NEURAL NETWORKS MODELING FOR METALLOGRAPHIC IMAGE RECOGNITION TO DIAGNOSE STEELS CONDITION

Abstract

The problem of the steel technical diagnostic automation was analyzed. The urgency to develop new methods and means of automated diagnostics steels is shown. The algorithm of metallographic images pre-processing was proposed. The neural network structures of the metallographic images recognition, which based on RBF and MLP paradigms, were developed. The neural network modeling results of the metallographic image recognition for diagnosis the steel technical conditions are shown.