Optimisation of welding parameters to mitigate the effect of residual stress on the fatigue life of nozzle–shell welded joints in cylindrical pressure vessels.

Abstract

The process of welding steel structures inadvertently causes residual stress as a result of thermal cycles that the material is subjected to. These welding-induced residual stresses have been shown to be responsible for a number of catastrophic failures in critical infrastructure installations such as pressure vessels, ship’s hulls, steel roof structures, and others. The present study examines the relationship between welding input parameters and the resultant residual stress, fatigue properties, weld bead geometry and mechanical properties of welded carbon steel pressure vessels. The study focuses on circumferential nozzle-to-shell welds, which have not been studied to this extent until now. A hybrid methodology including experimentation, numerical analysis, and mathematical modelling is employed to map out the relationship between welding input parameters and the output weld characteristics in order to further optimize the input parameters to produce an optimal welded joint whose stress and fatigue characteristics enhance service life of the welded structure. The results of a series of experiments performed show that the mechanical properties such as hardness are significantly affected by the welding process parameters and thereby affect the service life of a welded pressure vessel. The weld geometry is also affected by the input parameters of the welding process such that bead width and bead depth will vary depending on the parametric combination of input variables. The fatigue properties of a welded pressure vessel structure are affected by the residual stress conditions of the structure. The fractional factorial design technique shows that the welding current (I) and voltage (V) are statistically significant controlling parameters in the welding process. The results of the neutron diffraction (ND) tests reveal that there is a high concentration of residual stresses close to the weld centre-line. These stresses subside with increasing distance from the centre-line. The resultant hoop residual stress distribution shows that the hoop stresses are highly tensile close to the weld centre-line, decrease in magnitude as the distance from the weld centre-line increases, then decrease back to zero before changing direction to compressive further away from the weld centre-line. The hoop stress distribution profile on the flange side is similar to that of the pipe side around the circumferential weld, and the residual stress peak values are equal to or higher than the yield strength of the filler material. The weld specimens failed at the weld toe where the hoop stress was generally highly tensile in most of the welded specimens. The multiobjective genetic algorithm is successfully used to produce a set of optimal solutions that are in agreement with values obtained during experiments. The 3D finite element model produced using MSC Marc software is generally comparable to physical experimentation. The results obtained in the present study are in agreement with similar studies reported in the literature.