Model-based SHM of power plants rotating machinery

Vibration analysis has found widespread application for condition monitoring in a variety of applications and industries. With the continual development of cheaper and more powerful processing hardware, such systems have developed from utilizing simple checks on amplitude to those based around sophisticated spectral analysis. Most current systems that are employed for rotating machines operate by examining the gain and phase of the spectral components at both the fundamental frequency of rotation and a number of higher order harmonics. These components are then compared with normal signatures and any deviation is used to declare possibility of a fault. Differences in specific spectral lines, for example harmonics associated with blade passing or gear meshing frequencies, can be used to diagnose particular faults. Such analysis, however, has proven to be difficult to automate and relies on the off-line interpretation of a trained vibration expert. The paper presents application of the model based diagnostics method for early detection of faults in rotating machinery. The proposed diagnostic system based on two methods - modal analysis (OMA and OMAX methods) and non-linear signals models (NARX). The paper presents the application of the powerful class of non-linear signal processing tools known as NARX in rotating machinery diagnostic. The paper also presents an application of OMAX method in rotating machinery diagnostic. The OMAX method is formulated with an assumption that both measurable and immeasurable forces excite shaft vibrations. The proposed system was verified during research on a specialized test rig, which can generate vibration signals and on data recorded at wind turbine.