Empirical Validation of Turbine Engine Exhaust Measurements

Abstract

The reliability and validity of gas turbine engine exhaust measurements are of concern to all who measure these effluents. In the past, one measure of reliability has been a carbon balance between the fuel used and the C, CO, hydrocarbons, and CO₂ emitted which requires that a F/A ratio be known accurately and that the analyzed sample be representative. In addition to this carbon balance, we have considered the relationships between the concentrations of the several component species. For example, the plot of volume percent of CO₂ versus the volume percent of O₂, both properly corrected to complete combustion, is a straight line. The intercept on the Y-axis (zero oxygen) is a function of the H to C mole ratio in the fuel. This relation has been theoretically established and also established by empirical calculations. For all valid measurements, all points corresponding to CO₂ and O₂ analyses fall on the straight line. In addition, smooth functional relationships exist between other exhaust gas constituents as well, although the form is not a simply derivable linear relationship like that for O₂-CO₂. We have found it useful to employ the following correlations: CO₂ vs. log NO_x, log CO vs. log THC, log CO vs. CO₂, and log THC vs. CO₂- These relationships serve to indicate the probability of reliable data and point out sampling problems and instrument problems such as drift. Subtle changes in engine operation such as opening of the accessory bleed ports in the compressor discharge of the engine can be detected.

The use of a combination of these curves can serve to indicate which measurement is in error. If the O₂-CO₂ relation is linear and the CO-CO₂ relation is smooth, then problems with the total hydrocarbon analyses are indicated when the CO-THC relation is erratic. This would be confirmed by a similar CO₂-THC plot. No useful smooth relationship between smoke density measurements and the other constituents has been found to date. This may indicate poor data, lack of sufficient data, or possibly no correlation.