Better handle incompatible data type event in PI Vision trend
Scenario: Your company has a PI tag set up as string that primarily gets float values. It is a string because it sometimes contains messages explaining why a result was not obtained.
The tag is set up as a double attribute in AF and plotted in Vision. This works fine unless a non-numeric value is encountered within the Time Range you are looking at. If this happens, the trend is blank and No Data is displayed.
While PI annotations would have been a good use in this situation, it is difficult to go back and change this when you have many different tags with lots of historical data. It would be better if an incompatible data type event was handled like a system digital state and create a data gap instead.
Workaround: Create an Analysis that uses the Double() function to convert events to double. This creates a Calc Failed event when the event can't be converted to double.
Matt Voll commented
to reiterate a point above . . . one bad data point (coercion failure) will cause the ENTIRE trace to fail, whether its 1h or 30d. Regardless of whether that one data point should be handled differently, its seem rather problematic that one bad data point can cause the entire trace to fail