Abstract
Magnetic induction tomography (MIT) is a non-intrusive and non-invasive method that can image the cross-sectional conductivity distribution inside an object without contact. This paper proposes an MIT system utilizing a gradiometer coil as the front-end sensor for conductive fluid testing. Theoretical analysis and numerical simulations demonstrate that phase sensitivity to conductivity is significantly enhanced by eliminating the primary voltage through differential coils. The system achieves a sensitivity of 2°/(S/m) at 10 MHz with sub-5 m° phase noise and high long-term stability. Experimental phantom tests validate the system's capability to reconstruct conductivity distributions using a one-step Laplacian regularization algorithm.
Introduction
Magnetic Induction Tomography (MIT) is a "soft-field" imaging technique ideal for medical applications, such as brain stroke detection, and industrial processes, like oil-water flow monitoring. However, MIT faces a significant technical challenge known as the "primary field" problem. In standard MIT, the secondary signal induced by eddy currents in the target is often orders of magnitude smaller than the primary excitation field. For low-conductivity fluids like saline, this discrepancy leads to a limited dynamic range and high noise levels, making accurate imaging difficult.
Methodology and Sensor Design
To overcome the primary field problem, the researchers developed a front-end sensor based on a gradiometer coil configuration. This design employs differential sensing by using two identical sensing coils placed symmetrically on either side of an excitation coil. By connecting these coils in anti-phase, the primary magnetic field is naturally cancelled out at the receiver, allowing the system to focus on the much smaller secondary signal.
The hardware stack operates at 10 MHz, utilizing a National Instruments PCI-5122 data acquisition board (100 MS/s) and a PCI-5412 waveform generator. Phase measurements are extracted using Fast Fourier Transforms (FFT) within a LabVIEW environment. To transform raw phase data into 2D images, the system employs a Finite Element Method (FEM) to solve the forward problem and a one-step Laplacian regularization algorithm for the inverse problem.
Theoretical Insights: The Linearity Paradox
A key finding of this research is that a perfectly balanced gradiometer system is inherently non-linear. The authors discovered that introducing a small, deliberate asymmetry—resulting in a "residual voltage" ($V_{res}$)—actually linearizes the phase response. This makes the system more robust against mechanical vibrations and easier to calibrate.
There exists a "Goldilocks zone" for this residual voltage: if $V_{res}$ is too small, the system becomes hyper-sensitive but unstable; if too large, the benefits of the gradiometer's primary field cancellation are lost. By optimizing this balance, the team achieved a phase noise of less than 5 m°, significantly outperforming many industrial systems that struggle to stay below 100 m°.
Experimental Results and Stability
The system was evaluated using experimental phantoms consisting of saline-filled plastic bottles (2 S/m) in various configurations. Key performance metrics include:
- Sensitivity: 2° / (S·m⁻¹) at 10 MHz.
- Stability: Phase drift was maintained under 20 m° over a 4-hour period, essential for long-term industrial monitoring.
- Imaging Accuracy: The system successfully resolved multiple objects within a 12-cm sensing area. The use of an appropriate regularization coefficient (α) was shown to significantly improve the sharpness of the reconstructed conductivity maps.
Conclusion and Future Work
While the proposed gradiometer-based MIT system offers high sensitivity and stability, current limitations include acquisition speed due to serial channel switching (20 seconds per frame). Future iterations will focus on parallel data acquisition to enable real-time monitoring of rapid fluid dynamics and extending the 2D model to 3D volumetric imaging.
Summary updated