The MTIS Review of Historical Failures takes a published failure investigation, summarises what the original investigators found, and then shows what the MTIS Failure Investigation Assistant made of the same evidence, including where it went wrong. HF02 comes from a Saudi Aramco paper published in Hydrocarbon Processing in August 2026.
What happened
Two carbon steel elbows from the overhead system of an atmospheric crude distillation column were examined after seven years in service. The overhead ran at about 139 °C and 15.5 psi, carrying light hydrocarbons, with wash water and chemical injection upstream of the condensers.
- The 48-in. elbow near the top of the column had thinned locally along the path of the neutraliser amine injection, losing about 2 mm in two years.
- The 36-in. elbow at the inlet to the overhead condensers had thinned all round its circumference. At the worst point 4.95 mm of wall remained from a nominal 11.8 mm, a loss of about 59%.
What the investigators found
For the 36-in. elbow, the corrosion product read about pH 3 and contained up to 7.2 wt% chloride with no nitrogen. The missing nitrogen matters: it rules out ammonium chloride salt corrosion, the other usual suspect in a crude overhead. The steel had a normal ferrite-pearlite microstructure, so the material behaved as carbon steel should in acid condensate.
The authors put the thinning down to hydrochloric acid attack caused by poor chloride control. The process records showed why:
- wash water ran at 50 to 100 gpm, a ratio of roughly 1 to 2.5%, against the 4 to 6% the authors cite as recommended;
- the margin between operating temperature and water dew point fell below the recommended minimum of 15 °C on many occasions;
- the carbon steel wash water quill was not inserted towards the centre of the pipe.
The 48-in. elbow failed differently. The neutraliser amine stream was hitting the wall and causing erosion-corrosion. Its injection quill sat in the elbow itself, was inserted to 11% of the pipe diameter against a recommended 30 to 35%, discharged through a slit, and had no atomising steam.
What it means for asset owners
When an overhead line thins, a common first response is to upgrade the alloy. In this case the steel was normal, and both elbows traced back to operating settings and injection design. A metallurgical examination that stops at the sample will confirm the damage mechanism and miss the cause. Process records, injection details and chemical treatment history belong in the evidence from the start.
What an engineer should check
- Is the wash water rate measured and compared with the recommended ratio, or assumed from the design?
- Is the dew-point margin trended from operating data, so excursions below the minimum are seen?
- For every injection point: quill location, insertion depth, discharge direction and atomisation.
- On any removed section, are pH, chloride and nitrogen measured on the corrosion product, to separate HCl attack from ammonium chloride corrosion?
- Is the removed section kept intact for examination? See what to preserve after a failure.
How the MTIS Failure Investigation Assistant handled this case
Test on a published case, not an accuracy claim.
MTIS built a test case for the 36-in. elbow only, from our own summary of the paper. The incident report was written as a plant engineer would write it and left out chloride, pH, acid and dew point. We then gave the app the evidence in three steps. One run, engine version 87ba1de277, 6 October 2026.
| Evidence given to the app | App's leading answer | Score | Rank of hydrochloric acid corrosion |
|---|---|---|---|
| Incident report only | Hydrochloric acid corrosion | 11.1 / 100 (low) | 1st (runner-up 5.4) |
| + laboratory report and operating-data review | Uniform corrosion | 45.5 / 100 (moderate) | 2nd (40.1) |
| + answers to the app's own questions (14 asked, 3 answerable from the case) | Hydrochloric acid corrosion | 88.2 / 100 (high) | 1st (runner-up 54.9) |
What it missed or could not do
- The day-one lead scored 11 out of 100. That is a weak signal, not a diagnosis, and the incident report names a crude column overhead, which is a strong clue in itself.
- Adding the laboratory report made the ranking worse. The general "uniform corrosion" mode overtook hydrochloric acid corrosion by about 5 points. The app's follow-up questions put it back in front.
- 11 of the 14 questions could not be answered from the published case (flow velocity, welds and microbial activity among them) and were left as unknown.
- The app's root-cause step was not run. The root causes above (wash water rate, dew-point margin, injection design) come from the paper, not from the app.
- MTIS wrote the case knowing the answer. This is one case, one run.
Source. T. A. Al-Ghamdi, Y. A. Aldossary, K. F. Al-Qahtani and E. S. Al-Zahrani (Saudi Aramco), "Metallurgical investigation of crude distillation overhead system: A case study", Hydrocarbon Processing, August 2026. Read the original article. Figures quoted are the authors'. This article summarises the source in our own words; see the original for full details. It is general technical information, not advice on a specific asset.
Related reading
- How a corrosion failure investigation works
- Root-cause analysis in materials failures
- Materials selection for corrosive environments
- MTIS corrosion and integrity services
Follow MTIS on LinkedIn for new failure reviews and industry briefings.