Full-scale, 8-branch duct system in our laboratory

Future: working systems in factories

Proposed method produced less than 4% error on any branch airflow (see adjustable dampers).

Next stage of the study will consider more sets of conditions.

Full-scale, 5-branch system in our laboratory as well as 5 workings systems in factories and a local community college

Future: more working systems

Traditional methods work poorly.
Proposed methods work very well under most conditions.
Two perpendicular Pitot traverses can produce accurate measurements even under measurement poor conditions (see Pitot tube abstract).
Estimating airflow using pipe factors creates substantial error (see pipe factor abstract).

Next stage of the study is to apply methods to commissioning new systems.

Substantial differences on mannikins when facing downstream with source in hands (see lapel abstract).at waist level.

Preliminary results on human subjects suggest substantial deviations exist when facing a downstream source while doing various tasks (see lapel abstract).

Next stage: consider additional variables such as crossdraft velocity, different tasks, body posture, use of glasses, respirators, hard hats, etc.

 Two perpendicular Pitot traverses of velocity pressures can reliably determine mean duct velocity, but traverses are discouragingly time-consuming. It would be convenient if the shortcut method of multiplying the centerline velocity by the estimated "pipe factor" (ratio of the average to centerline velocity) was accurate.

This study compared estimates of mean duct velocity from three shortcut approaches to values obtained from two perpendicular traverses. In the "pipe factor=0.9 method" the pipe factor was given the commonly-used value of 0.9. In the "empirical pipe factor method," the pipe factor was determined from the first round of data. The "single traverse method" omitted the second traverse.

Two perpendicular ten-point traverses were taken repeatedly over several months on a total of 35 branch ducts on four working industrial exhaust ventilation systems using a calibrated digital manometer connected to a laptop computer. Duct velocities ranged from 1625 ft/min to 6477 ft/min.

The "pipe factor=0.9 method" produced deviations from "true" values ranging from minus 25% to plus 10%, with a median deviation of minus 4% and a standard deviation of nearly 7%. An attempt to improve the method by using the geometric mean of two measurements of the centerline value instead of a single value made no discernible difference.

The empirical pipe factor was 0.88 (R-sq=0.98), which was too close to 0.9 to change results significantly.

Omitting the second traverse rarely produced errors above 3% if the pipe factor was between 0.8 and 1.0. When the pipe factor exceeded unity, the error exceeded 3% in about 11% of cases. A procedure that required a second traverse if the pipe factor exceeded unity deviated from two-traverse results by more than 3% in less than 4% of cases.

Industrial Ventilation recommends that Pitot traverses be taken at least 7 duct diameter's length (i.e., distance=7D) from obstructions, elbows, junction fittings, and other disturbances to flows. That requirement is often difficult to meet in industrial exhaust ventilation systems, so the practitioner often must use the best location available. The inaccuracy due to use of "poor" measurement locations is not known. This study determined the deviations between Pitot traverses taken under ideal conditions to those taken at various distances downstream from various commonplace disturbances.

Two perpendicular ten-point, log-linear velocity pressure traverses were taken at various distances downstream of tested upstream conditions. The upstream conditions included a plain duct opening, a junction fitting, a single 90 elbow, and two elbows rotated 90 from each other into two orthogonal planes. The airflows determined from those values were compared to the values measured more than 40D downstream of the same obstructions under "ideal" conditions. The "ideal" measurements were taken on 3 traverse diameters in the same plane separated by 120 in honed drawn-over-mandrel tubing. In all cases, the Pitot tubes were held in place by devices that effectively eliminated alignment errors and insertion depth errors. Duct velocities ranged from 1500 ft/min to 4500 ft/min.

The results were surprisingly good if one employed two perpendicular traverses. When the averages of two perpendicular traverses was taken, deviations from "ideal" value were 6% or less even for traverses taken as close as 2D distance from the upstream disturbances. At 3D distance, deviations seldom exceeded 5%. With single diameter traverses, errors seldom exceeded 5% at 6D or more downstream from the disturbance. Interestingly, percentage deviations were about the same at high and low velocities.

This study demonstrated that two perpendicular Pitot traverses can be taken as close as 3D from these disturbances with acceptable (< 5%) deviations from measurements taken under ideal conditions