Identifying similar historical process patterns

Hi ​@TIM,

That's a textbook Similarity Search, and the part that makes or breaks the accuracy is the region weighting, so let me cover both the build and the tuning.

Creating it. Put your reference cycle in the focus chart select the temperature tag and the time window of the profile you want to match. Then in the search menu choose Similarity search → Create new, adjust the conditions, and hit Search. Results come back ranked by match score, best first (similarity search). 

Then tune it:

1. Anchor on the signature tag temperature here. Adding tags that don't define the shape just dilutes the match; keep the pattern to what actually characterises a cooking cycle.
2. Weight the part of the curve that matters — this is the lever most people skip. Highlight a region on the pattern (say the heat-up ramp or the peak-hold) and use the weight arrows to raise its importance the line thickens to show the weight level. Mismatches in that highlighted region are then penalised more heavily, so the search judges candidates on the feature that actually defines your profile instead of being dragged around by the flat or noisy stretches. Note the weighting applies to time-regions of the curve, not to tags.
3. Set the match threshold deliberately. Default is 70% match quality raise it for fewer, stricter results, lower it for a wider net. Start a touch below 70%, look at where the ranked list drops off into weak matches, then raise it to cut that tail.

Kind regards

Frederik