Data-driven prediction of critical collapse pressure of flexible pipeline carcass layer

Carcass layers in flexible pipelines are mainly used to resist the radial external pressure. The large hydrostatic pressure in deep-water environments often leads to collapse failure of the carcass layer, thus causing the destruction of the entire flexible pipeline. Therefore, the critical collapse pressure of the carcass layer under external pressure must be evaluated in engineering design. At present, the established equivalent theoretical model or numerical simulation method used to study the critical collapse pressure of the carcass layer has certain limitations, such as low accuracy or excessive calculation time. To overcome these problems, a Kriging-modified buckling model is proposed with radius R and thickness t as design variables, based on data-driven technology, to predict the critical collapse pressure of the carcass layer accurately. First, sample points are selected in the design domain based on the optimal Latin hypercube sampling, and the weight between the sample points and predicted points is calculated using the Kriging algorithm. Then, the above-mentioned weights are used to modify the buckling theory to predict the critical collapse pressure of the carcass layer. The numerical results indicate that the modified model has a higher prediction accuracy than that of the traditional buckling theory. Additionally, the effects of the number and pattern of sample points on the modified model are discussed. This paper presents an efficient design method and implementation technology for the design of deep-water flexible pipeline carcass. •Nonlinear buckling simulation of the carcass is carried out and verified by experiment.•The modified buckling theory is proposed based on data-driven method with Kriging algorithm.•The influence of the number and distribution of sample points on the prediction results is analyzed.