OCR Text |
Show produced substantially lower N O x emissions than the conventional burner without affecting organic H A P , C O or T H C emissions (Figure 5). Organic H A P emissions from the low-NOx burner lie well within the range seen with the conventional burner-low and near detection limits. PETROLEUM INDUSTRY FIELD DATA The data from the updated W S P A - C A T E F database were analyzed to determine if fuel gas type, criteria pollutant controls and design/operating parameters significantly influenced H AP emissions. Statistical tests were applied to screen for significant differences between source subcategories (t-test at 9 5 % confidence level using both normal and log-normal distributions), with engineering judgement then applied to confirm whether statistical differences were practically significant11. One limitation of the database is that all H A P s of concern were not measured in all tests. Therefore, the analysis was performed on a subset of the measured H A P s common to all tests so that a comparative analysis would be valid. Benzene, formaldehyde and the sum of benzene plus toluene plus xylene (BTX) were used as indicators of volatile organic HAPs. Benzene and formaldehyde are indicators of H A P s formed in different parts of the flame and thus m a y provide a clue to flame "failure modes." Total P A H (the sum of sixteen individual substances common to all tests), anthracene and benzo[a]pyrene were used to represent polycyclic organic matter (POM). Note, P O M is defined as a H A P in the C A A , and P A H are a subset of P O M . Naphthalene, which can be biased high due to sorbent contamination/decomposition issues, was included in the P A H total for this analysis to yield a conservative emission factor. Figure 6 compares B T X , formaldehyde and total P A H emission factors in pounds per million Btu (Ib/MMBtu) for boilers and process heaters. The central tendency of the data population (i.e., the 25th to 75th percentile) is represented in the figure by the vertical bar, and the range of data is expressed by the vertical line which shows the 10th to 90th percentile. Also shown in the figure is the range of detection limits achieved in the various tests. The height of the horizontal bars represents the 25th to 75th percentile of the detection limits. There are 36 records in the database for B T X for 32 different units tested. Comparing the B T X results for boilers and process heaters, there is no significant difference. The range of B T X emissions appears to be slightly broader for process heaters compared to boilers; however, a statistical t-test suggests the difference is not significant at the 95 percent confidence level. Thus, it is reasonable to combine the data for boilers and process heaters to produce a more robust emission factor. The resulting data for heaters and boilers combined are also shown in the figure. The range of formaldehyde emission factors shown in Figure 6 is considerably broader than B T X or P A H emissions. This is due in large part to the difficulty in making reliable formaldehyde measurements due to contamination. Formaldehyde is present at low concentrations in the ambient air as a natural product of transpiration and respiration. Many common materials such as carpeting and wood paneling are known to emit small amounts of formaldehyde. Thus, it is ubiquitous in many environments and contamination is sometimes unavoidable even with the best field and laboratory quality assurance practices. Nevertheless, formaldehyde is an intermediate 5 |