Most failure investigations begin with the thing that broke still in front of the investigator. Excavation collapse does not. The material that failed is disturbed, remixed and often removed during rescue; the void that remains is an open hazard on a public right-of-way; and within hours to a day or two the excavation is backfilled and the street repaved. Nothing about that sequence is improper. It does mean the reconstruction rests almost entirely on what was captured in a narrow window, and on records made before anyone knew they would matter.
There is no fracture surface to read
A cracked shaft or a burned conductor keeps its own record of what happened to it. Soil does not. The failure surface in a trench wall exists only as long as the wall stands after the collapse, and it is then walked on, dug through and buried.
That shifts the evidentiary centre of gravity. The physical record is thin and perishable; the documentary record carries a larger share of the analysis than it would in most other failure work.
Rescue legitimately destroys the scene
Where someone is buried, the response is an excavation of its own: vacuum trucks, hand digging, the trench deliberately widened, additional shoring installed by responders. The site an investigator eventually sees has been reshaped by people acting correctly under enormous pressure.
Distinguishing incident damage from rescue damage is therefore a routine first task. Responder photographs and video, incident command logs and the accounts of the crews who dug are what separate the two, and they sit outside the construction chain.
The window before backfill
Whatever is captured of the as-failed geometry has to be captured before the hole is closed. Laser scanning or photogrammetry of the excavation, its walls and the deposited material produces a measurable record anyone can re-examine later. Where no expert reaches the site in time, ordinary photographs do much of the work, provided they carry scale, orientation and enough context to locate them along the trench.
Sampling the spoil, the walls and the adjacent ground
Spoil is disturbed and remixed, so its value lies mainly in index properties: gradation, plasticity, and what materials were present at all. Samples from intact trench walls beyond the collapsed reach, and from undisturbed ground beside the alignment, retain more in-situ structure and are the better basis for strength testing.
Moisture content is the most perishable property on the site and should be sampled early and sealed. Where the excavation is already closed, borings along the alignment can still establish the profile and, through ASTM D1586 penetration testing, a depth-referenced measure of the material. None of this recovers the exact condition at failure; it bounds it.
The shoring hardware is a physical record
Trench boxes, spreaders, pins, hydraulic cylinders and timber all deform under load, and that deformation is readable afterwards. Bowed panels, bent spreaders, cylinder position and residual pressure, sheared pins and crushed timber describe the direction and rough magnitude of what arrived.
The hardware also identifies itself. Data plates tie a shield to a model, its rated depth and soil type, and the manufacturer's tabulated data, which Subpart P requires to be available while the system is in use. Preserving the components rather than returning them to the rental yard keeps that comparison open.
The inspection record and what the standard requires
Subpart P requires the competent person to inspect excavations, adjacent areas and protective systems for evidence of cave-in or other hazardous conditions before each shift, as needed during it, and after every rainstorm or other hazard-increasing occurrence. It requires the inspection. It does not, on its face, require a written record of it.
Many contractors keep daily excavation checklists voluntarily, and where they exist they are among the most informative documents in the matter — soil type, depth, protective system, water and weather, day by day. Where they do not, the inspection has to be established from testimony.
The competent-person file
Subpart P assigns specific duties to a competent person: someone capable of identifying existing and predictable hazards and authorised to take prompt corrective action, who classifies the soil and selects or approves the protective system. The supporting record is ordinary employment paperwork — designation, training certificates, prior classification decisions on the same job, and evidence of authority actually exercised. It is normally intact long after the trench is gone.
Machine, weather and locate data
Excavator and loader telematics place equipment in time. Weather data establishes rainfall and freeze-thaw. One-call tickets and utility as-builts fix what was known about buried installations before the dig. Permits, engineered shoring drawings and daily reports establish what the job was supposed to be.
These sources were created for reasons unrelated to the incident, and most sit with third parties. They also have retention windows, so the request needs to go early.
What preservation looks like
In practice: photograph and, if possible, scan the excavation before anything is moved; hold the shoring components rather than returning or repairing them; take and seal soil samples from the walls, the spoil and adjacent undisturbed ground; secure the day's paperwork before the crew disperses; and send preservation requests to the equipment, utility and rental parties promptly.
Rescue and reopening the street come first, and no investigation is worth delaying either. The distinction worth drawing is between the pressure to restore the site and the loss of a record a few deliberate minutes would have preserved.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.