Trench Safety and Support

The classification is the design, not the paperwork

Every protective system on a trench sits downstream of one call: what the soil is. Get that call right and the batter angle, the bench geometry, the shield selection and the strut spacing all fall out of it in sequence. Get it wrong — grade a ground as cohesive and firm because the face stood up overnight — and every number downstream is wrong by a margin no amount of good workmanship recovers. Classification is not paperwork filed after the dig. It is the design input, and it is the only input that the weather, the traffic and the previous crew's backfill can quietly change while the excavator is still parked.

A competent person makes that call, and the definition carries weight in both halves: someone able to identify existing and predictable hazards, and someone with authority to stop the work and have them corrected. The second half gets forgotten on busy sites. An operator who can name the soil but cannot pull a crew out is not competent in the regulatory sense. In the United States the framework sits in 29 CFR 1926 Subpart P, Excavations. Great Britain works to outcomes instead of tables under the Construction (Design and Management) Regulations 2015, which pushes the geometry onto a temporary works designer rather than a wall chart. Australia, Canada and the European states each publish their own regime. Confirm which one governs the site before quoting any angle to anybody.

Depth is the trigger, not the reason. Most regimes demand a protective system from roughly 1.5 m (5 ft) down unless the cut is entirely in stable rock, and demand an engineered design once the excavation passes about 6 m (20 ft). Below the trigger depth the soil still decides; it simply decides through the competent person's judgement rather than through a table. Fatalities keep occurring in trenches shallower than 1.5 m because somebody read shallow as safe. A cubic metre of soil weighs on the order of 1.5 tonnes. Chest-deep is deep enough to fix a worker in place and stop the chest wall moving long before a rescue team arrives.

Grading the face without a laboratory

Start with the visual pass, taken from the top and from a safe standing position, never from inside. Look for layering and the direction the layers dip — beds running down into the excavation are a different animal from beds running away from it. Look for fissures, tension cracks parallel to the edge, spalling off the face, and the tell-tale cone of dry material at the toe. Look for seepage, damp bands, and any change of colour that marks a water table or a perched layer. Look for evidence that somebody has been here before: a soft strip of different-coloured fill, old bedding stone, a service marker tape. Previously disturbed ground never returns to its parent classification, however long ago it was reinstated.

Manual tests confirm what the eye suggests. Thumb penetration gives a working feel for unconfined compressive strength: firm cohesive ground that resists the thumb and only indents under real effort behaves very differently from material the thumb sinks into. A pocket penetrometer or a torvane gives a number rather than an impression, and both belong in the site vehicle. Plasticity is checked by rolling a moist thread — cohesive soil rolls to a thin ribbon and holds together; granular soil crumbles and cannot be rolled at all. Dry strength on a broken lump separates cemented and cohesive material from soil that falls apart between the fingers. None of these replace a laboratory classification to ASTM D2487, Standard Practice for Classification of Soils for Engineering Purposes, but they are what is available at the face at seven in the morning.

The grades themselves are banded on strength and then overridden by condition. The strongest cohesive band commonly starts around 144 kPa (1.5 tsf) unconfined compressive strength, the middle band runs from roughly 48 to 144 kPa, and the weakest band covers everything at or below about 48 kPa. Overrides matter more than the bands. Granular material — sand, gravel, loamy sand — sits in the weakest band whatever a penetrometer says, because cohesion is what a penetrometer is measuring and granular soil has none to measure. Submerged soil, soil with water freely seeping from the face, and unstable submerged rock all land in the weakest band. So does any soil subject to vibration, and any soil in a layered system where a weaker bed dips into the cut.

What each grade buys you in batter

Slope ratios follow the grade directly. Under Subpart P the simple slopes are three-quarter to one for the strongest cohesive band, one to one for the middle band, and one and a half to one for the weakest, with vertical faces permitted only in stable rock, and simple slopes capped at about 6 m of depth before an engineer is required. Those ratios describe horizontal run per unit of vertical rise, and the run is what consumes the site.

Consider what the grade costs at ground level. A 3 m deep trench with a 900 mm formation width in the strongest band opens to roughly 5.4 m across the top. The same trench in the weakest band opens to close to 10 m. That difference decides whether the dig fits between a boundary fence and a live carriageway, whether the footpath stays open, whether a lane closure is needed, and whether battering is viable at all. Crews rarely lose a trench to a wrong ratio; they lose the programme to discovering the ratio after the barriers are set out.

Layered ground is where composite judgement bites. Where beds of different strength stack, the safe default treats the whole face as the weakest bed present unless a designer has separated the layers and slopes each to its own requirement. A firm clay cap over a running sand is not a firm clay trench. The cap will stand while the sand runs out from beneath it and drops the cap as a single slab.

What takes the soil down a grade mid-dig

Nothing about a classification is permanent, and the downgrades come faster than the upgrades. Rain is the most common. Water entering the face reduces effective stress, raises the mass being retained and turns fissured clay into something that sloughs. Regimes generally require re-inspection after every rainstorm precisely because the grade may have moved overnight while nobody was on site.

Vibration downgrades ground to the weakest band regardless of what the penetrometer read. Plant tracking along the verge, a roller compacting the haunch, sheet piling on an adjacent plot, a busy road twenty metres away, or a rail line further off all qualify. The test is not distance but whether vibration is transmitted, and a hand on the crest of the spoil heap answers it faster than an argument.

Time works against the face on its own. An open cohesive trench dries, shrinks and cracks; the crack fills with water at the next shower and acts as a hydraulic wedge. Freeze and thaw cycles loosen a face that stood clean in autumn. Adjacent excavation, a leaking main discovered during the dig, or a spoil heap parked too close all change the loading. Any one of these obliges a re-grade, and a re-grade obliges a change to the batter, the bench or the shield before anybody steps back down.

Benching, and the soils that refuse it

Benching is available only in cohesive ground. It cannot be used in granular soil, and it cannot be used anywhere the classification has landed in the weakest band, which rules it out of most utility work in made ground and river terrace gravels. The reason is mechanical rather than bureaucratic: a bench relies on a vertical rise standing unsupported for the life of the works, and material without cohesion has no mechanism to do it.

Where benching is permitted, the vertical rise of each individual step is capped — commonly at about 1.2 m — and the cap applies to each step, not to the total height of the arrangement. Multiple benching stacks those steps behind an overall slope that must still satisfy the ratio for the grade. Cut the benches level and cut them wide enough to walk; a bench too narrow to stand on becomes a place to fall from, and a bench that slopes back toward the face ponds water straight into the crest.

Combination systems earn their place on deeper jobs. A battered or benched upper section over a shielded lower section keeps the top width inside the boundary while giving the crew a rated box at the working level. The junction between the two needs care: the top of a shield used this way generally has to project above the base of the sloped portion, commonly by at least 450 mm, so that material off the batter cannot arrive inside the box.

When the boundary says no batter

Boundaries, services and live carriageways rule out battering on most urban work, and the alternative divides into two families that are routinely confused. A shield — a trench box — does not prevent the face collapsing. It protects whoever is inside it when the face does collapse. Shoring is different: it applies load to the face and holds it. Hydraulic aluminium struts and walers, timber and screw jacks, sheet piles with waling frames and props, and slide rail systems all belong to the shoring family and all resist active soil pressure that varies with the grade.

Selection runs off the classification twice over: once to establish the lateral pressure the system must carry, and once to establish spacing. Manufacturer's tabulated data covers standard cases, and it has to be on site, legible, and matched to the actual grade rather than the optimistic one. Outside the tabulated envelope — unusual depth, surcharge from adjacent plant or structures, layered ground, sloping surface — the system needs design by a registered engineer. EN 13331-1, Trench lining systems, covers product requirements for lining equipment in the European market; national codes elsewhere carry the equivalent.

Installation discipline decides whether any of it works. Boxes go in as the dig proceeds, not after. Over-excavation below the base of a shield is limited — commonly to about 0.6 m, and only where the shield is rated for the full depth and no material is being lost from behind or below. Voids behind sheets or panels defeat the system entirely, so pack them as they appear. Struts are installed and removed in a defined order, top down for installation and bottom up for withdrawal, and nobody works below a strut being released.

Water reclassifies the ground faster than anything else

Free water is the fastest route to the weakest classification and the hardest condition to argue with. Seepage from a face is not a nuisance to be pumped from the invert; it is evidence that the ground is losing fines and that the classification has changed. Sumping from inside the trench pulls water through the face and accelerates the loss. Dewatering belongs outside the excavation — wellpoints, ejectors or perimeter drainage — and it belongs in place before the dig reaches the water, not after.

Watch for perched water at a layer interface, which arrives without warning at a specific depth and turns a competent face into a running one within minutes. Watch, too, for the trench acting as a drain for a leaking service; a slow rise that does not correlate with rainfall usually has a pipe behind it. Where dewatering runs, somebody has to monitor it, and the excavation has to be re-inspected when it stops, including after an out-of-hours pump failure.

Spoil, surcharge and the zone of influence

Everything stacked near the crest loads the ground the classification was assessed on. Spoil is the obvious case and the usual offender: most regimes set a minimum setback of about 0.6 m (2 ft) from the edge, and on anything other than firm cohesive ground a metre or more is the sensible working figure. The setback exists for two reasons — surcharge on the face, and material rolling into an open trench onto somebody's back.

Plant, pipe stacks, arisings skips, crane outriggers and site traffic all surcharge the face, and none are covered by the tabulated case unless the tables say so. As a working screen, temporary works designers commonly project a line up and out from the trench base to define the zone of influence; anything with a footprint inside that zone loads the excavation. A building foundation inside it needs support designed before the dig passes it, not after the crack appears in the render.

The volume the classification just created

Cutting to a batter or a bench does not simply make the trench safer, it makes it bigger, and the volume arrives on the muck-away schedule and again on the backfill schedule. A battered trench is a prism, not a box, and the extra material is proportional to the square of the depth — which is why a modest depth increase on weak ground can double the arisings against the estimate drawn from the pipe trench on the section.

Backfill follows the same arithmetic in reverse, with two adjustments the take-off has to carry. Excavated material bulks on the way out, typically in the range of fifteen to thirty per cent depending on the soil, and compacts back down on the way in, so the volume leaving site and the volume of imported fill required are not mirror images. Bedding, surround and any capping layer occupy part of the trench and are usually imported to specification, so they come out of the general backfill quantity rather than adding to it.

Shields change the backfill method as well as the quantity. Backfill and compact in lifts as the box is drawn up, because pulling a box out of a completed backfill leaves a void the width of the panels along the full length of the run, and that void will find the surface as a trench-line settlement trough. Compaction in layers of the specified thickness, generally in the region of 150 to 300 mm depending on the plant and the material, is what keeps the reinstatement out of the defect list.

Getting in, getting out, and what is in the air

Access sits at 1.2 m (4 ft) in most regimes: at that depth and below, a ladder, ramp or stair must be within roughly 7.6 m (25 ft) of lateral travel for anybody in the trench. Ladders are tied and extend about a metre above the landing. Inside a shielded run the ladder goes inside the box, not beside it, because the exit route has to sit within the protection.

Atmosphere is the hazard crews under-rate. Where a hazardous atmosphere could reasonably exist — contaminated ground, made ground and landfill, work near sewers, fuel stations or gas mains, or any deep excavation with poor natural ventilation — testing is required before entry and monitoring during the work, covering oxygen, flammable gas and toxic contaminants such as hydrogen sulphide and carbon monoxide. Engine exhaust from a pump or a generator set on the crest will find its way into a trench.

Inspection is a daily obligation, not an event. The competent person inspects before each shift, as needed through the shift, and after every rainstorm or other occurrence that could increase the hazard. Anything found is either corrected before entry or the excavation stays closed. Edge protection, crossing points for pedestrians and lighting for night work belong to the same inspection.

Recording the call, and recording the change

Because the classification is the design input, changing it is a design change. Record the grade with the date, the time, the tests performed and the person who made the call, and photograph the face while doing it. When the grade moves — after rain, after a piling rig arrives next door, after a layer of gravel shows up at formation — record the new grade the same way and tell the crew before they go back in.

Two failures recur on incident reports. The first is a face that was classified once, on day one, and never reassessed through six weeks of open trench. The second is a crew that knew the grade had changed and had no route to stop the work. Both are failures of the same mechanism: the soil kept deciding, and nobody was listening for the decision.