Smart Jaw Chunks

The progressive development of machine tools and cutting materials leads to rising demands upon the gripping devices through continuous increase of cutting power and speed. In turning operation the gripping device must be able to transmit the drive power on the workpiece. To ensure a safe gripping of the workpiece at high spindle speeds it is necessary to determine the required initial gripping force depending on the cutting force and centrifugal force. Because of the insufficient models the existing approaches to determine the required gripping force lead to unreliable and very different initial gripping forces. Thereby the potential of modern gripping systems is not sufficiently used.

The main objective of this project consists in studying and developing a AI (Artificial Intelligence) based system for gripping force modeling in modern jaw chunks. The followed approach is a black-box one and in particular Neural Networks are used as modeling structures. According to such an approach the gripping force is suitably sensored and measured in different operating conditions, together with certain quantities which typically influence the gripping force value itself  (the chuck type, the gripping jaws, the workpiece geometry and the model, the angular velocity of the turning-lathe, the pressure of the hydraulic pump). Once all data are acquired, the parametric modeling system is trained through an offline procedure, so yielding a "true" gripping force value  in correspondence of certain inputs, which can be therefore used for control purposes or just real-time monitoring.

The data acquisition phase required the involvment of suitable sensors, both at output level (gripping force) and input ones (pressure and angular velocity of the turning-lathe) and therefore of conditioning signal processing board in order to make all available signals syncronized and compliant to the acquisition SW running on a PC. The modeling phase, as said above, has been performed offline by means of a third-party SW. Once trained, the modelling system has been also implemented and used in real-time so providing a gripping force monitoring service: the accomplished prototype is a PC-based one with inputs obtained with the same setup used for the acquisition phase.