Feedforward control of sheet bending based on force measurements

The accuracy which can be reached with feedback control in industrial metal forming processes is limited by product-to-product variations in the process. These variations may be controlled with a feedforward control system. To do so, a measurement system and a process estimator are needed to measure and correct for these variations. In this work, the use of force measurements for feedforward control of a steel sheet bending process is investigated. A multi-stage demonstrator process is developed with cutting, deep drawing and bending stages and a large number of force sensors. Several relations between process forces and the product geometry are studied in an extensive analysis of measurement data from the production process. It is proposed to build a moving window process estimator using the LASSO regression method. The parameters of the regression model are updated during production based on historical data of the production line. Several simulation runs are performed to estimate the effect of feedforward control on the process accuracy. These simulation runs are based on measurements from the real production line and minor assumptions. The proposed approach leads to an estimated decrease of 24% in the root mean square error of the final angle with respect to a simulation run with feedback control only.