How a mining operation in the Andes recovered 36 h on a SAG mill shutdown

The problem: dead time nobody had added up

When the maintenance superintendent showed me the SAG mill shutdown history, the number that jumped out was not the repair time itself — it was the invisible cushion between activities. A mining operation in southern Peru had run three consecutive shutdowns at 84 h actual against 48 h planned in Primavera P6. The gap: 36 h of dead time scattered across material waits, crew overlaps, and late shift starts.

Nobody had totalled it because each department reported its own window separately. Maintenance closed green. Operations closed green. The project closed red.

Diagnosis: where the hours were going

I mapped every activity by crossing the site access-control timestamps against the schedule exported from Primavera P6. The picture was unambiguous:

Root cause	Hours lost (average)
Waiting for parts at the central warehouse	12 h
Unresolved overlap between electrical and mechanical crews	9 h
Late night-shift start due to shared transport	7 h
Quality inspection with no assigned window	5 h
Equipment release held pending supervisor sign-off	3 h
Total	36 h

No single cause was catastrophic. Together they equalled 1.5 lost production shifts per shutdown.

The three interventions we applied

1. Pre-staging critical materials

The central warehouse sat 4.2 km from the mill. Standard practice was to pull spares at shutdown start. We changed that: critical parts — liners, hold-down bolts, seals — were pre-staged at the field store 72 h before the planned start. We defined "critical" as any item with an internal dispatch lead time above 2 h and no emergency stock on site.

Direct result: 10 of the 12 warehouse-wait hours eliminated.

2. Crew sequencing with an explicit buffer

The original schedule ran the electrical and mechanical crews in parallel with no buffer zone. In the field, any slip by one crew blocked the other because both accessed the mill interior through the same entry point.

We rewrote the sequence as follows:

Activity A — Mechanical crew (liner removal):          0–18 h
Transfer buffer:                                      18–20 h  ← inspection and clean-out window
Activity B — Electrical crew (drive inspection):      20–28 h
Activity C — Mechanical + electrical (assembly/test): 28–42 h

The 2 h buffer was not cosmetic float. It was the real time required to clean the work area and complete a visual inspection before the electrical crew entered. That time had always existed in the field; it just was not in the schedule, so nobody managed it.

3. Dedicated transport for the night shift

The night shift arrived late because it depended on the operation's general bus. I coordinated with the operator to assign a dedicated vehicle for the shutdown crew, departing 45 min ahead of the regular bus. Additional cost: zero — the vehicle already existed and needed only a formal assignment and a single communication to the transport coordinator.

That eliminated 6 of the 7 late-start hours.

Consolidated result

On the next scheduled shutdown, actual duration was 49 h against the 48 h plan. One hour of difference is statistical noise. What matters: the operator recovered 35 h of production in a single cycle.

The project manager asked me why we had not done this before. The honest answer: because the Primavera P6 schedule looked fine on screen, and nobody had crossed the real access timestamps against the planned activities. The data existed. It just had not been read.

What you can replicate today

If you manage rotating-equipment shutdowns in mining, check these points before the next cycle:

• Export the schedule and cross every activity against the actual area-access log.
• Identify the implicit buffers that exist in the field but are absent from the plan.
• Map warehouse items with an internal lead time above ???2 h??? and define a pre-staging protocol.
• Assign dedicated transport if the shutdown shift does not align with the general bus.

The hours do not vanish on their own. They are sitting in the data you already have.