Which reliability data do we use

[K Edwards]

We are probably in an unusual position of having many sources of good reliability data. We are a machine tool manufacture with observed data from machine studies, we also have individual component failure history and we also have plant CMMS uptime information. The problem is which do we try to model. Our aim is to try to build up the machine in block sim using our component data modelled in Weibull++ or supplier MTBF, then use FRED to analyse it. We know that the data we are inputting comes from the three stages in a components life and so the Beta value isn't realistic. If we use either our observed data or CMMS data in Weibull++ then we get good results but cannot go any further with the analysis. Which data should we use and how can we interpret the results

[Tarik]

Hi,

Check out this article
http://www.weibull.com/Articles/RelIntro/Data_Collection.htm

Each source of data has advantages and disadvantages

The customer data is good data that captures the various environments where your machines are being used. However, this data could be biased and could contain many errors or failures that happened due to customer misuses.

I would advice you to avoid only using MTBF values (many times you will have to assume an exponential distribution model, which is based on constant failure rate, i.e. no wareout), because it doesn't tell you much about the failure distribution model. If you have failure data, use them to construct a reliability distribution model instead of just relying on just an MTBF value.

Try to use your other sources of data to check or validate results obtained from your testing data.

Again, the answer is "it depends!". You will have to make a judgment on the validity and reliability of the data you are relying on.

I hope this helps