A machine learning approach to comfort assessment for offshore wind farm technicians

Current maintenance planning strategies in the operations and maintenance of offshore wind farms rarely ac-count for the comfort of technicians during transits. This creates uncertainties as transit from the vessel to structure might be unacceptable to technicians. Here, we model the welfare of technicians using the discomfort from the motions (three-dimensional accelerations) felt on crew transfer vessels (CTVs) during transits from port to wind farm. To explore technician exposure to vibration, acceleration data from vessel motion monitoring systems deployed on CTVs operating in the North Sea was synchronised with sea-state data from an operational ocean model. Processes of dimensionality reduction and machine learning (ML) were used to model the comfort of technicians from operational limits applied to models predicting Composite Weighted RMS Acceleration. Trained models were shown to provide estimations for the comfort variable with an R-2 value of 0.67 and an RMSE of 0.06 ms(-2). The comfort-based decision-making model is shown to be able to predict sail or not sail decisions for maintenance transits. The proposed model will have applications in maintenance planning for offshore wind farms, able to account for the comfort of technicians once identified limitations have been addressed to improve model predictions.