disclaimer: due to Pi system performance issues we had a lot of intermediary delays of several hours. In the end, all functionality was deployed, but we did not get to the point of solving the DEME use cases through and through. You will notice we've put down a solid ground work though!
At the PiWORLD EMEA Conference TrendMiner announces "ContextHub" that imports Event frames as a starting point for further analysis. Examples here are LIMS quality information or basic Batch metadata. Results from TrendMiner analytics such as deviating behaviour from a previously defined golden batch profile however generate context items of their own. At the Hack-a-Thon we wanted to demonstrate how "easy" it is to build an EF-writer and sync that data back into Event Frames as your master data store. This way other solutions can also take advantage of TrendMiner analytics through the PI software stack. As a cherry on top of the cake we also wanted to make the circle round (and turtles all the way down) by creating a PiVision dashboard where those events are listed and where you can open up your events back in TrendMiner.
Once TrendMiner got connected to the Pi Archive, we focused on Use Case 2 where the windmill building platform does not have context data about start and stop events like Jacking UP and Jacking Down phases. In a couple of minutes time, by using 1 example time, TrendMiner's pattern recognition capabilities (aka 'similarity search') could easily identify those periods. Once we had all these search results, we create the necessary Context Items.
In picture 1 above you can see that we have a start and end annotation in our trendview that indicates a context item.
In picture 2 we have created a view with all context items (events) for this asset. You'll notice that due to the lack of time we were only able to create 2 types but you'll get the idea.
You can also view these events in a gantt chart visualisation in case events are clustered in time. If we created different context types we'd get different rows in the gantt chart per stage.
Pi Event Frames
Using the PI AF SDK and webAPI we were then able to write the data back into Event Frames. You'll notice the exact same table view as we had in ContextHub.
Lastly we had to visualise these event frames in a table view for PiVision. Since we added a TrendMiner link as a variable of the event frame we could not use an existing Vison Symbol and created a Pi Vison Custom Symbol. For a future iteration we could definitely add a basic trendview for that specific event in this dashboard.
Here's that quick-and-dirty table view of those events. If you click on the link ...
it opens up that event in TrendMiner. Unfortunately for our Custom Symbol we had to use a single tag and could not use the asset but that's for a next iteration ;-)
For now we can just add a tag-set from a saved view.