Reduction of manufacturing and maintenance costs for offshore wind turbines through inline inspection and analysis of the fatigue life of welded joints using machine learning methods

Project description

Welded maritime support structures firmly anchored to the seabed are important for European industry and society, particularly in terms of energy generation and as navigation aids. They are needed, for example, for oil and gas extraction, for offshore wind energy production, for navigation aids such as lighthouses and channel markers, and in ports. Around 30 percent of the investment costs for the construction of an offshore wind farm are accounted for by the manufacturing costs for the foundations. These foundations are usually designed as so-called monopiles. Fatigue strength plays a decisive role in the mathematical design (dimensioning) of monopiles. The circumferential welds are particularly important in determining strength, as wind and wave loads are most likely to cause damage in the form of fatigue cracks during the service life of approximately 25 years due to the stress concentrations inherent in the welds.

The project is developing an automated and user-independent procedure for the local assessment of the fatigue strength of welds through the targeted use of optical 3D measurement systems. Particular attention is being paid to the development of a technical solution for the creation and evaluation of high-quality 3D scans of circumferential welds on monopiles. Another focus is on the use of artificial neural networks (ANN) to determine local and spatially resolved weld quality along the entire welded joint, represented by the local stress concentration (notch factor).

As part of this project, the project consortium will carry out the following fundamental work: Development and technical implementation of a suitable procedure for creating high-resolution digital surface models of welded joints, including the derivation of recommendations regarding the choice of scan parameters (e.g., resolution, etc.), as well as the recording of local and global internal stresses on small and large structures (monopiles). This will be followed by the development of an evaluation algorithm for the development of software tools for the automatic derivation of seam geometry parameters (seam transition radius and flank angle) and classification into evaluation groups according to DIN EN ISO 5817 from the surface model and the 3D scan data. An artificial neural network will be trained and validated for the direct derivation of notch shape factors from raw 3D data. The locally recorded influencing factors (residual stress, microstructure, notch shape factor) will be correlated with the crack initiation locations on small-scale test specimens and representative welded joints in the further course of the project. The final part of the project involves the development of a concept for location-resolved life assessment over the entire seam length on linear welded joints, taking into account the local geometry and notch effect, and the transfer to welded components with complex seam patterns. This link is generated using machine learning methods. The developed algorithm for service life assessment should be applicable both for in-line service life assessment (assessment of the circumferential welded joint during production) and for the inspection of existing offshore wind turbines.

Fraunhofer IWM subproject:

As part of the project, Fraunhofer IWM uses 3D scans and X-ray analyses to determine the spatially resolved geometry as well as the residual stress and microstructure of the welded joint. These factors are then correlated with local crack initiation locations from fatigue tests on small-scale test specimens using optical strain field measurement.