Load - Strength Interference model

[csasun]

For Mechanical Reliability, it is often said that L-S models are most appropriate in analysing esp. fatigue failures.

However, it is not clear as to how one can build the model for real life situation. Books like: Practical Reliability, O'Connor, Mechanical Reliability, ADSC Reliability Engineering Handbook, Keccicioglu etc. have given some basic fundamentals on the same. However, it is all 'focussing' on static analysis, or some level of simplified dynamic analysis.

But, in reality, as we all know, mechanical parts 'strength' deteriorates by time or usage; also, load spectrum too can change (esp. when we are talking about life targets of 10 years!) - this is so true esp. for developing countries like India.

In such scenario, how does one plan the modelling parameters of strength and Load distribution --> that can vary by number of occurence and time.
:)

[avanti]

If you take a look at "Fatigue Design" by Carl Osgood (pun by Wiley) it shows techniques for derating strength levels. You can then use this derated strength level in a L-S analysis.

MIL-HDBK-5 contains derating factors for metals based on fatigue (S/N curves).

[csasun]

If you take a look at "Fatigue Design" by Carl Osgood (pun by Wiley) it shows techniques for derating strength levels. You can then use this derated strength level in a L-S analysis.

MIL-HDBK-5 contains derating factors for metals based on fatigue (S/N curves).
---------
Thanks for that short but useful reply!

MIL-HDBK 5 is 'cancelled'.

However, for similar information, I felt, ASM handbook on 'Fracture & Fatigue' is more apporpriate and up to date.

On the other reference, Osgood is 'out of print'? `not available in stock' is the response I got.

Neverthless, I would like your comments on the following:
1. Derating - no doubt is good practical way of 'accounting' for dynamic behaviour - but still the analysis is 'static'.
2. These loads are typically 'sequence sensitive'. So simple derating may not be appropriate for assuming 'material strength properties'
3. These loads itself are not necessarily 'stationary' (or ergodic). Since over a period of say 5 years, the road conditions can change to worse. Therefore, the 'distribution' can get altered (mean shift) How do we account for this situation?

Therefore, both the distributions of load & strength are somewhat change over time. In such situation, how does one address the 'interference' modelling?

[avanti]

---------
Thanks for that short but useful reply!

MIL-HDBK 5 is 'cancelled'.

However, for similar information, I felt, ASM handbook on 'Fracture & Fatigue' is more apporpriate and up to date.

On the other reference, Osgood is 'out of print'? `not available in stock' is the response I got.

Neverthless, I would like your comments on the following:
1. Derating - no doubt is good practical way of 'accounting' for dynamic behaviour - but still the analysis is 'static'.
2. These loads are typically 'sequence sensitive'. So simple derating may not be appropriate for assuming 'material strength properties'
3. These loads itself are not necessarily 'stationary' (or ergodic). Since over a period of say 5 years, the road conditions can change to worse. Therefore, the 'distribution' can get altered (mean shift) How do we account for this situation?

Therefore, both the distributions of load & strength are somewhat change over time. In such situation, how does one address the 'interference' modelling?

MIL-HDBK-5 was replaced by the FAAs Metalic Materials Properties Development and Standardization (MMPDS) Handbook in 2003. However it is still commonly used in the aerospace industry.

Osgood's book is available used from many online book sellers including http://www.biblio.com/search.php?author=osgood&format=&title=fatigue&keyisbn=
The price is a bargain compared to what I paid for it when it was in print.
It is also available through inter-library loans.

Dynamic loads are usually converted to equivilent static loads (at least in my experience in the aerospace industry). For example acceleration loads (from vibration) are converted into usable static loads using Miles equation http://femci.gsfc.nasa.gov/random/MilesEqn.html

If the loads change with time one either has to make an estimate or just update the analysis as loads change. In aircraft for example it may result in a reduction in its total allowable flight hours. In aerospace structures are often designed with a factor of safety of 1.5 which results in a very high reliability prediction.

In the end it is the testing that confirms the design integrity of a product.

[csasun]

Thank you very much for that excellent 'web link'.

Some of my 'doubts' are definitely answered through 'Miles equation'.

I shall come back with more clarifications in a few days

[Charlie]

Hello. very nice and interesting post. No body can deny the importance of reliability. Neither do i ...

[avanti]

If you take a look at "Fatigue Design" by Carl Osgood (pun by Wiley) it shows techniques for derating strength levels. You can then use this derated strength level in a L-S analysis.

MIL-HDBK-5 contains derating factors for metals based on fatigue (S/N curves).

:o
This should have read (Published by Wiley)

[csasun]

Hi!

I have one another thing that is bugging w.r.t Factor of Safety (FOS)

When FOS is higher at 1.5, reliability value is no doubt good. Does that translate into a reliabilityvalue of 0.9990? (with a reasonable assumption of 9% COV for both Load & Strength).

If yes, now inorder to demonstrate this in lab, it calls for unrealistic sample size of 2236 at modest 90% CL (using binomial distribution)

Further, even at a very high Test Acceleration Factor of say 100, still the sample size is too large(in double digits).

Does this all mean that Reliability testing is of no practical relevance when designs are highly reliable?

Any comments/views please.

Sasun
----

[avanti]

Hi!

I have one another thing that is bugging w.r.t Factor of Safety (FOS)
.................................................. ......................
Any comments/views please.

Sasun
----

When FOS is higher at 1.5, reliability value is no doubt good. Does that translate into a reliabilityvalue of 0.9990? (with a reasonable assumption of 9% COV for both Load & Strength).

Yes in this case (however a FOS of 1.5 does not automatically equate to a high reliability). There are numerous papers that discuss reliability vs. FOS – here is an interesting one:
http://myweb.msoe.edu/~musto/Course%20Links/ME363/Anderson.pdf

If yes, now in order to demonstrate this in lab, it calls for unrealistic sample size of 2236 at modest 90% CL (using binomial distribution)

I often have to calculate high reliabilities at a high confidence level with test quantities small as 10 samples.
I use Z = (X1^2 – X2^2)/(k(s1^2 + S2^2)^.5) where k is obtained from the t-distrubution.

Does this all mean that Reliability testing is of no practical relevance when designs are highly reliable?

In my opinion reliability analysis should be conducted early in the development process to establish design features and test conditions. I've experienced devices that on paper have excellent reliability but failed tests so testing should always be conducted for critical devices or systems.

Any comments/views please.

I've seen or heard of the Load-Strength technique being used in many different areas including stress analysis, chemical reactions, pharmaceuticals, and even cost analysis. After all this is really simple statistics. Where numerous factors may be involved I've seen it used in combination with a Monte -Carlo analysis.

[csasun]

Hi! Avanti,

First of all thanks alot for this nice response.

I enjoyed the paper you provided as a link - quite interesting

In the formula for computing Z, what are X1, X2? Load and Strength Means?

Just to bring in new dimension to the discussion, as an alternative, for reliability demonstration using small samples, what is your experience with Beta directory?

Thanks in advance,
Sasun
-------

[avanti]

...............

In the formula for computing Z, what are X1, X2? Load and Strength Means?

Just to bring in new dimension to the discussion, as an alternative, for reliability demonstration using small samples, what is your experience with Beta directory?
...........................

Yes - but I like Ireson & Coombs definition of Stress (or Load) and strength:
"Stress is used to indicate any agency that tends to induce failure, while strength indicates any agency resisting failure" (Page 18-25).

I have zero experience Beta directory