Hydrogen distribution and redistribution in the weld zone of constructional steels

Abstract

The invention of electric arc welding revolutionized the steel construction industry, but also brought some problems when the welded region has inferior properties compared to the plate metal. A major cause of brittle failure was identi ed as hydrogen embrittlement of the weld zone, although a comprehensive understanding of this phenomenon is not, even now, available. Hydrogen in solution in the weld zone is found in arc welds, due to the intense conditions in the welding arc. There is invariably a sufficient source in the form of moisture and hydrocarbon residue to give a few parts-per-million (ppm) by mass of hydrogen in the weld pool, which is a sufficient concentration to bring the possibility of hydrogen cold cracking in the completed weld. Hydrogen is significantly mobile in steels at room temperature, which is certainly why a few ppm of hydrogen can concentrate on a microscopic scale and initiate cracks, but also means that on a macroscopic scale there is hydrogen dispersion, which can relieve the cracking risk or place hydrogen in hydrogen cracking susceptible regions. The understanding of solubility and mobility of hydrogen in steels of different compositions and microstructures is therefore paramount. The question investigated in this work is whether the characteristics of the weld hydrogen cracking tendency can be explained by the features of weld hydrogen transport, especially when steel selection is a variable. Plate steel ranging from a 0.22%C pearlitic steel to a 0.05%C thermo-mechanically controlled-rolled and accelerated-cooled (TMCR-AC) high strength low alloy (HSLA) steel with no pearlite, plus a 0.4%C non-plate steel, were included in the experimental program. Welds were made with rutile ux-cored-wire (R-FCW) at two hydrogen levels, together with rutile shielded-metal-arc (R-SMA) welds. In order to investigate the di usion rates, a novel experiment has been devised. The welded plate has been milled away at an angle from the underside of the weld to provide increasing distances between the fusion boundary and the plate under-surface. The formation of hydrogen bubbles in glycerol enabled the measurements of the time dependent diffusion distances. The results clearly show a square root time correlation, as expected from the Fickian mechanism and enabled the calculation of diffusion coefficients for different steels. A nearly four fold difference was found between the steels, with the fastest hydrogen movement in the TMCR-AC steel. To reveal the initial distribution of hydrogen some samples were frozen immediately after welding and machined under liquid nitrogen. This test ruled-out any signi cant hydrogen dispersion during the deposition of the weld and during the cooling down period. The experimental data were interpreted using a new numerical computer model, based on random jumps of hydrogen between equivalent lattice sites. It is shown that this numerical model gives identical results to the analytical Fickian approach, but has the advantage that it can be used for any boundary shape. When this model has been applied to the experimental data, some unexpected features have been found. The amount of hydrogen emerging at surfaces distant to the weld was higher than expected from a concentration-driven mechanism; suggesting that a di erent transport mechanism should be applied. The numerical model has also indicated a discontinuity in the hydrogen concentration at the fusion boundary. It is a consequence of the model that hydrogen solubilities and di usivities are inversely related properties of the metal; a feature which is supported by experimental evidence. The tendency of hydrogen cracking to appear in the weld metal rather than in the heat-a ected-zone (HAZ) can thus be explained by higher di usivity of hydrogen in the plate metal. It appears that there is a relationship between the diffusivity and the microstructure, particularly when the content and form of carbon is considered.