Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health, R01-OH-003165, 1999 Nov; :1-55
The purpose of this study was to determine the efficacy of current methods and methods proposed by the P.I. to "troubleshoot" industrial exhaust ventilation systems using measured pressures and flows. Exhaust ventilation systems for contaminant control often experience obstructions and other deleterious alterations to individual ducts. Those alterations can reduce airflow to hoods, reducing their reliability in controlling exposures to hazardous airborne contaminant. Since alterations often are hidden from sight inside opaque ducts, it is necessary to find them using indirect means, primarily by interpreting changes to observed airflows and pressures. The most commonly used method ("SPHone") assumes an obstruction has occurred if the magnitude of the hood static pressures (SPH) has fallen. The second method ("SPHtwo") was an obvious variation of the one-sided SPHtwo method wherein an obstruction was expected with a sufficiently large increase or decrease in SPH. The "IVM" method compares the design value of SPH to the observed value. If SPH has fallen and the next downstream pressure (SPend) has increased, it assumes that an obstruction has occurred. However, early tests quickly demonstrated that this method was unworkable for a longinstalled system. For that reason, an idealized version ("idIVM") was employed in this study in which the "before" observed values were substituted for design pressures. The proposed methods each employ the ratios of pressures in an effort to normalize changes in pressures due to events external to the duct being tested. The "Xduct" method employs the ratio of the dissipated energy rate to the kinetic energy rate (which can be computed from measured pressures), the "SPratbr" method employs the ratio of SPH to SPend for each branch duct, and the "SPratmain" method employs the ratio of SPH to a common reference pressure in a main, SPref. Six working systems in contaminant producing processes were challenged with combinations of serendipitous and deliberately inserted obstructions. For each round of measurements on a given system, hood static pressures (SPH) and velocity pressures were measured for each branch, and the static pressure a few duct diameters upstream of a duct's terminus (SPend) was measured for each branch and submain duct. All measurements were taken with standard Pitot tubes and a calibrated digital manometer. Custom written data acquisition software captured each measurement value from the manometer. Each troubleshooting method's value was computed from the appropriate measured values and compared to a range of thresholds for the test cases. If the method's value exceeded the threshold and an obstruction had been in that duct, the method was considered to be true positive for that duct, round of measurements, and threshold. False negatives, true negatives, and false positive also were assigned. The method with the highest value of AROC was judged to be the most effective. In addition to the field data, data collected for other reasons in four ventilation laboratory studies was analyzed in the same manner. From all of the cases of a particular data set, the sensitivity and false positive rate for that threshold and method were computed. The sensitivity was plotted against the false positive rate for each threshold for each method, and the area under the resulting "receiver operating characteristic curve" (Aroc was computed for each method. In addition, the thresholds that would achieve 10% and 20% false positive rates were determined for each method and the accompanying sensitivities compared. The results showed that for the laboratory conditions Xduct and SPratbr had nearly perfect detection of obstructions with nearly zero false positives (Aroc=1). The values of Aroc for idIVM, SPHone, and SPHtwo were substantially inferior. At the specific thresholds that would achieve either 10% or 20% false positive rates, the sensitivities for the traditional methods were substantially inferior to the those achieved by the proposed methods. All methods performed substantially less well under field conditions than in the laboratory systems, probably due to the laboratories' excellent measurement conditions and to fewer misc1assifications because of the higher degree of certainty of conditions in the laboratory. However, the results from the field studies showed even greater margins of superiority for the proposed methods in both Aroc values and in sensitivities achieved at 10% and 20% false positive rates. The idIVM and SPHone methods performed dismally in all tests, failing to detect 40% of profound obstructions and doing much worse on lesser obstructions. The SPHref_br, SPHref_main, and Xduct methods were roughly equal. They each detected at least 90% of the "substantial" and "profound" obstructions but less than half of the very light and light obstructions. The SPHtwo method was inferior to the pressure ratio methods but far superior to the SPHone and idIVM methods. At the threshold for Xduct selected to achieve a 10% false positive rate, airflows would shift by 4% to 7% under most conditions of practical interest. The SPHref_br and SPHref_main methods should perform similarly at their recommended thresholds. As one part of the study, observed equivalent loss coefficients for two woodworking systems were compared to values predicted using Industrial Ventilation loss coefficients and calculation method. The results were dismayingly poor. Only 30% of predictions had errors of less than 25%. Those results should be treated with caution since only two very similar systems were tested. Finally, analysis of data from nearly a thousand sets of perpendicular Pitot traverses collected for this study demonstrated that a second Pitot traverse seldom typically was needed to improve accuracy only if the ratio of the mean velocity to the centerline velocity of the first traverse exceeded unity. The study provides substantial validation for new methods, and it provides substantial evidence that current methods are inadequate for general use. Using the proposed methods and suggested thresholds, ventilation practitioners should find that troubleshooting ventilation systems is more reliable than when traditional methods. With fewer substantial obstructions overlooked, desired system airflows to hoods should be easier to maintain, making hoods more reliable in their protection of workers. With fewer false positive indications, practitioners should find that far less of their time is wasted in fruitless searches for alterations that are actually elsewhere in the system. This should encourage them to monitor and troubleshoot systems frequently enough to reduce the long time lags that generally occur before a problem is discovered and fixed.
University of Washington, School of Public Health and Community Medicine, Department of Environmental Health, Room F226D, Box 357234, Seattle, WA 98195-7234
University of Washington, Department of Environmental Health, Seattle, WA