Early Fault Detection at Gear Units by Acoustic Emission and Wavelet Analysis

To secure the availability of machines and facilities, the reaction to beginning damage must be as fast as possible. Thus, early detection of damage initiation is crucial to the quick planning and execution of reconditioning measures. This helps to minimize downtime and increase availability. Modern high performance transmissions in fabrication and energy industries more and more have to satisfy high requirements concerning their nominal load, running properties and operational stability. Helical cut gears with small modules are usually used to meet these demands. To assure an undisturbed operation, to avoid unplanned downtimes and consequential damages of high performance transmissions a condition monitoring and fault diagnosis is useful. Vibration analysis is a good tool for detecting faults and unacceptable operating conditions of machines in an early stage, due to its large dynamic range when using spectral and correlation analysis. Detecting a defect in a machine by vibration analysis is however not possible before the degree of damage is already affecting its vibration characteristics. Using classical vibration analysis, especially the detection of cracks and their propagation in rotating shafts and gear wheels, is possible only with relatively short forecasting times before failure. Early detection of cracks in shafts and gear wheels is possible by acquisition and analysis of acoustic emissions (AE). Unlike accelerometers that capture the physical behavior of the component, like lowor high-frequency vibrations up to 200 kHz with a linear frequency response, an AE sensor is very sensitive in its resonances at higher frequencies in order to detect ultrasonic impulses caused by changes within the material’s structure like crack formation and crack propagation [1, 2]. For the detection of gear defects in transmissions by AE analysis, the signal path should be as short as possible. This minimizes interferences, like the influences of ball bearings and other machine parts and emphasizes on useful information about faults. At a gear wheel test bench, acoustic emission sensors were placed on the gearbox casing and additionally on the ends of the rotating shafts in order to detect defects at an early stage as well as enabling their location and the determination of the affected component. Because of the superimposed emissions of the rolling element bearings and other components it is not sufficient to regard only the number and intensity of the emissions. Therefore an additional evaluation of the short-time excitations in frequency domain is necessary using wavelet analysis. Compared with an FFT, wavelet analysis offers a higher resolution in time domain especially for high frequent events. The crack initiation and crack propagation at the root of a tooth shows early changes in the wavelet plot, which are periodic with rotational speed. Propagating pitting shows a different behavior. Both different faults show a characteristic development during the lifetime of a gear wheel.