Insider Brief:
- Researchers at Shibaura Institute of Technology have demonstrated that machine learning, particularly the random forest algorithm, can accurately estimate the depth of the bearing layer—the stable soil or rock layer critical for building foundations in earthquake-prone regions.
- Analyzing 942 geological surveys from the Tokyo metropolitan area, the team compared random forest (RF), artificial neural networks (ANN), and support vector machines (SVM) using two test scenarios: one with location and elevation data, and another adding stratigraphic information.
- The study, published in Machine Learning and Knowledge Extraction, found the approach could replace time-consuming and costly standard penetration tests (SPT) in urban areas like Tokyo.
Tokyo researchers have turned to artificial intelligence to speed up and cut the cost of assessing soil conditions critical to earthquake safety.
A team from Shibaura Institute of Technology (SIT) has shown that machine learning models, particularly the random forest algorithm, can accurately estimate the depth of the “bearing layer” — the stable layer of soil or rock that supports a building’s foundation. Their study, published in Machine Learning and Knowledge Extraction, suggests the approach could help engineers in earthquake-prone regions design safer buildings without the time and expense of traditional testing.
In cities like Tokyo, where the risk of major seismic events is high, knowing the bearing layer depth is essential to preventing soil liquefaction — a condition in which the ground loses stiffness during shaking and behaves like a liquid. The standard penetration test (SPT) is the industry benchmark for measuring this depth, but it requires drilling and sampling, which can be slow and costly.
“The inspiration for this research stemmed from the pressing challenges in geotechnical engineering within earthquake-vulnerable urban landscapes like Tokyo,” noted Professor Shinya Inazumi, who led the SIT team. “As a region with a history of devastating seismic events, such as the 1923 Great Kanto Earthquake, accurate prediction of bearing layer depth is vital. Through our research study, we hope to empower urban planners and engineers with efficient tools for sustainable development, reducing costs and enhancing safety.”
The researchers analyzed 942 geological survey records from the Tokyo metropolitan area, pairing SPT results with three machine learning models: random forest (RF), artificial neural networks (ANN), and support vector machines (SVM).
The team ran two test scenarios. The first used only location and elevation data. The second added stratigraphic information — details about the composition and layering of underground soils. In both cases, the random forest model outperformed the others, with the second scenario delivering the highest accuracy. The RF model’s mean absolute error was 0.86 meters in the more data-rich case, compared with 1.26 meters when using location and elevation alone.
Adding more data points improved results further, researchers noted. When the researchers increased the spatial density of their input data to as much as three points per square kilometer, prediction accuracy rose noticeably. The finding suggests that denser regional datasets could make AI-based predictions even more reliable.
Machine learning offers several advantages over SPT, researchers said. Once models are trained, they can run quickly on existing geological and mapping data, eliminating the need for repetitive, on-site drilling. That could allow city planners to assess large areas rapidly and at lower cost, making it easier to identify safe building sites or prioritize areas for reinforcement, reseaarchers pointed out.
“Our findings highlight the transformative real-world potential of ML models in geotechnical engineering and urban planning, especially in earthquake-prone regions like Tokyo,” said Inazumi. “By combining ML with existing geological data, stakeholders can optimize site selection for resilient smart cities and other infrastructure projects, such as bridges or subways, with rapid, scalable simulations.”
