C
Hi All,
I have a question...
I received in a new component for the 1st time and "Goods In" did their inspection. On a batch of 3500, 75 samples were inspected. A dimension was not out of spec but failed to meet the inspection criteria according to ASQ/ANSI Z1.9 2003. Spec is 0.65-0.75 mm, values were recorded at 0.66-0.75. The reason for the failure was because the distribution was positively skewed and when I ran all the stats it turned out 1.9% of the batch could be above the upper spec limit, and therefore failed.
The thing is the supplier performed his inspection on a smaller sampler number (n=15) but his results were all within 0.68-0.70.
I haven't got their individual values and I imagine our inspection isn't up to scratch but my main question is as follows:
Can a poor measurement system introduce a skewness to the distribution?
I'm getting the supplier to measure a representative n=75 sample and I'm expecting it to be normal distribution as the validation up to now has shown it to be a normally distributed process.
So can this happen and is there anything glaringly obvious which could be the cause?
The inspection is on a tube wall thickness with a digital caliper. I agree it is not the most accurate technique, caliper flats and a curved surface but the supplier uses the same.
Thanks
Dan
I have a question...
I received in a new component for the 1st time and "Goods In" did their inspection. On a batch of 3500, 75 samples were inspected. A dimension was not out of spec but failed to meet the inspection criteria according to ASQ/ANSI Z1.9 2003. Spec is 0.65-0.75 mm, values were recorded at 0.66-0.75. The reason for the failure was because the distribution was positively skewed and when I ran all the stats it turned out 1.9% of the batch could be above the upper spec limit, and therefore failed.
The thing is the supplier performed his inspection on a smaller sampler number (n=15) but his results were all within 0.68-0.70.
I haven't got their individual values and I imagine our inspection isn't up to scratch but my main question is as follows:
Can a poor measurement system introduce a skewness to the distribution?
I'm getting the supplier to measure a representative n=75 sample and I'm expecting it to be normal distribution as the validation up to now has shown it to be a normally distributed process.
So can this happen and is there anything glaringly obvious which could be the cause?
The inspection is on a tube wall thickness with a digital caliper. I agree it is not the most accurate technique, caliper flats and a curved surface but the supplier uses the same.
Thanks
Dan