I was recently chatting with a number of experienced reliability practitioners that I’ve had the opportunity to come across in the industry.
They have been delivering projects using TARO (DNV’s downstream RAM tool) for many years. We discussed what makes a good RAM study and what an organization has to do to make a success of RAM analysis. Quoting what one of them said:
“RAM modelling doesn’t give magic answers like a crystal ball. To do RAM modelling you need to do some other things to understand how your systems behave.
Organizations preparing to do RAM analysis need to be tracking failures, failure causes, and failure impacts. It is not enough to only react to failures, do some Root Cause Failure Analysis and then move on.”
This is great advice!
With that in mind, I’ve decided to write down a few steps on how to perform an effective RAM Study. Please do not hesitate to step-in, comment or add stages that you believe are important:
Collate/treat/understand all required data
This step is fundamental and if not done properly, the study will not be effective at all. Prior to considering any RAM model, the analyst must collate required data. If the asset’s data is not available (for example, a new process plant), there are a number of data sources available in the market these days.
Once the asset is operational, the RAM analyst should start collecting the reliability data in a standardized manner. After enough data is collected, the RAM model could be re-run using the asset’s data and a compared to what was expected. My personal tip to you is that if you do benchmark your asset’s performance, publish the results and share with the industry so that we can all learn.
Define the objectives and scope
First of all – do not expect to populate a RAM model with all available data and an answer to fall out. The input data and model must be targeted for the study at hand and all unrelated data must be eliminated. For example, when evaluating the performance of different light bulbs we need to consider the reliability of the light bulb itself, but we need not consider the availability of power supply. Although the power supply does impact the performance of the light bulb’s ability to give light, its impact will be the same for all the bulbs evaluated so need not be included.
Define boundaries for the study
Define the analysis scope – it is extremely important to include all critical items but also important to not put effort into modelling items that are outside the sphere of influence. One failure mode might be enough to represent a system when information is not available and we would like to capture its impact on the overall performance.
Should I model a safety system? Safety systems may have direct impact on production; a unit may be shut down due to safety equipment behaviour. This might be true for an emergency shutdown valve which is designed to fail closed. This might not be true for a fire fighting water system which can be thought of as independent and parallel to the process, however even in this case, a matrix of permitted operations may not allow you to operate a unit if fire water is not available. (This is handled very well by the operability capabilities of Maros and Taro and their conditional elements!) Whether you model the safety system as part of your RAM study depends on the relationship to your process.
Prepare a study basis
What are the objectives of the analysis? This will impact on many stages of your analysis, even on the construction of the Reliability Block Diagrams (RBDs).
Register all data – the studies’ stakeholders should and will most definitely come back to you asking: “where did you get this data from?”
Compare overall results with expectations
Avoid “Garbage in – Gospel Out” by using comparative evaluation from previous work or your gut feeling. Typically, while preparing a RAM model, the RAM analyst gets in touch with engineers with vast experience on running process plants. They will know what to expect from a specific unit.
In addition, check the low criticality items too in case you have missed anything. Perhaps, something you expect to be there that might not be appearing?
Get married to your RAM model
The RAM model for a specific plant is something for life (call me romantic!). Do not divorce your RAM model as it can be used throughout the entire life-cycle of your asset’s life, increasing the benefits and value of your time and efforts.
Author: Victor Borges