Incident analysis
Found, reported, and still in service: what incident reports say about defects.
Six public reports from Australian regulators, courts and a coroner, read for one question: where did a known problem slip between finding it and fixing it? The answer is rarely that nobody noticed.
Published 7 min read7 sources
Written for safety managers · maintenance managers · site and project directors
Incident reports are usually read for their causes. Read them instead for what was already known, and a different pattern appears. In most of the reports below, somebody saw the problem first. A pre-start recorded it. A technician asked for the part. A procedure said what to do the second time it happened. The failure was in the step after.
These are public documents, and each is linked at the end. In the text we leave out the names of companies, sites and people, even where the source gives them; the sources are cited as published, without the names of anyone who died. The point is the pattern, not the parties.
Six reports, one pattern
The same faults, recorded almost daily, then filed
NSW mine processing plant · 2021 · regulator report 2023- The record
- A worker was buried in a conveyor chute after the emergency-stop lanyard failed to stop the feeder. The investigation found pre-start checklists recording the same 'high' faults almost daily, filed without corrective action, with nothing recorded about repairs or tagging the plant out. That the lanyard was stretched was 'common knowledge'.[1]
- Our reading
- The checks were being done and the faults were being written down. A checklist that goes into a folder is evidence that a hazard was known, not that it was controlled. What matters is what a 'high' fault does next.
A clearing system out of service for over a year
WA crushing plant · December 2016- The record
- A fitter was struck in the chest while a bogged cone crusher was being cleared by hand. The crusher's own hydraulic clearing system had been inoperable for more than twelve months, with seven of its twelve cylinders leaking. The repair had been planned for a shutdown and slipped when the parts were late.[2]
- Our reading
- A deferred defect changes the risk of every routine task done around it. The regulator asked for 'a timely response to loss of serviceability'. That starts with being able to see how long something has been unserviceable.
Worn parts, and repeated blockages in one shift
Queensland waste facility · 2021 · sentenced 2024- The record
- A worker died clearing a shredder on night shift. The court report records the shredder in poor mechanical condition, with most of its teeth non-operational and due for replacement, and repeated blockages through the shift. There was no documented procedure for clearing a blockage, and the only safety measure was an undocumented lock-out and tag-out. There was no supervisor on the shift.[3]
- Our reading
- Two signals were present: a known, deferred repair and a fault repeating within hours. Neither reached anyone who could stop the work.
Inspections that did not find the defect
WA drill rig · 2015- The record
- A 41 kg section of walkway mesh fell 18 metres from a drill mast and landed beside the occupied cab. Worn washers and non-locking nuts were found, and the regulator concluded that inspections were inadequate. It then set out the loop it expects: a competent person assesses the defect, it is repaired as specified, and the plant is inspected and tested before it is returned to service.[4]
- Our reading
- This is the regulator's own description of defect control, step by step. Every step produces a record. Only the people in it can do the inspecting, the judging and the testing.
Repair work, and what should have been in place
Queensland manufacturer · conveyor · sentenced 2024- The record
- A worker repairing a powder-coating line was caught in its unloading conveyor and died. Sentencing, the magistrate singled out three controls the company introduced afterwards as what should have been in place: notifying someone before starting work on the conveyor, completing a maintenance log, and isolating or tagging out the equipment first.[5]
- Our reading
- Two of those three controls are about information: who knows work is under way, and what was done. The third is physical. No record substitutes for it.
Five resets in seven days
Queensland theme-park ride · 2016 · coroner's findings 2020- The record
- Four people died when a water ride's pump failed and the conveyor kept running. In the seven days before, the ride broke down five times with the same pump-drive fault, and each time the drive was reset without a diagnosis. The written procedure said a repeat within 24 hours needed an engineering supervisor's authority. Staff were unsure whether that meant the second or the third time. The park had a maintenance system, but its planner was not told about ride breakdowns, which were logged on paper at the ride. The coroner found that 'shoddy record keeping was a significant contributor'.[6]
- Our reading
- The coroner was clear that the primary causes were missing engineering controls: no water-level monitoring and no interlock to stop the conveyor. Records were a contributor, not the cause. The lesson for maintenance is narrower, and still hard: a system existed, and the breakdowns never reached it.
Where the chain breaks
Put the six side by side and the same four failure points recur. None of them is a failure to notice.
- A finding that does not become an action. The fault is recorded, but recording it changes nothing about the machine: it still runs, and nobody is assigned. The processing plant's checklists and the ride's paper down-time sheets both did this.
- A deferred defect that nobody re-assesses. A repair is put off for parts or a shutdown, and the machine is worked around it. Each day the defect stays open, the risk of the workaround is taken again without being decided again. The crusher and the shredder both ran like this.
- A return to service without a competent check. The drill-rig report asks for plant to be inspected and tested by a competent person before it is used again. The ride's own procedure called for an independent functional safety test before safety equipment went back into service.
- Knowledge held in one place. An emergency-stop lanyard whose stretch was 'common knowledge' among workers, and was never fixed.[1] Engineering drawings and incident reports on one engineer's hard drive, and an incident history the head of engineering did not know about.[6]
What a connected record can and cannot do
Software did not cause any of these events, and it would not have been enough to stop them on its own. What it can do is narrower and still worth having: make each of those four failure points harder to fall through without anyone noticing.
- A critical failed check can gate the machine for everyone at once, so its next pre-start is refused rather than signed off.
- An unresolved finding stays open on the machine until a person resolves it, instead of being filed with the checklist that found it.
- The same fault coming back is counted against the machine, so 'is this the second time or the third?' has an answer.
- Returning a machine to service can be limited to people with that authority, with a record of who decided, when and why.
- The procedure, the manual and the last repair can be in front of the person doing the work, not on someone else's drive.
What it cannot do is set out in the same reports. It cannot isolate energy: the WA regulator's standard is 'lock, tag and try', done with hands and locks.[7] It cannot install an interlock, inspect a structure, judge whether a repair is adequate or supervise a night shift. The coroner put it plainly:
“Regardless of the training provided at [the park], it would never have been sufficient to overcome the poor design of the [ride], the lack of automation and engineering controls.”
Questions worth asking of your own site
- When a pre-start records a critical fault, what happens to the machine in the next ten minutes, and who knows about it?
- How many defects are open on your plant today, and how old is the oldest?
- If the same fault is reset three times this week, who will notice, and how?
- Who may return a machine to service, and where is that decision recorded?
- Could a new supervisor find the last five things that went wrong with a machine without asking anyone?
How Mainte carries a failed check through stand-down, repair and an authorised return to service is set out in From a failed check to a safe return to service.
Sources
- [1]Investigation report: serious injury of a worker at Thuddungra Mine · Resources Regulator, NSW Department of Regional NSW · May 2023 · Major safety investigation reportQuoted findings: executive summary (pp. 3–4) and section 6.6 (p. 19).
- [2]Mines Safety Significant Incident Report No. 252: fitter struck in chest by packing plate released under pressure · Department of Mines and Petroleum, Western Australia (now hosted by WorkSafe WA) · 12 May 2017 · Significant incident report
- [3]Waste resource company fined $140,000 after worker fatality · Office of the Work Health and Safety Prosecutor, Queensland · Decided 19 February 2024 · Court report (summary of sentencing)
- [4]Mines Safety Significant Incident Report No. 227: grid mesh falls from drill mast · Department of Mines and Petroleum, Western Australia (now hosted by WorkSafe WA) · 6 October 2015 · Significant incident report
- [5]Aluminium production company hit with $150,000 fine after worker fatality · Office of the Work Health and Safety Prosecutor, Queensland · Decided 19 April 2024 · Court report (summary of sentencing)The report gives the date of the incident as both June 2021 and June 2022, so no year is given for it here.
- [6]Inquest into the deaths at the Thunder River Rapids Ride: findings of the coroner · Coroners Court of Queensland · 24 February 2020 · Coronial findings and recommendationsQuoted at paras 340, 352, 387, 1010, 1028 and 1035. The findings name the people who died; this page does not.
- [7]Mines Safety Significant Incident Report No. 248: worker seriously injured when caught in moving conveyor · Department of Mines and Petroleum, Western Australia (now hosted by WorkSafe WA) · 12 September 2016 · Significant incident report