AI-enhanced casting defect detection in pump impellers: A transfer learning-based framework for quality control in smart manufacturing
Mukesh Sahu et al.
What the paper says
The viability of modern production systems hinges on maintaining high product quality and minimizing rejection rates. This has driven a shift towards digitalization in the industrial sector as a means of controlling and predicting product quality. For data-driven manufacturing methods to be effective, model performance must be high. The automated detection and classification of casting product defects is a difficult job. Recently, Convolutional Neural Networks (CNNs) have demonstrated improved efficacy in defect detection and classification. This research proposes a framework using deep learning techniques applied to a dataset of 8678 images of pump impellers. The quality of the cast is predicted using a custom CNN model and four pre-trained models (VGG16, ResNet50V2, InceptionV3, and MobileNetV2) with the help of a transfer learning method. Promising outcomes in predicting defective casting products have been achieved, with the VGG16 model demonstrating the highest accuracy at 94.31%. The proposed framework offers a significant practical impact by enabling manufacturing companies to identify defective products early in the production cycle. This early detection can substantially reduce inspection costs, lower resource loss from scrapping, and ultimately enhance the economic viability of the production system.
Evidence weight
Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40
| F · citation impact | 0.50 × 0.4 = 0.20 |
| M · momentum | 0.50 × 0.15 = 0.07 |
| V · venue signal | 0.50 × 0.05 = 0.03 |
| R · text relevance † | 0.50 × 0.4 = 0.20 |
† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.