I have a project running on a batch type machine. This machine iterates 4-5 runs a day, everyday. I’m looking to expand on my current project by integrating some form of machine learning. I’m looking for an ideal solution that’s easily integrated with Ignition, and capable of analyzing basic machine functions (currents, temperatures, pressures), recipe parameters in real-time, etc. With this running in the background, I’d like to leverage the analysis to improve predictive maintenance, predictive process quality, and even preventative QC if the knowledge base is large enough.
I don’t believe you understand the scope of the project you are attempting to accomplish. I believe essentially you are talking about building this from the ground up. There are no platforms that are recommended to achieve this and if there were they would most likely be a paid service, I believe you may want to go a different route unless this is a requirement
We used Aveva's Prism for basically this, feed in training data and then it would monitor our assets and produce an overall quality value that we would alert if it deviated too much. It worked relatively well and found multiple big >million dollar fail avoidances. However you have to understand that it's not a one person project, it took many stages (years) of designing the virtual model relationships, parsing training data, active monitoring and rework, and then working with site to handle alerts.
If you want to just play around with this on an individual level and learn, there is a machine learning package on the Ignition exchange. It can be a fun topic to play around with and in the era of more powerful AI computing it's an emerging field of opportunity. But I wouldn't commit to expecting any actionable results from it.
Even at a massive scale we had a ton of false positives alerting but the understanding was dealing with those was worth it for the couple things we caught that saved big dollars.