Exploring and supporting Ground Penetrating Radar Deployment
Guest User
Project type || PhD-project
Candidate || Ramon ter Huurne
Project duration || Feb 2019 - Feb 2023
In collaboration with: Alliander, GasUnie, MapXact
Description:
Excavation damages pose a significant challenge for the construction sector. In 2022 alone, nearly 47 thousand such incidents were reported. These excavation damages result in costly repair works, disruption of utilities, environmental damage, and safety risks. Although methods exist to detect cables and pipelines before excavation works commence, they often fall short. Utility maps can be inaccurate or incomplete, while trial pits only offer local insights and are also disruptive. Ground penetrating radar (GPR), on the other hand, provides a non-disruptive alternative that could potentially enhance the practice of cable and pipeline detection, thereby reducing the number of excavation damages.
Illustrative for the disruptive effect and high labor costs of a trial trench
However, to effectively utilize GPR, a unique combination of structural expertise and geophysical knowledge is required, which is often lacking in practice. This leads to unsuccessful applications of GPR, discouraging organizations from considering it for future construction projects. As a result, the potential of GPR within utility inspection practices remains largely untapped.
This thesis aims to improve utility inspection practices using GPR through the development, evaluation, and implementation of decision models powered by machine learning. It includes five research outcomes, divided into phases of problem exploration and solution development.
In the problem exploration phase, deeper insight was gained into decision-making regarding the deployment of GPR. The perspective of 'technology-in-practice' was used. By utilizing the theoretical frameworks of routine dynamics and Cultural-Historical Activity Theory (CHAT), the use of GPR was conceptualized, including considerations of when, where, and how to apply it. This has resulted in three scientific studies that:
... explain how the Dutch detection practice, including the role of GPR, looks;
... outline the 'triggers' that facilitate exploration and adoption of GPR;
... present how GPR can add value to future activity systems.
For the solution development phase, input was obtained from the identified applications of GPR. This phase involved preparing a dataset with all collected fieldwork data and developing and evaluating various types of machine learning-driven decision models. This has resulted in two scientific publications that:
... present a dataset of 125 detection activities in which GPR was used;
... outline which machine learning model (Case-Based Reasoning, Decision Trees, Random Forest, and Support Vector Machine) performs best in supporting ground radar use.
The findings of this thesis and the automation solutions can accelerate the adoption of GPR, improving the effectiveness, efficiency, and safety of cable and pipeline detection and potentially contributing to reducing damage to utilities due to excavation works.