Storms, swells, saltwater: Wind turbines at sea have to withstand a lot. Towers and support structures of the offshore plants are therefore under a lot of stress. Rust, for example, reduces the load-bearing capacity of the structures. Therefore, the manufacturers coat the surfaces of the components, for example the tower, with special protective layers to ensure the targeted lifetime. However, they cannot completely prevent the ageing of the plants. This makes it all the more important to detect changes on the surface of the towers at an early stage in order to prevent major damage or the shutdown of the plants. This saves time and costs. The researchers have already successfully implemented a new maintenance model in a small onshore wind farm.

Control surfaces autonomously

The intention of the scientists from companies and universities in the ISyMOO project was to record and evaluate the outer skin of the towers and their surface protection systems exclusively on the basis of digital data. To do this, they use autonomous inspection techniques, such as drones and cameras, as well as special sensors installed on the towers. These measure both meteorological and oceanographic data, such as air temperature and humidity, flow velocity and salt content, as well as corrosion data, for example the degree of degradation on the steel. With the help of the recorded measurement values and various methods, for example new algorithmic methods and machine learning processes, the relevant stresses and the condition of the surfaces are to be determined. The collected digital measured data and photos as well as virtual design data are transmitted to the control centre on the mainland and serve as the basis for the maintenance model. To document and manage the complex and large amounts of data, the project teams have developed an online platform.
 
In addition, the scientists take into account in their maintenance model what effects certain damage has on the wind turbine. To do this, they determine the type and extent of damage and link this to the function of the corresponding component. This allows them to classify the risks of damage into defined categories. Based on the methods developed, inspection and repair concepts can be derived from the data and integrated into the maintenance programme. At present, specially trained staff evaluate the data manually, which is time-consuming and costly.

Detecting and classifying damage automatically

Online image data platform
© Muehlhan AG – University of Bielefeld

In the online image data platform, the damage is identified by different labels, for example corrosion or coating damage.

Digital photos or videos of the wind turbine are automatically evaluated using machine learning. The computer program first learns to recognise and classify the type and intensity of the damage from a wealth of visual information fed into it. Inspection and repair concepts can be derived on the basis of the evaluated visual data and a temporal condition model. This allows the maintenance cycles to be continuously optimised. As repair costs at sea are extremely high, the system can significantly contribute to reducing operating costs.

Virtual twin shows defined areas of the tower structure

To view defined reference areas of the tower structure, corresponding to a photograph, the teams of scientists use virtual twins of the wind turbines. To do this, they used the "Virtual Twin Computation" method for the first time in the ISyMOO project. First, the towers of the wind turbines are scanned with commercially available digital cameras. The images are then combined to form a "virtual twin". To store and evaluate the measured sensor and image data for the virtual twins, the researchers developed a time series database system including a web analysis interface. Such a database system is specialised in storing and analysing sensor data regularly over a period of time, for example. It is used to monitor plant systems or components.

Monitoring and planning ahead

For the maintenance model, it is first necessary to apply the data obtained to defined virtual reference surfaces of the structures. The condition of each reference area is assessed and documented at specific time marks (for example, every two years) over the entire operating time of the plant. A risk of damage is defined depending on the structure and function of the reference area. Both the condition of the wind turbine surfaces and the respective risk are determined using visual data. The image data flows into a web-based online platform (BIIGLE 2.0) and is evaluated. The virtual twin is used for visualisation.

The existing and the permissible state of the surfaces are compared by means of a limit state analysis of the structures. This results in suitable monitoring or repair scenarios. The specially developed mathematical-empirical degradation model for coatings and steel surfaces provides support here. Furthermore, the researchers developed a simulation procedure that takes into account the resistance of corroded components to changing mechanical stress.

Maintenance model in practice

The maintenance model with the corresponding data interfaces is already available and can be used in pilot projects.

It was already used for the first time in a small onshore wind farm. The researchers successfully tested the methods developed to evaluate surface systems of steel structures in a largely automated manner. A little extra: The maintenance model can also be applied to structures with the same geometry such as silos, chimneys and supporting piles. (mm)

 

Last update: 13.06.2022

At a glance

Short title: ISyMOO
Funding Number: 0324254A,B,D-F
Topics: Offshore aspects
Project coordination: Muehlhan AG
Running time: June 2018 bis December 2021

Quintessence

  • Wind turbines: For the first time, the condition of the towers' surfaces can be assessed solely on the basis of digital data.
  • Virtual twins of wind turbines are used to define reference areas on the outer skin of the towers and to check for damage, such as rust.
  • A new simulation tool determines the lifetime of individual components under varying stresses. The results are incorporated into the maintenance model.
  • The "degradation model" for coatings and steel surfaces enables the integration of all data collected by sensors and cameras. It serves as a basis for decisions regarding maintenance.

Contact

Dr. Andreas Momber
Muehlhan AG

+49 40 75271-144

www.muehlhan.com

saltation GmbH & Co. KG

www.saltation.com

 

University of Bielefeld- RESEARCH GROUP Biodata Mining

www.uni-bielefeld.de/arbeitsgruppen/biodata-mining

 

Leibniz Universität Hannover - Faculty of Civil Engineering and Geodetic Science

www.stahlbau.uni-hannover.de/en

 

SubCtech GmbH

www.subctech.com

Among other things, EnArgus, the central information system for energy research funding, contains a database of all energy research projects - including this project.