The welding temperature fields and residual stress distribution of the Q345 I-girder with corrugated steel web were studied through numerical simulation and experimental methods. The thermoelastic-plastic three-dimensional finite element models of the I-girder with corrugated steel web were established by using the complete coupling method， and the welding temperature fields and stress fields were obtained. By defining multiple research paths in the finite element models， the distributions of longitudinal and transverse residual stresses of the I-girder with corrugated steel web under different welding paths were studied. In the welding process of the I-girder with corrugated steel web， the infrared temperature measurement method and resistance strain gauge method were used to study the welding temperatures and residual stresses. The results show that the numerical simulation results of welding residual stresses are in good agreement with the measured results， which verifies the correctness and feasibility of the numerical simulation method. During the welding process， the stable temperature of the heat source center is up to 140 1 ℃， and the temperature away from the heat source center decreases rapidly. When the distance from the heat source center is more than 25 cm， the temperature approaches room temperature. The maximum von Mises stress of the weld at the bending corner of the corrugated web is 395 MPa， which is much higher than the yield strength of the material. The longitudinal residual stresses on the web surface 0.5 cm away from the weld is up to 351 MPa， while the transverse residual stress on the web surface fluctuates in the form of parabola， with a maximum value of 48.5 MPa. The welding residual stresses on the I-girder with corrugated steel web are mainly longitudinal stresses， which are mainly distributed within 20 cm from the weld. The research results of this paper can provide a reference basis for the elimination of residual stresses of the I-girder with corrugated steel in practical engineering.