In mixed-model assembly lines, smooth operation of the assembly line depends on adherence to the scheduled sequence. However, during production process, this sequence is altered both intentionally and uninstentionally. A major source of unintentional sequence alteration in automobile plants is the paint defects. A post-paint resequencing buffer, located before the final assembly is used to restore the altered sequence. Restoring the altered sequence back to the scheduled sequence requires three distinct operations in this buffer: Changing the positions (i.e. resequencing) of vehicles, inserting spare vehicles in between difficult models and replacing spare vehicles with paint defective vehicles. We develop a two-stage stochastic model to determine the optimal number of spare vehicles from each model-colour type to be placed into the Automated Storage and Retrieval System resequencing buffer that maximises the scheduled sequence achievement ratio (SSAR). The model contributes to the literature by explicitly considering above three distinct operations and random nature of paint defect occurrences. We use sample average approximation algorithm to solve the model. We provide managerial insights on how paint entrance sequence, defect rate and buffer size affect the SSAR. A value of stochastic solution shows that the model significantly outperforms its deterministic counterpart.

This paper compares the performance of Deep Learning (DL) and multi-template matching (MTM) models for detecting small defects. DL models extract distinguishing features of objects but require a large dataset of images. In contrast, alternative computer vision techniques like MTM need a relatively small dataset. The lack of large datasets for small metal-surface defects has inhibited the adoption of automation in small-defect detection in remanufacturing settings. This motivated this preliminary study to compare template-based approaches, like MTM, with feature-based approaches, such as DL models, for small-defect detection on an initial laboratory and remanufacturing industry dataset. This study used You Only Look Once v5 (YOLOv5) as the DL model and compared its performance against the MTM model for small-crack detection. The findings of our preliminary investigation are as follows: (i) YOLOv5 demonstrated higher performance than MTM in detecting small cracks; (ii) an extra-large variant of YOLOv5 outperformed a small-size variant; (iii) the size and object variety of the data are crucial in achieving robust pre-trained weights for use in transfer learning; and (iv) enhanced image resolution contributes to precise object detection.

The high toxicity of materials in disaster waste poses a significant risk to the environment, including the air, water, soil, and all living beings. One of the commonly used disposal methods for hazardous solid waste is landfilling. The selection of sites for hazardous solid waste disposal requires extreme care and attention to multiple factors from environmental, social, and economic points of view. Considering the anticipated earthquake in Istanbul and the city’s excessive population and urbanization, experts estimate that debris waste will be approximately 25 million tons. In this study, we propose a Geographic Information System (GIS) based fuzzy Multi-Criteria Decision Making (MCDM) approach to select hazardous solid waste landfill (HSWL) locations within the scope of disaster waste management for Istanbul. First, the evaluation criteria were identified through a literature review and expert opinions. Next, criteria are prioritized using the Fuzzy Analytic Hierarchy Process (FAHP). Then, GIS data for the criteria are gathered from multiple resources and entered into ArcGIS 10.8 for spatial analysis. Last, the suitability map of Istanbul for the HSWL construction is built. Considering five candidates, the Analytic Hierarchy Process (AHP) is applied to select the most suitable locations for Asian and European sites in the city. Accordingly, Fevzipaşa/Silivri for the European side and Hasanlı/Şile for the Asian side were selected as the most suitable two options. Last, a sensitivity analysis was performed to investigate the impact of the highest weight criterion on the final solution.