Advanced Model Reference Thermal Sensing Technique for Detecting Rusty Gas Pipes Embedded Inside Concrete Walls
DescriptionThe Hong Kong and China Gas Company urgently needs an accurate, detailed, timely and cost-effective method of fault detection for short-ranged pipe sections embedded in the wall. The goal of this project aims at developing an innovative NDT technique based on the changes of thermal conductivity in the rusty gas pipes. There are 5 research tasks required to successfully accomplish this goal. Task1: finite element models will be developed using state-of-the-art ANSYS software to numerically investigate change in pipe thermal conductivity due to different fault scenarios, and pipe conditions. Please find in the attachment, a sample of the preliminary ANSYS models of this problem already being developed in our group at CityU. Task2: experimental investigation of these pipe and fault scenarios using high-speed Infra-camera and thermocouples will be carried out in our lab for validation purposes. After successful validation, further numerical simulation will be carried out to develop reference models by considering various pipe fault sizes, depths and locations as well as various ranges of room temperatures. Task3: post processing of the numerical results and development of software database and algorithm of the post processed results using MATLAB software. Task4 will involve the use of adaptive thermal excitation technique, and advanced signal processing technique such as Hidden Markov Model (HMM), for enhancing features extraction of generated signals due to change in thermal conductivity in a precise, well-defined manner to avoid feature mismatch and false judgment. To do this, the so-called quality factor of the actual signal will be obtained first by estimating the signal/noise ratio (SNR) to guarantee the effectiveness of our developed software. Then the HMM parameter estimation will be carried out based on the Baum–Welch method (expectation-maximization method) to uncover and maximize the underlying system statistical expectation model. Finally, in Task5, artificial intelligence (AI) scheme based on Fuzzy Logic which will characterize pipe temperature difference based on fault size, depth and location will be developed and utilized in our model. This will no doubt provide a cost-effective and user-friendly method for identifying the integrity of pipes going through wall.
|Effective start/end date||1/03/19 → …|