Quasi static error compensation of 5-axis large machine tools using on-board sensors and AI analysis

This paper summarize the first results of the research project SOMMACT (Self Optimising Measuring MAChine Tools), funded by the Seventh Framework Program of the European Commission, which proposes an innovative method aimed at the numerical compensation of these errors. The objective is the development of a controller with self-learning capabilities that, starting from historical information acquired by sensors systems, artefacts and finished workpiece measurements associated to the machine operative conditions, accumulates knowledge on machine performances, is able to predict errors that a machining process would present in different conditions and, therefore,can adapt compensation tables. Starting from the storage of a large data amount, a supervised learning method like Support Vector Machine, together with an inferential Fuzzy Logic-based system, is able to manage correspondences between historical accumulated information and a potential current situation, evaluating the influence of non-comparable physical dimensions, whose effects can’t a priori be added. This method can be applied on different machine tools typologies, also aside from the number of axes and is implemented inside a more complex software system aimed at supporting measurement procedures and provided of graphical user interface, able to calculate the volumetric error with 3D representations, to integrate different sensors systems, providing models for error functions calculation starting from measurements data and integrating communication processes with the CNC. This allows its effective application in real production sites, introducing relevant improvements in machine tools manufacturing field.