Paraconsistent Logic applied in the metallography of welds classification through morphological characteristics and entropy of digital images

Abstract

In the study of weldability, the metallography digital image analysis has become a tool of wide application. The morphological analysis and particle counting applied in digital images to classify patterns can produce contradictory data. In order to increase the reliability index in the results, this paper describes a model able enough to do the analysis for the final classification supported in an evidential non-classical logic, called Paraconsistent Logic (PL). The model used morphological analysis and entropy filter in metallography digital pictures. In generating three types of information data, it was used metallographic samples of the traditional process, the process for tempering and weld samples repair after tempering process. The actions of the model showed satisfactory results getting the classification of samples inserted to test. These results showed that the architecture of the proposed paraconsistent model is feasible and can be adapted according to the peculiarity of the analysis of the weld type.