A dryer that used to hold −70 °C starts tripping at −50 °C. The first instinct is to blame the material, and the second is to order a replacement bed. Both are usually wrong. Most underperforming beds are failing for a reason that a new load of the same sieve will reproduce within a year.
There are five ways a molecular sieve bed loses performance. They look similar from the control room and are completely different in the vessel.
The five failure modes at a glance
| Failure mode | Typical symptom | What causes it | Reversible? |
|---|---|---|---|
| Hydrothermal ageing | Slow, steady loss of capacity over years; cycles shorten gradually | Repeated exposure to steam at high regeneration temperature; framework dealumination | No — capacity is lost permanently |
| Coking / hydrocarbon fouling | Capacity falls, regeneration gas outlet shows heavy ends, bed may darken | Compressor oil carry-over, heavy hydrocarbons, glycol, insufficient regeneration temperature | Partly — steam or higher-temperature regeneration can help if caught early |
| Chemical poisoning | Capacity loss concentrated in the inlet layer; fast breakthrough | H2S, mercaptans, HCl, ammonia, mercury, siloxanes, oxygen in trace amounts | Depends on poison — sulphur species can be partly reversed, siloxanes cannot |
| Mechanical breakdown | Pressure drop rises, dust at the outlet filter, capacity drops | Liquid water carry-over, thermal cycling, fast loading, bed migration into support | No — replace the affected layer |
| Channelling / maldistribution | Sudden breakthrough while most of the bed is still dry; short cycles | Poor loading, failed distributor, cracked bed, wrong support grading | Yes — reload and repair internals |
Read the dew point trend, not the dew point
One number tells you almost nothing. The shape of the outlet dew point curve over successive cycles tells you which of the five you are dealing with.
- Steady rise over months, cycle length slowly shrinking. Capacity is being lost. Ageing, coking or a poisoning front advancing through the bed.
- Sharp step change, one cycle to the next. Something structural. A cracked bed, a failed distributor, a support screen rupture, or a valve passing.
- Normal at the start of the cycle, breaks through earlier each time. Usable capacity is falling while the kinetics still look fine — the classic signature of a poisoning front moving through the inlet layer.
- Never reaches design dew point from the first cycle after a changeout. Not a material failure. Loading, screening, regeneration setpoint or instrumentation.
Hydrothermal ageing: the one you cannot fix
Water at high temperature attacks the aluminosilicate framework itself. Over hundreds of cycles the crystal structure loses aluminium, the pore network partially collapses, and the equilibrium capacity falls. Nothing in regeneration brings it back.
The practical consequences are worth stating plainly. First, a bed that has aged out is a consumable, not a fault — five to ten years is a normal service life and longer is a bonus. Second, the rate of ageing is set by regeneration temperature and steam partial pressure, so running the regeneration hotter than necessary shortens life for no benefit. Third, if a bed has aged out well before its expected life, look at the regeneration control loop before ordering the replacement.
Coking and oil carry-over
Lubricating oil from a reciprocating compressor, glycol from a badly operated dehydration unit, and heavy ends from a gas stream all end up in the bed. They sit in the macropores, block access to the crystal, and slowly carbonise at regeneration temperature.
On natural gas service this is the single most common cause of premature capacity loss we see in replacement enquiries. The fix is upstream: coalescing filters, a carbon bed, and a regeneration temperature high enough to strip what does get through. If the bed is already fouled, a controlled steam treatment can sometimes recover part of the capacity, but the bed is on borrowed time.
Chemical poisoning: the inlet layer tells the story
Poisons are adsorbed strongly and irreversibly, and they are adsorbed first. That means the top of the bed — the first 200 to 400 mm — carries almost all of the damage while the bottom is still close to new.
This is why a layered sample taken during a shutdown is more informative than any certificate. Sample the top, the middle and the bottom, and measure residual water capacity, bulk density and crush strength on each. A top layer at 60% of its original capacity with a bottom layer at 95% is a poisoning or coking diagnosis, and a new bed will fail the same way unless the guard train changes too.
| Poison | Typical source | Behaviour on the sieve |
|---|---|---|
| H2S and mercaptans | Sour gas, biogas, refinery streams | Strongly adsorbed, partly reversible at high regeneration temperature, competes with water |
| Hydrogen chloride | Chlorinated compounds, catalyst carry-over | Irreversible; also attacks the binder and vessel internals |
| Ammonia and amines | Amine units, ammonia plants | Strongly adsorbed on acid sites, hard to desorb |
| Mercury | Natural gas, some LNG streams | Amalgamates and blocks the pore; requires a dedicated mercury guard bed |
| Siloxanes | Biogas from landfill and wastewater digesters | Burn to silica inside the pore on regeneration — permanent, and fatal to the bed |
| Oxygen (trace, with hydrocarbons) | Air ingress, blanketing faults | Promotes coke formation during regeneration |
Siloxanes deserve a sentence of their own because the failure is silent and total. They pass through a water scrubber, they pass through carbon if the carbon is spent, and they arrive at the sieve. Each regeneration cycle converts a little more of them to silica inside the pore. There is no regeneration that removes silica. If you run biogas, the siloxane guard is not optional.
Mechanical failure and channelling
These two are the easiest to fix and the most often missed, because the material tests look normal and the problem is in the vessel.
- Liquid water carry-over from a failed separator cracks beads. The fragments migrate down, blind the support, and raise pressure drop while reducing capacity.
- Fast loading breaks beads on impact. The damage is done before the vessel is closed and shows up as dust within months.
- Missing or wrongly graded support lets the bed migrate into the outlet, which shows as a sudden pressure drop rise and dust downstream.
- A cracked bed or lifted distributor creates a path of least resistance. Gas bypasses most of the bed, breakthrough is sudden, and the remaining material is still dry when you open the vessel.
Regenerate or replace: a decision that needs data
- Check the regeneration first. Verify the actual bed temperature at the outlet of the regeneration gas, not the heater setpoint. A bed that never reaches temperature never fully regenerates, and it will look exactly like a poisoned bed.
- Check for liquid carry-over and pressure drop history. A step change in pressure drop is mechanical, not chemical.
- Sample by layer. Top, middle, bottom, at least 200 g each, in sealed moisture-barrier bags. Measure water capacity, bulk density, crush strength and loss on ignition against the original certificate.
- Compare against the original specification. If you do not have the batch certificate from the last changeout, ask the supplier for it. A supplier who cannot produce one is telling you something.
- Then decide. Capacity recovered by better regeneration means no purchase. A poisoned top layer with a healthy bottom means a partial change plus a guard bed. Uniform low capacity across all three layers means the whole bed has aged out.
Send us a sample and we will test it
We test incoming samples for residual water capacity, bulk density, crush strength and loss on ignition, and report them against the original specification. To make that useful, send:
- 300–500 g from each bed layer, sealed in moisture-barrier bags, labelled by position
- The original grade and bead size, and the date of the last changeout
- Operating data: inlet water content, cycle time, regeneration temperature and gas
- Contaminants you suspect, with the analysis that suggests them
- A dew point trend export from the last six months if available
Send those to our technical desk and you get a grade recommendation, a datasheet for the exact grade proposed, and a quotation with packaging and Incoterms stated — not a generic price list.
Frequently asked questions
How long should a molecular sieve bed last?
Five to ten years is a normal service life in clean service with correct regeneration, and longer is possible. Beds that fail in two to three years are almost always suffering from coking, poisoning or liquid water carry-over rather than from normal ageing.
Can a poisoned molecular sieve be regenerated?
It depends on the poison. Water and some sulphur species can be desorbed at high regeneration temperature. Hydrochloric acid, ammonia, mercury and siloxanes cause irreversible damage, and siloxanes are converted to silica inside the pore during regeneration, which cannot be removed.
Why did my dew point fail suddenly after years of normal operation?
A step change points to something structural: a cracked bed, a failed distributor, a support screen rupture, a passing valve, or a sudden change in inlet conditions. Gradual change points to capacity loss. The two have completely different fixes.
What tests show whether a sieve is still good?
Residual water adsorption capacity, bulk density, crush strength and loss on ignition, measured on samples taken separately from the top, middle and bottom of the bed, and compared against the original batch certificate. A single mixed sample hides the most useful information.
Does higher regeneration temperature always help?
No. It helps only up to the point where the bed is fully regenerated. Beyond that it accelerates hydrothermal ageing and shortens the life of the material. Verify the bed outlet temperature during regeneration rather than trusting the heater setpoint.
Can I top up a bed with new material instead of replacing it?
Sometimes, and it is a legitimate short-term measure. But mixing new and aged material of the same nominal grade produces uneven flow and an unpredictable cycle. If more than a small fraction of the bed is spent, replacing the whole layer is usually the cheaper decision over the next cycle.
Related reading: regeneration and service life, how we test, and bead size selection.
Figures marked as typical ranges are indicative values compiled from published supplier datasheets and standard adsorption-engineering references. They describe the class of material, not our specification and not a performance guarantee for your unit. Bed design must be confirmed against your own flow, pressure, temperature and dew-point target.