| Rotor system,as an important component of rotating machineries,bears the impact of high thermal load,high speeds,interchange loads and other harsh environment during its service period,and so inevitably has various types of mechanical failures such as cracks,imbalances and collision faults.A crack fault changes slowly from an initiation status to deterioration status,but it could cause sudden fracture of the shaft and serious safety accidents when it expands to the deeper extent.The research of crack fault diagnoses of a rotor system has important academic significance and engineering value for the health detection of military,naval,civil and other rotating machinery.In this paper,a horizontally cracked rotor system with gravity and nonlinear spring characteristics is taken as the research object,and vibration response characteristics,time-domain and frequency-domain characteristic parameters,fault diagnosis effects of the system are innovatively compared from the perspectives of vibration displacements and vibration energies.The incremental two-dimensional principal component analysis(I2DPCA)and energy space method are first applied to fault diagnoses of the cracked rotor in the high-speed region,which provides a new solution strategy for online realtime health monitoring of large rotating machinery.Based on the proposed viewpoint of vibration energy,energy expressions of three typical rotor models including a Jeffcott rotor,a linear cracked rotor and a nonlinear cracked rotor are derived,and the differences in mechanical properties of different types of support bearings at the shaft end are compared.Combined with vibration energy analysis methods,vibration characteristics of above three different types of rotor systems in the vibration space and energy characteristics in the energy space are compared and analyzed.It is pointed out that they are consistent in reflecting vibration behaviors of the system,and energy space analyses have a greater amplitude advantage.The cracked rotor system with nonlinear spring restoring forces is further studied.Based on the vibration space and energy space,the identifiability between the vibration characteristics,energy characteristics,time-domain and frequency-domain characteristic parameters of the system at different crack depths,eccentric positions and rotating speeds are explored respectively.(1)Time histories,spectrum analyses,energy orbits and other graphical methods are used to intuitively show that vibration responses of the system with different fault parameters have greater differentiation in the energy space,and the energy space analyses strengthen high-order harmonic frequency components caused by the crack.(2)Using time-domain and frequency-domain characteristic indexes from a quantitative perspective finds that some indicators have lost the ability to represent the system and become invalid indicators in the vibration space,this phenomenon is more obvious in the high-speed area.While the values of each characteristic parameter still change with the vary of system parameters in the energy space,indicating that it has more abundant fault information,which provides the theoretical support for subsequent diagnosis results of applying the I2 DPCA algorithm.In the process of using the wavelet method to denoise vibration signals of the cracked rotor,the influence of different frequencies and threshold rules on the denoise effect is analyzed,and finally the intelligent fault diagnosis process is formed including signal acquisition,wavelet denoise,vibration-image conversion,I2 DPCA and KNN algorithm.The convergence,feature visualization,classification rate are analyzed by using experiments.Results show that the recognition rate of crack faults in the highspeed area is as high as 99.6% under the energy space based on the I2 DPCA algorithm,and it has good convergence and online learning ability in the case of small samples,and each performance index is better than the diagnosis effect in vibration space The experimental research is carried out on the platform of cracked rotor,which verifies the feasibility and efficiency of the proposed method. |