Machine learning assisted in forecasting the ice-induced vibration for jacket platforms

Abstract

Ice-induced vibration (IIV) is a severe threat to the safety of offshore platforms, especially to flexible structures such as jacket platforms in the Bohai Sea. Therefore, it is currently urged to accurate and rapid measure predictions of IIV responses, and this could be value-added feature of offshore structures' sea ice management system. This study optimized a short-term forecast model of structural IIV responses based on machine learning theory. The prediction model was developed using field monitoring data from the monitoring station on the Bohai Sea JZ20-2MUQ platform. Both the IIV responses prediction model based on sea ice parameters and the model based on environmental parameters were quite promising and had high accuracy. And the accuracy of 15.78 % and 12.76 % show that the latter model has a better training effect. This study proves that machine learning could assist in the method development of a fast, accurate, and real-time prediction model of structural IIV responses.