Analysis of Deficiencies in Current Leak Detection Methods for HEPA Filter Outlets
The Code for Construction and Acceptance of Ventilation and Air Conditioning Works (GB 50243-2002) specifies in Appendix B.3 “Leak Test for Air Filters“:
B.3.1 Leak detection for HEPA filters should be conducted using an optical particle counter with a sampling rate greater than 1 L/min. For Type D HEPA filters, a laser particle counter or condensation particle counter is recommended.
B.3.2 When using a particle counter to detect leaks in HEPA filters, a uniform concentration of atmospheric dust or air containing another aerosol should be introduced on the upstream side. For particles ≥0.5μm, the concentration should be ≥3.5×10⁵ particles/m³; for particles ≥0.1μm, it should be ≥3.5×10⁷ particles/m³. For Type D HEPA filters, the same ≥3.5×10⁷ particles/m³ applies to particles ≥0.1μm.
B.3.3 The scan method should be adopted, where an isokinetic sampling head of the particle counter is placed 20–30 mm from the downstream surface of the filter, and moved across the surface, edges, and sealing gel at a speed of 5–20 mm/s.
B.3.4 Leak detection must be carried out under conditions close to the design airflow rate. The leakage concentration measured downstream should be converted into penetration rate: for standard HEPA filters, this rate must not exceed twice the factory-certified rate; for Type D filters, it must not exceed three times the factory value.
B.3.5 During scanning, any area with a sudden spike in particle count should be further examined in place.
Notably, the specification does not clearly state whether the leak test should be conducted before or after the HEPA filter is installed. However, based on B.3.4 and Section 5.4.1 of Code for Construction and Acceptance of Cleanrooms (J71-1990), it can be inferred that the leak detection is meant to be performed after installation—as the scanning process does not specify scanning the mounting frame of the filter itself.
Leak detection after installation primarily checks two areas: the filter medium and the sealing interface between the filter and the air outlet. For cleanrooms with ISO Class 5 (equivalent to Class 100) or higher cleanliness requirements, leak tests must be conducted before installation, and only filters that pass should be installed.
Conducting leak tests on installed filters, however, presents practical difficulties. According to the requirements, the upstream concentration of particles must be at least 3.5×10⁵ particles/m³ for ≥0.5μm particles. Yet, in full-air cleanroom systems with dozens or even hundreds of HEPA terminals, achieving such upstream conditions is highly challenging. Should aerosol-laden air be introduced separately for each filter (highly impractical), or uniformly across the system? If uniform, it is impossible to scan all filters simultaneously.
This makes it highly likely that untested filters become contaminated by the introduced aerosol before they are even scanned—contradicting the intent of cleanroom protocols. In practice, the author has found that this method is difficult to implement and may violate fundamental cleanroom principles. For example, introducing unfiltered outdoor air into the upstream side of the HEPA filter in a biosafety laboratory contaminates the entire airflow path—from outer doors, clean zones, semi-contaminated zones, and buffer rooms—before it even reaches the exhaust HEPA filter (which is, in fact, almost impossible to test from the downstream side). The same applies to other cleanrooms, where introducing unfiltered air into upstream ducts may compromise the entire ventilation system.
Scanning leak detection with a particle counter involves placing the sampling inlet 20–30 mm downstream of the filter surface and moving it at a controlled speed of 5–20 mm/s along the surface, frame, and mounting edges. This process is time-consuming, and the longer the test, the higher the contamination risk.
These procedures directly contradict cleanroom construction practices, such as carefully wiping internal duct surfaces, sealing them with plastic film, and maintaining spotless panel surfaces.
Even assuming a leak is detected
(per B.3.4), the method makes it nearly impossible to pinpoint the actual leak location. The sampling hose is typically ≤1.5 meters long, and particles flow through it into the optical chamber, are amplified and sorted, then counted and displayed—this takes time. Due to this time lag and the scan speed, the actual location of the leak cannot be accurately identified. Even if a spike in count is observed, one cannot tell if the leak is from the filter surface, paper folds, or the scan side. There could be multiple possible leakage points corresponding to a single spike location (as shown in Figure 10-7, points 1 to 9), making accurate patch sealing nearly impossible.
For this reason, ensuring filter quality before installation is crucial—whether it passed factory testing, was transported properly, stored under suitable conditions, etc. Leak detection should be viewed as a supplemental step, not a primary one. Even if detection methods were perfect, sealing leaks would still be difficult.
Scanning the filter’s mounting frame to detect leaks poses the same problem—pinpointing the leak within the gasket is nearly impossible. In such cases, the filter must be removed, the gasket inspected or replaced, and the entire unit reinstalled. This again underscores the importance of proper installation techniques.
Therefore, leak prevention should be achieved through comprehensive control of the entire installation process—from unloading, storage, handling, unpacking, applying gasket seals, and securing bolts (in pressure-mount installations). Relying solely on post-installation leak detection is neither practical nor reliable.
