Chem/Process Industry

Query – Is there really any difference in MSA’s for the chemical/process industries? If the item being produced is not a discrete item, does this make conducting an MSA any more difficult or challenging?

@garyacone – Thank you for your input Mr. Cone. Your insight I think will provide a valuable addition to the forum. I hear you have some background in this stuff.

Some of my recent experiences have been with BB’s that believe that they have a destructive test (as might often seem to be the case in the process/chemical industry). What is your view on whether these can be done via nested vs. crossed study?

@MBBinWI Thank you for your question Mr. inWI.

I think the advice in Minitab’s help menu is quite good. Look under general advice for GR&R studies.

Too often, people have learned to use Destructive when it’s not appropriate because you will get a slightly better result.

@MBBinWI – Very few tests I’ve seen in the chemical process industry are destructive of the entire sample. Usually small aliquots are taken from the sample container, and then the test may destroy that subsample (an example might be flash point testing), but not the entire thing. So, most often Crossed studies are used.
However, some tests require the entire sample and are destructive. If it is an in-process sample, or a sample from a continuous unit, then there may not be additional sample material available. In this rare case I could see a nested study approach.
What about a test that is destructive, and requires all the sample volume in a single container, for a final analysis on a batch process? If a second sample of the same final product of that lot can be collected, should a crossed or nested Gauge R&R be used?

@garyacone – Thank you for your guidance. Often challenging to verify homogeneity, when existing reaction vessels have one sampling port. Sometimes, the low viscosity of the material and the presence of mixing within the vessel increase the likelihood of homogeneity.
Thanks again.

@garyacone

Thanks for the input. Homogenity of the batch is a problem that has not been adequately tested at all. Difficult to do at time, the viscosity of the material is so high that it is almost impossible to be homogenous, but customers still want the “data” on the COA. Because they use the product in such large quantities, the lack of homogenity is generally not a problem, but how do you determine the viability of running replications just to confirm that the homogenity is not changing over time and hence the small sample you have taken is not representative of the batch.

To clarify, you run a product in batch sizes of 50,000 lb. When you pull a sample you get a 250ml jar of product…. you can do this as often as is prudent. We do control the process parameters – heat, time, pressure, raw materials are all measured and controlled. Mix rates are “common” from batch to batch. I think everyone can agree it is difficult to determine if the process is changing from that single sample. The question becomes should you run more samples or simply rely on controlling the process parameters. If you do, at what frequency would you verify the relationship between input/process and output variables…especially given the amount of variation in the output measurement method.

@garyacone

Who said anything about this problem being unsolvable….

Give me and my colleague a break… we understand Y=f(x) very well. Also understand the existing constraints that while changable are financially constrained due to priorities in capital investments. The company has been assembled through acquisition and sometimes the equipment current configuration is not the “best” …e.g. where the sample points are… Certainly as was pointed out… getting consistency in the methods is a quick and easy fix. (There is a group working on this). Getting the customer to understand our sources of variation is another “longer term” issue. Haven’t you experience the “customer assumes no variation” scenario before?

Batch size is a function of the vessel. Customer demand is such that reducing the overall batch size would require a significant financial investment in additional equipment. Breaking the addition of ingredients into multiple stages has NOT been investigated, and is a great suggestion that we will look at to see if it will have an impact on final outcome.

General SIPOC is; Add ingredients, Heat, Cook, Cool, then Transfer….. so if breaking the add can improve the homogenity, without impacting the overall total cycle time, it will be worthwhile.

@rbutler Great point, Robert…. been there done that! Non-homogenity is a known issue.

Spec is wide enough that even with the measurement variation, we are confident in what we ship. It is detecting change in the process that remains the issue.

@ctickel A few posts ago ago you asked that I give you and your colleague a break. A break from what? I offered to help you solve the problem that you only served up and said it was difficult.

The group I called out can tell you how to answer your questions without capital, did you not want that offered?

I am told you are a good guy but it is not at all clear what you are actually seeking since your very first post. Please help me.

@MBBinWi Do recognize the opportunity that reducing batch sizes provides. However, multiple mixers to the “cooker” isn’t an option – capital and space constrained plus other issues(reaction requires heat to initiate, but requires mixing continue while “cooking”). So if you were “greenfield”, you would simply break the current large vessel into smaller units.. The vessel has a minimum batch size in order to cover the mixing blades, but your original idea of adding in stages, rather than all of A ingredients, then all of B ingredients, does have merit worth investigating. I do appreciate your comments.