Study: Fuzzy Logic Corrects Four SIL Misallocations at Ammonia Plant

Fuzzy logic has corrected four safety-integrity misallocations at a large ammonia plant, Nature reported this week.

A peer-reviewed study built around the plant’s primary reformer — part of a synthesis-gas unit rated at 3,000 tonnes per day — benchmarked a new framework against the original licensor’s assessment across 13 safety instrumented functions.

The stakes sit in three letters: SIL, or Safety Integrity Level. It is the reliability grade assigned to a Safety Instrumented Function (SIF) — the automatic system that shuts a process down when sensors detect it sliding toward danger. An under-rated SIL, the study reports, means a safeguard engineered to a lower reliability standard than the hazard actually demands.

That gap opens up exactly where the industry’s standard tool is weakest. Technology licensors apply the conventional Risk Graph method widely for SIL allocation, and it can return ambiguous or outright erroneous results when one hazard can be triggered by several independent failures rather than a single initiating event — the finding at the center of this case.

Three Findings in Plain Terms

  • SIL is a reliability grade, not a formality. The higher the level, the more dependable the automatic shutdown must be; SIL 3 carries a stricter reliability target than SIL 2.
  • Many-trigger hazards are where the old method misfires. In the study plant, two safeguards facing seven and eight separate initiating events were allocated SIL 2 when the framework showed SIL 3 was warranted.
  • The new method is calibrated, not radical. It reproduces the conventional answers at the centre of each risk category and diverges only where the traditional graph is ambiguous.

Inside the Synthesis-Gas Unit

The case study covers a 3,000-tonne-per-day ammonia plant and homes in on its synthesis-gas unit, where the primary reformer sits. Thirteen SIFs were assessed. Each was benchmarked against the baseline the technology licensor had originally allocated, Nature reported.

The verdict: the conventional method misallocated SIL in four critical cases. Two are spelled out in the findings — underestimation from SIL 2 to SIL 3 for two SIFs exposed to seven and eight initiating events respectively. The report does not detail the remaining two misallocations.

How the Fuzzy Risk Graph Works

The framework takes the probability intervals that feed a Risk Graph and fuzzifies them. Instead of forcing a number into one crisp category, it lets it belong partially to several at once, the study explains. Those graded inputs then pass through a Mamdani inference system — a rule-based reasoning method that weighs the inputs the way a seasoned assessor would, but in a repeatable, mathematical way.

Two design choices matter. The system is calibrated to reproduce conventional SIL assignments at the centres of each category, so it agrees with the old method wherever the old method is clear. And it explicitly aggregates demand rates — the frequency with which a safeguard is called on to act — across multiple independent initiating events, rather than leaving that to qualitative judgment.

Where the Licensor Method Fell Short

The four misallocations cluster where the study says qualitative methods fail: multi-initiating-event scenarios. A safeguard facing eight distinct ways to be challenged needs a higher integrity level than one facing a single failure mode, and the traditional graph’s coarse categories blur that distinction.

The reach of this matters beyond one facility. Because the conventional Risk Graph is the tool technology licensors use for SIL allocation, its failure modes can propagate through plant designs worldwide.

Ammonia gives the stakes physical weight: it is toxic and corrosive at high concentrations, and a primary reformer sits at the heart of a high-temperature, high-pressure process where an under-rated shutdown function leaves less margin between a deviation and a release.

The study’s contribution is to quantify where the standard assessment under-protects — not to assert that the plant in question was unsafe in operation. Nature’s report does not state the study’s funding arrangements.

When Reassessment Is Warranted

Per the study’s findings, the exposure concentrates in scenarios with multiple independent initiating events — the documented failures involved safeguards exposed to seven and eight such triggers. Operators relying on licensor Risk Graph assessments for hazards with many separate causes may, on this evidence, be carrying safeguards allocated one integrity level too low.

The remedy the paper validates is quantitative, not procedural: fuzzified probability intervals, a Mamdani inference core, and explicit aggregation of multi-event demand rates, benchmarked against the licensor baseline rather than replacing it wholesale.

Where This Leaves Plant Safety

The case study spans one plant and 13 functions, and the framework’s credibility rests on its calibration: it defers to the conventional method wherever that method is unambiguous. Its divergence is confined to the multi-trigger scenarios the study identifies as the traditional graph’s blind spot.

Nature, the journal carrying the study, is itself part of the story’s frame. Founded in the autumn of 1869 by Norman Lockyer and Alexander MacMillan as a public forum for scientific innovations, the British weekly publishes peer-reviewed research across the natural sciences and now operates editorial offices in London, the United States, continental Europe and Asia under Springer Nature.

Fewer than 8% of submitted papers are accepted, and the journal’s 2022 impact factor of 50.5 places it among the highest of multidisciplinary science journals.

The study concludes that the fuzzy approach provides a more accurate, discriminative and uncertainty-aware basis for SIL determination in complex multi-initiating-event scenarios — precisely where traditional qualitative methods fail.

References

  • “Fuzzy logic-based safety integrity level determination for multi-initiating-event scenarios: a case study of an ammonia primary reformer,” Nature.
Photo of author

Health Desk

Health Desk at Archyde.

Princess Aiko Wears Orchid-Motif Kimono for Singapore Visit on 60th Anniversary