Preventive maintenance plan: building it from real risk

Risk-based maintenance means putting the effort where a failure would cost the most, and easing off where a failure would cost almost nothing. Put like that, the idea sounds obvious. Yet it is rarely applied, because it assumes two uncomfortable things: accepting that you will do less on certain equipment, and having data on the real condition of that equipment.

What the approach replaces

On most sites, the maintenance plan has built up by sedimentation. One interval comes from the manufacturer, another from a regulatory requirement, a third from an incident that happened twelve years ago, a fourth from a habit whose origin nobody remembers any more. The whole thing works, but it is not prioritised: it gives the same attention to an equipment item whose stoppage blocks the line and to one whose failure is repaired in an hour with no consequence.

Moving to a risk basis does not mean redoing everything. It means laying over this existing plan a way of reading it that answers a single question: is the effort in the right place?

The answer is almost always the same. Part of the plant is over-monitored because it is easy to access and well documented. Another part is under-monitored because it is hard to reach, poorly documented, or simply forgotten. Rebalancing produces a gain without increasing the overall workload.

The two axes, and how to establish them

The severity of consequences

This is the axis that is simplest to establish and the most stable over time. It does not depend on the condition of the equipment but on its place in the installation.

Four families of consequence are worth assessing separately, because they do not answer to the same arbiters:

• The safety of people: it comes first, and it is not to be traded off against a gain in availability.
• The environment: releases, containment, the consequences of a loss of containment.
• Production: line downtime, product loss, restart.
• Compliance: the ability to respond in an audit, regulatory deadlines, keeping a certification in force.

This work is done in a meeting, with the production manager, the safety manager and the quality manager. It takes half a day per line and stays valid for several years, as long as the process does not change. Its practical construction is set out in building a criticality matrix.

The probability of failure

This is the axis that causes initiatives to fail, because it calls for data that nobody has to hand.

Probability is not estimated by guesswork. It is built from three sources: the history of failures that have already occurred, the condition observed during inspections, and the measured rate of degradation. The last two sit in your inspection reports, locked away in PDFs, which is why they are so rarely put to use. See centralise inspection reports and track equipment degradation.

Building the approach, step by step

What makes it fail

• A matrix built on opinions.Without a condition history, the probability axis reflects the perception of the participants, often dominated by the last memorable incident.
• Too wide a scope at the outset.A study of the whole site produces an impressive document that leads to no dated decision.
• No easing-off decided.If the approach only adds checks, the workload grows, the team disengages and the plan reverts to its earlier state.
• A manual update.Anything that depends on an annual review disappears at the first busy year.
• Forgetting compliance.An equipment item that is minor for production can be decisive in an audit; treating compliance as a consequence in its own right avoids surprises.
• An approach carried by maintenance alone.The trade-offs commit production and safety; without them, the decisions do not hold.

What the approach produces in practice

A risk-based strategy does not produce a matrix: it produces dated decisions.

Table scrolls horizontally on small screens.

The last line is the one most often forgotten, and yet it is the most honest. An equipment item about which you know nothing is not a low-risk item: it is one whose risk has not been assessed. Confusing it with the first is the most expensive mistake in the whole exercise.

Fitting risk together with regulatory obligations

This is the question that comes up in every meeting, and it deserves a clear answer: a risk-based approach does not reduce what is imposed on you.

Regulatory deadlines (periodic inspections, requalifications, checks required by the site's quality framework) form a floor. They apply regardless of your criticality analysis, and a matrix does not shift them.

What the approach organises is everything that sits above that floor: the checks you decided on yourself, the intervals you set for yourself, the additional inspections inherited from an old incident. On most sites, this voluntary part represents the majority of the inspection effort, and it is the part that has never been prioritised.

There is, however, a useful connection in the other direction. A well-kept history does not change the imposed interval, but it changes the quality of the case you present on the day of the deadline. Turning up in front of a certifying body or an auditor with a documented trend, tied back to its source reports, puts the discussion on a different footing than turning up with a pile of PDFs. This point is covered in preparing a technical or regulatory audit.

What the approach really costs

You have to be honest about the investment, because that is where projects are decided.

The main cost is not the tool: it is the time of the people who know the installation. Establishing severity draws on several functions for half a day per line. Reworking the history calls for a judgement on equipment tags that no one but your own teams can make. These two items cannot be compressed and they govern everything else.

The secondary cost is the data rework itself. This is the one that held these initiatives back for a long time, because it meant manually re-keying the content of hundreds of reports, work that nobody funds. It is the only item that automated reading genuinely brings down.

The gain, for its part, rarely shows up where it is promised. It appears first in the preparation of turnarounds and audits, where the time spent gathering documents falls markedly. It then appears in budget trade-offs, because a request backed by a measured trend defends itself better than one backed by a conviction. The effect on the number of unplanned stoppages comes later, and it is measured over several years. Claiming the opposite is the surest way to discredit the approach at the first review.

What a tool contributes, and its limits

Software does not establish severity: that is human work, contextual, engaging several functions. What a tool contributes is the upkeep of the probability axis over time: taking in inspection reports as they arrive, recalculating trends, and flagging the equipment whose condition is degrading faster than expected.

That is what tools for automatically reworking existing reports allow, with no re-keying and, in the best cases, without the documents ever leaving the site's network. The operational prioritisation that follows from it is covered in prioritise maintenance by real risk, and the effect on unplanned stoppages in reducing unplanned downtime.

For the support side of the work, see our Maintenance Intelligence pages and our AI strategy for maintenance. If your teams need to build up their skills in these methods, the AI training for maintenance managers works through these subjects on your own cases.