Hybrid EIT Method Combines Math and Machine Learning for Improved Building Structure Analysis
Researchers at the Tokyo University of Science (TUS) have developed a new hybrid approach that combines traditional mathematical methods with cutting-edge machine learning algorithms to improve the analysis of building structures. The method, called AND, aims to overcome the challenges of accurately reconstructing information obtained through electrical impedance tomography (EIT), a non-destructive imaging technique used to visualize the interior of materials.
EIT works by injecting an electric current between two electrodes, creating an electric field, and measuring distortions caused by the presence of foreign objects inside the material. Compared to other imaging methods, such as X-ray imaging and computed tomography, EIT is low cost and less cumbersome as it does not require large magnets or radiation. This makes it a promising technique for non-destructive structural health monitoring of cement-based building materials.
However, accurately reconstructing the obtained information as images has been a challenge in EIT. Existing mathematical algorithms used for reconstruction, such as one-step Gauss-Newton and iterative Gauss-Newton methods, often produce solutions with inaccuracies. To address this issue, researchers have recently started using machine learning algorithms, such as one-dimensional convolutional neural networks (1D-CNN). While these algorithms show promise, they struggle to handle previously unseen data, reducing their effectiveness.
To overcome these challenges, the TUS researchers developed the hybrid EIT approach called AND, which combines the benefits of the iterative Gauss-Newton method and 1D-CNN. The researchers tested their AND method on cement samples using both simulation and experimental data and compared its performance to that of the iterative Gauss-Newton and 1D-CNN methods.
The results showed that the AND method was more accurate in detecting small foreign objects inside the materials compared to the other two methods. When the cross-sectional area ratio of the foreign body to the sample was very small, the AND method reduced the size error to less than 1/6 of that using the conventional EIT method.
According to Associate Professor Takashi Ikuno from TUS, the new method has significant implications for disaster prevention and the analysis of existing structures. The improved application of EIT as a non-destructive testing method can help prevent building collapses and contribute to building health assessment.
The AND method performs 2D logical operations on multiple images obtained from EIT to detect small foreign objects. The researchers also identified another method to further improve the accuracy of EIT by changing the current injection pattern and combining the approach with other non-destructive evaluation techniques.
The proposed EIT reconstruction method, although not as high-resolution as other non-destructive evaluation methods, has the advantage of being cost-effective and easily deployable. It holds great potential for non-destructive foreign object detection, enabling regular building health assessments and rapid safety screening after earthquakes or explosions.
The researchers plan to continue their research to enhance the resolution and accuracy of the proposed EIT method by exploring different injection patterns and combining it with other techniques. They also believe that this technology can be easily adopted by inspectors and personnel.
Overall, the development of the hybrid EIT method marks a significant advancement in building structure analysis. With further improvements and refinements, EIT has the potential to become an important detection technique for preventing building collapses and ensuring the safety of structures in the future.
About Tokyo University of Science (TUS)
Tokyo University of Science (TUS) is the largest science-specialized private research university in Japan, with a mission to create science and technology for the harmonious development of nature, human beings, and society. Established in 1881, TUS has been at the forefront of scientific research and education, producing Nobel Prize winners and contributing to Japan’s development in various fields.
Funding Information
This study was financially supported by the AGC Foundation.
