The 4.2 m excavation with a copied risk assessment and a new crack in the neighbour's wall
An excavation 4.2 m deep is being opened for a basement parking level. The face nearest the boundary is 3.1 m from a 1950s double-storey house on strip footings, and the boundary wall now shows a 6 mm crack that was not present at handover. The drawings show an 11 kV cable crossing the excavation, but no cable locating or potholing has been done. The baseline risk assessment, copied from an earlier project, rates 'trench collapse' as likelihood 2 x severity 5 = 10 ('medium'), and the only control listed is 'workers trained and warned of the danger'. Thirty-eight millimetres of rain fell overnight and water is standing at the north corner. The site manager wants three labourers to go in and bail it out so that the concrete can be poured on Friday.
1. Critique the risk rating and the recorded controls, and explain what the rating has done to the decision-making on site.
Read the worked answer
The score is arithmetically consistent and substantively wrong. Three faults:
- Likelihood was rated against a hypothetical site, not this one. A 4.2 m face standing 3.1 m from strip footings of a 1950s house is a foreseeable catastrophic failure mode - the failure plane does not have to reach the excavation to undermine the footings. The new 6 mm crack is a leading indicator, not a cosmetic defect. Likelihood is assessed with the controls actually in place, and here there are none.
- 'Train and warn' is not a control against collapse. It sits at the bottom of the hierarchy and does nothing to the energy source. It only ensures the people who get buried understood the risk.
- A copied baseline is not a risk assessment. The Regulations require the assessment to be site-specific and done before the work starts by a competent person. Recycling one from another project removes the only step where someone looks at this ground, this cable and this neighbour.
The consequence: a '10 – medium' score authorises work to continue with no engineering input and no budget. The rating is not a measurement; it is a judgement that has to be defended afterwards. The defensible output here is high/extreme, triggering a design response, not a toolbox talk.
2. Nobody enters that excavation. Set out the controls you require before work resumes, applying the hierarchy of controls.
Read the worked answer
Eliminate / substitute (first, and cheapest in the long run):
- Re-sequence or re-engineer: a secant or contiguous piled wall, or moving the excavation line, removes the near-face problem instead of managing it. Ask the designer, in writing, whether the original batter/shoring assumption was ever valid at 3.1 m from the boundary.
Engineering:
- Temporary support designed by a competent engineer for a 4.2 m face, with a signed drawing and a stated installation sequence: shoring, struts, or a properly battered profile.
- Underpinning or propping of the boundary wall and footings, with the neighbour's written acknowledgement.
- Dewatering by pump and sump from the surface - never by workers in the hole - plus a diversion so the face is not saturated. Water is what turns a stable face into a run-in.
- Cable locating and potholing by the asset owner or an approved locator before the 11 kV cable is approached; the cable route marked and an exclusion zone set.
- Crack monitoring: tell-tales or a gauge fixed across the crack, read and recorded daily, with a trigger value and a stop-work rule.
Administrative:
- Documented inspection by a competent person before each shift and after any rain, before entry; a permit-to-enter system for the excavation; a stand-by person at the top; no entry by anyone working alone.
PPE: last, and it does nothing for a collapse. Anyone relying on PPE as the answer here should not be signing the risk assessment.
3. The site manager argues that the crack 'was there before we started'. How do you settle the factual dispute, and what happens if it cannot be settled?
Read the worked answer
Do not settle it by argument. Settle it with a record made now and, if possible, a record made before the work started.
- Look for the pre-construction record. A dilapidation or condition survey of the neighbour's property, dated and photographed, is the only thing that ends this discussion. If there is none, that is itself a finding - it should have preceded an excavation 3.1 m from a 70-year-old house.
- Make the record now. A dated, photographed, measured survey of the crack (width, length, location, crack gauge fixed across it) countersigned by the neighbour or their representative, and an independent structural engineer's opinion on cause and on whether the excavation is affecting it.
- Instrument it. Daily readings, a trigger value, and a written stop-work and notification rule that everyone on site knows.
If it cannot be settled: the burden sits with the contractor, not the neighbour. Until the cause is established, treat the face as unstable: the excavation is supported by design before anyone re-enters, and the monitoring is in place. Document the disagreement, the mitigation you required despite it, and the fact that the mitigation was required because the cause was unknown - that is what an inspector, or a court, will read.
A crack of 6 mm in older masonry near a deep face is a legitimate basis for the neighbour to demand a structural assessment, and the client should know that cost is already accruing.