Tensorflow Integration

Does Ignition support Machine Learning frameworks such as Tensorflow?

What would support for Tensorflow look like to you?

1 Like

So, for the moment it’s very basic: My primary question is how does Tensorflow, or any ML/AI framework, interact with the core logic of the Ignition platform supervisory capabilities? Other questions: Are there any limitations, that you (inductive automation) are aware of (Ignition <–> Tensorflow)? If so, can you (briefly) reply to the most important?
Basically, we are trying to implement a AI/ML platform for renewable energy applications, that means stand-alone power plants with minimal human interaction. We like what we see with the Ignition platform, but we would like to ensure compatibility before we move forward, or at least be reasonably sure that there are no major incompatibility issues.

I’ve read this 3 times and I still can’t really figure out what you’re asking.

Maybe you can define your question in terms of what kinds of data is an input to tensorflow and an output from Ignition, and/or the other way around.

1 Like

To go into that level of detail would get too involved (for the time being). Perhaps a better way of asking the question is do you see any major issues or incompatibilities with integration between Ignition and AI/ML frameworks such as Tensorflow? There is no information (that I could find, anyway) that covers Ignition integration or control under AI/ML. This was a round-about inquiry if it’s at all possible.

We have no Tensorflow or other ML-specific integrations, but the realtime and historical data in Ignition is easily accessible.

I can’t really tell you anymore than that if you can’t provide actual details. There certainly aren’t just doMachineLearning() hooks everywhere in the platform if that’s what you’re asking, but again, I really can’t tell.

I’ll circle back when I have more time to devote to this question. Thank you!

There’s also a module SDK that lets you hook into the platform and get access to a bunch of stuff you might not otherwise have access to, which could be another avenue for integration.

Thank you!! Could you point to that SDK? That may be the solution.

It’s not so much a thing you download as it is a set of outdated documentation and examples…


The actual API artifacts are hosted in a Nexus server that your project will automatically download when it resolves dependencies, assuming you use a sane build tool.

Thank you very much!

Hah! And where would one find this sane build tool? :stuck_out_tongue:


I would maybe check out the couple of Webinars that have been done on this topic by IA. They dont dive into the details too much but have some good examples. There are a couple of Apache Common Math libraries that have the ability to do some basic machine learning tasks. There is an example of a K-Means algorithm, and the library seems to have the ability to do some neural net type applications, but I haven’t had time to sort through how just yet.

I think the main issue that you’d run into trying to use a Tensorflow API is that the Java version is still experimental, so it may not be as supported as the other libraries they have.

@R.Alamsha posted this in another thread a while back. I’m not sure on it’s validity, but it may be worth looking into.

You may want to check out one of our strategic partners, Cirrus Link. They have modules to link your Ignition data to a few of the major cloud ML platforms (AWS, Azure, etc.)

We do have plans to add more ML integration in the future, but those plans aren’t yet to the stage where I can say anything meaningful about them. :slight_smile:

1 Like

I wonder. Are these plans more concrete now?
It surely would be interesting to be able to simply add the Python TensorFlow library and do the machine learning in Ignition.
I think with the easy visualisations in Ignition, you could really get some good value out of it.