Clustering Approaches to Automatic Modal Parameter Estimation

Most modal parameter estimation techniques utilize Stabilization/Consistency Diagram as a tool for distinguishing between physical system modes and mathematical modes. However, this process results in giving several estimates of the same mode and the task of choosing one estimate over others is left to the user. This task is highly judgmental, with user expertise playing a big role as to which estimate is being selected. It can be very tedious especially in situations when the data is difficult to analyze (low signal to noise ratio, closely spaced modes, heavily damped modes etc). One of the ways to get around this issue is to incorporate smart selection of estimates in the algorithm itself, so as to avoid the user interaction which, as stated previously, can be very subjective. In this paper two clustering based approaches are suggested for the purpose of automatic mode selection. These methods build upon the existing Stabilization Diagram tool; differing in the manner in which the stabilization diagram is constructed and clusters are being formed. Both approaches utilize a Euclidian distance based approach to automatically select the best estimate.