Proper MTBF calculation is a complex subject. In general it requires the design to be thoroughly analysed, giving each component a failure probability factor and multiplying by the quantity of each component to arrive at a total for the whole design.
To make this assessment easier, for electronic circuits the easy way is to use the established military standard MIL-STD-217-E (or earlier versions)to get the predicted failure rates for each type of component. You can then produce a predicted MTBF figure of x failures per y hours, etc.
I am not quite sure what you mean by "response time". Usually a mean time to repair (MTTR) figure is estimated, then MTBF-MTTR divided by the MTBF gives a "percentage availability" figure.
I am trying to find a simplier way to calculate MTBF.
Your response was more from a "R&D" point of view, ie to predict what would be the MTBF of a product based on the reliability of its complonenets.
I am talking from a "Service" point of view. You have the product already in the field. Hundreds of them are already installed. How to calculate the MTBF?
If you have product in the field then you are talking about "demonstrated MTBF" (rather than predicted MTBF). Assuming each product is not fitted with an elapsed time meter, you need to try and get an estimate of the likely hours each unit has run or figures from typical users), mutiplied by the total quantity shipped for that product. Divide the total hours run by the number of known failures (the last part should be easy from your service records!) hence: x failures per y hours, or y/x hours gives average MTBF.
Thanks, it looks that you are familiar in both the design and service of a product.
I was asked to calculate service response time.
My answer was: what is the percetage that the service rep responded within 48 hours. For example, if he had 20 service calls, to 19 of them he responded within 48 hours, and to one of them he responded only after 90 days, then I would say that 95% of his service calls were within 48 hours. Excelent response time. I was told, however, that I need to calculate the average response time, (and not the figure that I gave). In this case it would be:
(19calls x 2days + 1call x 90days) /20calls
which produce an average respone time of 5 days.
Very poor response time.
What do you think will be the right tool to measure this?
also, do u have a link to check MIL-STD-217-E?
Hi again. Yes, you are right, I have experience of both design and service. Now I understand your problem about the average response time, and I think your first analysis was reasonable, whereas the second case incuding the single 90-day repair distorts the average too much.
Do you have definite "service level agreements" in place with your management? You will never achieve an open-ended 100% "good" response time, but you may be able to persuade them to take a more realistic view and agree on a figure like 95% in "x" hours with the remaining 5% in less than "y" hours. You then need a controls to spot the odd repairs going beyond the "x" hours so that you can provide extra resources to keep the repair inside the "y" hours limit.
There are tools which can track these sort of figures and provide service department management, but I can't remember any names at the moment - I will think about it.
Your calculation is directed towards what is commonly called "maintainability" in the aerospace industry.
Your specific calculation is nominally a combination of the "logistical delay time" (LDT) and "mean time to repair" (MTTR)
Availability, which is the measure of the average "uptime" of a system is:
Availability=MTBF/(MTBF+MTTR+LDT)
where MTTR is the average time taken to physical troubleshoot a system, replace/repair the faulty part and bring the system back on-line.
LDT is the average time to get a part that's not immediately spared into the faulty system.
Maintainability, then, is the discipline by which a supplier would attempt to maximize the availability, given limitations on spares provisioning. Obviously, not every repair truck can stock all the possible spares; the truch would be enormous and the cost would be prohibitive.
There are alternate options such as "floating" spares, where expensive spares are centrally located and shared, thus increasing the logistic delay beyond a single service call, but significantly reducing the mean delay, since you do not necessarily need to wait for an ordered part from a factory.
