A customer recently had a large gas turbine driven multi-stage centrifugal compressor suddenly fail with no apparent prior indication of an issue. The compressor is one of three in parallel designed to compress natural gas from medium and high pressure to pipeline pressure. The turbine engine’s gas producer failed when one of the inlet air turbine blades worked loose and was ingested by the engine along with much of the subsequent stage components. The inlet guide vanes had been sticking which allowed the turbine blades to flutter excessively during changes in engine load. After the failure the customer requested Sygnology to review historical data to provide an indication of how long prior to failure was there indication of blade flutter.
Bearing vibration data and tachometer data with nominal 1 second sample frequency was processed going back three months. Signals were present six weeks prior to final failure for relatively short periods of time throughout the six weeks. These periods were correlated to other process and performance indicators and it was learned the periods where signals were strong occurred when engine load was declining.
Continuous automated proactive monitoring can identify and alert engineering, technical diagnostics and maintenance persons to issues like this before serious damage has occurred. Often, early indications will occur days and weeks prior to the catastrophic event, providing plenty of time for thorough diagnosis.
Turbine technicians and engineers used this information in their lookback review and to improve maintenance and monitoring procedures