| With the offshore oil exploitation, the security problem for the platform is becoming more and more important. Fatigue damage is regarded as one of the most important problems that cause the failure of tubular joints. The research of the fatigue of the tubular joints structure can predict the location and size of the fatigue crack and the residual life. It can improve the repair efficiency and reduce costs of the tubular joints structure.Firstly, specimens, which are under axial load, in-plane and out-of-plane bending, were respectively analyzed with MSC.MARC based on the numerical analysis in this thesis.Secondly, stress concentration factors were calculated with some numerical methods, which are at the location of the hot spot stress of the tubular joints structure. The mechanism was then presented in the thesis, that how the value of the stress concentration factors with the change of the geometric parameters, by analyzing various geometric parameters of the tubular joints structure. In the FE analysis, a total of 7100 8-node thick shell elements were created , including 84 weld elements.Based on the mesh density of this modle,two refined meshes were created by doubling and tripling the element density of the joint.The results indicated stress concentration factors were very close with the maximum discrepancy less than 5%. It was commented that 8-node thick shell with single meshes were sufficient to simulate the completely overlapped tubular joinFinally, stress concentration factors for various geometric parameters were predicated by regression analysis with the software MATLAB.Simulation results show that, for joint under brace axial loading, the stress concentration factor increases with increasingβCT,τTL,ξT andγC,but decreases with increasingτCT andθ. For joint under brace in-plane bending, the stress concentration factor increases with increasingβCT andτCT, but decreases with increasingθ. For joint under brace out-of-plane bending, the stress concentration factor increases with increasingβCT,τTL,ξT,γC, but decreases with increasingβTL,τCT,θ. Regression analysis indicated that the confidence interval of each regression coefficientβdid not include zero, all values of R were less than 0.8, and all values of p were less than 0.05. It showed that the proposed regression model is suitable for simulation. |