Study On Ultimate Capacity For Corroded Submarine Pipeline Based On GA-BP Neural Network

With the development of offshore oil and gas fields,submarine pipelines have been widely used as one of the most economical ways of transmitting oil and gas from the offshore oil fields to the terminals.Submarine pipelines can easily results in corrosion defects due to the effects of both harsh marine environment outside the pipelines and transporting material inside ones. Corrosion is one of the major reasons causing the damage of submarine pipelines.This research is funded by National Natural Science Foundation of China(Grant No.50439010). The objectives are to study the ultimate capacity of corroded submarine pipelines and assess the reliability of pipelines.Some codes and criteria have been published to predict the ultimate capacity of corroded pipelines,however,the results obtained by these codes and criteria are so conservative that maybe cause unnecessary maintenance and replacement.A new method combined with finite element method,BP neural network and genetic algorithm is proposed to predict ultimate capacity of corroded pipelines.The research is composed of the following parts.1.A review of the achievements obtained from both domestic and overseas researchers is summarized.Meanwhile,the aims of the thesis are presented firstly.2.Introducing the basic principles of neural network and genetic algorithm,and a GA-BP neural network model is proposed to predict the ultimate capacity of corroded submarine pipelines.3.The ultimate capacity of submarine pipelines with infinite-long and finite-long corrosion defects subject to internal pressure are studied by using GA-BP neural network.The effect of corroded length,corroded width,corroded depth and the ratio of outer diameter to thickness on failure pressure are discussed.In addition,the results predicted by GA-BP neural network are compared with ones calculated by ASME B31G and DNV RP-F101.4.The reliability of corroded submarine pipelines is research by a new method combined with GA-BP neural network,the response surface method and the optimization method.