Applying AI in acoustic-based sensor systems for gas leakage detection
dashboard Objectives
In the sRMC project, gas leakages are to be detected with the help of acoustic sensor systems using Deep Learning. Leakage scenarios are to be simulated at a small test facility. In addition, microphones and head noise sensors will be used to record acoustic data. In order to simulate the real plant case, disturbance noises caused by machines and pumps are to be simulated first. After the simulation, the test facility will also be positioned in real industrial plants in order to record leakage scenarios under real conditions. With the help of various machine learning approaches, the self-generated leakage database will be expanded with artificial training data to increase the probability of detection.
In the future, gas leakages could be detected across the board with the help of sensor systems and artificial intelligence. This is what we are researching in the sRMC project.
Figure: Leakage monitoring with and without sRMC.
dashboard MILESTONES
Vulnerability analysis
Definition of an early detection system
Development of an early detection system
Validation of the system at a test facility
dashboard Overview
Data collection
Data transformation
Machine Learning
dashboard Motivation
The current geopolitical situation and the resulting problematic energy dependency of individual countries are putting a further focus on minimising energy losses. The economic cost reduction is intensified by the energy saving package of the federal government.
In the future, more and more storage systems will be operated with hydrogen, or the pipeline network will be used with hydrogen instead of natural gas. The technology shift away from fossil fuels to renewable energies makes energy storage in energy carriers such as hydrogen necessary. Hydrogen is transported and stored at high pressures of up to 800 bar. In addition to the seemingly insignificant cost reduction due to energy losses, hydrogen leakage detection is of great importance because of the potential risk to human life. The research project will help to support and improve humans in the process of plant monitoring and replace them in the near future.
ashboard PUBLICATIONS
Publications coming soon.
PROJECT LEAD:
STAFF | CONTACT:
person Deniz Quick, M.Sc.
email Send email
phone +49 721 6699 4780
FOCUS AREAS:
More Projects
SmOP – Smart Overpressure Protection Device
Intelligent safeguarding of chemical reactors with adaptive safety devices.
RiIM – Risk-based Pipeline Integrity Management
Developing an AI-powered, risk-based pipeline integrity management system.
EuroValve – European Program on Evaluation of Safety Valve Stability
Development of a new safety criterion for safety valves.
Cover photo source "Tube Geometry", RiIM project: Harald Hoyer from Schwerin, Germany, CC BY-SA 2.0, via Wikimedia Commons
Cover photo source "Winstainforth", SafeDDT project, CC BY-SA 3.0, via Wikimedia Commons