We have worked with fleet management system. In general, there are lots of similarity of fleet data for many similar equipment. Instead of analyzing individual machine data, you might want to use similarity based data analytics. One example is to predict the equipment maintenance using the remaining useful life. Please see the tools from this link Conference Paper A similarity-based prognostics approach for Remaining Useful...
In fleet-based maintenance system, people normally use peer-to-peer approach to manage predictive maintenance. In wind farm fleet, a turbine-to-turbine technique has been developed, You can use this methodology as a reference. Patent Turbine-to-turbine prognostics technique for wind farms
Dear Professor Jay Lee . The paper is very interesting, thanks for sharing.
The similarity can help to identify various anomalies related to the maintenance and useful life of equipment.
During the research I have identified the generation of inconsistent data and the occurrence of false alarms that can compromise the results. What is the best way to identify / treat possible outliers?