Abstract
The industrial sector continuously demands innovation in sensing technology. Identification of flow conditions, gas/void detection, and porosity estimation are important factors in many liquid metal processes. This thesis aims to demonstrate the magnetic induction tomography (MIT) system in a liquid metal process model at a laboratory scale.The work designs MIT instrumentations which are able to acquire measurement data from eight coils system. The frequency range is between 100 Hz to 100 kHz enabling multi-frequency measurement. Its signal-to-noise ratio (SNR) reached 66 dB with a speed of 1.5 s/frame to produce real, imaginary, amplitude, and phase data and reconstructed images. On the software side, spatio-spectral image reconstruction algorithm has been formulated to do a spectrally correlated analysis identifying an object’s circumstances.
MIT sensors have been constructed for detecting and visualising liquid metal flow. Typical flow shapes have been successfully recovered with a correlation coefficient up to 0.9 and relative error as low as 0.2. In addition, a liquid metal shape classifier based on a neural network yields a test accuracy of 99%. As for interior voids in liquid metal, a convolutional neural network has been trained to quantify the number of non-metallic inclusions with 96% of test accuracy.
This research also develops a vector-based complex mutual inductance spectroscopic imaging and derives regional complex impedance diagrams. The resulting complex plots from the reconstruction comprehensively indicate the functional and structural characteristics of the metallic materials. Furthermore, this investigation for the first time demonstrates a novel thermal mapping system using eddy current based spectroscopic imaging data. Inductive based temperature mapping devices can have great potential applications where none of the existing thermal measuring devices could work noninvasively.
This study intends to contribute to the context of eddy current, imaging, and induction spectroscopy.
| Date of Award | 22 Jun 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Manuchehr Soleimani (Supervisor) & Ivan Astin (Supervisor) |