How to modernise a plant that cannot stop

In the appraisals we have run over the last three years, the factor that most often ruled out a technically sound option was not the capital cost. It was the downtime it required. A process plant that bills by the hour has an implicit price for every hour it is stopped, and that number is usually larger than the difference between the two options being compared.

The usual mistake is to treat the window as something you work out at the end: design the solution, calculate what stopping costs, and then start negotiating. Reversing that order changes the outcome of the project.

1. Define the window before the solution

The first question in an upgrade project should not be which technology suits, but how many consecutive hours the asset can be down and how far in advance the shutdown can be scheduled. With those two figures on the table, a good part of the catalogue of options disappears before an hour of engineering is spent on it.

At one of the plants we appraised, the maximum window was seventy-two hours once a year. Put on the table at the start, that constraint led straight to a modular solution which, in a scenario with no shutdown limit, would have been dearer and less elegant — and which was the only one that could actually be built.

2. Move work outside the window

Anything that can be built, wired, programmed and tested in the workshop reduces the risk of the shutdown twice over: it shortens the work on site and moves the mistakes somewhere fixing them costs no production. A prefabricated module that arrives tested turns weeks of installation into days of connection.

The same goes for work that can be done with the installation live or running. On the extension of a 138 kV substation, 84 % of the man-hours were worked with the substation in service; only what required an outage reached the night window.

3. Rehearse the sequence, do not just plan it

A switching plan reviewed in a meeting room and a plan rehearsed by the people who will carry it out are two different documents. The rehearsal — on a simulator, on a mock-up, or dry on the installation — is where the missing steps turn up, along with the keys nobody had and the two operations somebody had assumed were simultaneous.

• Rehearse with the people who will do it, not the people who planned it
• Time every step and add up the real durations, not the estimated ones
• Define the point of no return, and the criterion for aborting before you reach it
• Prepare the rollback in the same detail as the plan itself

4. Accept that the window may not be enough

The honest conclusion of some studies is that the work does not fit. When that happens there are three ways out: split it across several windows with a stable intermediate configuration, build temporary redundant capacity, or accept a longer shutdown and plan it a year ahead. All three are expensive. Finding out during the shutdown is far more so.

The cost of an unplanned shutdown is not measured against the cost of the works. It is measured against the production that never happened.

Upgrading an asset in service is a sequencing problem before it is a technical one. The projects that go well were almost always decided in the first few weeks, when somebody asked how much time there was and designed backwards from the answer.