Back to search

MAROFF-2-Maritim virksomhet og offsh-2

Autonomous Drone Based Survey of Ships in Operation

Alternative title: Autonom Dronebasert inspeksjon av skip

Awarded: NOK 10.0 mill.

Project Number:

282287

Project Period:

2018 - 2021

Organisation:

Location:

The Class Society DNV has performed production surveys in enclosed spaces using drones since 2016, demonstrating cost savings and increased personnel safety. The project goal has been to develop and demonstrate an intelligent, autonomous drone for inspections of ships and offshore vessels. The main objective is to reduce the need to enter tanks and enable remote inspection. The vision is a drone that can fly by itself, track where it is, collect information about the condition of the tank, and spot defects like rust and cracks and also measure steel thickness. It is expected that drone-assisted remote inspection will reduce survey costs for the clients, improve the safety for surveyors, reduce environmental footprint, and improve inspection quality, and thus ship safety. Several drone?capabilities are required to enable visual close-up inspection and non-destructive testing in confined, GPS-denied, and poorly lit environments. The ADRASSO project, a collaborative effort between DNV, Scout Drone Inspection, NEO, Jotun, Idletechs and NTNU, has researched and developed semi-autonomous drone navigation functionalities, AI-based computer vision for automated detection of defects, hyperspectral imaging and associated software for fast analysis of ?big? hyperspectral data, to automatically assess the condition of the protective paint used in steel tanks, and their chemical composition. The computer vision system automatically detects cracks in images and videos. The software is based on Deep Learning. Thousands of images from DNV?s databases were collected and prepared to train the Deep Learning algorithm. The detection performance is satisfactory, but the AI makes mistakes, sometimes not seeing a crack and at other times falsely reporting a crack. Therefore, a video inspection tool has also been developed to assist the inspector to review and correct the mistakes made by the AI. Preparing the images for training the AI is labor-intensive. An image labelling tool was developed to reduce this effort. Also, a corrosion detector was developed in parallel in a DNV-internal project. A real-time crack detector was developed that runs onboard the drone on a Jetson TX2 GPU (Graphical Processing Unit). The hyperspectral analysis software detects and classifies the chemical composition of paintings with high accuracy, and it can be analyzed on the fly. Several other uses of hyperspectral imaging were investigated in the project, for example corrosion severity, detection of corrosion under paint, assessment of the condition or age of the paint, assessment of remaining zinc coating thickness on galvanized steel, and detecting and classifying contaminants on the tank surfaces, to assess how well it has been cleaned. These use cases are still under investigation and not concluded. Two designs of a small, light-weight hyperspectral camera were developed, targeting use cases that have different optical requirements. The inspection drone is tethered, providing both power and data communication, and has unlimited operation time, unlike battery-powered drones. It has a robust physical design, and is equipped with a strong light and a 4K camera. Using an on-board 3D laser scanner, it performs in-door navigation and has anti-collision systems. It knows its location and can map the tank; it keeps a stable position when navigation controls are released, and it is operated giving high-level commands rather than by joy-stick and provides a graphical user interface for easy operation. The drone was demonstrated to be able to be fitted with an ultrasonic thickness measurement sensor and perform thickness measurements of the steel. A cloud system was developed for capturing and visualizing the data in 3D, including the drone?s position inside the tank in real-time. This opens up for navigating the drone remotely, from outside the tank, so called BLOS (Beyond Visual Line of Sight). Two successful demonstrations onboard Floating Production, Storage and Offloading ships (FPSO) were carried out during the project. The ADRASSO project is currently continued in the NFR-funded project REDHUS which started in January 2021. REDHUS extends ADRASSO, making another stride towards truly remote inspection, i.e. no human entry in tanks. Additional anomalies are to be automatically detected, e.g. deformations, pitting, and leaking. Additional autonomous drone capabilities are to be developed to enable safe navigation in cargo tanks from outside the tank, BLOS. A micro-drone is to be developed for inspecting also ballast water tanks in double skin hulls. Inspection planning, execution, and reporting is to be more automated and based on 3D tools. Guidelines for automated drone-based inspections are to be developed in conjunction with new workflows and infrastructure for digital survey process. The commercial focus will be stronger since ADRASSO explored technical feasibility and demonstrated technical proof-of-concept.

DNV has demonstrated the technical feasibility of navigating intelligent drones in GPS-denied environments and detect cracks and corrosion. NEO has developed two optical designs of a compact, light-weight HSI camera suitable for a low-payload drone. Jotun has demonstrated that HSI can detect chemical binders in paints, and chemical changes over time. Idletechs has demonstrated how hyperspectral images can be analyzed in real time on a computational platform for drones. Developed new versions of their EMSC module. ScoutDI has developed a tethered drone for GPS-denied confined spaces, flying on position control for easy operation. It has unlimited flight time, robust data communication, and a cloud service. NTNU has demonstrated Proof-of-concept for identifying pinholes and binders with HSI. Integration of do-it-yourself hyperspectral imager with Scout drones; Developed approach for using Independent Component Analysis to identify spectral signatures in hyperspectral data.

Inspections of ships and offshore structures is mandatory (imposed by national governmental bodies as well as international bodies) in order to ensure vessel safety, and planning of maintenance and repair. Such inspections are time consuming, expensive, and potentially dangerous. Inspections today require extensive human effort. The underlying idea is to use autonomous drones equipped with RGB and HSI cameras and onboard image recognition/computer vision systems to survey the tanks and other structures. In order to be successful the project will perform research and development within diferent disciplines and research areas, such as imgage recognintion/computer vision, compact hyperspectral imaging systems and multichannel pattern recognition (hardware and software), and integrate the equipment with an autonomous drone. We believe that the proposed consortium consisting of DNVGL (one of the world's leading Classification Societies), Norsk Elektro Optikk - NEO (a world leader in Hyper Spectral Imaging Sensors and image analysis software), Idletechs (which is at the forefront of handling big quantitative data, providing tools for analysis and compression of multidimensional, dynamic data), NTNU (Norway's leading technical university; in this project NTNU and contributes with competence within real time image analysis hardware and software (together with Idletechs), drone autonomy (in cooperation with Scout DI), and compact hyperspectral camera - together with NEO); Jotun is one of the world´s major coating manufacturers with more than 200 people working with R&D on professional coatings.

Publications from Cristin

No publications found

No publications found

No publications found

Funding scheme:

MAROFF-2-Maritim virksomhet og offsh-2