Aplicación de sistemas inteligentes al control de calidad de la producción de piezas en serie mediante la reconstrucción de imágenes

Abstract

Introduction

Faced with the current need for an industrial update to the Industry 4.0 model, the need arises to achieve an intelligent system capable of controlling a production chain of parts made of high-value materials to minimize defective parts. For this reason, a network project is created, based on Machine Learning that acts as distributed control over inferior processes. These nodes in self-control through Machine Learning, interact with each other.

objectives

* Use image reconstruction methods that allow varying the geometric parameters necessary to avoid discarding materials. Previous works and theses have focused on this (Mula, 2016 "Three-dimensional reconstruction of conical perspectives" UPCT and Conesa, 1998 "Reconstruction of pieces in a gentleman's perspective" UPCT).

* Apply Machine Learning techniques for the prediction of results, and create the superior control network.

Results

Self-control through Machine Learning has been successfully implemented at the node level, predicting possible variabilities with a very efficient hit rate.

Conclusions

This new system translates into time and cost savings, in addition to ensuring the almost nullity of failures once the system performs supervised learning, and is reinforced with validations of new cases.