Prediction Of Remaining Strength Of The Corroded Marine Risers

Marine riser, main connection of the offshore platform and subsea wellhead, is also the vulnerable spot in marine oil and gas transportation, as is effected by the transmission medium and outside environment, the corrosion damage is extremely easy to initiate. When corrosion occurs, riser’s thickness reduces, loading capacity decreases, that will eventually lead to the riser’s rupture and leakage accident, which means severe environment pollution and huge financial loss. However, limited by the manpower and material resource, it is impossible to replace or repair all corroded marine risers, therefore, the residual intensity assessment of these risers is of great necessary.Now many evaluation criterions and standards of the corroded pipeline were set at home and abroad, but their conservation may cause unnecessary maintenance and replacement. In recent years, the method of using artificial neural network to assess residual intensity aroused wide concern. In this work, weight and threshold value of back propagation neural network (BP) was optimized by artificial neural network (BP) and GA-BP neural network was built to make the prediction of marine risers that contains corrosion defects.For residual intensity prediction of marine riser that contains single corrosion defect, GA-BP neural network model was built between the pipe diameter, thickness, length and depth of corrosion defects, ultimate tensile strength of pipeline and residual intensity of marine riser, influence of radius-thickness radio, length and depth of corrosion defect to marine riser’s residual intensity was discussed. For residual intensity prediction of marine riser that contains the interaction of corrosion defects, GA-BP neural network model was built between pipeline diameter, length of long and short corrosion defects, depth of corrosion defects, distance of corrosion defects and residual intensity of marine riser, influence of length, depth, distance of corrosion defects to marine riser’s residual intensity was also discussed here. Result shows:it is feasible to use artificial neural network to make prediction on the residual intensity of marine riser, meanwhile, the GA-BP neural network can effectively improve the convergence and accuracy of the network. It is a more scientific and effective prediction model.