Interpreting Type 1 MSA (Measurement Systems Analysis) Results

G

grantmate

Hi Guys

I am a novice at MSA studies but i have been tasked with implementing them into our company, but anyway here goes.
I have been doing GR&R studies on our equipment and appraises but it's been taking longer than expected. So i thought a Type 1 study(less accurate but acceptable by customer) would be quicker.
The issue is, i have started with 2 - 3" Micrometer and got the exact same result 30 times measuring 1 part by 1 appraiser(obviously got 1000000% repeatability) and then tried introducing more parts (5) and measured them in sequence 6 times each(30 results) and got this result(see attached file). now the repeatability is barely acceptable but the capability presented is out.

any suggestions on how best to do a MSA study with only one appraiser to give a repeatability percentage (my current constraints from manager)

any advice would be appreciated

thanks

Grant
 

Attachments

  • msa 2-3 mic.xlsx
    363.2 KB · Views: 390

Bev D

Heretical Statistician
Leader
Super Moderator
Re: Type 1 MSA Help intrepreting the results

OK I'm really sorry your boss thinks that someone who is untrained/uneducated in MSA can 'just do the job'. While they may be able to fulfill the cookbook stuff to make a Customer 'happy', they can't do anything else very well. this isn't 'easy' or intuitively obvious. you did the right thing by seeking out help from others who have experience and knowledge in this field.

first - you have to randomize the readings for every part when multipel parts are involved. there is a clear bias towards greater repeatability at teh end of the 6 readings (the resutls are much closer to each other)...if I read you correctly. and of course there is almost no part to part variation. This results in an overinflated SD that makes the Cg ratio look smaller than it really should be.

If you only have one operator and you are really only intereseted in repeatability then my suggestion is to use 30 parts each measured twice. Remember to randomize which parts are measured adn ensure that the second reading of any given part is not sequential to the first reading. ( operators have a human nature response to try to get the same resutls when they 'know' what they got the first time)

You can calculate the measurement error as the average range of the two readings for each of the 30 parts divided by 1.128.

then you can calculate the total observed SD directly.

see this MSA Spreadsheet for templates and references.
 

rickpaul01

Involved in HankyPanky
Re: Type 1 MSA Help intrepreting the results

Hi Bev, The spreadsheet doesn't load correctly from here. Can you email it to me? Do you have my email address?
Thanks
 

Bev D

Heretical Statistician
Leader
Super Moderator
try the attached file
 

Attachments

  • MSA tools for ELSMAR.xls
    590.5 KB · Views: 829
Last edited by a moderator:
G

grantmate

thanks guys this really helped. will try some a few tests to see if i finally have a method nailed down.:applause:
 
Last edited by a moderator:
Top Bottom