After many times of maintenance of an offshore platform it is found that there is crack propagation at the corner which connects the living building to the deck. The idea of crack length identification based on the multi-dimensional strain around the crack is proposed in this paper. A finite element model of offshore platform with initial crack is built. Multi-scale strain data and corresponding crack length are used as feature input and output for the machine learning model respectively. The crack length is predicted by gradient boosting regression tree (GBRT) model. Test results show that the value of MSE and R2 can reach 0.0006 and 0.9991, respectively. At the same time, the model is proved to have good anti-interference to noise.

ocean platform  /  crack extension  /  machine learning  /  gradient boosting regression tree