MTTR - Quick and dirty way to calculate Mean Time To Repair - Solenoid valves

[Michael Busha]

Does anyone know a quick and dirty way to calculate MTTR? We have two kinds of solenoid valves, one is ISO rated, the other is not. I want to prove that the MTTR of the ISO rated valve is not significantly different from that of the non ISO rated valve. We have the capability of actually disassembling and reassembling both valves to get an accurate measure. Is there a formula I can plug the values into? Thanks for any help. The results of this analysis could save my company $40,000.

 

[Marc]

I would relate it to MTBF and do an accellerated life test. I don't know how to do MTTRepair.

Can you state the difference between MTBF and MTTR?

 

[Al Dyer]

Would the data involved be attribute or variable data?

Would also like to know if what you are looking for is Mean Time Between Failure?

We do alot of tool life studies (drills, insers etc...) which is basically MTFB. It's a good tool and can have a quick payback on investment time (cost).

ASD...

 

[Abe]

Yes,

a.Take 5 of the ISO valves and 5 of the other type
b. Define significant repair task(s).
c. make the repair task(s)
d. record repair times
e. Make a t-test for significance

Make this test twice with one tecnician repairing ISO velve and one the other one than switch tecnicians and repate

Good luck

 

[Michael Busha]

Thanks Abe. The procedure looks great but, I'm embarrassed to say I cannot find "T-test" in any of my references. Could you elaborate?

 

[Abe]

Originally posted by Michael Busha:
Thanks Abe. The procedure looks great but, I'm embarrassed to say I cannot find "T-test" in any of my references. Could you elaborate?

Hi Michael,
T-test is a statistical test of hypothesis that compare two samles and make a statement with some confidence whether the two samples
came from one (No MTTR difference)or two populations (significant MTTR difference).
You may find infronation on this subject in any statistical textbook
If you need more information or help you may email me at [email protected]

Avi

 

[Roger Eastin]

Just a thought: Some MTTR data are not normally distributed, so be careful when applying your t-test. Check your data first to determine whether it is normally distributed or not. If it is not,do a transformation (for instance, take the square root of the data) on the data to approximate normality and then do your t-test. This is a typical technique used in designed experiments and it works well.