<![CDATA[In the context of Industry 4.0, automotive manufacturing increasingly adopts artificial intelligence (AI) to improve quality assurance and reduce dependency on manual inspection. One critical issue in transmission assembly is improper installation of torsion spring components on the parking lock pole, which can lead to transmission malfunction and severe quality defects. Traditionally, this inspection process relies on human visual checking, which is time-consuming and prone to human error due to fatigue and varying operator conditions. This research proposes an AI-based visual inspection system to automatically detect incorrect torsion spring installation in a car transmission production line. The proposed system utilizes Convolutional Neural Network (CNN) models for image classification, deployed on an edge computing device integrated with Programmable Logic Controller (PLC) interlocking. Three CNN architectures (MobileNet, EfficientNet, and ResNet) are evaluated to identify the most suitable model for this application. The dataset consists of production images captured directly from the factory environment, with data augmentation applied to enhance robustness under varying lighting conditions. Model performance is evaluated using accuracy and K-Fold Cross-Validation. Experimental results show that the ResNet model achieves the highest performance, with an average accuracy of 99.66%, demonstrating its effectiveness in detecting improper torsion spring installation. The implementation of the proposed system eliminates the need for human visual inspection and reduces processing time in the production process. This study confirms that AI-based edge vision systems can significantly enhance quality assurance in automotive transmission manufacturing.]]>
Copyrights © 2025
