MARIOW – Maritime AI-Guided & Remote Operated Welding System

Project Overview and Motivation

The joint project MARIOW focuses on the development of an intelligent, semi-autonomous underwater welding system for the maintenance of maritime infrastructure. The objective is to improve the quality, efficiency, and safety of underwater welding operations through the use of AI-supported image processing, advanced robotics, and an automatable flux-cored arc welding (FCAW) process.

Underwater welding on port facilities, offshore structures, pipelines, and metallic constructions is complex, high-risk, and difficult to reproduce consistently. Traditional welding performed by divers is costly and technically limited, while demand exceeds the availability of qualified specialists. MARIOW addresses these challenges by combining robotics, sensor technology, and artificial intelligence with the underwater flux-cored arc welding (UW-FCAW) process. Unlike conventional stick electrodes, this method uses a continuous welding wire, enabling more stable and consistent weld seams underwater and providing the foundation for process automation.

The project was carried out by an interdisciplinary consortium consisting of industrial and research partners, including companies such as AMT GmbH and UNTERWASSERKRAUSE-MUTZECK, as well as research institutions such as TH Köln, DFKI, and Fraunhofer IGD. This close collaboration across the entire value chain enabled the development of an integrated system combining welding technology, robotics, sensor systems, and artificial intelligence.

Scientific and Technical Objectives

Fraunhofer IGD’s primary contribution focused on the development of a robust visual sensor system for direct integration with the welding torch. The system incorporates a stereo camera setup with AI-based real-time image enhancement using autoencoder and U-Net architectures.

The goal is to ensure reliable monitoring and analysis of the welding process even under challenging visibility conditions, such as water turbidity or strong lighting contrasts. In addition, marker-based positioning methods and 3D distance measurement techniques are being developed to accurately capture the welding head, workpiece, and weld seam. These capabilities provide the basis for path planning, robotic control, and quality assurance.

Technical Implementation and Innovations

Fraunhofer IGD developed and tested various camera and housing configurations capable of operating in real time on edge-computing platforms such as NVIDIA Jetson and Raspberry Pi systems. The AI-based image enhancement models were trained using domain-specific datasets and integrated into the overall system architecture.

In close cooperation with the project partners, algorithms for 3D distance measurement and interfaces for integration into a ROS2-based robotics and welding system were also implemented. The combination of image and process data enables continuous monitoring and evaluation of the welding process, supporting both automated weld path detection and quality control.

Results and Benefits

The project resulted in a functional prototype of an AI-assisted underwater image processing system that supports the planning, execution, and documentation of welding operations. During the final demonstration, the camera system, weld path recognition, robotic arm, and UW-FCAW process were successfully integrated into a complete operational system.

The developed solution features a modular design and can be transferred to other application areas such as offshore wind farms, pipeline inspections, and port infrastructure maintenance. It provides both scientific and economic value by advancing the automation of maritime maintenance processes while reducing the physical burden on skilled personnel.

Future projects are planned to validate the system under real harbor conditions, including exposure to saltwater, currents, and wave action.