Enhancing circular economy via detecting and recycling 2D nested sheet waste using Bayesian optimization technique based-smart digital twin

Abstract

The recycling process is controversial on a worldwide scale since it is based on the concept of sustainability and minimizing the environmental footprint. From this perspective, a practical framework for managing the 2D nesting waste is presented in this paper, which is a crucial phase in many production processes. Deep Learning (DLs) models are used to detect (i.e., segmentation) and categorize (i.e., classification) waste area sizes to enhance the circular economy via recycling procedures. Moreover, the apparel industry for having huge 2D nesting waste is implemented in this paper. Furthermore, data augmentation is performed in this paper to overcome the lack of datasets. The segmentation stage is applied based on the SegNet deep convolutional neural network (DCNN). In addition to, the classification stage based on the DLs, ResNet34, InceptionV3, and DenseNet121 is implemented to classify our datasets. The experimental results demonstrate that the proposed framework outperforms other existing techniques, where the Area Under the Curve (AUC) of 99.87 %, detection success rate (DSR) of 99.88 %, sensitivity of 99.98 %, precision of 99.98 %, F1-score of 98.99 %, mean square error (MSE) from 0.01 % to 2.49 %, efficiency of 97.41 %, and computational time of 3.9 s. Therefore, our proposed framework achieves the best performance compared with the literature. The future work will focus on enhancing the circular economy by utilizing the 3D nested sheet waste.

Keywords

• Bayesian optimization technique
• Circular economy
• Classification
• Deep learning
• Segmentation
• 2D nested waste