Abstract:

The periodic modulation of the rotational speed of a machine element, such as a bearing element or a gear, is recognized as a valuable feature in characterizing the physical condition of the machine component. A time-domain template-matching algorithm is introduced as a robust nonlinear frequency modulation estimation technique for extracting this feature from perturbations in the acoustic signal emitted from a damaged machine component. The technique is applied to both synthetic signals as well as real accelerometer and microphone data taken from damaged railroad car roller bearings. The results are compared with those obtained using a standard baseband phase demodulation technique. Although the template-matching algorithm is computationally more expensive, it can be applied directly to a limited-duration, raw input signal, unlike the baseband demodulation technique, which relies on bandpass filtering and synchronous averaging over a lengthy time duration. [See NOISE-CON Proceedings for full paper.]