Fourteen hydraulic systems, one rising iron trend, and a pump bearing caught three weeks before it grenaded.

The smelter has run a quarterly oil-condition program with our Jebel Ali lab for the past four quarters. Fourteen hydraulic systems — cast-house tilting cylinders, pot-room rectifier cooling skids, anode-handling press HPUs — sampled on-site by an Indus engineer, shipped to the lab, screened for cleanliness and wear. The standing instruction from the operator’s reliability team is short: "tell us what is wrong before it stops the line." A cast-house line stop on a smelter is measured in molten metal and missed shipment slots, not in hours.

When the Q1 2026 round of bottles landed in the intake bay on a Wednesday morning, sample #7 — the anode-handling press HPU — looked unremarkable on the first pass. Visual: clear, no haze, no free water. Viscosity at 40 °C: in spec for ISO VG 46. Water by Karl Fischer: 96 ppm, well under the 200 ppm action ceiling. ISO 4406 cleanliness coding came in at 20/18/15 against a target of 20/18/15 — green by every standard metric the operator tracks on their CMMS dashboard. On paper, sample #7 was a non-event.

This is the case for not buying oil analysis on cleanliness coding alone. ISO 4406 tells you how dirty the oil is. It does not tell you what is wearing. A pump can be days from a bearing failure with cleanliness still inside spec — because the wear particles being shed are mostly in the sub-4 µm range that ISO 4406 does not count, and because a working filter is pulling the larger particles out faster than the bearing is shedding them. You need ferrography and SOA to see the trend before the filter loses the race.

The lab issued a single-page PDF verdict against sample #7 on day 5 after intake. Cleanliness coding: green at 20/18/15. Water content: green. Viscosity: green. Ferrography: amber — rising bearing-wear signature, recommend pump-down inspection within 30 days. The SOA panel was attached as a trend chart over the four prior quarters showing iron at 11 / 14 / 17 / 38 ppm and chromium at 2 / 3 / 4 / 9 ppm — both crossing the 2× baseline trigger that fires an amber flag. Verdict went to the smelter’s reliability lead before lunch.

The operator pulled the pump on day 23 — 18 days after verdict release. The shop teardown report came back the same week. The main pump bearing race was 75% spalled, with the cage starting to deform. The pump would have seized inside the next four to six weeks of running. A seized main pump on an anode-handling press HPU does not stop politely — it grenades, takes the pump body and the discharge line with it, and on a smelter line that runs three-shift seven days, replacing it under emergency mobilisation runs into a replacement HPU plus a multi-day production hold.

Cleanliness coding tells you how dirty the oil is. Ferrography tells you what is wearing. The two are not the same question. A pump can be three weeks from a bearing failure with ISO 4406 still green — because the bearing is shedding sub-4 µm steel faster than your code can see it. Oil tells you what is failing before the equipment does, but only if you are reading the right slide.— Omar Al Maktoum · Ferrography Lead · Jebel Ali

Quarterly oil-condition rounds at the smelter follow the same four phases every quarter. The first two are time-boxed by the on-site sampling window; the last two are run in the lab on a strict 5-day per-sample TAT clock from intake to verdict.

Below is the working subset of SOP-LAB-007 — our standing procedure for a quarterly oil-condition round running ISO 4406 cleanliness coding plus ferrography plus SOA. The full revision runs 31 lines; the 22 below are what was ticked off and signed against the Q1 2026 round.

Below is the actual Q1 2026 round result against ISO 4406 target codes, water-content ceilings and the four-quarter rolling baseline on the SOA panel. Highlighted findings are the four that did not pass cleanly — three amber, one red on the wear-trend column for system #7.

Pump came down on day 23 — 18 days after our verdict landed on the operator’s desk. The teardown shop confirmed what the ferrogram had said: the main pump bearing race was 75% spalled, the cage was starting to deform, the pump was inside the four-to-six-week window before a hard seizure. Replacement parts were ordered through the operator’s standard procurement window, the pump-down was scheduled around a planned production maintenance slot, and the line never stopped. Estimated savings against the unplanned-failure case: ~AED 800K in avoided unplanned downtime and an emergency replacement HPU. System #7 has been moved to a 14-day monitoring cycle in the operator’s CMMS while the rebuilt pump beds in.