Condition based maintenance is a strategy that, if implemented effectively can extend the operational life of rotating machines. The philosophy originates from more traditional, time-based maintenance strategies where activities are scheduled based on the running hours of a machine. Instead, with condition based maintenance the scheduling is optimized such that these activities are only carried out when the condition of the machine indicates certain maintenance or repairs are required. The condition can be determined with a variety of monitoring technologies including those where the data is recorded when the machine is off-line (while not in operation) or on-line (when the machine is in operation). Generally, off-line testing is performed infrequently at larger time intervals compared to on-line monitoring because of the lost revenues when the machines are not in operation and the added expense of disassembly. For on-line measurements periodic or continuous monitoring can be employed and usually depends on the technology and resource availability. Ideally, multiple technologies are incorporated into a common software platform for easier analysis and data correlation. For condition assessment there are a variety of internationally accepted standards for absolute limits as well as acceptable changes to trend. Any trend analysis should only be considered when the machine is operating in similar conditions. This underlies the importance of trending operating parameters along with the monitored data. Additionally, some failure mechanisms can only be recognized with data collected at a variety of operating conditions (e.g. changing load) often done during troubleshooting activities.
This paper will focus on on-line monitoring technologies for large generators driven by steam/gas turbines and hydraulic turbines. Several case studies will be presented in a way to emphasize the importance of operating conditions for both trend analysis and when troubleshooting the machine condition. The technologies will include magnetic flux for rotor shorted turns, stator endwinding vibration, shaft and bearing vibration, and rotor to stator air gap (for hydraulic turbine driven generators). It will be seen that depending on the technology different operating parameters will affect the monitored data differently, some with a significant influence. If the data is collected periodically at predefined intervals there should be a mechanism to manually input these values for future reference. Conversely with continuous monitoring it is often more convenient to use modern communications protocols (e.g. Modbus over TCP/IP) due to the potentially large number of data sets to input the operating parameters at the time of data collection.