刘刚

个人信息Personal Information

教授

博士生导师

硕士生导师

主要任职:船舶工程学院党委书记

性别:男

毕业院校:日本广岛大学

学位:博士

所在单位:船舶工程学院

学科:船舶与海洋结构物设计制造

办公地点:综合实验二号楼412房间

联系方式:0411-84706091,13898403510

电子邮箱:liugang@dlut.edu.cn

扫描关注

论文成果

当前位置: 中文主页 >> 科学研究 >> 论文成果

Optimization design on the riser system of next generation subsea production system with the assistance of DOE and surrogate model techniques

点击次数:

论文类型:期刊论文

发表时间:2021-02-01

发表刊物:APPLIED OCEAN RESEARCH

卷号:85

页面范围:34-44

ISSN号:0141-1187

关键字:Next generation subsea production system; Riser system; Optimization design; Design of experiments; Surrogate model

摘要:The Next Generation Subsea Production System (NextGen SPS) is a new concept for petroleum development in ultra-deep water (UDW) areas. It can improve the structural performance of riser as well as provide several operational benefits to subsurface well completion (SWC) equipment. The design of NextGen SPS's riser system which includes rigid riser and flexible jumper-like the free standing hybrid riser (FSHR), is a very important issue for the definition of NextGen SPS. This paper details an optimization design on the NextGen SPS's riser system, with the assistance of the design of experiments (DOE) and surrogate model techniques. The optimization model pertaining to riser system is formulated firstly. The DOE is a statistical technique that guides a sensitive study on the behavior of the riser system before the optimization analysis. Structural responses are obtained by the fully coupled methodology. Through such a preliminary study, the effective contribution of each design variable at the riser performance will be known and some general conclusive remarks will be obtained. Based on the DOE results, design variables are screened to improve the efficiency of optimization process. Particle swarm optimization (PSO) method is employed to conduct the optimization analysis. In this analysis, surrogate models, which are developed by back propagation neural network (BPNN), replace the time consuming dynamic analysis to predict structural responses. Latin hypercube sampling (LHS) method is adopted to generate training sample and testing sample for the BPNN. NextGen SPS that operates at a depth of 3000 m is used as the case for this investigation. The efficiency of optimization design is improved by DOE and surrogate techniques, and a reduction of approximately 46% for the riser system cost is achieved. The obtained conclusions have applicability in reference to the engineering design of FSHR and the study procedure will provide reference for study on other new structure concept.