Start-Up Is Where the Design Assumptions Stop Working

Almost every plant is designed around a single operating point, and almost nothing that decides whether it works happens at that point. The equipment is sized for steady state. The plant spends a large part of its life getting there and leaving it.

THE DESIGN POINT IS NOT WHERE THE PLANT LIVES

A design basis states a feedstock, a throughput, a temperature, a pressure and a gas composition. Every vessel, fan, filter and heat exchanger is then sized around those numbers.

The plant does not run there. It warms up, it cools down, it turns down when the offtake drops, it trips when a pump fails, and it restarts. Those are the hours in which a plant is either quietly accumulating damage or consuming consumables far faster than the design model predicted.

Steady state is also the easiest condition to model, which is why it gets modelled first. It is the transient that decides most of the maintenance bill.

WHAT ACTUALLY CHANGES

The first change is the temperature profile. A hot vessel is not uniformly hot. The refractory face heats long before the steel shell behind it, and the difference between those two temperatures is a mechanical load. Heat a lining too quickly and it spalls; heat it too slowly and the start-up burns time and fuel. The permitted heating rate is usually set by the lining, not by the process.

The second change is gas velocity. At low load the gas moves slowly, so every square metre of wall has more time to pull heat out of it. Somewhere downstream there is a point that is coldest, and if that point falls below the water or acid dew point, liquid forms on a surface that was specified for gas. Condensate is where corrosion starts, and it is also where particulate starts sticking.

The third change is tar behaviour. A compound that is a gas at full operating temperature is a sticky liquid at half that temperature. During a low-load period the cleaning stages can be asked to handle material that is no longer in the phase the design assumed. Filters load faster. Sorbent beds consume their capacity in a fraction of the expected time.

The fourth change is the control system. Loops are tuned in the band where they were validated. A controller that behaves well at full throughput is frequently sluggish when the load is low, and a sensor near the bottom of its range is least accurate exactly when the control loop depends on it most. During start-up the model has no history to work from, which is the one condition under which a model-driven controller has nothing to offer.

The fifth change is mechanical. Expansion joints, flanged seals and refractory anchors all see differential movement during a thermal cycle that they do not see at steady state. Gas containment problems are usually found here rather than in the vessel body.

WHERE THE COST AND THE DAMAGE LAND

Damage appears later than the event that caused it. A lining cracked during a rapid heat-up fails during the following quarter. Consumables spend their budget early, so the sorbent and catalyst replacement interval assumed in the operating cost is quietly shortened by every cycle.

There is also a permitting consequence worth stating plainly. In many jurisdictions, the permit treats start-up and shutdown differently from normal operation, and continuous monitoring does not count those periods in the same way. That means the hours with the most variable gas quality are also the hours with the least recorded data. Whether that is a design problem or a reporting problem depends on the plant, but it is uncomfortable either way.

WHAT THIS ARGUMENT DOES NOT SOLVE

Transients cannot be designed out. Every plant has to start, and any plant that sells heat or power has to follow a demand that moves.

Designing for them costs money in ways that do not show up in a performance guarantee. Heat-up systems, low-load-stable burners, instrumentation that stays accurate at the bottom of its range, and thicker linings all raise capital cost. Thicker linings also raise thermal inertia, which makes the next start-up slower. A plant optimised for start-up can end up a worse plant at full load, and there is no free version of this trade.

The harder problem is evidence. Instrumentation is usually specified for the design point, so the transient period is often the least measured part of the operating year. A designer who argues that transients dominate the maintenance cost is making a claim that the plant, as instrumented, may not be able to prove. That is a reason to instrument for it, not a reason to distrust the argument, but it does mean the honest position is an argument rather than a measured result.

And the exposure is not uniform. A plant running one consistent feedstock into a steady industrial offtake sees far less thermal cycling than a waste plant following a variable supply. The same design can be adequate in one case and expensive in the other.

QUESTIONS TO ASK BEFORE COMMITTING TO A DESIGN

What is the design heat-up rate, and what sets the limit? If the answer is the refractory specification rather than the process, the lining is controlling your operating flexibility.

At minimum load, is the gas above its dew point at every point downstream? Name the coldest point in the system and say what protects it.

How many start-ups and shutdowns a year are assumed, and where in the operating cost are they counted?

What does the control system do outside the load band it was validated on, and who checks that it still behaves?

Which measurement would show that the plant stayed inside its window during a transient, and who reads it after the fact?

What does the permit say about start-up and shutdown, and are those periods in the compliance record?

Talk to us at zero-x.co