A new approach for data analysis and prediction in Capping Machines field

Utilizzo di tecniche di Machine Learning e Data Science in ambito industrialePurpose of this work is to analyze the behavior of a machine that performs electronic driven capping to plastic bottles – typically used in water & soft drink market. After having understood the working principles that drive the machine, it is necessary to extract – from all the collected information – relevant data able to describe and characterize the system. These data are needed in the attempt of finding a correlation between the machine setup and its performance. The core parameter that defines this performance, intended as the quality of the production result, is the torque required to unscrew the bottle cap. Desire of AROL S.p.A. is specifically to understand and control the machine parameter that are more involved in the capping process, to gain as much control as possible over the machine outcome. The available information is of two kinds. The first concerns the algorithm and setup parameters of the machine that performs the capping operation. The second regards the sensor data, saved in datalog files, that express the values of the motor variables, such as position, torque, velocity and current. The work steps required to accomplish the study are mainly: -??the comprehension of the machine algorithms and extraction of "discrete" information from the available data. -??the use of prediction models and, in general, of analysis strategies aimed at finding a correlation between the process data and the final Reopening values. The results of this study show only a partial understanding of the effect of all the machine behaviors upon the final result, that is seen in (relative) imprecise prediction outcomes. The remaining level of uncertainty is probably related to data not in our possession– that can be under or beyond our control – or to random conditions of the system, of environmental or structural type.