An Efficient Knowledge Based System for the Prediction of the Technical Feasibility of Sheet Metal Forming Processes

This paper presents an efficient knowledge based method using data mining and multilayer perceptrons (MLP) for the prediction of the technical feasibility of sheet metal forming processes. The stored forecast models can be applied to similar geometries and forming processes using the digital fingerprint to identify the most suitable MLP. Moreover, we establish an algorithm to detect extrapolation which is a priori applied to each new design in order to avoid probable high errors in the forecast model caused by extrapolation. We demonstrate the benefits of the method on a model problem specially designed to investigate a wide range of industrial relevant forming characteristics.