Production Process Quality Inspection with Machine Learning Approach

Abstract

Technological developments in the industrial world encourage innovation in the inspection process, one of which is the application of artificial intelligence with machine learning. CV. XYZ is a palm oil machine component fabrication workshop that still applies manual quality inspection. Manual inspections are prone to errors, depend on human skills, and take a long time. This research aims to develop an automated inspection system using the YOLO (You Only Look Once) model which is a convolutional neural network (CNN) based algorithm for product defect detection. The manual inspection used is considered inconsistent, error-prone, and time-consuming. The use of machine learning is able to identify product defects such as geometry defect, porous defect, and surface defect. Evaluation of model performance using confusion matrix, loss graph, and precision recall curve. The results obtained show that the model has detection accuracy with a mAP50-95 value of 74.5%, mAP50 of 88.5%, and detection time of 0.0084 seconds per image.