Development of gas leak detection system using fuzzy logic, optical flow and neural networks

• Main Author: Carrillo, Edgar, II
• Format: text
• Language: English
• Published: Animo Repository 2015
• Subjects: Gas leakage; Neural networks (Computer science); Fuzzy logic
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/5833
• Institution: De La Salle University

Natural gas is increasingly in demand worldwide and also here in the Philippines. The Philippines has a lot of natural gas reserves and as the economy is improving, energy demand is also improving. Meeting the energy demand of the country is a challenge and securing for the efficient transfer of raw materials for the production of energy like in the pipeline system is also a challenge. Managing a pipeline system is critical to the growth of the Philippine Natural Gas industry as a mismanage pipeline causes a lot of worries from companies owning the oilfields. Aside from possible financial loses, the natural gas leaks are also known to be harmful to the environments as it contains methane. There are a lot of existing technology for pipe detection but most of them are found to be expensive especially the hardware based detectors. The challenge is to create a natural leak detector that is cheap and effective. This can be done using image processing approach as image processing is proven to be cheap and only the algorithm should be furnished. The proposed method of the researcher in detecting leaking is a leak detector that can detect the severity of leak in terms of no leak/low severity, average/medium severity and high leak severity. This is done by processing the image by the use of fuzzy logic, optical flow and neural network. Specifically, the study aims to create a prototype set-up in an aquarium based system for bubbles detection, development and implementation of a fuzzy logic algorithm, optical flow and neural networks for detecting leak severity and determination of the accuracy and robustness of the algorithm. The concept was to get data from prototype experimental set-up and capture it using a camera. The camera will then transmit the signal to the computer. The computer will then processed the data to detect severity of leak in terms of no leak/low severity, average/medium severity and high leak severity. The output of this proposed method was the algorithm that is cheap and accurate.