Neural network identification and control in metal forging

The success of a typical forging requires satisfying both final shape and final microstructure objectives. The final microstructure is often very sensitive to poorly known process parameters. Closed-loop control would be desirable to compensate for this sensitivity. The task of designing a suitable controller is made difficult by the lack of analytical process models, and the limited measurements. This paper proposes a neural network identification and control scheme for a simple extrusion. Some preliminary results are given