Flow Measurement by Wavelet Packet Analysis of Sound Emissions

Hüseyin Göksu

doi:10.1177/0020294018768340

Abstract

Fluid, when running through pipes, makes a complex sound emission whose parameters change nonlinearly with respect to flow speed. Especially, in household pipe systems, there may be spraying effects and resonance effects which make the emission more complex. We present a novel approach for predicting flow speed based on wavelet packet analysis of sound emissions rather than traditional time and frequency domain methods. Wavelet packet analysis, by providing arbitrary time–frequency resolution, enables analyzing signals of stationary and non-stationary nature. It has better time representation than Fourier analysis and better high-frequency resolution than wavelet analysis. Wavelet packet analysis subimages are further analyzed to obtain feature vectors of norm entropy. These feature vectors are fed into a multilayer perceptron for prediction. Prediction accuracy of 98.62%, with 3.99E−04 L/s mean absolute error and its corresponding 1.85% relative error is achieved. Time sensitivity is ±0.453 s and is open to improvement by varying window width. The result indicates that the proposed method is a good candidate for flow measurement by acoustic analysis.

Highlights

Flow of a fluid can produce complex sound emissions in household pipe systems because of the spraying effects, resonances in the pipe structures and other mechanical interactions including friction
They defined the signal noise as the standard deviation of the frequency averaged time-series signal and presented experimental results that indicate a nearly quadratic relationship between the signal noise and the mass flow rate in the pipe; Safari and Tavassoli[7] showed empirical results that indicate that there is a relationship between the output signal of microphone in frequency domain and the flow rate in a pipe; Medeiros and Barbosa[8] presented a method for measuring flow in pipelines based on the frequency domain values of vibration caused by the flow of water; and Hu et al.[9] used accelerometer on a pipe to recognize the type of household water use activity
We look for a mapping tool from the feature vectors to the flow speed

Summary

Introduction

Flow of a fluid can produce complex sound emissions in household pipe systems because of the spraying effects, resonances in the pipe structures and other mechanical interactions including friction. An average error of 15% was determined; Kim et al.[11] introduced a nonintrusive autonomous water monitoring system based on pipe dynamic theories to household plumbing in the form of an optimization problem and achieved flow measurement with 7% mean absolute error; and Kim et al.[12] used parameter estimation via numerical optimization technique to estimate flow rate with 5% average error

Methods

Results

Conclusion