Cleanroom Airflow Measurement Certainty

This article examines the benefit of statistical forensics on cleanroom certification reports. Two certification efforts were conducted in successive years on the same client facility performed by two certification efforts. Airflow CFM was examined for ceiling-mounted terminal hepas whereby both total capture flow hood readings and air velocity readings were taken.

We were contacted by a medical device client to assume the cleanroom certification responsibilities previously being conducted. The client’s cleanroom facility consists of 98 ceiling hepa filters of type A, disposable, ducted supply, serviced by two rooftop air handling units. The facility airflow was also supplemented by a handful of recirculating fan filters, Type B, to achieve design air exchanges. This discussion concerns only the Type A hepa filters.

As is often the case, some of the hepa filters were inaccessible by a total capture flow hood due to process equipment obstruction. Of the 98 Type-A hepa filters, 88 were flow hood accessible. The client presented our testing group with the previous certification test report which included reflected ceiling drawings, hepa filter locations, hepa filter identification numbers, and the most recent certification data. In the interest of consistency, our group would plan to adhere to the hepa filter identification scheme.

The owner provided airflow data indicated with a footnote in which 10 of the 98 filters were measured with a velocity instrument and, consequently, had their volumetric flow calculated (Ak method) rather than being directly captured with the flow hood. The measured velocities were not explicitly stated, nor were the value(s) for the corresponding area, Ak, given.

This facility was particularly well-designed, balanced, and maintained with regard to airflow. The client maintenance personnel and primary mechanical contractor were on-site during the off-hours testing effort to assist in remedying any shortcomings in real time. The hepa filters all had very healthy airflow with an average of 630 CFM and a relative standard deviation of just over 15%, with no anemic hepa filters.

After we completed the airflow measurement effort consisting of direct capture, backpressure-compensated flow hood measurements, and corresponding airflow velocities with a multi-point pitot array grid at 6” from the filter face, we compared the ‘as-found’ readings with the results provided by the owner. The values were very similar, with an average CFM of 651 (in contrast with the 630 CFM also backpressure-compensated). The difference in the average velocities and flows was less than 4%.

We used a multipoint pitot array grid on the same ten filters that had been used because of the equipment obstructions. The Ak value was determined as the average of 24 accessible filters with contemporaneous backpressure-corrected flow hood readings and multipoint pitot array velocity readings. Equations 1 and 2 below describe the proper application of the velocity method for determining airflow CFM, which we utilized:

EQ.1: Qmeas = Vmeas * Ak EQ.2: Qmeas = Ameas * Vmeas * Kv

Where: Qmeas – measured volumetric airflow with capture flow hood (ft3/min) Vmeas – Air velocity measured with multipoint pitot grid at a prescribed distance (ft/min) Ameas – Area of exposed filter media measured with a tape measure (ft2 ) Ak – the corrected exposed filter area which reconciles Qmeas with Vmeas. Kv – the velocity correction, which reconciles Qmeas with Ameas. Ak – can be expressed as the product of Vmeas and Kv.

In evaluating the data, a particular phenomenon in the previous results was noted whereby the ten obstructed hepa filters appeared to have higher CFM readings than the other 88 filters. The previous obstructed filter measurements averaged 760 CFM in contrast to the unobstructed filters, which averaged 615 CFM.

The present testing group determined an average of 661 CFM on the obstructed filters and 651 CFM on the unobstructed hepa filters. Since the placement of the process equipment within the facility had no correlation with the air balancing results, an obstructed hepa filter has an equal likelihood of delivering less-than-average airflow as it does deliver higher-than-average airflow.

Was this apparent higher airflow phenomenon imaginary, illusory, or simply coincidental? Is it possible to determine, statistically, whether the obstructed filters and, therefore, the method used to measure them was significantly different? A two-sample, student’s T, difference-of-means test was conducted on the two data sets, the previous and the present test results.

Under the previous methodology, the t-test result of 4.83 indicates a >99.995 statistical significance that the set of 10 readings came from a different population, i.e., the test and calculation methodology was definitely in error. In contrast, the t-test done on our data (0.53) shows no statistical significance between direct capture flow hood readings and velocity-calculated CFM readings.

In our tests, it can be seen that 90% of all the velocity-calculated readings were above the average CFM of all filters. In fact, 90% of the calculated CFMs were higher than 83% of their flow hood measured CFMs. 50% of the velocity-calculated readings were below average, and the other 50% above average, as would be expected if a valid CFM calculation method were used.

This analysis demonstrates that, for airflow CFM, when it is necessary to calculate the volumetric flow from measured average velocity and measured area, using the NEBB CPT protocol detailed in equation 2 results in statistically significant data. The difference-of-means method of statistical analysis is a useful and even indispensable tool for revealing contradictory methodology.

This is especially important where detailed procedural disclosure is lacking and incomplete test data is provided. Statistical methods can indeed be used forensically to evaluate cleanroom test reports, as this study shows.