Predicting the Porosity in Selective Laser Melting Parts Using Hybrid Regression Convolutional Neural Network

Assessing the porosity in Selective Laser Melting (SLM) parts is a challenging issue, and drawback of using existing gray value analysis method to assess difficulty subjectivity selecting uniform grayscale threshold convert single slice binary image highlight porosity. This paper proposes new approach based on use Regression Convolutional Neural Network (RCNN) algorithm predict percent CT scans finished SLM parts, without need for subjective difficult thresholding determination image. In order test algorithm, as training RCNN would require large amount experimental data, this proposed efficient creating artificial images mimicking real scan slices part with similarity index 0.9976. Applying improved prediction accuracy from 68.60% binarization 75.50% RCNN. The was then further developed by optimizing its parameters Bees Algorithm (BA), which known mimic behavior honeybees, hybrid (BA-RCNN) produced better 85.33%.