The Defects Behind Most LEV Test Failures

Airflow starvation: how a system slowly stops breathing

Starvation is the commonest mechanism and the most gradual. An LEV system is a network of resistances, and every month of operation adds a little more: dust settling in horizontal runs where transport velocity sagged, filter cake building faster than the cleaning cycle sheds it, a flexible connector crushed a little flatter behind a machine, a damper nudged by a ladder, grilles furring over. Each restriction steals a fraction of the airflow, the fan works against a steadily rising resistance, and velocities everywhere drift down together.

The signature is exactly that togetherness. Where one hood fails alone, suspect geometry or its own branch; where every reading on the report sits five to fifteen per cent below benchmark, the system is being throttled somewhere common: the filter, the main duct, the fan inlet. The examiner's velocity table is a diagnosis in rows if you read it as a pattern rather than a list.

Starvation is also the most preventable mechanism, because it is visible weekly for the price of a glance. Static pressure gauges across the filter tell you cake is winning. Airflow indicators at hoods tell you branches are dying. Dust settling on surfaces near a hood that used to run clean tells you capture is sagging. A weekly check that records these three signals catches starvation months before it becomes a red entry, and the fix at that stage is housekeeping rather than remediation.

One starvation cause deserves its own caution: cleaning settled dust out of ducts treats the symptom. The dust settled because transport velocity fell below what that dust needs; find why the velocity fell, or the ducts refill on schedule.

Capture geometry: the failure that lives in centimetres

Capture is a short range weapon, and most geometry failures come down to distance. The air velocity a hood generates falls away brutally fast in front of it; move a plain hood one duct diameter from the source and the capture velocity has collapsed to a small fraction of the face figure. A hood that captured perfectly at fifteen centimetres can be useless at forty, with the system behind it in flawless health.

Geometry fails in mundane ways. A machine shifted for a delivery and never shifted back. A workbench rearranged so the task now happens beside the hood rather than under it. A flexible capture arm parked out of the way and left there, because repositioning it per job was never made anyone's habit. A damaged hood lip, a missing side baffle, a flanged hood with the flange removed in a repair. Cross draughts from a new door, a pedestal fan or a relocated compressor blowing the contaminant out of a capture zone that used to be sheltered.

The tells are visual and immediate, which makes geometry the mechanism operators catch best. Smoke, dust or fume visibly escaping around a hood, deposits growing on surfaces the system used to keep clean, operators smelling the process through a running system. A smoke pencil in a trained hand finds a geometry failure in thirty seconds, which is why it belongs in the weekly check for any system with moveable hoods or arms.

Fixes are correspondingly cheap when caught: return the machine, re rig the bench, mark the arm's working position on the floor, refit the baffle, manage the draught. The expensive version of the same failure is the one that waits fourteen months for an examiner to measure it, having dosed the operator all the while.

Plant degradation: the machinery ageing underneath the airflow

Behind every capture zone is rotating machinery, and it ages like all machinery. Fan belts stretch and slip, and a slipping belt can shed a large share of fan speed while sounding almost normal. Impellers erode in dusty airstreams and cake in sticky ones, losing blade profile and moving less air per revolution. Motors are swapped in a weekend breakdown and reconnected with reversed rotation, after which a centrifugal fan still blows, badly, in the right direction, and a shop can run for months on a fan delivering a fraction of design. Bearings wear, dampers seize where a balancer left them or where a helpful hand moved them, filter cleaning mechanisms fail so cake builds unshed, and gaskets and access doors leak air into the system before the fan that never passed a hood.

Degradation's signature on the report mimics starvation, readings down across the system, which is why the two are worth separating at diagnosis: starvation is resistance rising, degradation is the mover weakening, and the pressure picture tells them apart. Rising pressure drop across the filter with falling flows says starvation; falling fan pressure with falling flows says the fan itself.

The defence is ordinary maintenance discipline applied to a system people forget is plant: belts checked and tensioned on a schedule, rotation verified after any electrical work as a standing rule, bearings on the greasing round, filter mechanisms function tested, dampers locked and marked once balanced. None of it is exotic. LEV fails here mostly because extraction sits on nobody's maintenance schedule, being neither production machinery nor building services, and a system that is nobody's plant degrades on schedule.

The missing benchmark: the failure that happens on paper first

One failure mechanism involves no dust and no decibels: the system for which nobody can produce the commissioning data. A TExT is a comparison, measured performance against the performance the system was designed and commissioned to deliver, and when the benchmark side of the comparison is missing, the examiner is reduced to judging against generic ranges and their own experience. On a marginal system that uncertainty resolves against you, and the honest report says so: adequate control not demonstrated, insufficient data.

Estates inherit this failure rather than commit it. The system came with the building, the installer folded years ago, the commissioning report lived in a filing cabinet three tenancies back. It feels like a paperwork gap and behaves like an engineering one, because without design figures every future judgement, every rebalance, every modification is guesswork anchored to nothing.

The fix is reconstruction, and it is worth doing properly once. A competent LEV engineer assesses the process, establishes what capture each hood needs for the contaminant and task, measures what the system delivers, and produces a new benchmark: effectively recommissioning the system as found, with target figures for every test that follows. Fold the exercise into remedial work or an uprating if one is due, because the measurement overlap makes the marginal cost small.

Then protect the benchmark like the asset it is: the commissioning data kept for the life of the system, copies with the logbook and the scheme of tests, and every modification closed with updated figures. The estates that never suffer this mechanism twice are the ones that treat benchmark data as part of the system, transferred with it, updated with it, and named in the handover of any building, contract or provider.

Turning the mechanisms into a weekly check that actually predicts

The five mechanisms earn their keep when they become the structure of the weekly user check, because a check built on mechanisms predicts the test instead of merely preceding it. Generic ticklists ask is the system working, which invites yes; a mechanism check asks five specific questions that each have a number or an observation as the answer.

For starvation: the filter pressure reading, written down, against its marked clean and dirty limits. For geometry: a smoke pencil pass at the working position of each moveable hood or arm, and a glance for new deposits. For drift: has anything about the process changed this week, a question that costs nothing and catches the tee in before it beds in. For degradation: the fan note on start, belts on the monthly extension of the check, damper marks intact. For the benchmark: readings recorded against the commissioning figures, so the log accumulates the very series that makes every future judgement cheap.

Write the check to fit on one laminated page per system, make the answers numbers and words rather than ticks, and route anything out of range to a named person the same day. Ten minutes a week per system is the entire cost, and the return is that failures surface as maintenance items in quiet weeks rather than red entries with retest fees and exposure questions attached.

Close the loop with the examiner: show them the log at each TExT, and ask them to mark the readings they most want tracked. Examiners write gentler remedial lists for systems whose owners visibly watch them, not from charity but because the evidence changes the engineering judgement about what will stay adequate until the next test.