Prediction of Load—Displacement Curve of Flexible Pipe Carcass Under Radial Compression Based on Residual Neural Network

• Published in: China ocean engineering 2023-02, Vol.37 (1), p.42-52
• Main Authors: Yan, Jun, Li, Wen-bo, Vaz, Murilo Augusto, Lu, Hai-long, Zhang, Heng-rui, Du, Hong-ze, Bu, Yu-feng
• Format: Article
• Language: English
• Subjects: Accuracy; Artificial neural networks; Carcasses; Coastal Sciences; Compression; Deformation; Displacement; Elastic deformation; Engineering; Equivalence; Experimental data; Flexible pipes; Fluid- and Aerodynamics; Marine & Freshwater Sciences; Mechanical properties; Neural networks; Numerical and Computational Physics; Oceanography; Offshore Engineering; Original Paper; Plastic deformation; Plastics; Simulation; Stainless steel; Stainless steels
• ISSN: 0890-5487; 2191-8945
• DOI: 10.1007/s13344-023-0004-8

The carcass layer of flexible pipe comprises a large-angle spiral structure with a complex interlocked stainless steel cross-section profile, which is mainly used to resist radial load. With the complex structure of the carcass layer, an equivalent simplified model is used to study the mechanical properties of the carcass layer. However, the current equivalent carcass model only considers the elastic deformation, and this simplification leads to huge errors in the calculation results. In this study, radial compression experiments were carried out to make the carcasses to undergo plastic deformation. Subsequently, a residual neural network based on the experimental data was established to predict the load–displacement curves of carcasses with different inner diameter in plastic states under radial compression. The established neural network model’s high precision was verified by experimental data, and the influence of the number of input variables on the accuracy of the neural network was discussed. The conclusion shows that the residual neural network model established based on the experimental data of the small-diameter carcass layer can predict the load–displacement curve of the large-diameter carcass layer in the plastic stage. With the decrease of input data, the prediction accuracy of residual network model in plasticity stage will decrease.