Flare Recovery Margin Consumption Rate (FRMCR)

GMI-OG-013Oil and GasIntegration-Required

Definition

Flare Recovery Margin Consumption Rate, or FRMCR, is the percentage of available flare and relief recovery system capacity consumed by AI-directed flare minimization during normal operations.

The Measurement Gap

Hideo is the process safety engineer at a Gulf Coast refinery running a fluid catalytic cracker, three crude distillation trains, and a pair of hydrotreaters. Last year the refinery installed an AI-driven flare minimization system designed to route as much continuous flare gas as possible into the recovery compressors instead of the flare stack, reducing methane emissions and aligning with the operator''s ESG targets. The system has been a public-relations win. The corporate sustainability slide deck quotes a fifty-three percent reduction in routine flaring against the pre-AI baseline. The recovery compressors run hotter and longer than they used to, but they are sized for it and the trends look stable.

The morning of the upset, the FCC riser pressure climbed faster than the operating crew expected. A regenerator slide valve had stuck partway open and the catalyst circulation was running too rich. The unit was tripping toward its emergency depressurization sequence. The flare and relief system is designed for exactly this. When the FCC trips, the relief valves lift and the flare burns whatever the unit can''t hold. The relief design basis is a worst-case scenario where the recovery compressors are off and the entire relief load goes to the stack. That number is on a wall in the safety office. It is what the relief and flare system was sized for.

What the relief and flare system was not sized for was a worst-case scenario where the recovery compressors were already running at eighty-seven percent of rated capacity at the moment the FCC tripped, because the AI flare minimization system had been loading them up across the previous twelve hours of normal operation. When the FCC dumped, the relief load arrived on top of the AI''s normal-operations recovery load. The compressors couldn''t take more. The relief valves lifted to the stack. The flare did its job. But the unit''s control room had no warning that the relief margin going into the upset was thirteen percent of capacity instead of the full hundred percent the original design assumed. The AI had been quietly consuming the safety buffer for a year.

Hideo''s post-incident review traced this carefully. The flare minimization system had been doing exactly what it was asked to do: maximize recovery, minimize routine flare. The recovery compressors had been running inside their rated capacity. The DCS had no alarm on recovery loading because the loading was within rated and was the AI''s normal operating state. The safety case for the flare and relief system, last reviewed four years ago, did not contemplate the AI''s presence. The relief design basis was a worst-case upset; the AI had moved the system''s normal-operations baseline so close to the relief design basis that the upset overlap was the dangerous part.

Legacy emissions and process metrics do not see this. Flare emissions are down. Recovery compressor uptime is normal. Process safety performance indicators show no precursor signals on the FCC. The AI''s optimization metric (recovered MSCF per day) is at record highs. The dashboard reports what the AI was asked to optimize. It does not report how much of the relief margin the AI is consuming to do so, or what relief margin is left for the upset that the relief system exists to handle.

What is missing is a metric that watches the safety buffer the AI is silently spending.

Formula

Flare Recovery Margin Consumption Rate, or FRMCR, is the percentage of available flare and relief recovery system capacity consumed by AI-directed flare minimization during normal operations. The complement of FRMCR is the relief margin available for process upsets. The KPI exists to answer one decision: how much of the safety buffer the relief and flare system was designed to provide is the AI''s emissions-reduction logic currently using up?

The structural form is straightforward. FRMCR equals the actual throughput of the flare gas recovery compressors during normal AI-optimized operations divided by the actual tested capacity of those compressors, expressed as a percentage. The actual tested capacity is the operative number: design basis is a starting point, but every compressor in service has a derate from design due to fouling, wear, ambient conditions, and equipment age. FRMCR computed against rated capacity will systematically understate the consumed margin; the metric has to use a recently tested actual capacity.

The complement, relief margin, is the percentage of recovery capacity available to absorb upset relief load before the flare stack lights off. A facility with FRMCR at sixty percent has forty percent of recovery capacity in reserve. That reserve is the buffer between a routine relief event becoming a flare stack event versus being absorbed quietly by the recovery system. The number that matters operationally is not FRMCR itself; it is the reserve and how it compares to the relief loads the facility''s historical upsets have actually produced.

Proposed formula, pending validation pilots.

At a Glance

Category
Emissions Model Integrity
← Post-Autonomous Health Impact (PAHI)PEMS Drift Detection Index (PDDI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment