Next-Generation Robotic Electromagnetic Non-Destructive Evaluation for Welding, Additive Manufacturing, and Automated Fiber Placement

About the Project

Conventional electromagnetic (EM) NDE methods are well established for surface and near-surface inspection of metallic components, but their deployment remains largely manual, offline, and poorly integrated with modern manufacturing automation. As a result, current EM inspection practices struggle to address complex geometries, as-built variability, and dynamic environments typical of Industry 4.0. These limitations are particularly critical in emerging processes such as metal additive manufacturing (AM), where defects can arise during fabrication and must be detected and addressed in real time to avoid costly scrap and rework.

Robotic EM NDE offers a clear pathway to overcome these challenges by enabling consistent sensor positioning, adaptive path planning, and closer coupling between inspection and manufacturing processes. However, achieving fully autonomous robotic EM inspection requires advances beyond sensor technology alone, including robot perception, adaptive control, coordination, data interpretation, and intuitive visualisation. Traditional industrial robots, which rely on pre-programmed paths and static environments, lack the flexibility needed for inspecting previously unseen components or evolving geometries.

This project therefore targets the core technologies required to deliver a fully autonomous EM NDE pipeline, integrating sensor-enabled robot perception, manufacturing automation, collaborative and mobile robotics, and real-time machine learning–based data interpretation. The work spans both off-line inspection of complex components and in-line, layer-by-layer EM NDE during metal AM.

Aims and Objectives

The overarching aim of this project is to develop and demonstrate a fully autonomous, robotic EM NDE framework capable of inspecting complex metallic components and in-process manufactured parts with minimal human intervention. Specifically, the project aims to:

1. Enable seamless real-time integration between EM NDE, robotic control, and manufacturing processes, supporting both off-line inspection of complex components and in-line inspection during AM.
2. Develop a fully autonomous robotic EM NDE system that can perceive, inspect, and adapt to complex components and evolving geometries without reliance on pre-programmed paths or prior CAD models.
3. Advance sensor-enabled robot perception and adaptive control, allowing EM NDE sensors to maintain optimal positioning, stand-off, and orientation over complex and as-built surfaces through real-time feedback.
4. Deliver real-time interpretation and quantification of EM NDE data, integrating machine learning and analytical methods to detect, localise, and characterise defects during robotic inspection.

The research will aim to pursue the following objectives to achieve a fully automated EM NDE deployment, data interpretation and reporting workflow:

Objective 1 – Robotic Integration of Eddy Current NDE Systems

• Develop a fully integrated robotic eddy current inspection platform for autonomous deployment on complex and in-process manufactured metallic components. Integrate modular EC sensors with industrial and collaborative robots, including end-effector tooling, compliance, and force/stand-off control, with real-time synchronisation of sensing, motion, and data acquisition. Validate inspection repeatability, positioning accuracy, and stability on complex, non-planar geometries.

Objective 2 – ROS-Based Robotic Navigation and Adaptive Inspection Planning

• Develop a ROS-based autonomy framework enabling adaptive navigation and inspection without reliance on pre-defined paths or prior CAD models. Implement sensor-enabled perception, surface reconstruction, and localisation to support adaptive coverage planning and real-time trajectory adjustment for optimal sensor orientation and stand-off. Demonstrate inspection of previously unseen components and dynamic environments, including in-line AM scenarios.

Objective 3 – Machine Learning for Automated Defect Detection, Characterisation and Sizing

• Curate labelled robotic EC datasets and develop ML models for robust defect detection, localisation, and quantitative sizing under variable lift-off, noise, and surface conditions. Integrate real-time ML inference within ROS to support autonomous inspection decisions and adaptive scanning. Quantify uncertainty and benchmark ML performance against conventional threshold- and rule-based EC analysis.