Using combined temporal and spatial information in multi-hypothesis based damage detection

Damage detection is very important in today's fast-turning light weight rotating machinery. High standards being put in front of designers necessitate the incorporation of new emerging technologies exploiting low-cost embedded sensors and microprocessors. These new tools make it possible to include sophisticated damage detectors that report on the condition of the machinery. The diagnosis system will provide valuable information about the state of the machine and report when damage is developing. Given this knowledge the extent of the damage can be assessed and the remaining life time can be determined. Careful modeling of rotating machines allows one to predict the influence of each failure mechanism on the dynamic response. However the effect of this damage may often be buried within the measured information (due to the complex dynamic behavior of rotating machines). In this paper it is proposed to enhance the delectability of the various faults by using a two stage approach. First we employ our physical understanding of the dynamic behavior in presence of a fault. This understanding is used to magnify certain aspects of the dynamics most attributed to the sought malfunction. At a second stage, we propose to utilize the already demonstrated idea of Kalman-filtering-based multi-hypothesis testing to detect the location and extent of the damage. The physical understanding on which this work is based upon is the fact that different faults manifest themselves in different ways in the spatial, temporal and frequency domains. Appropriate processing of the measured data coming from several sensors allows for detection of the damage. The initial steps of the proposed diagnosis scheme is developed for a rotor with a cracked shaft. It is shown that the the total response is not a good indicator for the presence of a crack more sentitive measures for the presence of a crack exploit forward and bacward separation of the rotor's whirl and inspection of high harmonics of the unbalance response. Those high-order harmonics are a result of the parametric excitation caused by the time-varying nature of the crack. These points are demonstrated by a numerical simulation of a model of a crack and by measured data and thus provide sgood foundation for further research.