Parametric analysis and spectral whitening of signals generated by leaks in water pipes [articol]

Abstract

This paper presents a way to determine the best modeling algorithms for working with signals generated by water pipe leaks. Three methods of parametric modeling are presented in this paper: auto-regressive modeling AR, moving average MA modeling and auto-regressive – moving average ARMA modeling. From these methods, the auto-regressinve modeling is the best one for analysing signal sequences from water pipe leaks. A special MATLAB Toolbox was used in order to work with the signals and the parametric models. The name of the Toolbox is ARMASA. Several programs were written in order to work with ARMASA functions and with the leak signals. The influence of signal length and number of estimation coefficients, are studied in order to show which parametric modeling method works best with signals generated by pipe leaks. The conclusion is that for these type of signals, the AR autoregressive model is the optimal solution. With the help of the obtained spectral distribution values, we can further analyze the signals in order to find with precision the position of the leaks. After the exact choice of a parametric modeling algorithm, in ths case the AR model, we are able to see the benefits of this choice when dealing with spectral analisys. Signal whitening can be used in order to improve the quality of the Cross Correlation Function (CCF).