The purpose of this work is to investigate the factors influencing the resistance upset butt welding process to obtain an understanding of the metal behaviour and welding process characteristics, so that new automotive steels can be welded with reduced development time and fewer failures in production. In principle the welding process is rather simple, the materials to be joined are clamped between two electrodes and pressed together. Because there is an interface present with a higher resistance than the base material, heat will be generated at the joint position once the welding current and voltage are applied. As soon as the material is warm and soft enough, an extra (upsetting) pressure is applied and a weld is formed. For making wheel rims, after welding, the upset is removed by means of chisels and shear cut- ting the edges of the weld, followed by rolling the remaining material into the rim. After the upset has been removed the wheel rim is cold formed in several steps to its final shape. For the purpose of this research 2.0 mm thick S460MC and 1.65 mm thick HR45 steel were examined at a test welding facility at Fontijne Grotnes BV. Both steels were specially developed for cold forming and deep drawing purposes and have a mainly ferritic microstructure. An initial research into the crack formation in wheel rims during production was performed, proving that there were no hard and crack susceptible microstructures present in the weld area of the researched materials. The idea that the wheel rims crack due to for example to the presence of bainite and martensite was rejected. There were two main types of cracks discovered, large cracks which are formed due to initial misalignment of the sheets, grease and/or oil at the interface during welding and an uneven heat distribution during welding. The other type of cracks were micro-cracks, which originate from notches introduced during the trimming and rolling of the wheel rim edges after welding. A good alignment of the chisels and cutting tools is necessary to minimise the amount of material that is pushed into the wheel rim Because it was not feasible to do welding experiments in a full wheel rim welding machine at a production facility, the Fontijne Holland Welderr was employed, this is a scaled down version of a production welding machine, but without the material feed or post weld handling facilities. During welding it was observed that the weld sample did not heat up evenly, the emitted light at the interface was brighter at some places than at others. Also after welding the oxidation of the joint area showed that there was a non-uniform heating, with warmer and colder areas. This non-uniformity results in non-uniform weld characteristics and weld quality differences over the length of the weld. Tensile tests of samples taken both from the edges and the centre of welded samples showed that the maximum tensile strength can be as high as 95% of the tensile strength of the base material. However, the strain to failure of the welded samples was only a maximum of 60% with a 30% difference between the central test sample and the edge samples. Erichsen testing was also employed to provide an indication of weld quality. Only the centre of samples was tested due to the way the Erichsen tester is constructed. The Erichsen height was upto 85% of the base material height for the welded samples and it was found that if there are no edge defects present, the Erichsen test is a quick and valid weld quality test. Due to observed non-uniform weld quality and heating of the weld area, the focus of the research shifted to understanding why there was a non-uniformity present. Temperature measurements both by thermal camera and thermocouples showed that there is indeed a non-uniform temperature distribution present during welding over the joint interface. The outer edges of the joint line have higher temperatures than the central part of the weld. Local difference can be as high as 300 K over just 20 mm distance. Also there are hot spots present at either end of the joint line approximately 10 mm from the edges. These hot spots can give rise to local melting in steels with low melting temperature segregation bands. The causes of the uneven temperature distribution were sought. Employing local current measurements during welding was not possible, however when the welding process is just started and there has been no signifcant heating of the weld sample, it is possible to link the voltage distribution directly to the current density distribution. Measuring voltage distributions is relatively easy, voltage probes were spot welded to the surface of the weld samples and measurements showed that there was a non- uniform voltage distribution present during welding. This non-uniformity had the same distribution as the temperature distribution, low voltages at the places were the temperatures were low and high voltage signals where the temperatures were high. Because the non-uniform voltage distribution was measured for both 1 and 2-piece samples (samples with and without an interface) it can be concluded that the boundary conditions governed by the welding equipment have a major influence. The influence of the clamping system and the contact surface of the electrodes were researched to determine their its influence on the heating of the weld area. It was shown that the clamping pressure of the samples was not high enough to prevent slip of the samples during welding and that the samples had to be positioned in the centre of the electrodes to minimise non-uniform heating. At the places of the highest local pressure, the contact resistance is the lowest and heating will consequently be slower at those locations. The highest edge temperatures were measured at the far end of the electrodes, where the lowest clamping pressures are expected. To obtain more uniform heating it is necessary to obtain a more uniform current distribution. This can be accomplished by changing the contact interface between the electrode and the sample surface or the interface between the weld samples. Adapting the electrodes has a small influence but will also result in longer welding times due to a decrease in contact surface area (decrease of 10% results in a 5% increase in weld time). The interface change can decrease the welding time by up to 20% due to a local faster initial heating. Unfortunately the change in interface surface resulted in less repeatable welds. Residual stress in the welded samples was also examined. Both samples with the upset present as well as samples where the upset was removed were measured. The samples with the weld present showed tensile stresses at the centre of the joint line, but compressive stresses at the edges of the sample. When the upset was removed the overall stress state of the sheet became more neutral, the compressive and tensile stress peaks become lower. The stress distribution was explained conceptually with the aid of a modified bar model. Modelling with both MS.Marc and ABAQUS was employed in this research to obtain a better understanding of the complex electrical, thermal and mechanical relationships as observed during real experiments. Three different models were developed, a thermal model, a thermo-mechanical model and a local strain model. The local strain model showed that the Erichsen testing method is a valid method for weld quality assessment. The modelled total strain observed during flaring of wheel rims was lower than the modelled and measured strain during Erichsen testing. Both the thermal and thethermo-mechanical models are predictive. The first one is based on input parameters directly coming from measurements during welding experiments. The thermo-mechanical model is based on the temperature dependent material parameters. The initial thermal model was designed to determine if the non-uniform current density distribution, as measured during experiments, could result in a non-uniform heating of the weld area. The model did not include any temperature dependent parameters nor a mechanical coupling. This showed that the non-uniform current input resulted in a non-uniform heating of the sample. By obtaining a more uniform current distribution over the length of the joint area, a more uniform heating can be obtained. It was also indicated that around contaminations at the interface, either conductive or insulating, a locally higher temperature exists. The mechan- ical model is a more realistic model as it includes the upsetting force. The model predicts the formation of the upset when there are misalignments of the sheets during welding. Because this model includes the temperature dependent material parameters, it is possible to predict the residual stresses after welding, which are in qualitative agreement with the measured residual stresses in real welds.