Enhancing Urban Resilience Against Liquefaction through Machine Learning Models

Researchers at Shibaura Institute of Technology in Japan have developed machine learning models to enhance urban resilience against liquefaction—a significant risk in earthquake-prone areas. Their study produces accurate 3D soil maps detailing the depth of bearing layers, which help city planners identify stable construction sites. This innovative approach uses artificial neural networks and bagging techniques, incorporated with data from 433 locations in Setagaya, Tokyo, to predict soil stability effectively, ultimately contributing to safer urban development.

Urban resilience against liquefaction—an issue of critical importance for earthquake-prone regions—is being enhanced through advanced machine learning models developed by researchers at Shibaura Institute of Technology in Japan. Their innovative approach involves presenting contour maps that delineate the depth of soil bearing layers, which significantly aids city planners in identifying zones susceptible to liquefaction. In Japan, where seismic activity is a persistent concern, there exists an urgent need for precise predictions regarding soil stability to preemptively address liquefaction risks. Leveraging machine learning techniques, including artificial neural networks (ANNs) and ensemble learning strategies such as bagging, researchers analyzed data collected from 433 testing sites within Setagaya, Tokyo, to formulate accurate 3D representations of soil bearing layers. These models diligently pinpoint stable construction sites, bolster disaster preparedness, and ultimately promote safer urban development, thereby fortifying cities against liquefaction threats. As urbanization intensifies, the risk of natural disasters escalates, making it imperative for urban planners and disaster management authorities to mitigate potential hazards. Liquefaction poses a significant danger to infrastructure; it occurs when intense seismic activity causes water-saturated, loose soils to lose their integrity, thereby acting like a viscous liquid. This dangerous phenomenon can lead to numerous structural failures, including the sinking of buildings, fracture of foundations, and damage to essential utilities such as water mains. The destructive ramifications of liquefaction are historically evident, as seen in the aftermath of major earthquakes. For instance, the 2011 Tōhoku earthquake triggered liquefaction that imparted damage to approximately 1,000 homes. Similarly, the Christchurch earthquake of 6.2 magnitude in 2011 saw liquefaction obliterate 80% of the water and sewage systems, whereas the 2024 Noto earthquake led to widespread liquefaction affecting 6,700 houses. To enhance cities’ resilience, Professor Shinya Inazumi and his student Yuxin Cong have harnessed machine learning to anticipate how soil behaves during seismic events. Their initiative utilizes geological data to construct detailed three-dimensional maps of soil layers, distinguishing stable zones from those highly susceptible to liquefaction. This methodology surpasses traditional manual soil testing capabilities, offering a comprehensive assessment of spatial soil behavior. Published in the journal Smart Cities on October 8, 2024, their recent study utilized ANNs and ensemble learning to accurately predict the depth of soil bearing layers, a pivotal factor in assessing soil stability and liquefaction susceptibility during earthquakes. Prof. Inazumi asserts, “This study establishes a high-precision prediction method for unknown points and areas, demonstrating the significant potential of machine learning in geotechnical engineering. These improved prediction models facilitate safer and more efficient infrastructure planning, which is critical for earthquake-prone regions, ultimately contributing to the development of safer and smarter cities.” The researchers compiled bearing depth data from 433 sites in Setagaya-ku, Tokyo, employing standard penetration tests paired with mini-ram sounding tests. This consisted not only of the depth of the bearing layer but also pertinent geographical details including longitude, latitude, and elevation. Using the compiled data, the research team trained an ANN to forecast the bearing layer depth at 10 different locations, then cross-referenced these predictions with actual site measurements to evaluate their accuracy. Employing bagging—a technique that iteratively trains the model on varied data subsets—resulted in a notable enhancement of prediction precision by 20%. Illustrating their findings, the researchers generated a contour map to visualize the soil bearing depths in a 1 km radius surrounding four selected sites in Setagaya Ward. This map serves as an invaluable resource for civil engineers in locating suitable construction sites characterized by stable soil conditions while simultaneously aiding disaster management professionals in identifying regions at greater risk of soil liquefaction, thus enabling improved risk evaluation and response strategies. Envisioning their methodology as instrumental to the evolution of smart cities, the researchers advocate for data-driven approaches that inform urban planning and infrastructure development. Prof. Inazumi optimistically states, “This study provides a foundation for safer, more efficient, and cost-effective urban development. By integrating advanced AI models into geotechnical analysis, smart cities can better mitigate liquefaction risks and strengthen overall urban resilience.” Looking toward the future, the researchers plan to augment their model’s predictive accuracy by including additional ground condition variables and tailoring specialized models for coastal and non-coastal regions, accounting for the influence of groundwater—an essential element influencing liquefaction dynamics.

The topic of urban resilience against liquefaction is particularly relevant in the context of Japan, which experiences frequent seismic activity. Liquefaction is a significant hazard in these areas as it undermines the stability of infrastructure during earthquakes. The phenomenon involves loose, water-saturated soils behaving like a liquid under intense shaking, leading to devastating structural damage. Machine learning, specifically models that generate 3D maps of soil layers, presents a novel and effective strategy for understanding and predicting how urban environments will respond to seismic events, thus facilitating better planning and disaster management efforts. The work conducted by Professor Shinya Inazumi and his student Yuxin Cong exemplifies the integration of advanced technology in enhancing geotechnical engineering and urban safety.

The research executed by Professor Shinya Inazumi and Yuxin Cong highlights the substantial advancements that machine learning can bring to urban planning, particularly in earthquake-prone regions. By generating detailed 3D soil maps and improving prediction accuracies for soil stability, this study not only enhances our understanding of liquefaction risks but also provides critical tools for city planners and disaster response teams. As urban areas continue to expand in the face of climate change and seismic threats, such data-driven methodologies will prove invaluable in fostering safer, more resilient cities.

Original Source: www.preventionweb.net