Detection and elimination of strain reading anomalies in distributed strain sensing readings

Distributed strain sensing (DSS) is increasingly used in structural health monitoring (SHM). Unlike conventional measurement technology, DSS allows for quasi-continuous measurement of strains with high spatial resolution, enabling the recording of global strain changes in the structure. However, DSS readings are subject to process-related interference, such as dropouts and strain reading anomalies (SRA). SRA can hinder data analysis during postprocessing or even falsify the results. Because of high amount of data produced by DSS, it is crucial to detect and remove SRA from DSS readings during automated preprocessing. Outlier removing methods from literature were analysed for their potential application to DSS readings. Based on this analysis, a sliding median z-score had been proposed, implemented in the fosanalysis framework and applied to various DSS readings for validation and to determine suitable parameters.