Real-Time Monitoring of the Inoculation Process in Cast Iron Production Using Deep Neural Networks
Open Access
- 1 January 2026
- journal article
- Published by International Digital Organization for Scientific Information (IDOSI) in International Journal of Engineering
- Vol. 39 (4) , 1033-1046
Abstract
Grain refinement in cast iron is a critical operation in the casting process, achieved through the addition of inoculant powder to the molten metal. This process enhances mechanical properties such as strength, hardness, and wear resistance. The type, quantity, and method of adding the inoculant powder significantly influence grain size and distribution. Typically, the powder is added to the molten metal at the final stage before it enters the mold. However, challenges such as nozzle blockages and uneven powder distribution reduce efficiency and lower the quality of the casting part. Traditional monitoring methods, which rely on human operators, are often inefficient. Moreover, environmental risks around the casting equipment further complicate manual supervision. To address this issue, the present study employs machine vision techniques and deep neural networks for real-time monitoring of the inoculation process at the casting production line. This automated approach enables precise measurement of parameters such as inoculant quantity and spray angle, thereby enhancing production line efficiency. In this research, the U-Net architecture was utilized for image segmentation, while ResNet34, InceptionV3, and VGG16 networks were employed for image classification. The highest segmentation accuracy, 0.9841, was achieved using U-Net with an InceptionV3 encoder, and the best classification accuracy, 0.9463, was obtained using ResNet34. Compared to traditional monitoring methods, this approach demonstrates superior accuracy and has a positive impact on the casting process.Keywords
This publication has 21 references indexed in Scilit:
- A novel plant arrangement-based image augmentation method for crop, weed, and background segmentation in agricultural field imagesComputers and Electronics in Agriculture, 2025
- EU-Net: Automatic U-Net neural architecture search with differential evolutionary algorithm for medical image segmentationComputers in Biology and Medicine, 2023
- Attention-UNet architectures with pretrained backbones for multi-class cardiac MR image segmentationCurrent Problems in Cardiology, 2023
- TDIP: Tunable Deep Image Processing, a Real Time Melt Pool Monitoring Solution2023
- Efficient Neural Net Approaches in Metal Casting Defect DetectionProcedia Computer Science, 2023
- Modality specific U-Net variants for biomedical image segmentation: a surveyArtificial Intelligence Review, 2022
- U-Net and Its Variants for Medical Image Segmentation: A Review of Theory and ApplicationsIEEE Access, 2021
- Automation in Agriculture by Machine and Deep Learning Techniques: A Review of Recent DevelopmentsPrecision Agriculture, 2021
- Pretreatments of gray cast iron with different inoculantsJournal of Alloys and Compounds, 2016
- Rethinking the Inception Architecture for Computer Vision2016