Determining the influence and correlation for parameters of flexible forming using the random forest method
Luka Sevšek, Sandi Baressi Šegota, Zlatan Car, Tomaž Pepelnjak
Abstract
Single-point incremental forming (SPIF) enables the forming of fix-clamped sheet metal by moving a relatively small geometrically simple tool along the trajectory, producing the desired shape of the final product. Excessive thinning of the sheet results in fracture, determining the limit of formability. This characteristic of the forming process can be improved by upgrading the basic SPIF process to two-step forming, whereby a more even distribution of the sheet thickness can be achieved by pre-bulging with a hemispherical punch. This study focused on analysing the SPIF process and a hybrid two-step forming consisting of sequential bulging and SPIF. The analysis focused on the output parameters of sheet metal thinning and maximum forming force components and was conducted with Abaqus simulation software. An innovative new approach for influence analysis of technological, material and geometrical input parameters and correlation analysis between the mentioned parameters was performed using the random forest (RF) method, which allows the determination of individual parameter influence by analysing tree-shaped models obtained through the training process. The analysis results show a significant influence of the workpiece wall angle and part depth on thinning and initial sheet thickness on values of the forming force components. The results also show a great correlation between the parameters of the bulging depth and the part depth after SPIF and the significant influence of the appropriate choice of the backing plate geometry for the target product geometry.