Two studies, one goal
RAM - reliability, availability and maintainability - predicts how much a plant will actually produce once failures and maintenance are accounted for. RCM - reliability centred maintenance - decides what maintenance to do so that prediction holds.
They are sold and delivered together because one without the other is incomplete: a RAM model with no maintenance strategy is a forecast nobody owns, and an RCM programme with no availability model has no way to prove it was worth the cost.
RAM modelling
- Build the reliability block diagram from the process flow, capturing redundancy and sparing
- Assign failure and repair data from plant history where it exists, and from OREDA or equivalent where it does not
- Monte Carlo simulation over the plant life to predict availability and production loss
- Identify the bottleneck equipment - typically a small number of items driving most of the loss
- Test the value of options: a spare pump, a larger intermediate tank, a different turnaround interval
The value is in the ranking. Availability spend is routinely directed at the equipment that fails most often rather than the equipment whose failure costs most, and those are seldom the same list.
RCM strategy
- Define the function of each asset and what constitutes functional failure
- Identify failure modes and their consequences - safety, environmental, operational or economic
- Select the maintenance task by consequence: condition-based, scheduled restoration, scheduled discard, failure-finding, or run to failure where that is genuinely the right answer
- Set intervals from the failure characteristic rather than from convention
Run to failure is a valid strategy
RCM is often assumed to mean more maintenance. Frequently it means less. For assets whose failure is cheap, obvious and safe, planned maintenance consumes resource that a critical asset needed - and time-based overhaul of an item with no wear-out pattern introduces infant-mortality failures that would not otherwise have occurred.
Frequently asked questions
What data do we need for a RAM study?
Plant failure and repair history if you have it. Where records are thin we use OREDA or equivalent industry data, and the model documents which figures came from where so the result can be challenged sensibly.
What availability improvement is realistic?
It depends entirely on the starting point and the bottleneck. The honest answer is that the study identifies where improvement is available and what it costs - anyone quoting a percentage before modelling is guessing.
How does RCM relate to our existing PM schedule?
We start from it. Most schedules are inherited from vendor manuals and never revisited. RCM usually removes some tasks, adds a few condition-based ones and changes several intervals.
Does this connect to functional safety?
Yes. Failure-finding intervals for safety instrumented functions come out of the same analysis as SIL verification proof-test intervals, and they should not be set independently.