To Joe Petersen, I have seen the aerospace example used before, However, a petro-chemical plant is not an aeroplane. I believe the success in the aerospace industry is because of the numbers of similar equipment, added to the multiple locations. This is the situation where statistics scores, IMO.
Best regards
Joe McCormack

Re garys example, I would need to know if there were other components installed that hadnt failed also were they new items or repaired items
an example I collected 25 years ago same bearing same application had failures from 100 to 4000hours the weibull parameters said random but on closer investigation and classifying bearings into original brgs as received and then replaced brgs showed 2 distributions one wear in (poor purchasing, repairs( and 2nd dist wear out Softwear like ReliaSofts Weibull7 can account for complete and censored data

Hi Joe,
Weibull analysis is used sucessfully in the aviation industry - It is used in the design phase when they prove the reliability of components. Reliability in aircraft design is like quality in manufacturing - it starts at the design feasibility stage. Component manufacturers have to prove the reliability of thier components and they use Weibull analysis as one of the tools to prove what they call reliability growth. The other thing abot aviation failures is that they are not fatal like lots of people think. The redundancy built into aircraft means that few single component failures cause crashes on thier own. You should see an aircraft log book to understand how frequently aircraft components fail - or travel by air a lot.
Rgds
Steve

quote:

I agree with you that 2-3 points may not be enough in some cases, but for others it maybe enough data. This all depends on the situation and circumstances it is being applied to and I guess is the real difference between a "Mathematical Statistician" and a "Reliability Engineer".

2 or 3 data points is better than none - (well no failures is better than some)- anyway you know what I mean.
With 2 or 3 data points you can start to figure out what the causes might be - but you need to understand the equipment and if you can, talk to the guys who did the repair or get hold of the broken bits etc. Knowing beta and eta is not realy going to help you solve the problem...

Going back to Gary's original excercise

Component A:
1st installation lasted 10 hours
2nd installation lasted 100 hours
3rd installation lasted 190 hours
Component B
1st installation lasted 99 hours
2nd installation lasted 100 hours
3rd installation lasted 101 hours

The statistics tell you nothing unless you GoTalk.
Both of these patterns could have come from the roulette wheel (figuratively speaking). Both components could be the wrong component for the application and be infant mortality patterns... 100 hours is not a great component life.
Steve