An automatic inspection approach for remanufacturing components using object detection

Abstract

Remanufacturing is the process of restoring a used product to the specifications of original manufactured product with a matching warranty. This process benefits the remanufacturers to a greater extent as it just requires the replacement of worn-out or obsolete components, thereby providing significant economic, social, and environmental benefits. Remanufactured components should meet the customer’s demand as new products and achieving this is a tough task due to uncertainty involved in the quantity and quality condition of the returned product. Inspection is one of the critical tasks in remanufacturing as it determines the quality of End of Life (EoL) product on arrival for remanufacturing, also inspection plays a crucial role in making the most appropriate decisions to proceed or scrap the product. Due to the drawbacks in the traditional manual inspection process because of its unreliable, subjective, time-consuming, and error-prone nature, attempts to adopt automatic inspection techniques gained attention in the industry. This study examines the capability of optical inspection techniques to increase productivity, profitability with higher reliability in remanufacturing. In the study, object detection methods are implemented to classify and locate defects on metallic surfaces using an open-source dataset GC10-DET. The detailed image pre-processing techniques, class imbalance techniques, and their effects on the model performance are also discussed. The YOLO (You Only Look Once) V4 algorithm with CSPDarknet-53 was used to locate and classify the defects. The performance of the algorithm is compared with other state-of-the-art techniques published in the literature for recall, average precision, and mean average precision (mAP) metrics. Our model demonstrates effective defect localization with a mAP of 66.78% and 93.63% for original and augmented datasets, respectively, which shows promise for the development of automated inspection technology.