MSA for Automated Equipment

Hi,

I am trying to determine whether the measurement systems in an fully automated line are stable enough to be approved. There are 4 different measurement systems that I need to look into.

Normally the MSA that I do are type 1 gage studies and gage R&R but in this case the type 1 gage study only gets me so far and the gage R&R isn't valuable since there aren't any other operators or measurement systems to compare against.

I took a class with Minitab and it seems like a gage linearity and bias study is what I want but I am having trouble determining what criteria I should use to evaluate the results. Looking at what Minitab gives me, it seems like there is some issues with bias and linearity but it's all very small. I don't know how to tell whether this is enough of an issue that we can't trust the system to measure within the tolerance window.

Can someone give me some advice? I have tried to give all the details I can think of below and have attached a minitab file and excel sheet with the raw data.




3 of the measurement systems are for length. For these, we had calibration bars created to measure and check the system. We wanted 5 for each but for one reason or another we lost one on the BID and Spindle checks. Each bar has been measured by an external calibration company using an optical comparator.

The current measurement was created by connecting an e-load and setting a static current. I did this at 2 settings (perhaps more is needed, but the measurements are extremely consistent).

The tolerances for the parts we are measuring are:
Outer Tube +/- 1mm
Spindle +/- 1mm
BID +/- 2mm
Current +/- 0.5A

Thanks,

Crev64

 

Hi,

Ok, what follows makes some assumptions about your data but hopefully will give you a steer on the best way of doing this.

The first thing I would do is break your process down into rational subgroups. My analysis involves just the Outer Tube data,
you have five reference artefacts with ten measurements observations against each presumably in time series order, more observations would of cours be better but we can use these if thats all we have.

Create Individual moving range charts against each reference size and firstly check whether the measurement process is consistent. Do this by ensuring that all the data is within the control limits on each chart, whilst you are at it check for chunkiness, i.e. you will need at least 4 possible values within the moving range chart.

If the data is chunky fix it by sampling over a longer period of time (your data shows no signs of this). If the data is not consistent then find the causes and fix the issues and re run the study. There is no point in characterising the study with any statistics until the data demonstrates consistency. In your case, 3 out of five of the charts demonstrate consistency therefore there is no reason why the others dont behave in the same way.

Summary of Your Outer Tube Data - tolerance on length = +/- 1mm

In all cases there is no significant threat to the tolerance once the measurement process is consistent, but there are nuisance elements of bias which in a couple of cases require furtehr investigation.

Ref value 397.356 - Inconsistent, something happened after observation 4

Ref value 323.509 - Consistent, there is a detectible bias of 0.05 mm which is more than 13 times the size of the probable error therefore this requires investigation, i.e. why does the process consistenly measure the artefact around 0.05 mm smaller than the reference?

Ref value 250.185 - Inconsistent, something happened after observation 8

Ref value 176.497 - Consistent, there is a bias of 0.004 mm (rounded), in my opinion this is so small as to be inconsequencial. Use the measurements to imporve the manufacturing process

Ref Value 103.11 - Consistent, there is a bias of 0.055 mm which is 55 times more than the probable error.This will require investigation

T1 - I don't know what this represents but when its charted in the same way as above it demonstrates a lack of consistency

I have attached the analysis file below

Hope this helps