Constraint-Proximity Equipment Life Index (CPELI)

GMI-OG-016Oil and GasIntegration-Required

Definition

Constraint-Proximity Equipment Life Index, or CPELI, is the cumulative percentage of operating time a process unit spends within a defined proximity band of each of its constraint limits under APC optimization, compared against the same measurement on pre-APC baseline operation. The KPI exists to answer one decision: is the AI optimization earning margin today by quietly consuming equipment life that will be paid back as shortened turnaround intervals, accelerated maintenance, and extra unplanned outages?

The Measurement Gap

Stanton is the reliability engineering manager for a midwestern complex refinery: one CDU train, an FCC, two hydrotreaters, and a delayed coker, all tied through a common APC layer that the company installed six years ago and upgraded with a machine-learning optimization module two and a half years ago. The upgrade was sold as a smarter optimizer: same constraints, same operating envelope, but tighter trajectories against the limits and more aggressive use of available margin during favorable market conditions. The first year''s economic case held up. Gross margin against the planning case improved by about thirty-five cents per barrel on the FCC, and the CDU pushed an extra three percent in throughput on the slate the planning team thought was the binding constraint.

The trouble showed up in the maintenance numbers. The next FCC turnaround had been planned for a four-year interval based on the unit''s historical wear and the inspection findings from the previous run. The reliability engineering team began their pre-turnaround inspection planning at year three, expecting to find the usual wear patterns. What the borescope inspections revealed was different. The riser refractory was further along than the team had ever seen at three years. The slide valve seats showed cumulative erosion that matched the four-year-end-of-run baseline from previous turnarounds. The catalyst regenerator cyclones had wear patterns suggesting accelerated solids loading. The unit''s integrity was running ahead of its calendar, and the turnaround was going to need to come forward by at least nine months.

Stanton''s team pulled the operating data and ran a comparison against the pre-APC-upgrade era. The unit''s operating temperatures had not exceeded any limit. No safety system had tripped. No alarms had cleared. But the time-at-proximity-to-limit picture was a different story. Before the optimizer upgrade, the riser outlet temperature spent about eight percent of operating hours within five degrees of its metallurgical limit. After the upgrade, it spent forty-three percent of operating hours in that band. The catalyst circulation rate spent about eleven percent of operating hours within five percent of the slide valve''s erosion threshold before the upgrade and thirty-seven percent after. The combined CDU charge pump differential pressure was running closer to fatigue-curve thresholds than the original design assumed. The unit had not been pushed outside its limits. It had been pushed against its limits, much more often, for much longer. The equipment was wearing on a faster clock as a result.

The economics of the optimizer''s thirty-five cents per barrel of margin improvement had to be netted against the cost of the turnaround coming forward nine months. The CAPEX impact of one extra turnaround across the asset''s remaining life, plus the lost margin during the unplanned outage window between turnarounds, plus the additional inspection and maintenance work the accelerated wear was producing in the between-turnaround period, was a substantial figure. When it was netted against the optimizer''s margin contribution, the program was still positive but by less than half of what the reported margin number suggested.

Legacy refining metrics do not catch this. Unit availability is high: the unit is running. Yield is improved: the optimizer is delivering. Safety performance is clean: no excursions, no trips, no alarms. The standard reliability KPIs report condition and outcomes; they do not report how much margin between operation and limits the unit is consuming continuously as a tax on equipment life.

What is missing is a metric that asks how close to its limits the optimizer is running the unit, and how that compares to the pre-AI baseline that the equipment''s wear life was calibrated against.

Formula

Constraint-Proximity Equipment Life Index, or CPELI, is the cumulative percentage of operating time a process unit spends within a defined proximity band of each of its constraint limits under APC optimization, compared against the same measurement on pre-APC baseline operation. The KPI exists to answer one decision: is the AI optimization earning margin today by quietly consuming equipment life that will be paid back as shortened turnaround intervals, accelerated maintenance, and extra unplanned outages?

The structural form is constraint-by-constraint. For each constraint limit on the unit (temperature limit on a reactor outlet, pressure limit on a vessel, throughput limit on a pump, catalyst activity limit on a reactor bed, erosion-rate-based throughput limit on a slide valve) define a proximity band as a percentage of the limit. Starting recommendation: ninety to one hundred percent of the constraint limit. Measure the percentage of operating time the parameter falls inside that band. That is CPELI on that constraint. Compute the delta against the pre-APC baseline. CPELI_delta is the metric that drives action; the absolute CPELI is the supporting evidence.

The constraint set is what determines whether the metric is honest. Every process unit has many constraints: design constraints, safety constraints, operational constraints, control constraints. CPELI should be computed on the constraints whose proximity affects equipment life: temperature constraints tied to metallurgical degradation, pressure constraints tied to vessel fatigue, throughput constraints tied to erosive wear, catalyst activity constraints tied to coking and selectivity, fouling-driven constraints on heat exchangers. The unit''s reliability engineering function defines the constraint set, supported by the original equipment design basis, vendor recommendations, and the unit''s own degradation history. Constraints that exist for control or safety reasons but do not affect equipment life (instrument range constraints, control loop saturation constraints) are excluded.

The proximity band is the second judgment call. A wide band (eighty percent of limit and up) produces a metric that captures a lot of activity but loses sensitivity to truly aggressive optimization. A narrow band (ninety-five percent of limit and up) captures only the most aggressive operation but may miss meaningful drift. Starting recommendation: ninety to one hundred percent on most constraints, with the option to define a narrower band for specific constraints where equipment vendors have provided guidance about the wear acceleration that occurs above particular thresholds.

The baseline period is the third judgment call. The pre-APC-upgrade era is the right comparison if it exists and is recent enough that operating practice was otherwise similar. The pre-AI human-operator era is acceptable if the optimizer upgrade is recent. The challenge is finding a baseline period long enough to be statistically meaningful (at least three months of stable operation, ideally six) without going so far back that other operational changes confound the comparison.

Proposed formula, pending validation pilots.

At a Glance

Category
Refining & Process Optimization Trade-offs
← Shut-In Restart Cost Index (SIRCI)Crude Blend Cross-Functional Cost Index (CBCCI) →

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