Having Trouble with Basic MSA Concepts

Six Sigma – iSixSigma Forums General Forums New to Lean Six Sigma Having Trouble with Basic MSA Concepts

This topic contains 3 replies, has 3 voices, and was last updated by  Ashley Leonzio 1 year ago.

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
  • #55907


    Hello everyone!

    I need your help regarding basic measuring system analysis concepts and implementation.
    Ive read the MSA 4th manual, but I still have some trouble to understand a couple of things when it comes to practice.

    I dont really know where to start so Ill just try to give real life-ish examples and whats my problem with implementing MSA on it.

    I dont know when to do GRRs and when to do process capability studies. When or what do they apply to? In my head, I thought that GRRs are for measuring devices (micrometers, scales, calipers, handheld thermometers and so on). And process/machine capabilities apply when you have a process that alters a part and a parameter (or parameters) of a part needs to be within a certain range.

    Let me try and give an example. Its not an actual process of ours and may make no sense, just bear with me.

    Lets say I have a machine that sprays chocolate coatings on cakes. A worker puts a cake on a conveyor belt, the cake travels through a booth, where a spray gun sprays the chocolate on it, then it travels through a cooling tunnel so the coating solidifies.
    The ultimate process control characteristic here is the amount of coating added to a single cake. We check the amount by measuring the weight of a cake on a scale before putting it on the belt, and then measure it after the cooling tunnel. However we have quite a few parameters on the machine that affects the added amount, like the conveyor belt speed: the faster it is, the less the added amount is. The pressure of the spray gun: the higher it is, the more the added amount is. And the temperature of the cooling tunnel: the higher it is, the less the added amount is as it drips down from the cake.

    In my opinion here I can do a GRR only with the scale I use to mesure the weight of the cakes. The coating machine’s characteristics arent handheld gages, they are built in sensors (pressure is displayed on a digital panel and can only be changed there too, same goes to the conveyor belt speed and tunnel temperature). So, we use the added amount as the parameter for the process/machine capability study. And on the reverse, we cant usemachine/process capability studies on measuring devices, right?
    Is this the correct approach?

    Additionally, we can do stability, bias and linearity tests for the weight scale, but how could we do a linearity or bias test for built in machine sensors? Let the machine operator stop and set the conveyor belt 10 times to a specific speed and check it with a handheld tachometer? How could we check bias and linearity for a resistance temperature detector? Open up the machine, take the RTDs out and perform the tests?

    Thats all for now, I know these are very basic concepts, but I need your assistance, and I very much appreciate any help.



    GR&R is essentially for a specific measurement instrument, which can be isolated from other variables, and for which you have a sample of known value. For instance, you can’t do a GR&R on a weight scale unless you have something of known weight that you can weigh repeatedly without influence from other variables. GR&R is not the only way to do MSA. We like to use it because it’s relatively simple.



    I should add, be careful not to conflate MSA with RCA. Measurement system analysis concerns how well we can trust our measurements — how much of the variation we’re seeing is due to measurement systems. Root cause analysis is about finding the reasons for variation from desired results. It sounds like you should be should focus on RCA not MSA.


    Ashley Leonzio

    @Staydog is spot on. You were asking about overall understanding, but what are you trying to do/solve for? I’m adding to elaborate on what has already been stated…

    MSA is performed to validate the measurement system (in your case the scale), not the process itself, whereas Process Capability is performed to determine the ability of a process to meet specifications (in your example the spec for added weight).

    MSA should come first. This is so that you know if you can trust your data as measurement systems are imperfect – they themselves have variation! You can use MSA to determine the accuracy & precision of the measurement system. As Straydog said, GRR is one of a few tools. You can use GRR to assess the scales. (Note that your measurement system should follow what I call the ‘ten bucket rule’ for resolution meaning that you need the measurement system to measure at 1/10th increments relative to the specification. If you want the weight difference to be 1oz, you would need a scale that accurately measures to a 10th of an oz or better.) You can also do Attribute Studies and Audits to conduct an MSA. For example – in your cake coating scenario, if you were also looking for the operator to judge imperfections in the coating to determine ‘good’ vs ‘bad/defect’ you may perform an attribute study and perform a Kappa Analysis.

    Performing process capability assessments applications are: (1) Routine process performance audits, (2) Judge the effects of process adjustments, (3) Evaluation of new equipment, and (4) Assigning equipment to products. You can perform your study on the output y-value of ‘added weight’ to determine if the process is in control and meets specifications. By doing this you can identify both common & special causes of variation and make decisions on how to proceed. (See graphic below) I personally use capability assessments to determine if the goal we set is achievable or if we need to consider re-engineering the process overall.

    Regarding your last question, it sort-of sounds like you are just starting to touch on looking for optimal settings. This is definitely more in RCA territory. In this example you may want to consider design of experiments.

Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.