International Journal of Industrial Engineering: Theory, Applications and Practice

Trajectory Optimization of Tobacco Seedling Planting Mechanism Based On NSGA-II Algorithm and Response Surface

Liqiu Yang — 2026-07-28

In order to ensure the uprightness of tobacco seedlings during transplanting and improve the transplanting quality, this paper proposes a hybrid optimization framework that integrates the NSGA-II multi-objective genetic algorithm and response surface methodology (RSM). Taking the planting depth and lateral offset as the optimization objectives, nine key mechanism parameters affecting the planting trajectory are selected as the initial design variables. Firstly, with the global search capability of the NSGA-II algorithm, we locate the feasible region that fits the constraints of the mechanism and obtain the optimal solution set. Then we screen out five significant influencing factors based on the Plackett-Burman test, and use the Box-Behnken response surface design to carry out the localized refinement modeling. Based on the regression model, an optimal combination of parameters is obtained; the mechanical trajectory under this set of parameters meets the agronomic requirements for transplanting.

Edge Computing Terminal Deployment in Power Internet of Things Based on Cultural Ant Colony Algorithm

Zhengrong Wu — 2026-07-28

This paper addresses the dynamic deployment of edge computing terminals in the heterogeneous Power Internet of Things (PIoT), where large-scale devices, fluctuating concurrency, and spatial heterogeneity challenge the balance among coverage, resource efficiency, and cost. We propose a dynamic deployment strategy based on the Cultural Ant Colony Algorithm (CACA). A multi-dimensional terminal model is constructed, covering candidate locations and hardware parameters. With minimization of annual average deployment cost as the objective, constraints including communication uniqueness, hardware adaptability, and delay are incorporated. CACA integrates swarm intelligence and cultural evolution via pheromone-guided search, elite collaboration, and global update to yield near-optimal deployment plans. Experiments show that our method achieves full coverage under varying states, saves over RMB 110,000, reduces response time by >22%, and lowers energy consumption and failure rate by >15% and >18%, respectively. It demonstrates superior engineering applicability and optimization efficiency.

Class-Imbalance Aware Learning for Enhanced Noninvasive Neonatal Jaundice Diagnosis – Using Convnext-Tiny with Focal Loss (CNXT-FL)

Sathyaseelan Krishnaraj — 2026-07-28

Early diagnosis of neonatal jaundice carries significance as it could prevent serious neurological problems. Conventional diagnostics based on assessment of total serum bilirubin (TSB) in blood samples are invasive, painful, time-consuming, and require a laboratory setting. As such facilities are often obscure in resource-limited remote locations, AI-based image analysis was frequently investigated as an alternate solution to optimize clinical workflow. However, the need for an optimized healthcare decision support system remains a research gap. This research introduces a Transformer-inspired ConvNeXt-Tiny architecture with Focal Loss (CNXt-Fl) for rapid, scalable, and accurate diagnosis of neonatal jaundice. The proposed framework includes (i) Albumentations-based augmentation for ensuring model robustness to image acquisition conditions, (ii) ‘focal loss’ for class-weighted learning to address class imbalance, and (iii) a pretrained ConvNeXt-Tiny model to ensure insensitivity to small datasets. Experimental results from CNXt-Fl evaluated on a public neonatal jaundice dataset consisting of 760 neonatal images achieved an accuracy and F1-score of 0.9912 and 0.99, respectively. Further, comparison with conventional CNN, other pre-trained models, and state-of-the-art deep learning models revealed that the proposed model achieves better results comparatively. These results highlight the potential of lightweight transformer-inspired architectures with class-imbalance aware learning for reliable early detection of neonatal jaundice in real-world healthcare environments. Thus, this work holds significance in the optimization of healthcare decision support systems.

A Novel Multi-Expert Multi-Criteria Decision-Making Method Integrating Fermatean Fuzzy Sets with CRITIC Weighting

Kuei-Hu Chang — 2026-07-28

Due to the involvement of multiple experts and evaluation criteria, the multi-expert multi-criteria decision-making (MEMCDM) problem becomes a highly complex decision-making task. The MEMCDM problem includes both qualitative and quantitative evaluation criteria and accounts for experts’ differing cognitive information. The subjective and objective weights of the evaluation criteria must also be considered. However, conventional MEMCDM methods cannot simultaneously and effectively process experts’ cognitive information while considering both subjective and objective criterion weights. Therefore, this paper proposes a novel MEMCDM approach that integrates the Fermatean fuzzy set (FS) with the criteria importance through the intercriteria correlation (CRITIC) weighting method. The proposed method employs Fermatean FS for its capacity to simultaneously handle both qualitative and quantitative evaluation criteria in MEMCDM problems. In the numerical example, an illustrative case of electric vehicle charging station location selection is employed to demonstrate and validate the effectiveness of the proposed MEMCDM approach. The calculated results are also compared with those obtained using the intuitionistic FS method, the conventional Fermatean FS method, and the CRITIC method. The numerical results for the location selection of the electric vehicle charging station demonstrate the rationality and effectiveness of the proposed MEMCDM approach.

Cluster Validity Indices for Uncertain Data Based on A Non-Parametric Kernel Method

Changwan Ko — 2026-07-28

Cluster validity indices are essential for determining the true number of clusters and validating clustering results. Most existing indices represent each data object as a single point and therefore fail to capture uncertainty inherent in the data. Even indices that consider uncertainty often rely on predefined probability distributions, such as Gaussian distributions. In this study, we propose non-parametric cluster validity indices derived from conventional indices, namely the Dunn, Calinski-Harabasz, and Davies-Bouldin. The proposed indices do not require a specific prior probability measure and can handle arbitrary forms, sub-clusters, noise, and high dimensionality. These indices compute compactness (within-cluster dispersion) and separability (between-cluster separation) in a reproducing kernel Hilbert space using a kernel function to assess cluster validity. Experiments on both artificial benchmark and real-world astronomical datasets demonstrate that the proposed indices are superior to existing cluster validity indices for validating the results of clustering problems.

Design and Implementation of An Automated Scheduler for Pharmacists in A Local Hospital

Lotfi Hidri — 2026-07-28

To deliver adequate healthcare to the community, it is crucial to lower expenses and enhance the working environments for healthcare professionals, such as nurses, doctors, and pharmacists. Traditionally, pharmacists were not considered a significant component of the healthcare system, as their role primarily involved medication delivery. However, over the past few decades, the role of pharmacists has evolved significantly, making them a crucial and integral part of the healthcare staff due to their involvement in new missions. Consequently, the pharmacists' community must be considered within the broader healthcare system. Pharmacists now face new challenges, such as unbalanced workloads and burnout. This paper examines a case study involving pharmacists at a local hospital. Pharmacist scheduling was traditionally performed manually, a process that proved both time-consuming and error-prone. Moreover, the resulting schedules often failed to meet pharmacists’ preferences, leading to dissatisfaction. A decision support system (DSS) for optimizing pharmacist scheduling in a hospital is developed in this paper. This DSS aims to automatically and efficiently generate well-balanced schedules that satisfy the pharmacists as much as possible. Balancing workload, considering pharmacist preferences, and meeting daily coverage requirements are challenging tasks that are difficult to accomplish manually without the assistance of a computer. Therefore, a mathematical model, specifically a Mixed Integer Programming (MIP) model, was developed to generate an optimal schedule for pharmacists. This model considers individual pharmacists’ preferences, their ability to fulfill duties, workload balancing, and annual leave requirements. The model has been implemented, and a graphical user interface has been developed. This interface allows users to input the required number of pharmacists needed each month, the preferences and also allows automatic generation of the schedules. An experimental study was conducted to evaluate the proposed DSS's performance. As a result of the automated schedules, overtime has been eliminated completely, and pharmacists' preferences have been accommodated. The proposed DSS automates pharmacist scheduling while considering human preferences and fairness criteria, offering a practical tool for hospital administrators to enhance efficiency and staff satisfaction.

Multi-Skilled Worker Matching Method Considering The Changes of Serus in The Seru Production System

Kunyuan Huang — 2026-07-28

In the seru production system (SPS), matching multi-skilled workers and serus is a crucial problem. In this paper, a multi-skilled worker matching method in SPS considering the changes of serus is proposed. It can be used to generate the initial matching scheme for multi-skilled workers and serus before the production begins, and to adjust the matching scheme under the changes of serus caused by the changes of production tasks during production. Firstly, according to the preference information of multi-skilled workers and processes in the SPS, the satisfaction degree calculation methods of multi-skilled workers and processes in the SPS are proposed. Furthermore, a multi-objective programming model for matching multi-skilled workers and serus is constructed. Meanwhile, based on the classical multi-objective immune algorithm (MOIA), an improved multi-objective immune algorithm (IMOIA) is designed to solve the model. The experimental results demonstrate that IMOIA has advantages over other algorithms in set size, dominance, convergence, and distribution uniformity of Pareto solutions. Finally, the effects of different matching scheme adjustment methods and different algorithms are compared, and sensitivity analyses are carried out.

Research on E-Closed-Loop Supply Chain Decision-Making Considering Service Competition and Government Under Different Platform Recycling Modes

Yuntian Xia — 2026-07-28

Considering an E-closed-loop supply chain consisting of a producer, the competing resell e-commerce platforms and a third-party platform, under the government subsidy mechanism, by constructing game models under four scenarios, the influence of different platform reusing modes and facility level competition intensity on the optimal tactical choice of the E-closed-loop supply chain is compared and then analyzed. The study found that: (1) Under the mode of resale e-commerce platform recycling-government subsidy to resale e-commerce platforms, the profits of producers and resell e-commerce platforms are the highest, while under the mode of third-party platform recycling-government subsidy to the manufacturer, the profit of the producer is lowest. (2) When the degree of competition in the service level of resell e-commerce platforms is low, the service level is highest in the mode of government-subsidized recycling by resell e-commerce platforms. When the degree of competition is high, the service level provided by resell e-commerce platforms is the highest in the form of third-party platform recycling-government support to the producer. (3) As the recycling cost of producers increases, the order amount of products, the recycling rate of products and the number of recycled used products continue to increase.

A Novel Fuzzy Approach for Solving A Fully Fuzzy Multi-Objective Transportation Problem Using Triangular Fuzzy Numbers

Priyanka Sharma — 2026-07-28

The transportation problem (TP) is a classical optimization model in operations research and logistics and represents a special class of linear programming problems. In practical decision-making scenarios, multiple conflicting objectives often need to be optimized simultaneously under uncertain conditions. To address this uncertainty, fuzzy concepts are used to model the imprecision in transportation costs, supplies, and demands. Although several methods have been developed for solving fully fuzzy multi-objective transportation problems (MOTP), classical solution techniques such as the North-West Corner Rule have rarely been extended to fully fuzzy environments. This paper proposes a new fuzzy extension of the North-West Corner Rule to obtain a compromise-efficient solution for a fully fuzzy multi-objective transportation problem, in which all parameters are represented by triangular fuzzy numbers. The fuzzy multi-objective problem is transformed into a fuzzy single-objective transportation problem (SOTP), and the proposed approach is used to generate an initial fuzzy basic feasible solution, which is further improved using the fuzzy generalized modified distribution (FG-MODI) method. Numerical examples demonstrate that the proposed method yields fuzzy optimal solutions comparable to existing techniques while requiring fewer computational steps, thereby highlighting its efficiency and simplicity. The method is easy to implement and suiTable for real-world transportation problems characterized by uncertainty. Additional analytical metrics, such as sensitivity and compromise analysis, further validate the robustness and effectiveness of the proposed method.

Scheduling Model for Complex Equipment Development Considering Project Robustness

Chen-Hao Wang — 2026-07-28

To deal with the cooperative scheduling problem of complex equipment development task duration, cost and project stability, considering the uncertainty of development process, the cost limitation of development task chain and project stability, from the perspective of project robustness, using Graphical Evaluation and Review Technique (GERT), AoN network and the idea and method of milestone, we construct a scheduling model for complex equipment development considering task rework. Firstly, we construct a hybrid GAN network based on milestones and an embedded AoN network, considering rework problems caused by overlapping tasks and use robustness to measure the stability of the project. Then, aiming at maximizing project robustness and minimizing project cost, establish a robust scheduling optimization model for complex equipment development projects, and design a genetic solution algorithm. Finally, we verify the validity and rationality of the model by an example; the new model is completed 7 months earlier than the average duration without considering the sequential execution of task chains.

Circular Economy and Lean Production in Taiwan's Green Energy Industry: A Hybrid MADM-QFD Approach

Po-Yen Lai — 2026-07-28

The urgent need to implement and expand the Circular Economy (CE) has made it essential to develop an integrated evaluation model. Existing studies indicate that industries face numerous challenges in their transition to CE. This study conducts a series of quantitative analyses focusing on CE performance measurement, the relationship between Lean Production (LP) methods and CE, using Taiwan's green energy industry as the empirical research subject. The selection of this industry is based on its frontline role in sustainable development. The primary objectives of this study are to define critical CE performance indicators and investigate applicable LP methods. By applying the Quality Function Deployment method, particularly the House of Quality, this research analyzes the interaction between LP and CE, employing Multi-Attribute Decision-Making techniques to quantify LP’s contribution to CE performance. While prior research suggests that LP and CE can complement each other, their practical integration remains an area worth exploring. The results indicate that Visual Management emerged as the highest-impact LP method for enhancing CE performance, followed by Identification of Equipment Failure Points and Root Causes, Poka-Yoke, Production Management Kanban, and the Elimination of No-Load Operations. The findings of this study provide valuable insights for government agencies and corporate managers seeking to promote CE implementation.

How Does Cross-Screen Visual Search Influence Visual Processing and Eye Movement Behavior Under Different Perceptual Window Conditions?

Biao Yan — 2026-07-28

The increasing adoption of multi-screen systems in complex human–machine environments highlights the importance of understanding cross-screen visual search behavior for improving operational efficiency. However, the mechanisms underlying changes in visual processing and eye movement behavior during screen transitions remain insufficiently understood. This study employed a moving window paradigm to systematically investigate visual search performance and eye movement behavior during cross-screen visual search. A total of 20 participants completed visual search tasks in a simulated multi-display interface, while reaction time, fixation behavior, and area-of-interest (AOI) eye-tracking metrics were recorded. The results revealed significant effects of the perceptual-window condition on reaction time (p < 0.001), fixation duration (p < 0.001), and fixation count (p < 0.001). Reaction time decreased substantially from the 3° condition to the 5° condition, whereas only limited improvement was observed between the 5° and 7° conditions. Similarly, fixation duration and fixation count were significantly reduced under larger perceptual-window conditions, while no significant differences were found between the 5° and 7° conditions. These findings indicate a plateau effect in eye-movement-based visual processing efficiency beyond approximately 5°. Significant spatial orientation effects were also observed for reaction time (p < 0.001), time to first fixation (p < 0.001), and AOI visit count (p < 0.001). The east condition produced the longest reaction time and time to first fixation, whereas the west condition yielded the shortest reaction time. In contrast, the north condition showed the shortest time to first fixation and the highest AOI visit count, suggesting more efficient attentional engagement and visual exploration in the north region. Participants showed faster visual engagement and higher visit counts in the north region, followed by the west, while the east and south regions exhibited reduced visual sampling, indicating a stable spatial asymmetry in cross-screen visual search behavior. Although the north condition facilitated faster attentional engagement, the west condition yielded the shortest overall reaction time, suggesting partially distinct mechanisms between early attentional orientation and overall search efficiency. Overall, these findings indicate that cross-screen visual search dynamically influences visual processing and gaze allocation patterns under different perceptual-window constraints. The results provide empirical evidence for optimizing multi-display workstation design, control room interfaces, and complex monitoring systems, and offer a behavioral basis for improving information layout strategies in cross-screen operational environments.

Optimizing Resilience Using Network Structural Parameters for Reconfigurable Supply Chains

Hrishikesh Choudhary — 2026-07-28

Resilience of supply chains is considered a desirable feature owing to the disruptions and uncertainties at various levels and operations. Several strategies to enhance resilience are practiced with supply chain network designs, multisource decisions, inventory and capacity buffers, near-shoring, etc. In the present work, the assessment and optimization of resilience under reconfigurations of a generic supply chain is conducted utilizing network structural parameters like size, density, clustering coefficient, centrality, and effective arc length. The effect of reconfigurations on resilience is analyzed separately for the upstream and downstream of the supply chain since the practical implementation on each side is different. A bi-objective optimization model is developed for maximizing resilience and minimizing transportation cost, considering several potential configurations using the network structural parameters. Pareto optimal solutions are found through NSGA II for model data sets of these combinatorial optimization problems representing upstream and downstream separately.

Evolutionary Game Analysis of Integration Decisions for Informal WEEE Processing Channels under Government Intervention

Ziyan Zhang — 2026-07-28

The rapid growth of waste electrical and electronic equipment (WEEE) has become an important environmental issue. China’s WEEE recycling and treatment system consists of both formal and informal channels. Informal channels have long played an important role, but they are associated with serious pollution, low efficiency, and disorderly development. Based on China’s dual-channel recycling governance context, this study examines the integration strategy choices of upstream informal recyclers and downstream informal dismantlers under government support. First, this study constructs payoff matrices under different strategy combinations based on a Stackelberg game. It then applies evolutionary game theory to analyze the strategy evolution process of both parties and uses MATLAB simulations to examine the effects of key parameters. The results show that the behavioral evolution of informal participants exhibits a follow-the-other effect; that is, both parties tend to choose either “integration” or “non-integration” simultaneously. Government subsidies significantly affect the direction of equilibrium evolution. When the subsidies satisfy S1∈(0,0.97) and S2∈(0,1.93）, the system may evolve toward either “non-integration, non-integration” or “integration, integration” as stable states. When the subsidies exceed these ranges, the system is more likely to evolve toward “integration, integration. Numerical simulations further show that the recycling competition rate has a U-shaped effect on integration strategy choice. In addition, the lower the recycling utilization rate of informal dismantlers and the stronger the cooperation willingness of informal participants, the more likely the system is to evolve toward “integration, integration.”

Optimizing Supply Chain Performance: A Hybrid Approach that Integrates Artificial Intelligence and Mathematical Modeling with Stochastic Models to Reduce Costs and Improve Efficiency in Manufacturing

Junhua Wang — 2026-07-28

Supply chain management plays a role in guaranteeing cost-efficiency and the smooth running of production and logistics. Existing methods, such as deterministic optimization models and traditional machine learning approaches, often face limitations in handling uncertainties, temporal dependencies, and high-dimensional structured data, which restrict their ability to achieve robust forecasting and optimization. To address these concerns, an integrated framework using the TabNet and Stochastic Mixed-Integer Linear Programming (SMILP) is proposed with an aim to cost reduction and increasing efficiency in supply chain systems. The methodology proposed herein initiates with the preprocessing of data, with missing values in the dataset imputation process using K-Nearest Neighbor (KNN) and mode imputation methods, including Min–Max scale normalization. Next, feature engineering is conducted, where rolling window statistics are applied to provide seasonal and temporal demand trend information. The TabNet model is used to learn feature importance and provide demand forecasting, which is then routed into SMILP to optimization production, inventory, and logistics in anticipation of uncertainty. The hybrid orientation encapsulates the predictive power of TabNet, as well as the power to make decisions in the presence of stochastic optimization, to provide resiliency and adaptability. After evaluating the effectiveness with Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE), we conclude that the cost-optimizing power of the model affects management in the supply chain.

Integrating Sustainable Supply Chain Management, Industry 5.0 and Circular Economy: A Bibliometric Review of Research Directions

Prakriti Gajar — 2026-07-28

This study aims to discover the convergence of sustainability imperatives, digital transformation, and human-centric industrial paradigms in the context of Sustainable Supply Chain Management, Industry 5.0, and the Circular Economy. Recognizing the current state of literature is theoretically disjointed with a lack of integrative synthesis on these topics, this research seeks to fill the theoretical divide by conducting a bibliometric and SLR of 429 peer-reviewed articles. The aim is to explore the intellectual structure, thematic development, and evolving research streams at their intersection. Utilizing the PRISMA guidelines in conjunction with bibliometric analysis in Biblioshiny and VOSviewer software, the results reveal key themes, collaboration networks, and uncharted conceptual intersections. Along with descriptive mapping, results are integrated using the socio-technical systems perspective, general systems theory & resource-based view to develop an integrative conceptual framework. This conceptual framework helps to explain how human-centric digital technologies can be leveraged for the transformation of circular and sustainable supply chains.

An Economic Bayesian Model for a Modified Group Chain Sampling Plan under the Weibull Distribution

Faten S. Alamri — 2026-07-28

This paper presents an economic model for the Modified Group Chain Sampling Plan (MGChSP) under the Weibull distribution using a Bayesian framework. The MGChSP integrates features of group and chain sampling, providing higher accuracy with smaller sample sizes, particularly valuable in situations involving costly or destructive testing. By incorporating prior information, the Bayesian approach enhances cost estimation and supports more reliable decisions when limited data are available. The total cost of the proposed plan is compared with existing sampling plans, demonstrating that the Bayesian method yields more cost-effective acceptance decisions. Comprehensive charts, tables, and computations are included to illustrate the performance and comparisons of the proposed model.