Abstract: The accuracy of ultrasonic residual stress testing was affected by many factors, which limited the further promotion of this technology. The effects of surface oxide scale, coupling agent type, coupling agent film thickness, equipment accuracy and tester operation difference on the test results were systematically studied, and the effectiveness of the optimization measures was verified by comparative tests. The results show that the method of removing surface oxide scale and floating ash by sandpaper could significantly improve the test accuracy. The use of oil as a coupling agent not only had good rust resistance, but also provided excellent coupling effect. When the thickness of the coupling agent film was controlled at 0.5—1.0 mm, the discreteness of the residual stress test data was only 5%. When the sound velocity measurement accuracy of the device was increased from 1 ns to 0.2 ns, the standard deviation of the single-point repeatability test could be controlled in the range of -5—5 MPa.