TrendMiner supports fingerprint deviation monitoring.
By applying this monitor feature you will be alerted to batch runs that deviate from golden batch behavior. You can also be alerted to process transitions that do not match a reference behavior.
To learn more about fingerprints, please refer to the fingerprints article.
Set up fingerprint for deviation monitoring
- Create and save a fingerprint of the standard process behavior of interest.
- This can act as the golden batch profile for a given product, or the golden transition behavior of a process.
- You can also save a search criterion that will serve as the trigger condition for reference monitoring.
- In the golden batch case, this could be a value based condition search that returns all batches of the corresponding product. In the golden transition case, this could be a step search that returns all transitions of interest.
Note: the trigger is used to align the fingerprint with the actual data. The trigger and the fingerprint are aligned at the start of both the fingerprint and the defined trigger.
To set up a process monitor:
1. In the monitoring menu 
, select the fingerprint of interest and select the deviation option by clicking the corresponding tile:
2. Select the appropriate trigger from the saved searches list.
3. Define a proper detection threshold.
Note: The detection threshold is a measure of similarity. This measure is calculated based on the sum of the residuals to the fingerprint. This is very similar to the similarity score, used in the similarity search. I.e. when the detection threshold is 100%, no deviation at any point is allowed. The data has to be within the ranges of the fingerprint at every point. More information on how to interpret the similarity scores, can be found in the article: how should I interpret similarity scores.
- Select one or more notification options (email, inbox).
- Toggle on the switch for the fingerprint pattern to start monitoring.
Alerts may be delayed by up to 2 minutes. Email alerts contain a direct link to the detected behavior.