Automated dynamic strain gage data reduction using fuzzy c-means clustering

This paper describes fuzzy c-means (FCM) applied to the automation of large data reduction and review tasks. A data processor has been developed which determines the number of distinct structural mode responses of airfoils in a turbomachine and groups all similar responses in order to facilitate the analysis of test results. Successful implementation of the processor has demonstrated a reduction of data analysis time by a factor of ten while eliminating much of the subjective interpretation and error resulting from the manual data review process. Cluster validity measures from unsupervised optimal fuzzy clustering methods have been incorporated such that no a priori assumptions about data set structure (e.g., number of clusters, range of responses) are necessary. An application to high pressure compressor rotor blade data is presented. The paper concludes with a discussion of future work to enhance processor performance