Design Failure Rates - Mechanical 3

[5wp]

How are design failure rates determined for equipment that is produced? I know that this is a rather broad question, but this is not my field. Thanks, any assistance would be greatly appreciated.

 

[IRstuff]

What exactly do you mean by "design failure rates?"

TTFN

FAQ731-376

 

[MikeHalloran]

In order to design to a failure rate, I'd need to accrue some field failure data... which never arrives soon enough to save your job, and is usually wildly inaccurate, incomplete, or synthetic.

Usually, I got trouble enough just getting it to last through the warranty period.




Mike Halloran
Pembroke Pines, FL, USA

 

[TheTick]

Of course, some failures are by design.

 

[tbuelna]

"...and is usually wildly inaccurate, incomplete, or synthetic..."

MikeHalloran,

Based on my personal experience, I would disagree with your comment. I work in aerospace, and we do analysis for both failure rates and modes. A failure rate analysis is usually just a straight forward statistical calculation, based on historical failure rates for each given component in a system. It's uncanny how accurate they usually are.

 

[BrianPetersen]

But aerospace takes years or even decades to design major projects. When you work for company X and your competitor Y came out with a new innovative product that trumps yours, that sort of timeframe for testing and evaluation ain't gonna happen. If you work in automotive and have access to enormously powerful analysis and simulation tools, that can offset the time constraint somewhat. But in the world of mostly one-off custom projects built by mostly smaller companies without such resources, the most expedient thing to do is estimate how much steel you need based on whatever simple calculations can practically be done TODAY, then double or triple it to make sure. Then at least the failures won't be catastrophic structural ... they'll be the little things that nobody had the time to go through, and the short-cuts and substitute (cheaper) materials that the shop or the purchasing department did without telling you. (Been there)

I think MikeHalloran is in the same world as I am ...

 

[5wp]

I guess if I were going to purchase product X, is there any info available any where,ie.independent lab tests, etc. that would give me a better idea if product X is worthy of purchase. I agree getting thru the warranty can be a victory in itself. But other than reputation, are there other ways?
I'm asking the question in general, and I guess by your answers I can see that it all depends.
Thanks for the replies, much appreciated.

 

[prost]

If you are talking about failure rates of widgets, then Google "Weibull analysis". You are essentially fitting failure data using a Weibull distribution. It sounds fancy when you see it, for instance, in a job description, but it's really pretty straightforward.

 

[IRstuff]

For a specific consumer product, short of doing your own reliability test, there is no way of knowing or even predicting. There are, of course, generic MIL-HDBK-217, or Bellcore, reliability predictions, but there is no way to know whether your particular product choice fits the prediction.

In terms of structural materials, you have the manufacturer's material specifications, and you make a trade between how much you allow the material to be stressed, and how much cost or weight you can tolerate. Just bear in mind that no one is willing to pay for, in cost or weight, a system that will never fail.

TTFN

FAQ731-376

 

[MikeHalloran]

I've worked a lot in corrective design of commercial and industrial products, where there may be no formal mechanism for collecting failure data, and there's no time or budget or will to create one, or a mechanism exists but the existence or nature of the failures may be concealed for a number of nontechnical reasons.




Mike Halloran
Pembroke Pines, FL, USA

 

[tygerdawg]

I think you may interested in the MeanTimeBetweenFailure MTBF and MeanTimeToRepair MTTR data and values. These values are an area of Reliability Engineering. They're calculated from impirical data and/or actual real field data. Tread lightly in this area of the dark arts laddie, there be beasties about.

I'm certainly no kind of expert, only had some exposure to it with a former company and our products. Our product was an industrial device advertised as "20,000 hours MTBF". But one customer challenged that and demanded to know what the heck that meant, and how the data was derived. Research with the Engineering Department showed that they had tabulated "whatever data was available" for all the myriad components that made up the device. Then they somehow magically came up with "20,000 hours" as MTBF. Further investigation determined that the 20K hours was for "maximum loading and duty cycle" of the device. If the loading and duty cycle were less (it ALWAYS was), then the 20K value increased. But then again, there was no real value to quote the customer for max MTBF. Now, after several years of data under their belts, the company quotes 80,000 hrs MTBF.

MTTR is the time to repair / replace a component, and has stayed static due to component choices.

TygerDawg
Blue Technik LLC
Advanced Robotics & Automation Engineering

http://www.bluetechnik.com

 

[5wp]

Thanks tygerdawg, advice well taken. I guess I'm quickly learning the subject area is more complicated that it first appears. Some very good points were brought up here, much appreciated.