The magnetostriction measurement can measure the internal stress of steel material nondestructively by utilizing the magnetostriction effect of steel materials. In the magnetostriction method, the calibration curve, which is the correspondence between the output voltage depending on probe characteristics and the principal stress difference, significantly influencing the analytical results of both applied and principal stress. In this study, the influence of loading history on the calibration curves was clarified by tensile tests including repeated load within the elastic range to realize more accurate magnetostriction measurement. Additionally, the adequacy of the residual stress distribution was discussed based on magnetostriction measurement results and stress analysis using the shear difference method for T-beam with 2 welding lines. As the results, the calibration curves are considered unlikely to change significantly by loading history within the elastic range. In stress analysis, the analyzed and estimated results almost matched when averaging the stress distributions obtained along the same measurement line for outbound and inbound route.