Where collision-prevention projects lose their value — and the fix.
Once a system meets the intervention standard, the risk shifts from specification to deployment. These are the patterns that turn a compliant installation into one the operators work around.
| What goes wrong | How you notice | The fix |
|---|---|---|
| Detection tuned in a workshop | Nuisance alarms on curves, berms and near reflective steel | Commission in the real environment. Dust, grade and reflective surfaces change the behaviour that looked clean on a test pad |
| Interventions quietly overridden | The event log is full of warnings and short of interventions | Audit the intervention rate, not just the event count. A system that is always overridden has been switched off in practice |
| Pedestrian tags worn inconsistently | Recorded interactions fall while near-misses do not | Make tag issue part of the lamp-room routine so non-wear is visible before the shift, not after an incident |
| Braking calibration never re-checked | Intervention distance drifts with load, grade and brake wear | Re-verify after brake work and with representative loads — the stopping distance in the commissioning report is one machine on one day |
| Mixed fleet, mixed systems | Machines from different suppliers do not detect each other | Make cross-fleet interoperability a specification item and test it before rollout, not during |
| Events logged but never reviewed | Compliance evidence accumulates while incident rates hold steady | Put a short weekly event review on a named person’s agenda; the log is only useful once somebody reads it |
| No interaction baseline | Layout changes get argued rather than targeted | Use the first month of events to find the three places worth re-engineering. Most of the long-term gain is in layout, not in stopping machines |
The two worth watching hardest are nuisance alarms and the unread log, and they are connected. A system tuned somewhere other than the site produces alarms operators learn to dismiss; once dismissal is habitual, the intervention data stops describing the mine and starts describing the workaround. Tuning on site and reviewing the log weekly are cheap, and together they are what keeps the system trusted.