Application of Ground Penetrating Radar (GPR) Combined with Artificial Intelligence to Eliminate Cable Breaks and Subsidence During Urban Construction

The "Black Labyrinth" Beneath Urban Grounds and Millions of Dollars in Damage: The project's client is a general contractor for urban transportation and water supply and drainage infrastructure, undertaking a contract to excavate utility trenches and install 35 km of underground pipelines running through densely populated central districts.

The biggest challenge hindering all urban infrastructure projects is the lack of accurate data about the "underground world" beneath the asphalt. Over decades of overlapping development, the urban subsurface is a complex network of high-voltage power cables, telecommunications fiber optic cables, clean water pipes, abandoned old sewers, and geological sinkholes. As-built drawings from 20-30 years ago are severely inaccurate. Traditionally, contractors have to send workers to dig trial pits every 50 meters. This method leads to two serious consequences:

  1. A series of cable and pipe rupture accidents: Excavators working in sections between two exploratory pits repeatedly struck national fiber optic cables or ruptured high-pressure water pipes, resulting in contractors incurring billions of dong in compensation fines and having construction suspended for weeks to rectify the problems.
  2. Risk of road subsidence: Failure to detect underground water pockets or geological voids beneath the road surface beforehand can cause the asphalt to collapse unexpectedly when heavy trucks pass over it, directly threatening traffic safety.

MAPPING THE GROUND USING GPR RADAR AND DEEP LEARNING To make the dense underground appear "transparent" before lowering the excavator bucket, the geophysics engineering team deployed a 3D Array Ground Penetrating Radar (GPR) solution integrated with Artificial Intelligence:

  1. Data collection using a towed multi-frequency radar system: Instead of blocking roads to dig exploratory trenches, a dual-polarized radar antenna array is mounted on the rear of a survey vehicle moving normally with traffic at a speed of 40 km/h. The device emits ultra-high frequency electromagnetic pulses (from 200 MHz to 2 GHz) that penetrate deep into the ground to a depth of up to 8 meters. When the waves hit objects with different dielectric constants (such as metal, PVC plastic, concrete, air voids, or water pockets), they reflect back to the receiver along with millimeter-level accurate RTK-GPS satellite positioning coordinates.
  2. Decoding Hyperbolic Spectral Images with Artificial Intelligence (AI Radargram Interpretation): The biggest bottleneck of traditional GPR technology is that the data returned consists only of confusing black and white parabolic curves (hyperbolas); an experienced geologist would spend months staring at a screen trying to guess which curves are rocks and which are plastic pipes. The technology team fed all the raw wave data into a convolutional neural network (CNN) model trained on millions of subsurface cross-sections. In just a few hours of processing, the AI ​​algorithm automatically filtered out noise from damp clay, accurately distinguished the material of each pipe (cast iron pipe, HDPE plastic pipe, or copper cable bundle), and, most importantly, precisely pinpointed the voids (Voids) forming beneath the asphalt concrete layer.
  3. Creating a 3D underground BIM model and guiding the excavator: The entire AI analysis result is automatically built into a 3D spatial model (3D Subsurface BIM). This model is loaded directly onto a tablet screen located in the excavator driver's cab. When the excavator's bucket approaches an underground high-voltage power cable within a 50 cm radius, the system in the cab immediately sounds a red alarm and automatically applies hydraulic pressure to the boom.

BREAKTHROUGH IN SAFETY AND INFRASTRUCTURE CONSTRUCTION PROGRESS After scanning the entire 35 km project route in just 10 days and nights, the GPR-AI solution has delivered outstanding economic efficiency for the main contractor:

  • 95% Reduction in Underground Construction Accidents: Throughout the 14 months of excavation and pipe installation, the project did not experience any telecommunications cable breaks or main water pipe ruptures, saving tens of billions of VND in compensation and administrative penalties.
  • Early detection of 18 potential sinkholes: Thanks to the discovery of geological voids caused by long-standing underground sewer leaks, the contractor proactively injected grout to reinforce the road surface in time before it collapsed.
  • Shortening the overall project timeline by 4 months: Completely eliminating 80% of the manual excavation work, doubling the speed of mechanized construction compared to traditional standards.


Need a reliable technology partner for your business? See DUDI's dedicated tech team & white label.