Detection of breathing crack in rotor system from the directional frequency response functions and dynamic stiffnesses

Abstract

Recently, the significance of timely fault diagnosis of rotating machinery is growing because it can enhance the reliability and safety of machine operation, and cut the machinery maintenance costs. A considerable amount of the rotor system failure over the past four decades can be attributed to the shaft cracks. The presence of a crack may lead to a catastrophic failure of rotating machinery and cause great economic loss. Thus, the incipient detection of a rotor crack will not only avoid severe damage and expensive repairs due to the malfunction of rotor system but assure the safety of personnel. Among other diagnosis methods, a vibration-based method, especially modal testing, has been considered as a reliable method for many years because it can perform fault diagnosis of machines without dismantling them. However, an effective method to extract physical information that shows both degree and location of crack from modal testing results has not been researched. Therefore, the primary objective of this thesis is to develop effective methods that can detect a breathing crack from complex modal testing of rotating machinery. For the approach to identify fault information, this thesis employs the directional frequency response functions (dFRFs), which can evaluate anisotropy and asymmetry of a rotating machine, and the directional dynamic stiffnesses (dDSs), that is the inverse of the dFRFs, in the complex coordinates. The anisotropy and asymmetry indicate non-axisymmetry of stationary and rotating parts of rotating machines, respectively. As such, the dFRFs and dDSs can be used as an excellent indicator to represent the faults of rotating machines. When a rotor system possesses both anisotropy and asymmetry, conventional coordinate transformation cannot eliminate the periodically time-varying coefficients in the system equation of motion. A finite degree-of-freedom time-varying equation can be transformed into an infinite degree-of-freedom time-invariant on...