Predicting the trenching speed is a task of paramount importance, in particular to correctly estimate the time (and hence the costs) required for the subsea operations.
The project aimed at developing Machine Learning models to predict subsea trenching activities. The models have been trained and tested on real operational data, acquired during an off-shore campaign in the Mediterranean see. The project mainly focused on predicting the trenching speed, as a function of the trencher configuration and soil characteristics. The results of the project show how the ML models outperform the results given by the available analytical models, provided that input data are consistent with the range where the ML models have been trained.