An optimized virtual beam-based event-oriented algorithm for multiple fault localization in vibrating structures

Abstract

This paper investigates a new multiple fault localization method using the virtual beam-based event-oriented algorithm. The virtual beams, represented by some sensor chains, are recently proposed to isolate the potential faults according to the distribution of sensors to certain regions and their associated nonlinear dynamic feature characteristics. To improve the accuracy of fault localization and isolate the multiple faults at more specific and detailed regions, in this paper, the virtual beam-based method is combined with a specially developed feature characterization algorithm named as the subtract on negative add on positive (SNAP). This newly proposed virtual beam-based subtract on negative add on positive (VB-SNAP) method is to take advantage of the virtual beam-based algorithm to narrow the search scope for the potential fault events using fewer sensors and adopt the event estimation strategy of the binary-based SNAP method to enhance the accuracy of fault localization by providing smaller and specific regions containing the potential faults. Two likelihood matrices are constructed according to the alarm states of sensors and the response of sensors from the virtual beams, respectively. Fault localization of the VB-SNAP is, therefore, simplified to estimate the overlapping cells obtained from those two likelihood matrices. The performance of the proposed virtual beam-based event-oriented algorithm is demonstrated through real experimental results. It is shown that the proposed VB-SNAP is superior to the original virtual beam algorithm and SNAP method alone through the localization of multiple bolts loosening in solar panel of a satellite-like structure.