Standard Deviation Results

[Robbie]

Can I assume that the standard deviation results can be used to calculate levels of confidence from simulation e.g. if the expected numbers of failures is given as 14.8 and the std deviation is 2.5, can we say that the the number of failures is between 12.3 and 17.3 with 68% confidence (using standard normal tables) or is this incorrect use of the std deviation results?

[David]

Hi Robbie,

This is a common application and is valid.

[Robbie]

Further to last message. I have modelled 16 identifical, but independent pumps in our facility using blocksim - I have linked them in series, simulated over 1, 2 and 3 yrs to yield the number of expected failures. It works well apart from the std deviation puts the lower bound into the negative domain for the shorter time frames.

I have done a similar job in Weibull, using real data, calcuated the reliability parameters, and then using the spreadsheet function have tabulated for each pump the conditional probability of failure over 1, 2 and 3 years (given their current run time). From this probability of failure, I sum the results and get an overall number of expected failures in the given period. I've also added the 10% and 90% estimated for each pump. Can I sum these aswell the same way I can for the mean estimate?
Also, the mean values are similar for both Weibull and Blocksim, but not the confidences (presumably because I;ve taken the blocksim mean and std dev and assumed normality, and used tables to calculate them?

[Pantelis]

First see … http://www.weibull.com/LifeDataWeb/fisher_matrix_confidence_bounds.htm (http://www.weibull.com/LifeDataWeb/fisher_matrix_confidence_bounds.htm)
To compute the bounds you would need to formulate the reliability equation as your function, and then follow the steps shown.

Weibull++ does this automatically using the RBD functionality. Did you try that?