Bridging the Gap: Integrating Neural Radiance Fields and Micro-drones for Enhanced 3D Volumetric Finite Element Analysis

   Civil Engineering

  , Dr Elisa Bertolesi  Applications accepted all year round  Self-Funded PhD Students Only

About the Project

Objective: The central objective of this project is the innovative deployment of Neural Radiance Fields (NeRFs) and micro-drones in the process of 3D volumetric Finite Element modelling of bridges. This approach seeks to address and alleviate traditional difficulties encountered in bridge modelling and inspection.

NeRFs Overview: NeRFs, stemming from the domains of machine learning and computer vision, present an exceptional technique for synthesising new perspectives of intricate 3D scenes from a limited set of initial images. The method involves training a deep neural network to convert 3D coordinates into a colour and opacity, thereby learning a continuous function of the volumetric scene. This function can subsequently be sampled from new viewpoints to produce further images of the scene.

Application of NeRFs to 3D Volumetric Mesh Element Model Generation: In this project, we propose using NeRFs for the generation of 3D volumetric mesh element models of bridges. Micro-drones capture a series of images from various vantage points, which serve as the data input. These images are then used to train a NeRF, resulting in a continuous 3D function representing the bridge. This function can be sampled to create a detailed 3D volumetric mesh of the bridge, which is a crucial prerequisite for any finite element analysis.

Existing Bridge Modelling Challenges: Developing accurate models of existing bridges is a multifaceted challenge, as it necessitates extensive data on the bridge's geometry and material properties. Obtaining such data can be a complex and resource-intensive task, particularly for large or intricate structures. Traditional data collection methods like manual inspection or laser scanning often involve significant costs and can fail to capture all essential details. Transforming this data into a Finite Element model further requires specialised knowledge and software.

Micro-drones: The utilisation of micro-drones offers a novel solution to the aforementioned data collection challenges, by accessing difficult-to-reach bridge sections and capturing high-resolution images from diverse perspectives. This approach can significantly decrease the cost and time required for data collection, while simultaneously enhancing the quality and comprehensiveness of the data.

Addressing the Issue: This project presents a novel strategy for bridge modelling by combining the strengths of micro-drones and NeRFs. The intention is to develop a methodology for creating accurate 3D volumetric finite element models of bridges that is faster, more detailed, and more cost-effective than conventional approaches. Drones will capture the required data, and the NeRFs will process this data into models. This opens up new possibilities for bridge inspection and maintenance.

Potential Impact: The proposed project has far-reaching implications. By improving the efficiency, accuracy, and cost-effectiveness of bridge modelling and inspection, it could significantly enhance the way we manage and repair our infrastructure. The methods developed could also be applied more broadly within structural engineering, potentially reshaping the way we model and analyse various structures.

Ideal Candidate: The successful candidate for this project will be an innovative thinker with a robust background in Structural Engineering or related fields. Proficiency in Finite Element Modelling is essential, and familiarity with IoT, drone operations and image processing would be advantageous. The candidate should possess strong analytical skills, the capacity to work autonomously, and have a keen interest in connecting traditional engineering practices with emerging technologies. Experience or a willingness to learn machine learning methods would also be desirable.

Computer Science (8) Engineering (12) Mathematics (25)

Register your interest for this project