Research On Numerical Control Ultrasonic Shot Peening Straightening Process And Prediction Model

To meet the demand of high speed and mobility of aircraft in aeronautics and astronautics industry,integral structural parts are becoming larger and more complicated, and requirements of mechanicalproperties like manufacturing accuracy and fatigue performance are more difficult to reach. As anadvanced forming and strengthening process, numerical control ultrasonic shot peening(USP) is barelylimited by the size, complicated structure and material of integral parts, and highly improves thecomprehensive property of the peened parts. So it is significant to research the USP straightening andforming process. Research on USP straightening process of single curve2024-T351aluminum alloytest specimen has been done and the process prediction model has been established and validated. Themain work can be concluded as follows:(1) Physical process and residual stress introduction of USP straightening were analyzed, thetheoretical formula of USP residual stress and straightening value was deducted, and influence on USPresidual stress field of angle of peening impact was studied. The research shows that non-normalimpact decreases the conversion ratio of impact energy to spring-back energy and the straighteningvalue depends on the property of peened material, peening intensity and peening path.(2) USP straightening process on single curvature deformed aluminum alloy specimen has beenstudied, and the influence of peening parameters on straightening performance has been quantitativelystudied. Research has shown that the midpoint straightening value M0has apparent effect on thestraightening performance, which increases with the increase of peening width B, feed speed F, trackpitch T and peening electricity I, but when F, T and I reach to certain values, the M0will decrease.Local bulge and spherical bulge will take place with improper parameters, but the former can beeliminated by using appropriate B, T, M0or I, and the later can be suppressed with reasonable F.(3) Influence of the process parameters on USP straightening value and the coincidence ofbetween simulation and experiments results were studied by using ABAQUS. Resultant averagecoincidence rate is5.5%with average error0.026mm, which means adopted simulation method isreasonable and reliable.(4) USP straightening prediction model by using BP neural network was explored, and hiddenlayer nodes number and network generalization ability were studied. The results show that optimalhidden layers nodes number is4, and prediction mean error is downgraded to be0.02mm from0.05mmafter datasets expansion by using ABAQUS. Prediction model was verified in the deformed ribbedstructural part straightening.