Measurement system analysis using variance comparison
Now we come to the actual topic. Is the measurement setup even capable of measuring what it is supposed to measure? I can only answer that with a conditional yes. And I will now show you why this is the case.
These are all the volume flow measurements that you have already seen on the previous page. The boxplot basically shows me that all these measurements should not be lumped together under any circumstances. I can do without the variance analysis. I think the reason is clear: if you can install the fan in different positions, this leads to other variances. The radiator issue must also be considered separately. Because the radiator is an additional variable. And I’ll show you that now.
The null hypothesis states that if all variances overlap, you can assume a certain measurement stability. This is what the p-value says at the end. Nice saying: If p is low zero has to go! Meaning: If the p-value is less than 0.05 then the variances do not overlap. Which ultimately tells me that the different measurements are not comparable and therefore there can be no measurement system capability. You don’t even need to try an MSA 3 because it would fail. Not because of the measurement technology, but because there are too many variables in the measurement setup.
A comparison with and without honeycomb is also not possible. What you can already see, however, is that there are two groups where the variances overlap. So let’s look at the group with honeycomb.
There you go, that fits. You can live with that and also derive a basic measurement system capability. This should also work for the group without honeycomb.
That also works. Who else notices something here? Right, we still have a variable in it. The apex sometimes on the left sometimes on the right. This means that I use ANOVA to see whether the mean values are approximately the same. The same null hypothesis applies here too.
That was obvious, you could already see that on page two in the comparison of the mean values. In conclusion, I can say that the measurement technology can resolve or measure what we might want to see. Due to the lack of plausibility checks using ANOVA, the measurement setup of the box is far too variable to ensure error-free measurement. You can also consider this as part of the FMEA, which I have shown you today. Ultimately, there can only be one way to install the fan for the test. There is also the issue of honeycomb and radiators, as well as push or pull. Difficult starting position!
One last comment on the subject of pressure measurement. If you look at the measured values, the question is not whether the measurement technology can measure stably. That’s not the problem, I couldn’t reproduce a measurement setup that could produce plausible values. So we can come to the end. Last page please.
407 Antworten
Kommentar
Lade neue Kommentare
1
Urgestein
Urgestein
Veteran
1
Veteran
Moderator
Moderator
Mitglied
1
Veteran
Urgestein
Moderator
Veteran
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →