International Journal of Industrial Optimization

Effect of return rate on optimal order quantity for single-period products: a modified newsboy problem approach

Wen-Feng Hsiao — 2026-02-25

This paper proposes a modified newsboy problem for single-period fashion products, sold by e-stores, for the case where product returns are allowed during the sales season. This is because, even if an e-store sells high-quality products, returns may still occur due to unpredictable consumer behavior in online sales. In the developed model, the assumption from the literature that the number of returns during the sales period is unlimited has been modified to align with real-world conditions. In formulating the problem, the number of returns is assumed to be bounded by a random variable. The corresponding mathematical model is established and the optimal order quantity investigated. The traditional newsboy model is shown to be a special case of the proposed model for which the return factor is ignored. The numerical results show that the optimal order quantity determined when the return rate is ignored is always overestimated and the resulting expected total profit is thus always underestimated. A sensitivity analysis is performed to determine the effects of the model parameters on the optimal solution. The numerical example demonstrates that, in the case of the product return rate, the order quantity predicted by the traditional newsboy model is overestimated by approximately 10%, which results in a profit reduction of about 1%. The numerical results have shown that when the return rate is ignored, the estimated optimal order quantity is higher than the true optimal order quantity and causes the total profit to be slightly underestimated.

Semantic brain tumor segmentation from 3D MRI using u2-net with custom dilated and residual u-block

Elvaret — 2026-02-25

Segmentation of brain tumors in volumetric medical images is challenging due to the complexities of the tumor structure, the types, and the heavy-weight 3D data processing. In contrast, 2D-based segmentation methods on the slice data reduce the amount of information due to the anisotropic shape of the tumors and lead to poor segmentation results. This study proposes a 3D network structure combining ReSidual U-Block (RSU), custom dilated block, and U2-Net for automatic segmentation of brain tumors from MRI images, namely 3D RSU U2-Net+. The RSU and custom dilated block are embedded and joined in the nested U-Net structure to obtain multi-resolution features and global information, enhancing segmentation accuracy while reducing computational overhead. The proposed method outperformed the segmentation results of the standard U-Net, on brain tumor data in the medical segmentation Decathlon (MSD) dataset. The proposed model achieves an average validation soft dice loss of 0.1320 and dice score coefficient of 78% and intersection over union of 64% for testing. Although having 3 times parameters, the model requires less GPU time (397.7 minutes) than U-Net (433.6 minutes), demonstrating improved computational efficiency resulting from the effective use of residual and dilated blocks. Moreover, the model achieves 75.4% average sensitivity and 99% specificity for edema, enhancing, and non-enhancing tumors. These experimental results show that the 3D RSU U2-Net+ has been able to outperform the U-Net. However, the model’s performance on non-enhancing tumors remains relatively lower compared to other tumor types, indicating on opportunity for further optimization.

Occupational health risk assessment of manufacturing workers using the hand activity level method

Silvi Widodo — 2026-02-25

Repetitive manual handling tasks in industrial settings often expose workers to musculoskeletal risks, particularly when performed without ergonomic consideration. At PT. ABC, production workers are routinely engaged in lifting and assembling heavy components, raising concerns about their long-term health and safety. To address this issue, a structured assessment using the ACGIH TLV for Hand Activity Level (HAL) was conducted to evaluate biomechanical exposure and identify ergonomic risks. This study contributes by applying a quantitative, evidence-based framework to assess real workplace conditions and offer actionable insights for intervention. It also demonstrates how HAL and Borg CR-10 metrics can be integrated into practical ergonomic evaluations in industrial environments. The research involved five workers from the concrete production division. Data were collected through direct observation and video analysis to determine hand movement frequency and peak force levels. The HAL values and Borg CR-10 scores were used to calculate the Exposure Ratio (ER) for each worker, serving as the main indicator of ergonomic risk. Results revealed that all five workers had ER values ranging from 1.18 to 1.30, exceeding the ACGIH TLV threshold of 1.0. This indicates a consistently high risk for work-related musculoskeletal disorders (WMSDs). Frequent lifting of 12–13 kg loads combined with moderate-to-high hand activity and poor posture contributed to elevated strain levels. These findings confirm that the existing work system places employees at risk and highlight the need for immediate ergonomic improvements. Moreover, the HAL-ER assessment framework used in this study provides structured data that can be utilized in simulation-based planning or optimization models. By integrating these metrics into ergonomic redesign scenarios such as task reallocation, force-load balancing, or layout planning future studies can enhance both worker safety and operational efficiency.

The role of mathematical formulation in solving the unbalanced assignment problem

Francis J. Vasko — 2026-02-25

In a 2019 paper, the authors claim to have developed a modified Hungarian method that performs better than a number of other solution methods for the unbalanced assignment problem (UAP) based on the solution of one UAP instance that has been discussed in the literature. The purpose of this short paper is to demonstrate that the math formulation used in the 2019 paper was not as restrictive as the standard one commonly used in the literature and therefore the comparison is not valid. The commonly used UAP math formulation not only tries to minimize cost, but also tries to level load the jobs onto the machines. The formulation from the 2019 paper allows many jobs to be assigned to a low-cost machine. Hence solutions (not even optimums) to the 2019 formulation can be better than the optimal solution using the standard UAP math formulation. Additionally, it will be shown that the Modified Hungarian method proposed in the 2019 paper does not generate guaranteed optimums to the math formulation used in that paper (let alone the standard UAP formulation). An 8-job and 5-machine assignment problem that appeared in the literature will be used to illustrate the points mentioned above.

Robust parametric optimization of cyclone separator by means of probabilistic multi - objective optimization

Maosheng Zheng — 2026-02-25

In this article, robust parametric optimization of cyclone separator is done by means of robust probabilistic multi - objective optimization (RPMOO). In RPMOO, the optimal attributes (objectives) are essentially divided into two types, i.e., both unbeneficial and beneficial types, which devote their partial preferable probabilities with equivalent manner quantitatively; especially the averaged value of the experimental data of each attribute and its dispersity are evaluated individually in accordance with its corresponding type. The total preferable probability of each scheme alternative is formed from the multiplication of all available partial preferable probabilities, which is the uniquely decisive indicator of an alternative in this assessment; the optimum scheme is with the highest total preferable probability. For the parametric optimization of cyclone separator, the inlet velocity, helical angle, and outlet diameter are as the variable parameters, while the pressure drop and separation efficiency are the evaluated responses of the cyclone separator to get optimization, the former is an unbeneficial type of attribute and the latter is a beneficial type of attribute. The orthogonal array L9(33) was employed to arrange the experimental scheme alternatives. The evaluated results indicate that the optimized experimental scheme is alternative 6, which yields the optimal responses of a pressure drop of 0.3 mba and a separation efficiency of 98.95 % at an optimum inlet velocity of 13 m/s, an outlet diameter of 72 mm, and a helical angle of 5. This work reveals the independent contributions of the averaged value of the experimental data and its dispersion to an attribute response in the optimization process, and the irrelevance of pressure drop and separation efficiency in the system.

Optimization of real-time forest monitoring system using yolo v9 object detection and 2.4 ghz wireless network: resource allocation, energy efficiency, and industrial deployment strategies

Rachmad Andri Atmoko — 2026-02-25

Large forest areas are increasingly exposed to illegal activities and environmental threats, while conventional monitoring systems suffer from limited coverage, high energy consumption, and delayed response. To address these challenges, this study proposes an optimized real-time forest monitoring system designed for industrial-scale deployment in remote environments. The primary objective is to enhance surveillance efficiency by integrating AI-based object detection, long-range wireless communication, and resource-efficient system design. The proposed system employs ESP32-CAM sensor nodes integrated with 2.4 GHz CPE wireless links and a gateway-based YOLOv9 object detection framework. Bandwidth utilization is optimized through selective transmission of processed detection metadata instead of raw images, while deployment parameters are optimized using simulation-based planning. A web-based monitoring platform with an optimized REST API supports real-time visualization and alert generation. Experimental results show that the system achieves reliable communication up to 500 m with packet loss below 5% and latency under 50 ms at distances up to 300 m. Human detection accuracy reaches 98.5% under optimal conditions, with performance degradation observed in dense vegetation and low-light environments. Energy evaluation confirms sustainable operation, with ESP32 nodes consuming 160 mA and the gateway operating at 3.7 W. Comparative analysis indicates reductions of 37% in deployment cost, 24% in energy consumption, and 51% in latency compared to similar systems. This study concludes that the proposed architecture effectively balances accuracy, scalability, cost, and energy efficiency. The novelty lies in the integrated optimization of edge-based AI detection, selective data transmission, and simulation-driven deployment for industrial forest monitoring.