We have developed a methodology for AI-based monitoring and tested it using real vibration measurement data. What sets this methodology apart is that it evaluates the reliability of each individual processing step and passes the results on to the subsequent analysis steps. This makes it possible to output classification results from the AI diagnostics along with metrics indicating their reliability. The procedure was developed methodically and tested and validated using measurement data from the real-world laboratory, without incorporating the specifics of the use case into the method itself. This makes it possible to apply the method to other use cases.
more info