ABSTRACTThe analysis and design of joints constitutes one of the most important aspects when designing steel structures. It is a common practice to model the joint behaviour by zero‐length rotational springs placed between the beam and corresponding column, or by mechanical models composed of springs and rigid bars. These models have serious limitations when considering the interactions between the shear, bending and axial forces and deformations of the panel zone.In this research, a new approach based on the characterization of modal components is presented. First, the steel joint is modelled as a cruciform finite element of 12 degrees of freedom. Then, the stiffness matrix is decomposed in eigenvectors and eigenvalues. The eigenvectors provide information about the joint deformation modes, while the eigenvalues give information about the stiffness associated with each mode. To identify the main modes and their characteristics, a parametric study has been carried out for welded joints with and without stiffeners. For both cases, all combinations of IPE 120 to IPE 600 profiles for beams and HEA 160 to HEA 1000 profiles for columns have been studied. The only restraint imposed in the combination process is that the beam width be smaller than the column width.First, the cruciform finite element stiffness matrices are obtained. Subsequently, a spectral decomposition is carried out using MATLAB, and then Python scripts are developed to classify the modes and eigenvalues. The statistical analysis of the results shows that after removing the three rigid solid modes the most relevant deformation modes are the shear and the two possible bending modes including their interaction. The remaining six modes correspond to deformation modes with high stiffness values that are not excited under common loading situations.