filters (DPF). The disparity in PM and CO levels between these two groups enhanced the fit, and very poor correlation was found (R₂ = 0.024) when only the non-DPF buses were considered. A group of antique trucks showed high PM/CO ratios, with a negative value of R₂. Regression on a dataset (units of measure: g/bhp-hr) for seven engines, six of which were nonroad engines, yielded a value of R₂ of 0.110. Final examination of five datasets used in this paper show that the simple average of the PM/CO ratio for all the trucks in each set varied from 0.117 to 0.224. The highest and lowest PM/CO ratios encountered for all of the data reviewed were 1.69 and 0.004. The median was always below the average, and showed a variable relationship with the average. The standard deviation of the PM/CO ratios for the datasets varied from being half of the average to nearly twice the average. Overall conclusions are that the relationship between CO and PM was weak for the datasets explored, that the best fit from one dataset did not accurately characterize the data from other datasets, and that the units of measure used to establish the regression were important for cases where varying cycles, loads or test weights were used. There was one exception to the overall poor fit. A fleet of buses, some with high PM and CO, and some with low PM and CO (due to diesel particulate filter (DPF) aftertreatment) showed a reasonable fit. However, when the buses were separated into DPF and non-DPF equipped buses the fits were poor.