AN APPLICATION OF AN ARTIFICIAL INTELLIGENCE CAMERAS FOR DEFECT DETECTION IN STAMPING PRODUCTION PROCESS

Abstract

This research aims to (1) study the application of artificial intelligence cameras in defect detection during the stamping process, and (2) propose guidelines for the application of AI cameras in defect detection in the stamping process by using Hardware: Vidi deep learning (Cognex) and Software: 1) Keyence, 2) JM viste, 3) CCS SP.vision, 4) FOFA for model control, with experiments designed based on the Design of Experiments (DOE) concept.

The research findings revealed that (1) applying AI cameras for defect detection in the stamping process, particularly using Keyence AI cameras together with Neurocle software, significantly improved the accuracy and consistency of detecting defects such as dents and scratches on workpieces compared to human inspection. This resulted in a higher defect detection rate and effectively reduced loss due to defective products. (2) Regarding the proposed guidelines for maximizing the effectiveness of AI cameras in real-world applications, based on experimental results and data analysis, the researchers suggest establishing standardized installation methods by positioning and angling the cameras appropriately according to the nature of the work. Controlling the illumination intensity in the inspection area to maintain a constant level is crucial to allow the image processing software to perform optimally. Systematic training for staff on how to use the Keyence software should be provided to enable proper configuration and parameter adjustment, minimizing human errors. Additionally, using high-end GPUs compatible with Keyence software specifications, such as NVIDIA RTX 4080/4090, is recommended to enhance learning and image processing speed and accuracy. Periodic inspection and updating of the AI model are necessary when defect patterns change to ensure sustained high accuracy. Furthermore, system maintenance including checking the readiness of AI cameras and the surrounding environment, such as cleaning lenses and controlling dust and lighting, is essential to prevent interference with defect detection during actual production processes.