Investigation of methods for the localization and reconstruction of the wave impact on a floating wind turbine pontoon

• Inversion methods for localization and reconstruction of the wave impact were proposed. • The presented methods effectively predict the position of the wave impact load and reconstruct the time history of the load. • Machine learning methods are used to build data sets and train models. • These methods show a great potential to reverse identification of various loads in offshore structures. Wave impact under extreme sea conditions poses significant risks to offshore structures, such as the pontoons of floating wind turbines. Monitoring the structures and inverting the impact loads are crucial for structural safety design. In this study, an indirect approach is proposed for the localization and reconstruction of waves impacting flat plates, curved plates and floating wind turbine pontoons. For impact load localization, the correlation dimension method and the long short-term memory (LSTM) neural network method are employed. The correlation dimension method achieves accurate localization results only for flat plates, whereas the LSTM method shows good performance for all structures. For load time history reconstruction, the Tikhonov regularization method is applied. Reconstruction is achieved with very high accuracy. This work can provide guidelines for inverse estimation of environmental loads acting on offshore structures.