Application of ANFIS to predict springback in U-bending of nickel-based alloy

Most reinforcement plates in aircraft engines comprise nickel-based alloys and are shaped using bending. However, the nonlinear behavior of materials makes it difficult to predict the springback following component bending, thereby leading to quality issues. Increasing the accuracy of springback prediction is therefore a crucial issue in the field of stamping. This study examined the bending and springback behavior of nickel-based alloy HAYNES 230. First, we obtained the minimum bending radius and forming properties of the nickel-based alloy at room temperature. Next, we used the finite element analysis software DEFORM to collect springback data related to the U-bending of nickel-based alloy HAYNES 230. The primary forming factors were punch round radius (R_p), cavity round radius (R_c), die clearance (D_c), and forming velocity (F_v). We performed an 81-combination full factor experiment to establish the adaptive neuro-fuzzy inference system (ANFIS) prediction model and then used the median of each factor parameter to perform a 16-combination full factor experiment to validate the prediction model. We trained the ANFIS prediction model using the following four membership functions: a triangular membership function, a trapezoidal membership function, a bell-shaped membership function, and a Gaussian membership function. To find the correlation between each membership function and the data and confirm the error of the prediction model, we calculated the coefficients of the correlation between the results derived from the four membership functions and the analysis or experiment results. The results derived from the Gaussian membership function were most highly correlated with both the analysis and experiment results. ANOVA of the full factor experiments indicated that the contributions of R_p, R_c, D_c, and F_v were 65.554%, 2.072%, 29.585%, and 0.768%, respectively. The factor with the greatest impact on springback was R_p.