Prove it on your operation.

Start with historical sensor data. We train on an early healthy interval, score later operation blind, and review whether Athena identifies known or previously unexplained degradation.