Experiments-based surrogate modelling of the forming load evolution in robotic single point incremental forming of conical parts

The finite element simulation of SPIF processes is complex and computationally intensive, making data-based prediction models very attractive. A very fast data-based surrogate model was developed to predict the forming forces required to manufacture conical parts by SPIF. Robotic SPIF experiments were performed following a full-factorial design of experiments where the tool radius, trajectory axial step and sheet thickness were varied. The force evolution during each experiment was measured and the resulting dataset was significantly reduced using the Principal Component Analysis. Polynomial and Gaussian Process Regression models were developed in order to predict the maximum force as well as the force evolution, while additional experiments were performed to validate the predictions.