Condition health monitoring and its application to cavitation detection/characterization within hydropower turbines

Abstract

Hydroelectric power has been the number one renewable energy source in the U.S. since the beginning of the industrial revolution and continues to be today. Hydroelectricity is a critical component in the power production grid to keep greenhouse gas emissions and pollution minimized. As such, it is crucial that unexpected shutdowns and unplanned maintenance of hydropower turbines be kept to a minimum, so as to maximize hydroelectricity production. This thesis aims to investigate condition health monitoring (CHM) methods specifically designed for non-intrusive cavitation detection within hydropower turbines. Cavitation is a highly damaging phenomenon common within turbines. When allowed to continue undetected over an extended period of time, cavitation can lead to severe and crippling effects for efficient operation. The application of CHM will lead to less downtime and ultimately more electrical production from hydropower turbines, resulting in the maximization of the U.S.'s number one renewable energy source's potential. An instrumented cavitation inducing apparatus was designed and built for laboratory testing. The goal of the cavitation inducing apparatus was to produce both non-cavitating and cavitating flows within the available flow range. Also, it was critical for the apparatus to be simple and allow the instrumentation utilized to be placed as close as possible to the cavitation within the flow. Instrumentation including pressure transducers, accelerometers and acoustic emission sensors were used to non intrusively record cavitation signals from the cavitation apparatus. Multiple signal processing techniques, spanning both the time and frequency domains were utilized to develop methods and metrics to quantify the cavitation monitoring data. Most of the techniques are well documented, including analyzing the root mean square values of the signals and utilizing the Fast Fourier Transform for frequency domain analysis. There were also some signal processing techniques developed throughout this project, specifically for cavitation monitoring. The metrics and methods developed proved successful at identifying volatile flow rates and subsequently the onset of cavitation state change with the flow. It was also determined that time domain signal processing techniques were more successful at cavitation characterization than frequency domain techniques. There is confidence the methods developed for non-intrusive cavitation monitoring through this thesis could be easy transferred to on site operational test data received from a cavitating turbine and successfully diagnose the onset of cavitation with the flow range.