Every filtration system fails in one of two ways. It blinds, meaning the media fills with solids until flow can no longer pass. Or it bypasses, meaning fluid finds a path around the media and continues downstream carrying everything the filter was bought to remove.
Blinding is expensive but honest. Bypass is cheap and silent, and it destroys downstream equipment while every instrument on the skid reports normal operation.
The curve shape is the diagnosis
A blinding filter produces a differential pressure curve that rises continuously, gently at first and then steeply as the remaining open area shrinks. Downstream quality holds throughout. This is the curve you want, because it is predictable and schedulable.
A bypassing filter produces a curve that rises normally, then flattens — or falls. The temptation is to read the plateau as a stable operating point. It is not. It means the fluid has found a lower-resistance path, and the only paths available are around the seal, through a seam, or through a hole in the media.
Where bypass actually originates
In our field investigations over the last four years, the distribution of root causes has been remarkably consistent:
- Seal lift under differential pressure — a bag or cartridge collar that unseats as the pressure across it rises, typically at 1.2 to 1.8 bar on poorly retained elements
- Sewn seams — needle perforations in bag construction that are larger than the media rating, present from the first minute of service
- Element collapse — an under-rated cartridge core failing structurally, opening a direct path through the middle of the element
- Housing internals — a warped seal plate or a missing spring plate allowing flow around the element bundle entirely
- Wrong element fitted — a 222 element in a 226 seat, or a DOE element with a knife-edge seat that never engages
What to do about it
Trend the differential pressure continuously rather than reading it at change-out. A plateau in the curve should raise an alarm, not a sense of relief.
Sample downstream on a schedule that does not depend on the filter appearing to work. Particle counts or turbidity on the filtrate side is the only measurement that distinguishes a filter that is working from one that has stopped.
And when specifying, treat retention hardware as seriously as media rating. A 1 µm absolute element in a housing that permits bypass at 1.5 bar is a 1 µm filter for approximately three days.