A team of engineers at Purdue University has created a patent-pending approach for reducing the risk of hazardous strikes to underground utility pipes during construction projects.

Designed to potentially lower associated financial losses, service disruptions, injuries and fatalities, the device is expected to improve upon traditional ground-penetrating radar (GPR) data to better determine the location, orientation and radius of underground pipes.

Source: Purdue University/Yuxi Zhang

The researchers explained that although GPR offers a non-destructive approach to locating pipes by analyzing the electromagnetic wave reflections and resulting hyperbolic reflections, it tends to overlook uncertainties and issues related to data quality.

“Our uncertainty-aware model compensates for this gap by robustly quantifying uncertainty in order to create a buffer zone,” the team added. “With this improvement, construction contractors, utility-locating service providers and excavator manufacturers can better interpret GPR data, supporting safer and more effective underground utility mapping.”

The team used a Bayesian framework to quantify uncertainty in estimating underground pipe depth, horizontal position, orientation and radius.

In doing so, the team created a forward model that forecasts electromagnetic wave travel times from 3D pipe geometries and a statistical inference framework to estimate pipe parameters and quantify uncertainty. The system also includes diagnostic metrics for assessing GPR data quality. Validation showed strong predictive accuracy, with a 0.454 ns root mean square error (RMSE) and high coverage rates — 97.5% on simulated data and 77.5% on field data — thereby demonstrating reliable pipe detection and estimation.

An article detailing the work, “Uncertainty-aware localization and orientation estimation of underground pipelines from GPR data,” appears in the journal Advanced Engineering Informatics.