Machine learning manager exchange resource and how manufacturing can be helped by machine learning

I would like to be more informed about the machine learning manager exchange resource.
Is there a good location I can learn more about it without installing?
Is the exchange resource in a finished form, or is it in a kind of testing phase?


I am trying to become more informed on trade-offs on ways to achieve the benefits of machine learning.
I would also appreciate information on ways that in manufacturing, artificial intelligence is paying off.
I read many examples of artificial intelligence working in marketing, recommendations, and searching.
I haven't read about examples where artificial intelligence has helped production facilities specifically.

When I try to imagine the artificial intelligence helping manufacturing, I think maybe an AI would reduce all the manufacturing to copies of one production line that is most profitable, and many copies of that.
I think of Ford selling only Trucks, or McDonalds offering breakfast all day during the pandemic.

Particularly if there is some accessible chunk of machine learning performance that I can integrate to some part of maybe supply chain optimization, or something that I can immediately asses that an AI can do as a task that I can't, or that it does very very fast, that would very appreciated.

Thanks in advance for the help.

Hi I was one of the Application Engineers that worked on this resource and am happy to give you more information on it.

Other than installing the resource and using it the best way to learn more about the resource would be to go to the description of the resource or to watch the 2022 ICC session where this resource was first introduced, Using Ignition with Machine Learning Libraries | Inductive Automation. The resource is complete but more algorithms or features could get added in the future.

Ways that machine learning is being used in manufacturing are commonly predictive maintenance or quality analysis but there are certainly many more uses in manufacturing out there. This resource would be a good place to start using for supply chain optimization but that depends on what you are trying to optimize and how much data you have to train the model against.

Let me know if that answer’s your questions or need any more information on the resource and I would be happy to help!

4 Likes

Thanks, I appreciate the video, watching now.


I hope to give useful insight from my user side. I didn't see a video link.
I looked harder after your post. I noticed the watch button today.
I use a 22" wide screen 1920x1080.
Maybe I will always notice it now.

When I click the watch button, using Chrome, I get an error.


Would you be so kind as to provide me the link to the documentation that is mentioned in the video ?
I appreciate it if so. My local is tied up with a lower version testing, and I am not sure about having more than one local.

I liked the cluster example, but my process is so fast for my lines that I can't utilize that example practically.
So I wonder about the other algorithms.

The guide mentioned in the video is still being worked on, I will provide the link once that has been released.

As for the other algorithms there are clustering, regression, decision tree, and a few simple neural networks currently built into the resource.

1 Like

Conner,

Can you send me the runScript syntax to use for a expression tag to call the processModelByTagPath function? I tried making it from looking at the video but it does not seem to work.

Hi Bryan,

The expression syntax should look like: runScript("exchange.ml.hook.models.processModelByTagPath", pollRate, modelName, tagPath1, ..., tagPathX)

You will also need to have the Gateway Scripting Project property configured to access these scripts from an expression tag.

Let me know if this allows you to use this function in an expression tag.

I use this:

runScript("exchange.ml.hook.models.processModelByTagPath"
, 1000
, '6 Group Valve Spray Pickling Model'
, "[default]tag1"
, "[default]tag2"
, "[default]tag3"
, "[default]tag4"
, "[default]tag5"
)

Don't know how to format for this forum but there are indentations for all rows after runScript.

I imported the entire ML project to my global project(Gateway Scripting Project) so the libraries are in there.

That syntax looks correct to me. What model algorithm & type are you using? I can look into the model to see if this issue is specific to the model or a problem with that function.

clustering with k-means++. I can make the model no problem. I originally imported the ML 1.0.1 project as a new project. Set a path to the main so I could get to it. Made some models. All seemed good. Went to my tag browser and made the expression tag and now I am stuck with the error. I went in to my project that I have set for "Gateway Scripting Project" and imported everything from the ML 1.0.1 into that project. Still got the error. Changed my "Gateway Scripting Project" to MachineLearning_2022-11-18_1550 and still get the error.

Ok so made another tag from my main project(that does not have the scripts in it as they are in the global) and it works.

Made an expression tag with Dataset as the data type and here is the expression:

runScript("exchange.ml.hook.models.existing()")

Works like a champ. I get a 3x7 dataset which has all my models that I have made so far when i drill into the metadata for this new tag and go to value and then name.

Tech support remoted into my box last night and seen my issues. They have updated my ticket. Seems I am just having problems with the processModelByTagPath function.

Next Level: Here is an interesting article on Quantum Machine Learning

1 Like

I don't think any quantum computers have made it to market yet.

1 Like

They've only been 5 to 10 years out for the past 30 years, so we should be seeing them half decade or so. lol

1 Like

Can't wait for debugging where the act of looking for a bug determines if it exists or not!

1 Like

Not for sale, no. The Qubits have to be kept very close to absolute zero, so infrastructure is a problem. You can purchase quantum compute services from cloud providers, though, for big $$$. IBM and Microsoft are players.

These players offer emulators for you to practice quantum techniques at a more affordable scale.

1 Like

I believe with quantum states the proper word should be decatting

3 Likes

With this tangent topic, a cat joke had to get out of the box sooner or later.

3 Likes

Or two, or three.... Wait! Are quantum cat jokes even countable?

3 Likes

Counting changes the results according to Heisenberg.

I didn't know you could already pay to get your modeling done on quantum computing from those big players already @pturmel. I need to research if IBM has something competitive to AWS and Azure,
or if they use one of those or both in their cloud services.

How were the results @BRYAN_MOBLEY ?
I want to know if it saved you time getting some data, did something more difficult to do from queries, or showed something from the data that was hard to see.