Asymmetric deformation prediction and process optimization of T- joint of medium and heavy plate

Abstract

Aiming at the deformation and stress control of T-joint welding of large and thick plates in the frame of mobile crushing station, the time series simulation and process optimization are carried out by using the thermal elastoplastic finite element method. A three-dimensional local constraint model of Q355B steel is established, and a double ellipsoid heat source is introduced. The temperature field, stress field and macro morphology of simultaneous welding, double-sided cross welding and single-sided alternating welding were compared and analyzed. The results show that the residual stresses of the three time series all present a “M-type bimodal” high tensile stress distribution with the peak value close to the yield strength in the weld zone; The single side alternate welding is affected by the “asymmetric rigid constraint effect” on the first welding side, which leads to the serious imbalance of bilateral stress distribution and forces the reverse deflection of the bottom plate, which is the fundamental reason for the macro three-dimensional deformation of the frame; Double sided cross welding realizes the offset of thermal bending moment through the reverse movement of heat sources on both sides, so that the component presents a small and symmetrical arch shape, and it is easy to compensate through the reverse deformation process, which is an optimal process scheme. Quantitative analysis revealed that single-sided staggered welding resulted in a noticeable 0.22 mm indentation in the central region. Conversely, optimized double-sided cross welding eliminated this localized indentation, effectively controlling the maximum deformation to 0.78 mm and improving the structural flatness. These extracted parameters established a reliable database of inherent strain, enabling rapid simulation of a full-scale 10-meter frame, providing significant engineering value for deformation control in heavy machinery manufacturing.