Inspection Interval Optimization

[walte27f]

Would anyone be able to help, using Excelt, to calculate an inspection interval optimized to catch a certain percentage (95% in this case) of failures?

Below is a small portion of my data set, but for example if I have failures occurring at the cycles listed below, how do I calculate the inspection interval that would give me 95% confidence or allow only 5% of failures to go unpredicted? In this case cost isn't a factor, only preventing the interruption.

Cycles at which a failure occured:
49
273
574
854
1077
1354
1704
2009
2075
2221
2331
2428
2519
2571
2633
2777
2915
3016
3080
3374
4286

Thanks in advance, please let me know if I can provide any more info

 

[ehbadger]

Have you heard of the "Crystal Ball" software for Excel? Sounds like it might do exactly what you want. This link probably explains it better than I ever could.

http://www.oracle.com/technetwork/middleware/crystalball/overview/index.html

 

[IRstuff]

Alternately, you might consider modeling it with a Weibull distribution:

http://www.weibull.com/hotwire/issue14/relbasics14.htm

http://www.weibull.com/itools/index.htm

There are companies that sell that type of software

TTFN
I can do absolutely anything. I'm an expert!
homework forum: //

http://www.engineering.com/AskForum/aff/32.aspx

faq731-376 forum1529

 

[IRstuff]

There are a couple of Excel spreadsheets designed for Weibull analysis:

http://www.cpkinfo.com/files/Weibull_calculator.xls

http://www.barringer1.com/jan09prb_files/Weibull-Analysis-In-Excel.xls

http://connected-knowledge.com/2013/08/01/how-to-match-to-weibull-distribution-in-excel/

http://www.samos.aegean.gr/actuar/dlekkas/environmental stats/Using Excel for Weibull analysis.pdf

http://www.lifetime-reliability.com...n-plot-model/Weibull_Distribution_Example.pdf

http://www.qualitydigest.com/print/23237

This one requires a purchase

http://www.lifetime-reliability.com...n-plot-model/Weibull_Distribution_Example.pdf

http://www.lifetime-reliability.com...ot-model/weibull-distribution-plot-model.html

TTFN
I can do absolutely anything. I'm an expert!
homework forum: //

http://www.engineering.com/AskForum/aff/32.aspx

faq731-376 forum1529

 

[walte27f]

Thanks for the replies. I have a Weibull calculator that I use for setting a hard/soft time limits. In this case we want to calculate how often we should inspect a seal, that can be repaired on wing rather than replaced, any thoughts or similar experience? Thanks again, I look forward to learning from your experience.

 

[IRstuff]

Note that your presented data suggests that the 95% life is on the order of 150 cycles, which is not very good, given that you have up to 4286 cycles on other parts

TTFN
I can do absolutely anything. I'm an expert!
homework forum: //

http://www.engineering.com/AskForum/aff/32.aspx

faq731-376 forum1529

 

[walte27f]

Yes, there are low cycle failures, but in this case they are repairable and are not limiting the life of the part.

 

[BigInch]

You have only given failure statistics.
From which I deduce that,
5% of the items fail within 1700 cycles 95% last longer than 1700 cycles.
10% of the items fail within 2000 cycles
67% of the items fail within 2600 cycles
90% of the items fail within 3000 cycles

If you test every item for 1700 cycles you can be reasonably sure that you've caught enough failures that 95% will last longer than that.

Now it appears that only 50% of those that didn't fail during the test will last for up to 3000 cycles. Problem is that you've already expended a little more than 50% of that life cycle in testing. Not a very reliable product once past that testing phase.

However if you establish that the lifetime is 1500 cycles,it seems that 90% of all items will last that long.