German, Swiss, Austrian Discussions
The task of this presentation is to make a large volume profile assessment in a large datatable with data coming from database (100.000 profiles).
The task is carried out by using JMP Pro Functional data explorer (FDE). The profile is separated in different features called “functional components” (e.g. FC1)by FDE, saved into a column for each feature. So a feature like a dip is converted into a scalar representing the existence of this feature (e.g. FC1 >> 0 = dip;FC ~ 0 = flat,<< 0 = hill). The functional component FC1 can then be used as a label for training with a neural net. The prediction from neural net isimplemented as a formula in the final data table to assess that curve.
Finally the engineer can easily update data from database, with having assessments updated as formula evaluation. The assessment seems to be very robustfor the data created in this case from a measurement device, as the Rsquare of neural net keeps quite high, even with new data for some month. Additionallya sensitivity analysis is carried out with artificially generated data by means of Copilot.
