Measurement Of Bauschinger Behavior For Sheet Metals Based On Three-point Bending With Pre-strained Specimen

Springback is an unavoidable phenomenon in sheet metal forming,which directly affects the dimensional accuracy of parts.With the development of modern numerical simulation technology,the process of sheet metal forming could be simulated well by finite element analysis.The precise description of material mechanical behavior from constitutive model is the key to guarantee the accuracy of springback prediction.The accuracy of the finite element simulation could be promoted by the hardening model considering Bauschinger effect.However,the compression of sheet is severely restricted by buckling.Bauschinger behavior of sheet is difficult to quantify.The three-point bending with pre-strained specimen is a new method to measure Bauschinger behavior of sheet metal,which is simple to implement without special testing machine.Nevertheless,the research on measurement strategy of this method is not mature at present.In this paper,aiming to establish a relatively complete system of.three-point bending with pre-strained specimen,improvement and innovation are put forward for the measurement strategy combined with several reverse algorithm and deep learning modeling method.The research work and main conclusions are as follows:(1)The tension-compression stress vs.strain responses of DP980 and AA6022 sheets are obtained based on in-plane tension-compression test.The parameters of Y-U model are identified by single element simulation and the influence of different pre-strain levels on parameter calibration is studied.The inverse result with multi-prestrain levels capture the Bauschinger behavior under different prestrains and accurately predict the bending springback of sheets.The parameter groups measured at different pre-strain levels differ greatly,so that the choice of pre-strain levels is a decisive factor for parameter calibration.Considering the bauschinger behavior with various pre-strains is a necessary condition for the applicability of the parameters obtained.(2)With AA6022 as the reserach object,this paper achieved to take the information of load vs.displacement curves and springback angles as the targets for the multi-objective inverse strategy.Based on that,the influence of the dominance of inverse objective on the quantitative research of the Bauschinger behavior is studied.The load-dominant solution has a better prediction effect on load and springback Angle and the error of springback prediction angle is only 0.126°.In order to further confirm the adequacy of load information as a target,the stamping process of CR3-G140/40-U sheet is simulated.The prediction error of critical dimensions based on single-objective inverse is controlled at about 0.2mm,which is reduced by nearly 86% compared with the isotropic model.The importance of accurately describing Bauschinger behavior to reduce the dimension error of sheet metal stamping and the feasibility of load information as a single objective optimization are verified.(3)Based on time convolutional deep learning neural network,in this paper,real-time calibration model for friction coefficient and real-time calibration model for kinematic parameters are established for the first time.The calibration error of the real-time calibration model for friction coefficient is within 0.028 and the goodness of fit of the real-time calibration model for kinematic parameters is entirely acceptable.In the three-point bending simulation test,the prediction error of springback angles is less than 0.3°.The springback tests of AA6022,DP980 and CR3-G140/40-U sheets are carried out and the prediction error of the springback angle with the model calibration parameters is less than 0.5°,which verified the feasibility of the real-time calibration model.The data-driven modeling method provides a new idea for measurement of Bauschinger behavior and solves the drawbacks of the traditional inverse method which is time-consuming and not universal.To sum up,this paper overcomes the barriers of parameter identification for the advanced constitutive model and improves the ability of accurately accurate characterization for the bauschinger behavior of sheet metal.