Regression model for prediction the thickness of parabolic part under uncertainty of sheet metal properties

Abstract

This research aims to predict the thickness loss of parabolic forming by using hydromechanical deep drawing (HMD) process. The methodology used is the finite element method on the uncertainty of the mechanical properties for 1 mm thickness SPCC sheet metals. The properties consisted of three aspects: strength coefficient (K), strain hardening exponent (n-value), and plastic strain ratio for all three axis (r0, r45, and r90). Multiple simulations were completed by randomly selecting K, n, r0, r45, and r90 values following the distribution obtained from the uniaxial tensile test. The results would be analyzed by using the regression equation to create the equation to predict the thickness loss of the test pieces obtained from the formation process using the K, n, r0, r45, and r90 variables without the use of finite element method. This would results in a decrease in the simulation time. Based on 30 experiments of uniaxial tensile test, it shows that the sheet metal properties are normally distributed. After that, 60 cases of the five variables were randomly selected to simulate the formation by using the finite element method. Finally, the results obtained from the aforementioned simulation is analyzed and generated into a regression equation. From the comparison between the value predicted from the regression equation and the result obtained from the formation process, it can be seen that the regression equation can effectively predict the thickness loss when the five variables are under the same distribution properties. However, the efficiency will decrease when the values of the five variables are beyond the distribution properties obtained from the experiment. In addition, the uncertainty of mechanical properties of sheet metal in the simulation displays the chances of discovering test pieces with over 40% thickness loss. The simulation resulted in 5 damaged test pieces, accounting for 8.33% of the total process.