Casting Fault Detection by Deep Convolution Neural Networks

The gold standard for fault identification in the casting process is the human visual inspection though that includes human error and it is time-consuming. In this work, to automate the inspection process in casting production three deep learning-based artificial intelligence systems for detecting manufacturing flaws in submersible pump impellers are created. The suggested casting fault detectors intend to create a system for fault identification in casting goods that are automated, reliable, and accurate. Here on the top view image data of submersible pump impellers, two transfer learning-based convolution neural networks and one lightly structured convolution neural network are created and verified. The performance of the suggested models is measured in terms of accuracy, sensitivity, precision, and f1-score, and then compared with the state-of-the-art works. The best performance among the proposed casting fault detection systems are 99.88% accuracy, 100% precision, 99.77% recall, and 99.88% f1-score in binary classification and this outperformed previous related work.