“FaDO does Fault Detection in realtime (‘Online’).”
Fault detection lies at the very core of machine learning, and by extension to AI at large. Think like this: if a machine encounters a fault, it comes at the edge of what it learned thus far. So there are two ways forward:
- Update the learning with the new, unexpected piece of information (‘constructive fault‘), increasing the power of the machine. In short: we found room for improvement!
- Accept the fault as an unwanted deviation (‘mistake‘), aligning with the very purpose of the fault detector.
FADO is about trading (1) and (2) in a suitable way.
This brings us with the key advantage and uniqueness of this approach: FADO starts from the very first step, and does not wait to be effective until suitable (‘training’) data is found. This is crucial in many applications where normal behavior is not necessary old/historic data, and hoarding old data would simply not fit the business. In engineering lingua, the approach is real-time or online.