Yansheng Du

On-site multi-pass welding in building steel structures often induces transverse shrinkage, affecting assembly accuracy and final geometry. Pass-by-pass thermo-elasto-plastic simulation can capture the welding process but is computationally expensive for large steel components. This study proposes an efficient framework for predicting welding-induced transverse shrinkage in H-shaped steel beams by integrating merged weld-pass modeling, a shrinkage function model, and an equivalent indirect action method. A Q355B single-sided V-groove butt-welded plate was simulated to compare the original weld-pass model with three merged schemes. The balanced three-pass scheme preserved the main mechanical response while reducing computation time from 18.863 h to 5.528 h. Based on flat-position and vertical-position plate-welding simulations, a layered shrinkage function model was developed using a plate baseline term and a geometric coupling correction term, with LOOCV RMSE, MAE, and maximum absolute error values of 0.2855 mm, 0.2424 mm, and 0.5123 mm, respectively. The predicted shrinkage was then converted into an equivalent temperature load. The equivalent indirect action method achieved an MAE of 0.1883 mm, RMSE of 0.2713 mm, MAPE of 9.2685%, Pearson coefficient of 0.9730, and R² of 0.9042. This framework supports rapid transverse-shrinkage prediction and construction accuracy control for site-welded H-shaped steel beams.

The exterior frame of super-high-rise steel structures with inclined columns and dense beams contains numerous welded joints and complex constraints, making field welding prone to residual stress and cumulative displacement. Taking the 19th erection segment of the Tianjin China Overseas City Plaza main tower as an example, this study monitored welding-induced stress and column-top displacement using vibrating-wire strain gauges and BeiDou GNSS, and established a finite element model based on the equivalent indirect action method to evaluate different welding sequences. The results show that stresses near welds evolve in stages during lower-flange, web, and upper-flange welding. Adjacent welding causes limited disturbance after the completed joint has stabilized, whereas lower-layer welding can induce secondary shrinkage in upper-layer welded regions. The simulated residual stresses and column-top offsets show reasonable agreement with the monitoring results in terms of overall magnitude and distribution trend, with mean values 2.50% and 6.22% lower than the measurements, respectively. Welding sequence significantly influences deformation. Horizontal grouped skip welding improves construction efficiency by about 30%, and vertical reverse welding further reduces the mean column-top offset by 53.31% and controls the maximum column-top offset within the project-specific limit. Therefore, under the studied structural configuration and construction conditions, the combined sequence of horizontal grouped skip welding and vertical reverse welding is recommended as a practical welding-sequence scheme.