When the spread is this high, every refiner wants more barrels through the FCC, the crude unit, or shorter coker drum cycles.
The Trading Economics Crack Spread index, which tracks margins for the average US refinery, hit a record $75 per barrel in September 2026, up nearly 200% since mid-February.
The real question isn’t whether you want to push rates or defer a turnaround, it’s whether you have the mechanical analysis to show it can be done safely. Both problems trace back to the same root cause: the engineering answer usually takes longer to get than the margin window lasts. With today’s computing power and AI, that answer can now come back in seconds instead of weeks.
1. Push throughput on constrained units with the data to back it up
With US refiners already running above 100% of nameplate capacity, the constrained units set the ceiling on how much of that margin a site can capture.
This is exactly the problem a client faced with its coke drums, which undergo persistent thermal cycling. Traditional FEA analysis can take a month or more, and without a clear view of the coker’s real structural health, potential production gains were going unclaimed. The plant was operating on conservative estimates rather than known limits. Using a live structural twin, the team uncovered more available structural performance during specific thermal cycles, safely intensified coker throughput, and quantified the impact of faster preheating on cycle variability. All without new infrastructure or a fresh FEA study every time they wanted to test a change.
That’s the model: instead of commissioning a new study every time you want to test a rate increase, engineers run throughput, pressure, and temperature scenarios against the asset’s real fatigue and stress limits and get answers in seconds, because the structural twin is already built, already validated, and already live with current sensor and inspection data.
2. Defer the turnaround that would take a high-margin unit offline safely
The flip side of pushing rates is protecting the units you already have running. A scheduled turnaround doesn’t know or care what the crack spread is doing. But integrity assessments go stale the moment they’re written, and a lot can change on a coker or a fired heater between the last inspection and the date a T/A was originally scheduled.
This is the same reasoning behind a live fatigue-tracking approach: rather than relying on periodic, static assessments, that structural twin stays continuously updated with as-maintained inspection data, applying API 579 fatigue assessment procedures to calculate accumulated damage and predict remaining life on an ongoing basis.
That’s the difference between deferring a turnaround because “we think it’ll be fine” and deferring it because current data says it will be, with an engineering trail behind the decision. It doesn’t mean pushing every T/A indefinitely. It means making that call with real information instead of a stale inspection analysis, and capturing more of the current margin window before the unit has to come down.
How Akselos closes both gaps
Both scenarios come down to the same constraint: speed of trustworthy engineering analysis. Akselos’s structural twin closes it, solving full-asset-scale structural models in seconds instead of weeks, using data that’s already live rather than a snapshot from the last scheduled study.
For scenario 1: engineers test rate increases against the coker’s real fatigue and stress limits in seconds, because the twin is already built, validated, and live. No new FEA study for every throughput test.
For scenario 2: the twin stays continuously updated with as-maintained inspection data, running API 579 damage and remaining-life calculations on an ongoing basis, so the deferral decision is backed by current numbers, not a stale inspection.
The question is whether your engineering infrastructure lets you capture high margin spikes, or you just watch it pass by while the study gets scoped.