Vision-Based Damage Localization Method for an Autonomous Robotic Laser Cladding Process

Abstract

Currently, damage identification and localization in remanufacturing is a manual visual task. It is time-consuming, labour-intensive. and can result in an imprecise repair. To mitigate this, an automatic vision-based damage localization method is proposed in this paper that integrates a camera in a robotic laser cladding repair cell. Two case studies analyzing different configurations of Faster Region-based Convolutional neural networks (R-CNN) are performed. This research aims to select the most suitable configuration to localize the wear on damaged fixed bends. Images were collected for testing and training the R-CNN and the results of this study indicated a decreasing trend in training and validation losses and a mean average precision (mAP) of 88.7%.