WASHINGTON — Researchers have developed an intelligent monitoring pipe that combines optical sensing with machine learning algorithms to monitor and predict 3D soil settlement. With more development, the system could help provide early warnings of risks from soil settlement, helping prevent pipeline displacement, accidents from building cracks and even structure collapse.
"Soil settlement directly endangers the safety of engineering structures like buildings, bridges, pipelines and slopes," said the research team leader Dandan Sun from Shanxi University in China. "Our device overcomes the problems of traditional soil monitoring by providing precision 3D measurements and capturing dynamic changes using a simple structure that can be directly buried in soil."
In the Optica Publishing Group journal Optics Express , the researchers describe their intelligent monitoring pipe, which was made by fitting a PVC pipe with 3D-printed protective structures, temperature compensation components and fiber Bragg grating (FBG) arrays. They show that machine learning models can detect early soil collapse and predict progression by analyzing data collected from the intelligent monitoring pipe system.
"One application of this technology could be in old urban communities, which are often built on soft or unstable soil," said Sun. "By monitoring the 3D settlement trajectory of a building's foundation in real time and predicting, in advance, whether it will enter a dangerous stage, problems could be fixed before they become dangerous. The method could also be useful for landslide detection and monitoring the structural health of bridges as well as railway or highway subgrade settlement in challenging environments."
Capturing 3D soil changes in real-time
Although monitoring soil settlement can help detect risks early, current methods either take measurements from a single point, don't provide real-time tracking of the dynamic evolution of settlement or can't account for complex and realistic local soil conditions.
A type of soil known as loess, which is found in various regions around the world, is particularly challenging for today's soil settlement methods. This wind-blown silt is very loose and highly prone to erosion, collapse and settlement, especially after rain or construction.
To create a better monitoring device that could also be used with loess soil, the researchers designed a sensor based on arrays of FBGs. FBGs are extremely sensitive to tiny deformations caused by soil extrusion and settlement while also being immune to electromagnetic interference and robust in harsh environments such as soil. To further protect the sensor and ensure stable measurements, the team fabricated 3D-printed structures that shield the FBGs from soil when buried.
A FBG is a small section of optical fiber containing tiny, periodic variations in its refractive index. FBGs reflect light strongly at a specific wavelength, and when the fiber is stretched or compressed by soil movement, the spacing of the grating changes. This shifts the reflected wavelength, which can be converted into soil deformation signals.
"Instead of a single grating, we used two sets of five-point grating arrays attached to the pipe at a 45° cross angle, plus an independent temperature compensation grating to eliminate the influence of temperature fluctuations," said Sun. "This makes strain measurement more accurate and enables the capture of multi-directional deformation signals."
Combining the single-point signals from individual FBGs using the Frenet-Serret frame — a way to describe how a curve moves and bends in 3D space — allowed the researchers to reconstruct the overall 3D morphology and trajectory of soil settlement.
Detecting simulated soil settlement
To verify the sensor's performance, the researchers conducted indoor air tests that showed that the wavelength shift of the gratings exhibited a perfect linear relationship with strain and reliably detected very small deformations. The sensor also accurately captured strain signals under different angles and displacements, verifying its multi-directional perception ability.
Next, they tested the sensors using laboratory-based soil burial tests. This involved burying the intelligent monitoring pipe in loess soil in a test chamber with six water bags placed in the soil. Soil voids were created by gradually draining the bags to simulate soil settlement. The Frenet frame was used for 3D morphology reconstruction, and monitoring data was fed into machine learning algorithms for stage prediction.
The sensor system accurately identified various stages of soil settlement induced by drainage and captured the nonlinear deformation of soil particles from ordered arrangements to sudden sliding when drainage exceeded 8,000 milliliters. Additionally, the 3D morphology reconstruction results were highly consistent with the actual settlement morphology, clearly reconstructing the 3D trajectory of soil settlement.
The researchers experimented with various machine learning models, finding that the Random Forest algorithm performed best. It produced a settlement stage classification accuracy of 95.65% and a relative error of only 4.02% in predicting drainage-induced settlement volume.
To further optimize the technology, the researchers plan to carry out field tests around foundations of rural and urban buildings on the Loess Plateau in China, slope mining areas of open-pit coal mines and along municipal pipelines. They are also working to optimize the device by making it smaller and more integrated, adding wireless and remote functionality and making the device more affordable. They also want to develop software that could be used for real-time visualization of 3D settlement trajectories, automatic early warning of settlement stages and long-term data storage to make the system easy for engineering personnel to use.
Paper: L. Xie, M. Liu, J. Mao, H. Liu, Y. Yu, P. Chen, Z. Zhao, Y. Fu, D. Sun, J. Ma, "Fiber Bragg Grating-Integrated Soil Settlement Three-Dimensional Trajectory Pipe Sensor: Dynamic Soil Subsidence Evolution and Stage Prediction," Opt. Express, 34, XXXX (2026).
DOI: 10.1364/OE.589254
About Optics Express
