How to see what Spall has learned about the plant
Wiki: Improve concepts, ML insights and value opportunities
When to use this
You want a single, plant-wide answer to “what has Spall actually learned so far”, not just one machine’s card: which machines already have a working model of normal behavior, which are still building history, and what’s missing before the rest catch up.
Steps
- Click Improvement, then Learning (
/learning), right next to Opportunities. - Read the summary line at the top: how many machines Spall has learned normal behavior for out of the total, and which machines are still building history.
- Scroll the machine table for the detail behind that summary. Each row shows:
- the models trained for that machine, grouped by family (normal behavior, precursors and the rest) with a count for each, click a family to expand it and see every model and its sample count
- when it was last retrained
- whether a golden run (its best sustained pace) has been found
- how many leading-indicator precursors it has found
- how many open findings it currently has On a phone, tap a machine name and the same detail stacks underneath it.
- Click any machine’s name to open its own page, where the same models and findings appear alongside its live status.
- Check What it needs to learn more, below the table. It lists, for each machine still building history, exactly what hasn’t been learned yet, plain language, and once Spall has tried and not yet succeeded, the real reason why (for example, how many readings it has seen against how many it needs), not a guess at the underlying cause. Spall retrains every hour from live floor data on its own. Nothing on this page is configured by hand, and nothing here is invented: a machine with too little history simply isn’t shown as learned yet.