Rate of change processing of acoustic data from a permanent monitoring system for pipe crack early identification: a case study

Abstract

A permanent “leak-before-break” type acoustic monitoring system has been deployed by the South Australian Water Corporation across the Adelaide City Central Business District (CBD) since early 2017. Several analytic approaches applied to the collected acoustic data and results, in terms of reducing the number of uncontrolled pipe main breaks, have been reported. New statistical approaches to the analysis of the data are being continually developed with a statistical method based on the rate of change of the acoustic data within measured spectrums reported in this paper. The method involves the determination of short and long-term noise power level benchmarks, using noise power levels in an empirically determined number of windows with different frequency bin widths, to determine the rate of change of power levels across the frequency spectrum. The rate of change is quantified via the determination of normalized power level changes over weekly and monthly periods and the calculation of noise ratios relative to earlier weekly and monthly periods. Thresholds are applied to determine whether a weekly or monthly change alert should be raised for the investigation of potential cracks