Abstract

A framework to increase the level of automation of manufacturing processes towards the paradigm of intelligent manufacturing is proposed. The case study considered is the milling of aluminum alloys through a three-axes Computer Numerical Control (CNC) machine tool. The developed controller is called Evaluation and Perception Controller (EPC), and it con- sists of three layers: sensor, perception, and cognitive. A set of sensors displayed in the working volume collect the information necessary to reconstruct the state of the system. In the perception layer the acquired data are processed and learned, thus maintaining updated the models of the process which drive the process optimisation in the cognitive layer. The Optimal Control Problem (OCP) method is utilised to calculate the controls of the process that optimise a target function, defined accordingly to the specific context in which the operation is executed (namely roughing or finishing). In the general case, the objec- tive function takes into account: productivity, quality, and costs of the process. The material removal rate along the tool path is used as index for process productivity. The quality is quan- tified through measurements of roughness, therefore it is improved by limiting the static and dynamic displacements (i.e. the vibrations) of the tool. The costs of the process here con- sidered are those associated to tooling (i.e. wear of the tool), and energy absorbed by the spindle during on-air free movements. In order to ensure the feasibility of the solution, algebraic and differential constraints are im- posed on the dynamic response of drives and spindle. The OCP is then solved through an efficient optimisation library developed by the group of Mechatronic Engineering of the Uni- versity of Trento. The EPC has been designed as a portable system that can be integrated into any CNC ma- chine, once it has been calibrated, and a dedicated communication interface with the NC has been implemented. The tests necessary for the OCP calibration, and requirements for the communication layer between EPC and CN, are described and discussed. Finally, the EPC is tested on real milling processes. The validation of its performances is done by comparing the outcomes of the process with respect to a nominal case, in which the process is set up according to guidelines given by tool manufacturer. It worth be noted that this work represents not only a step foreword in increasing the level of automation of machining, but rather it proposes an architecture and an approach which can be generalised to several manufacturing processes.