Reliability Modelling

I have been working full on until New Year building component library for a consultancy and running reliability modelling. I have run into a lot of issues with the model and would like to share them to promote an understanding.

The very first thing you do as an engineer is always question the validity of those data you acquire. In summary, I would not recommend doing reliability modelling. My personal opinion is, it is a waste of time, effort and money. If you are looking at doing reliability modelling, chances are, your existing plant reliability is not great and your reliability knowledge is not comprehensive. For the accuracy you get, you are better off with a 0.9 factor of industry average reliability figure. No modelling out there I have seen is accurate enough for any good use. If you are doubtful on the quality of people you are able to hire into the maintenance team, use a factor of 0.7 and you should have a somewhat conservative availability figure. Yes it looks ugly, yes it looks unrealistically low, but I'm sorry to say, that is reality of the availability and reliability figure you should expect for saving cost hiring cheap people. I could not emphasize enough, good asset management and reliability starts with good people.

Back onto the topic I was suppose to be writing about - Limitations of reliability model. Firstly, ALL reliability model I have seen is designed in series. It is all well and good if your process is in series like a simple production line of a simple mine site, if you have a complicated processing plant, your reliability model will not do. In fact, there's so much work trying to design the model to fit your plant, it is just not worth the effort. Unless there's a free template already setup similar to your plant and takes just a little bit of effort to patch up, there's no point going down this path.

Secondly, in a complex process plant, you will have varying equipment MTBF and MTTR. Every plant's figure is unique. For the model to be accurate enough to be of any use, it has to be from your plant, your production forecast, you historical availability, and reliability. This is because as the errors build up in the reliability model, the final result is again of not much use to you as the owner.

Thirdly, a complex equipment in a complex plant will have a long list of failure modes to prevent. Some of these failure modes will be attended to in one work task and reset their likelihood of occurrence and budgeted life. None of the reliability model I seen cater for this.

With this three fundamental issues in modelling unresolved, I would not recommend any company looking at carrying out the modelling without understanding the limitations of it.