Skip to main content

Predicting and managing weld induced distortion in thin-walled, steel structures

Periodic Report Summary - DISTORTION (Predicting and managing weld induced distortion in thin-walled, steel structures)

Many metal structures are assembled from thin plate with welded supports for stiffness to resist local loadings. However, welded joints, which require large heat input, may incur significant distortion in the finished plate. Although the causes of distortion are known and have been the focus of number of studies, there is still a lack of fundamental understanding of process and physical parameters in causing distortion. The overall aim of this work was to identify the interaction of process and physical parameters in causing distortion of welded ferritic thin steel plates.

Experimental measurements and the finite element method were used to identify the relationship between distortion and the influence of pre-existing, i.e. residual, stresses in the plates. The effect of onset of transformation temperature on distortion was examined. An improved comprehension of the mechanisms causing distortion and a readily useable model to explore alternatives had significant potential in a wide range of industries and thus was a major driving force for continued research. The ability to predict with reasonable certainty the geometry of distortion would enable users to evaluate alternative design and production parameters.

A programme of experimental work on fabricated high strength alloy steel butt welded plates was conducted. This included fabrication of the experimental welds, temperature measurements during welding using gas metal arc (GMA) welding, residual stress measurements by neutron diffraction and microstructural characterisation using optical, scanning and transmission electron microscopy.

The residual stress measurements were performed using the neutron diffraction technique at the SALSA diffractometer at the Institut Laue-Langevin (ILL), Grenoble, France. The effect of two weld consumables on the residual stress in the welded plates was examined. The different weld consumables generated different residual stress patterns. The conventional weld metal produced tensile longitudinal stresses of approximately 400 MPa, whereas in the low transformation temperature (LTT) weld metal compressive stresses of -100 MPa were induced in the vicinity of the weld. General features of the macrostructures and microstructures of the fabricated welds were also documented.

The findings could be summarised as follows:
1. in the fully transformed solidification microstructure of a conventional weld metal, grains of proeutectoid ferrite were found at the former austenite columnar grain boundaries, while interlocked and randomly orientated acicular ferrite and cementite were seen in interior of the former austenite grains,
2. the examination of the microstructure of the fusion zone in the LTT weld metal revealed that it only consisted of the acicular ferrite and gave some indication of alignment of acicular ferrite plates. Further characterisation of the microstructural development during rapid cooling from the austenitic state was investigated using time resolved confocal scanning laser microscopy (CSLM), providing useful information about the onset of the transformation temperature and the kinetics of transformation which could be utilised in finite element modelling.

A time efficient and accurate numerical approach for prediction of distortion in thin walled structures was proposed and implemented in a thermo-elastic-plastic analysis of the welded plates. A sequentially uncoupled approach was applied where a thermal analysis was first completed to solve for thermal profiles. Subsequently, a mechanical analysis was carried out which read in the temperature profiles and solved for displacements, strains and stresses distributions. This algorithm calculated strain caused by both the thermal expansion and the solid-state phase transformation of the weld metal. It was used to study the influence of the transformation start temperature (Ts) of the filler and base material on the distribution of the residual stresses in the fusion zone and its vicinity. The results of the model showed that the majority of the camber distortion in the welded plates took place towards the end of the cooling after the welding was completed. As the weld cooled down to the room temperature, tensile stress was developed in the welded plates in the fusion zone due to the contraction of the latter. The magnitude of this stress was determined by the difference between the transformation finish temperature and room temperature. On cooling, compressive stresses were also developed in the region near the fusion zone due to phase transformation from austenite to ferrite. The compressive stress competed with the tensile strains which resulted from the thermal contraction of the fusion zone. If the phase transformation strain was greater than the thermal contraction, the weld material would have compressive residual stresses at room temperature and vice versa.

Distortion was studied by applying the presented model. It was found that a weld wire with smaller Ts produced lower residual stresses in the weld region, thus resulted in a weld with a reduced distortion. The results of the model agreed well with the experimental results with an exception of transverse stresses calculated for the tow transformation temperature welding (LTTW) wire, whose magnitude was found to be significantly lower than that of the experiment. The algorithm could be considered as a useful tool for selecting material for the weld wire with known Ts and for obtaining a desired stress state of the final weld, while it also provided a strong conceptual, theoretical and computational framework to interpret experimental results and plan new experiments.

In addition, the macro-texture and micro-texture of the welds fabricated using the conventional ferritic weld filler metal and the LLT wire were compared. The microstructure of the welds from both wires showed the presence of the ferrite phase and some allotriomorphic ferrite was also seen for the conventional weld wire. On the macro-scale the texture of the welds differed in intensity. For the LLT wire the orientation of the grains was found to be more aligned in (001) direction than that for the conventional wire. Electron backscatter diffraction (EBSD) analysis of the LTT fusion zone showed evidence of variant selection, consistent with an applied compressive stress in the welding direction; no such evidence was observed for the conventional welding consumable.

The in situ EBSD of the material from the LTT weld fusion during the thermal cycling zone, i.e. through the ferrite to the austenite and to ferrite transformations, demonstrated a change in the orientations of the grains in a small subregion. These variants all obeyed the crystallographic orientation relationships expected for a purely displacive transformation, despite the diffusive nature of the transformation of the austenite phase to ferrite. It was suggested that the minimisation of the total Gibbs energy through the balance of the chemical and mechanical contributions would provide a simple route to the design of welding consumables, which could make use of the variant transformations to minimise the residual stress and the distortion in GMA welds.

Regarding the project potential impact, it is estimated that a typical shipyard spends 20 to 30 % of labour hours on rework due to plate distortion, which can have significant financial implications. It is estimated that distortion costs EUR 2.5 million per ship. While it is not expected that distortion and subsequent rework can be eliminated completely, a reduction in the magnitude of the phenomenon would significantly reduce the costs of the industry.