Welding joint is a key component in steel structure engineering, which is often subjected to complex stress and becomes the weak link of building structure, and different fracture modes may appear under different stress. In this study, the ductile fracture of Q345 steel welded joint was investigated by selecting the base material and heat-affected zone (HAZ) material. Based on the J–C (Johnson–Cook) fracture model of stress triaxiality factor, a series of round bar specimens were designed for the tensile test, and the variation of minimum diameter and gauge distance was tracked in real time by digital image correlation (DIC) technology, and the cloud map of strain distribution in the whole loading process was obtained. The constitutive model and fracture criterion model of Q345 steel welded joint base material and HAZ material are established through experimental measurement. The results show that the maximum triaxial stress is at the center root of the minimum diameter of the notched round bar tensile specimen, the sudden change of the notched edge will lead to stress concentration, the stress change of the notch section is the most significant, extending from near the notch, the maximum strain is at the root of the notch, and the fracture position appears at the minimum section of the notch. According to the true stress-strain curve of a standard tensile specimen and the least square method, the values of parameters A, B, and N of the J–C (Johnson–Cook) constitutive model of base material and HAZ material are 379.80, 452.89, and 0.37 and 392.90, 540.61, and 0.49, respectively. The fracture strains of the base metal and HAZ material under different triaxial stresses were measured experimentally, and the J–C fracture model was fitted. The values of failure criterion parameters D1, D2, and D 3 were 1.025, −0.008, and 3.617 and 0.678, −2.689E-5, and 7.683, respectively. This provides experimental parameters for mechanical properties and fracture models of welded joints in structural engineering, refined finite element models considering material failure for structural numerical analysis, and basic data for evaluation and design of building structures, which have good social and economic benefits.