Degradation Data Analysis, Ea<0, is there a problem?

[#1]

..JMP\JMPPRO\18\Samples\Data\Reliability\Device B.jmp, Degradation Data Analysis, Ea<0, is there a problem?

[#2]

Have you read JMP documentation about this,  Reliability and Survival Methods > Degradation > Model Specification > Nonlinear Path?

[#3]

I have read it, but I still don't understand the above questions.

[#4]

Hi @Rily_Maya ,

In the approach you're doing, it isn't appropriate to try to fit the model per system ID - there is not sufficient information or variation around the parameters to correctly define the activation energy (Ea) for each system, because each system hasn't had a number of temperatures tested against them - instead the intent is to apply the model to the whole data set to find a suitably fitting model to the data (see below) that can generalise against the variation. The data isn't perfect (I suspect intentionally), so in real world settings you would maybe consider applying a different formula, or refining the data that is available. Another option would be to take the formula into the Non-linear platform to add more restrictions and perform things like SSE grid searches to better refine the parameters. 

Thanks,
Ben

[#5]

If using Device B.jmp as an example in "Reliability and Survival Methods > Degradation > Model Specification > Nonlinear Path" is misleading, then the corresponding example is inappropriate.

[#6]

Hi @Rily_Maya,

The Device B set is intended to show a modelling approach that separates the reaction rate model per temperature, allowing for estimation of an average parameter estimate for the Ea and other parameters - in this context the goal isn't to look at each device as they aren't tested that way - the approach is appropriate and includes noise and other variations that you would expect in the real world.  What's the goal for going through the analysis? For education or for practice?

Thanks,

Ben

[#7]

Set the lower specification limit to -0.5 to calculate the pseudo-failure time of each path after 130,000 hours under the operating condition of 80°C, and this is also for learning and research purposes.

[#8]

Hi @Rily_Maya ,

In that case, you want to use the Destructive Degradation platform instead, so that you can relate the prediction to the temperature that is used and not the specific equipment. Open up the platform and fill it as so.

Once that’s run, go to Red Button > Prediction Settings and apply your baseline to the process, then run the model as you have done.

Then you will see your inverse predictions for 80C in the report tabs below, you can click the red button here and ‘save crossing time’

The 'after 130,000 hours' part doesn't play into here because you're looking for an inverse prediction of the temperature, to provide a failure time.

Thanks,

Ben