Wear is traditionally measured offline. A new methodology for online detection and monitoring of wear has been investigated in this thesis. This methodology consists of design of an online wear testing apparatus and development of techniques for online wear detection and monitoring using imaging methods. A simple, cost-effective, laboratory-use apparatus for online wear testing and monitoring was designed, which combines a simplified Pin-on-Disk wear tester and an imaging system. Using the designed wear apparatus, dry sliding wear test for three contact modes (point, line and conforming) can be conducted, meanwhile images of a surface in motion can be acquired in real time. Online wear detection and monitoring can be performed by analyzing a sequence of images of a wearing surface. This wear apparatus offers more accurate results than traditional offline wear measurement instruments in the sense of neither interrupting the dynamic wear process nor changing wear environment or conditions. Furthermore, it is fast, noncontact, cost effective and easy to operate. Considering the crucial role illumination plays in an imaging system, a fiber optic ring light, according to the choice of illumination strategy, was chosen as a proper illuminator for the imaging system, under which the wear features of interest were revealed. An illumination model and reflectance model were then developed for ring light illumination. Both models were numerically evaluated and validated by experiments. Good agreement was found between the models and measurements. A wear process alters surface parameters, and results in changes of surface radiance, image irradiance and image texture thereafter. The proposed models helped to interpret wear patterns present on the specimen’s surface images and provided theoretical basis for detection and monitoring of wear by analyzing image sequences. Wear detection on machined surfaces can be regarded as a texture segmentation problem. Two filtering approaches, namely unsupervised detection scheme using multichannel Gabor filters and supervised detection scheme using optimized filters were developed for online wear detection. Both approaches can successfully detect wear patterns present on textured surfaces, machined surfaces for instance. Experiments conducted on various real wear samples confirmed the usefulness of these two approaches. Monitoring of a dynamic wear process can be conducted by analyzing a sequence of images from a wearing surface. Wear behavior of an ad hoc dry sliding wear process was investigated under different operating conditions. Fractal analysis technique was also applied to estimate the state of wear in dry sliding situation. Experimental results showed that fractal values, namely fractal dimensions and vertical intercepts computed from power spectra of a sequence of images of a wearing surface are potential good indicators of state of wear.
|Award date||11 Sep 2006|
|Place of Publication||Enschede|
|Publication status||Published - 11 Sep 2006|