Loading...

Designing, Manufacturing and Evaluating Water Flow Measuring System Based on Audio Signal Analysis

Akbari, Alireza | 2024

152 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57452 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Aryanpour, Masoud
  7. Abstract:
  8. The ever-increasing applications of fluids in residential areas as well as in main industries require investment to advance and optimize fluid control and their real-time flow rate monitoring. Numerous methods have been developed for measuring fluid flow rates, particularly for water. This study focuses on the design and evaluation of a sound-based water flow measurement system. As the first test, a microphone was placed on the water pipe, and through time-domain analysis, a second-degree relationship was found between the pressure amplitude of the signal and the flow rate to increase the accuracy of signal processing. Analysis of the signal-to-noise ratio revealed that the vibration of the water valve handle caused the dominant frequency in the whole recorded spectrum. Then the acoustic measurement device was installed on the water valve, and the results were collected to train a Machine Learning models and predict flow rate. The acoustic data were transformed using the MFCC algorithm into the frequency coefficients, which were the input data for Classification Machine Learning models. Efficiencies of 6 Machine Learning models were estimated including: LDA, QDA, LR, SVM, DT and KNN. Logistic Regression Classifier was identified ,with an accuracy of 93%, as the best model for estimating the water flow rate in the current research. Furthermore, effect of 8 parameters in the fluid flow plus 2 parameters of the pipe and the valve, namely the internal and the external factors, on the received sound signal were studied and estimated quantitatively. Estimations of the natural frequencies the faucet and the pipe revealed that the dominant frequency in the sound signals stemmed from the vibrations in the internal structure of the water valve. Using two models of CFD simulations and literature-based analysis, we studied and compared effects of the internal factors including: flow velocity, turbulence intensity, surface roughness, elbow redirect, forward-facing step height, turbulence length scale, and cavity dimensions. We showed that using simple 2D simulations, one can identify the location of the strongest flow fluctuations in piping parts such as elbow and investigate mechanisms of sound generation
  9. Keywords:
  10. Flow Measurement ; Flow Induced Vibration ; Structure Borne Noise ; Detecting Voice ; Machine Learning Applications ; Acoustic Device ; Flow Detection ; Fluids Flow Signal Processing ; Non-Intrusive Flow Measurement

 Digital Object List

 Bookmark

No TOC