A smart low-cost non-invasive water flow sensor for IoT applications

Project Team

Dr. Mohamed F. Abdelkader

Dr. Mohamed F. Abdelkader

Port Said University

Principal Investigator

Dr. Amr Elbanhawy

Dr. Amr Elbanhawy

Ain Shams University

Co-PI

Sponser​

Project Objectives

  • Develop a new generation of smart non-invasive hydraulic flow sensors that can be integrated with smart water management initiatives in smart cities and smart home applications.
  • Achieve low-cost, acceptable levels of precision, non-invasive ease of installation, and minimum calibration procedure for the developed smart sensors using recent advances in machine learning and signal processing

Specific Objectives

  • Develop a new integrated digital non-invasive water flow sensor for IoT applications.

  • Develop a new machine learning and signal processing approaches for water flow estimation form single or multiple sensor modalities.

  • Build the necessary human and infrastructure capacity for the national interest area of smart water applications.

  • Establish a smart water flow monitoring test-bed for further research and technology development

  • Establish a multidisciplinary multi-institutional collaboration link between the machine learning and signal processing research team at Port Said university and the mechanical sound and vibration lab at Ain Shams university.

Research Methodology

  • Physical and Numerical Testbench: building a flowmeter rig with computational fluid dynamics (CFD) simulation software

  • Experimental analysis of different sensor modalities for velocity and pressure monitoring

  • Develop a new data sampling approach to reduce power consumption: 

    • use of novel data sampling approaches such as Compressive sensing for reducing the hardware and power constraints on the IoT nodes connected to the new sensors.

  • Develop a novel data-driven machine learning flow estimation approach

    • investigate the effectiveness of applying deep learning approaches to build a data driven flow regression algorithms from different low-cost sensors. The hydrodynamic models will be used to generate test sequences for initial model training, further model tuning will be conducted with real measurement obtained through the developed testbench.