Experimental modelling of a top-tensioned riser for vibration-based damage detection

Damage incidents in subsea risers may lead to catastrophic economic, environmental and human safety consequences. Therefore, early detection of damage is of paramount importance. Vibration-based damage detection methods have been extensively studied in the past for damage detection in a vast variety of structural types, however, their application to risers has hardly been explored to date. In the limited existing literature of vibration-based damage detection in risers, the proposed methods have mostly been verified by numerical studies or by experimental models which lacked realistic representations of ambient excitations and similarity of the model to a full-scale structure. This paper presents, for the first time, experimental modelling of a top-tensioned riser at a laboratory scale for vibration-based damage detection studies, in which the similarity of the experimental model to a full-scale structure, the multifaceted effects of damage and the dynamic ambience are explicitly addressed. An experimental riser model is built and subjected to random wave loading in a flume according to the JONSWAP spectrum to simulate the ambient excitations realistically. The experimental model consists of a tensioned rod made of acetal plastic and several small lumped steel masses attached to the rod such that they reduce the natural frequencies of the model to the range of wave frequencies that can be generated in the flume. An analysis is conducted to account for the similarity between the experimental model and a full-scale structure with the priority given to the modal characteristics of the structures. Corrosion is considered as the cause of damage and it is simulated non-destructively on the experimental model by partial removal of the attached lumped mass at the assumed damage location and reduction of applied tension load. The amount of removed mass and reduced tension load are calculated such that the simulated damage on the experimental model is equivalent to an actual cross-sectional area loss due to corrosion damage in the full-scale structure. An application of a vibration-based damage detection method is demonstrated using experimental data obtained for several different damage scenarios. An auto-regressive (AR) model is fitted to the acceleration response of the riser and coefficients of the AR model are used as the damage sensitive features (DSFs). Damage detection is achieved by conducting outlier analysis on the DSFs.