Hello,
a customer is asking the following " I m currently working on a project related to reducing stock volumes. In order to do so, we would like to estimate the reliability of a certain process. On one hand we want enough stock to make sure that we can keep the second process running if the first process has an unexpected shutdown, on the other the cost of a lot of stock is huge. Question is if JMP would be able to help me with the simulations…. Many of these components are in series, if one fails the entire plant is in shutdown. On top of that, also operators can make mistakes what would cause production errors and thus no end product. As soon as the plant is in shutdown, the buffer volume after the process will be used so that the next process can continue. Let’s say maintenance takes 8h to replace the equipment, a time frame we can cover with the buffer tank. After fixing that pump, the process needs several days to get back to the 8h stock level. If within these days a second failure occurs, we don’t have sufficient buffer volume to operate the process. à also the second process will go in shut down à financial loss. The question we have is thus how big the intended buffer volume should be… ".
As a matter of fact process equipements are mixer and pump, we can model failure using RBD and make simulation using RSS. However I'm wondering 1) if minimal buffer volume could be part of the RBD/RSS model and how to or 2) has to be estimated based on the output of a sub-process RBD/RSS ? 3) Can human error be part of model(s) ?
I found great material on RBD/RSS here: Reliability Analysis Using JMP® Pro - JMP User Community, J38 JMP Pro RBD Standby Extensions (Wright) - JMP User Community , J39 System Reliability in JMP Pro RBD (Wise) - JMP User Community
Is there any improvement in RBD/RSS since JMP 12 I should be aware off ? more recent material than the one listed above I should look at ? In case the conversation becomes deep anyone can help on this topic ?
Thank you
Emmanuel
Hi Emmanuel! RBD and RSS do come close to being useful for your customer, but I think they are really needing "discrete event simulation software." As you point out, handling a depleting buffer volume is something that I have not seen in RSS. (It might be possible, I just haven't seen it done.) It does seem to me that human error could be modeled as a reliability "block" in RSS.
Others may have more experience with this, but if you need more assistance, I would be glad to help.
Jerry
Hi Emmanuel,
The case can be solved in the following way.
- Assume the buffer tank has unlimited storage.
- Construct and run the simulation. And analyze up/down time of each simulation as follows:
- The time segments are either uptime or downtime (due to failure)
- Uptime is refill time, downtime is depleting time, assuming both rates are constants. The volume in the buffer tank is a piecewise linear function, assuming a huge initial value.
- One needs to construct that function and record the minimum in one simulation.
- Do that for all simulations.
- The difference between the huge initial value and the global minimum is the solution to the problem.
- But this is simulation, the solution is random. One may need to analyze individual solutions to see whether there is a need to extrapolate.
