Application of a data fusion based methodology for online detection of health status and defect type in the bearing of a grinding machine is presented. In practice, knowing the exact defect status and type is infeasible. Information regarding the current health status and defect type of a bearing may help in building prognosis models. As the proposed detection methodology is based on data fusion, dependence on a single damage identification parameter is obviated. The fused data parameter takes into account the correlation among all the damage identification parameters considered. Diagnosis of a bearing with naturally induced and progressed defect may have multiple complexities. Typically used condition monitoring parameters, such as R.M.S. and peak may not have monotonically increasing trends. In the case of natural defects, one type of the defect may be prominent in the initial phase and later on, another type of defect may outgrow the first one or both may exist simultaneously. The methodology is verified with the help of a dataset acquired from a naturally induced and progressed defect on an accelerated test rig. The bearing is dismantled after the experiment to confirm the defect type identified through the method.
展开▼
