Amplitude distortion of measured leak noise signals caused by instrumentation: Effects on leak detection in water pipes using the cross-correlation method

Abstract

A common way to detect and locate leaks in buried water pipes is to use leak noise correlators. Vibration or acoustic signals are measured on or in the pipe using sensors placed either side of the leak, and the difference in the leak noise arrival times (time delay) at the sensors is estimated from the peak in the cross-correlation function of these signals. Over many years, much effort has been spent on improving the quality of the leak noise signals with the aim of improving the time delay estimate. In this paper it is shown that even if the signals suffer from severe amplitude distortion through either clipping or quantization, then an accurate time delay estimate can be obtained provided that the zero crossings in the noise data are preserved. This is demonstrated by using polarity co-incidence correlation on simulated and measured data. The use of random telegraph theory is also used as an approximation to allow the derivation of approximate analytical solutions for the cross-correlation function and cross spectral density of clipped noise to facilitate further insight into the effects of severe clipping.