Situation: 1.) Machine's measuring performance has tendency to drift (i.e wear and tear)
2.) Product's characteristic to be measured degrades when measured repeatedly.

Question:
1.) Can i run stability study with this situation? :frust:
2.) if so, how?:confused:


thanks in advance

>>reynald

2.) Product's characteristic to be measured degrades when measured repeatedly.

I'm not sure what you mean by "Product's characteristic to be measured degrades when measured repeatedly."

Situation: 1.) Machine's measuring performance has tendency to drift (i.e wear and tear)
2.) Product's characteristic to be measured degrades when measured repeatedly.

Question:
1.) Can i run stability study with this situation? :frust:

2.) if so, how?:confused:


Possibly. Information about the product, the characteristic, the process and the measurment device will help.

1) Does the product characteristics degrade immediately after one measurement, or can you obtain several repeats before degradation?

2) Is the product/process autocorrelated for short periods of time? That is, is the product very homogeneous for short periods of time. An extrusion or drawing process would be an example of this.

If the process is highly homogeneous for short periods of time, that is, there is little or no product variation until drift over time introduces variation, and all short term variation is assignable to measurement error, obtain consecutive parts sufficient for the entire study.

If you can measure the characteristic 2-3 times before degradation, measure 1 part 2-3 times, then the next part 2-3 times, and so on for the duration of the stability study. Plot on an Xbar/R chart to analyze.

If you cannot measure more than once before degradation, then measure multiple parts, then repeat for the duration of the study. Plot on an Xbar/R chart to analyze.

The success of this approach depends on minimal short term variation of the process. The control limits of the stability chart should be based on measurement error. The more part to part variation exists in relation to the measurement error, the less effective this test will be.

I'm not sure what you mean by "Product's characteristic to be measured degrades when measured repeatedly."

Thanks Marc,
what i mean is sort of "destructive testing". But product is not really totally destroyed, only "degraded".

Thank you Miner,
Let me clarify some points:
Possibly. Information about the product, the characteristic, the process and the measurment device will help.

1) Does the product characteristics degrade immediately after one measurement, or can you obtain several repeats before degradation? >>#1 Yes, can measure about 6 to 8 times before it is totally destructed, but some products are usually destructed even after only 4 time measurements.
2) Is the product/process autocorrelated for short periods of time? That is, is the product very homogeneous for short periods of time. An extrusion or drawing process would be an example of this. >>The process (or machine that measure) is autocorrelated in a sense that it gives lower and lower value over time.
If the process is highly homogeneous for short periods of time, that is, there is little or no product variation until drift over time introduces variation, and all short term variation is assignable to measurement error, obtain consecutive parts sufficient for the entire study.

If you can measure the characteristic 2-3 times before degradation, measure 1 part 2-3 times, then the next part 2-3 times, and so on for the duration of the stability study. Plot on an Xbar/R chart to analyze.>> I get this one, yes this is possible, though i cant really tell if the drift is due to the product, or due to the machine
If you cannot measure more than once before degradation, then measure multiple parts, then repeat for the duration of the study. Plot on an Xbar/R chart to analyze.>>:confused: Even after brain processing, i still can not get this one.


But thanks anyway, i'll do #1 first. :thanks:

To clarify one point, collect all of the samples that you will use for the entire study at one time to minimize the effect of the process. Ideally, all of the variation should come from the measurement device.