Machines are a key element in the production system and their failure causes irreparable effects in terms of cost and time. In this paper, a new multi-objective mathematical model for dynamic cellular manufacturing system (DCMS) is provided with consideration of machine reliability and alternative process routes. In this dynamic model, we attempt to resolve the problem of integrated family (part/machine cell) formation as well as the operators’ assignment to the cells. The first objective minimizes the costs associated with the DCMS. The second objective optimizes the labor utilization and, finally, a minimum value of the variance of workload between different cells is obtained by the third objective function. Due to the NP-hard nature of the cellular manufacturing problem, the problem is initially validated by the GAMS software in smallsized problems, and then the model is solved by two well-known meta-heuristic methods including non-dominated sorting genetic algorithm and multi-objective particle swarm optimization in large-scaled problems. Finally, the results of the two algorithms are compared with respect to five different comparison metrics.

The production planning problem of flexible manufacturing system (FMS) concerns with decisions that have to be made before an FMS begins to produce parts according to a given production plan during an upcoming planning horizon. The main aspect of production planning deals with machine loading problem in which selection of a subset of jobs to be manufactured and assignment of their operations to the relevant machines are made. Such problems are not only combinatorial optimization problems, but also happen to be non-deterministic polynomial-time-hard, making it difficult to obtain satisfactory solutions using traditional optimization techniques. In this paper, an attempt has been made to address the machine loading problem with objectives of minimization of system unbalance and maximization of throughput simultaneously while satisfying the system constraints related to available machining time and tool slot designing and using a meta-hybrid heuristic technique based on genetic algorithm and particle swarm optimization.The results reported in this paper demonstrate the model efficiency and examine the performance of the system with respect to measures such as throughput and system utilization.

The cell formation (CF) is one of the mostimportant steps in the design of a cellular manufacturingsystem (CMS), which it includes machines’ grouping incells and part grouping as separate families, so that thecosts are minimized. The various aspects of the problemshould be considered in a CF. The machine reliability andthe tool assigned to them are the most important problemswhich have to be modeled correctly. Another importantaspect in CMS is material handling costs that they consistof inter-cell and intra-cell movement costs. Moreover, setup and tool replacement costs can be effective in CFdecision making. It is obvious that CF cannot be completedwithout considering the number of demand. With consideringof all of the above aspects, an extended linear integerprogramming is represented for solving the cell formationproblem (CFP) in this study. The objective is to minimizethe sum of inter-cell movement, intra-cell movement, toolreplacement, machine breakdown, and setup costs. In theother terms, for states that cost of movement is higher thantool-changing cost, although a part can have the inter-and/or intra-cell movements, the model tries to find a solutionwhich part is allocated to one cell and with changing thetools, processes of that part is completed. In addition, tovalidate the model and show its efficiency and performance, several examples are solved by branch and bound(B&B) method.

The main contribution of the present study is to develop a novel hybrid machine learning model to enhance the defect prediction in industrial manufacturing processes. In this work, the model integrated four base models of XGBoost, LightGBM, CatBoost, and an artificial network, whose features are modeled with Random Forest (RF) as the metamodel using a stacking ensemble approach. For this study, the industrial data from Kaggle were used, and for their extensive and detailed hyperparameter optimization with Optuna, we greatly improved the prediction performance with the model. In the context of this study, key challenges like the data imbalance and the selection of the important features were solved using data balancing techniques like SMOTE and random forest-based analysis for selecting the most important input features. The hybrid model generated great results, which were quite better than the traditional single models, with an accuracy of 96.06% and precision, recall, and F1 scores of 95.10%, 97.32%, and 96.20%, respectively. The real-world applications of this model can be many by accurately and timely predicting defects in industrial environments. All results are reliable and interpretable due to the usage of robust data preprocessing methods, including feature standardization and correlation analysis. This study's results will have a significant impact on such tasks as defect management in manufacturing, as it provides a very scalable solution to enhance product quality, minimize operational cost, and improve process efficiency. This research illustrates the promise of hybrid machine learning methods in tooling manufacturing process optimization and the performance of industry.

Through advanced analytical models such as machine learning (ML) and, conceptually, Large Language Models (LLMs), this study explores how Industrial Internet of Things (IIoT) applications can transform educational experiences in the context of smart steel production. To mitigate the shortage of authentic industrial datasets for research, we developed an industry-validated IIoT educational dataset drawn from three months of operational records at a steel plant and enriched with domain-specific annotations—most notably distinct operational phases. Building on this foundation, we propose an IIoT framework for intelligent steel manufacturing that merges ML-driven predictive analytics (employing Lasso regression to optimize energy use) with LLM-based contextualization of data streams within IIoT environments. At its core, this architecture delivers real-time process monitoring alongside adaptive learning modules, effectively simulating the dynamics of a smart factory. By promoting human–machine collaboration and mirroring quality-control workflows, the framework bridges the divide between theoretical instruction and hands-on industrial practice. A key feature is an interactive decision-support dashboard: this interface presents ML model outcomes and elucidates IIoT measurements—such as metallization levels and H2/CO ratios—through dynamic visualizations and scenario-based simulations that invite risk-free exploration of energy-optimization strategies. Such tools empower learners to grasp the intricate multivariate dependencies that govern steel manufacturing processes. Our implementation of the Lasso regression model resulted in a 9% reduction in energy consumption and stabilization of metallization levels. Overall, these findings underscore how embedding advanced analytics within IIoT education can cultivate a more engaging, practice-oriented learning environment that aligns closely with real-world industrial operations.

The fundamental function of a cellular manufacturing system (CMS) is based on the definition and recognition of a type of similarity among parts that should be produced in a planning period. Cell formation (CF) and cell layout design are two important steps in implementation of the CMS. This paper represents a new nonlinear mathematical programming model for dynamic cell formation that employs the rectilinear distance notion to determine the layout in the continuous space. In the proposed model, machines are considered unreliable with a stochastic time between failures. The objective function calculates the costs of inter- and intra-cell movements of parts and the cost due to the existence of exceptional elements (EEs), cell reconfigurations, and machine breakdowns. Due to the problem’s complexity, the presented mathematical model is categorized in NP-hardness; thus, a genetic algorithm (GA) is used for solving this problem. Several crossover and mutation strategies are adjusted for GA and parameters are calibrated based on Taguchi experimental design method. The great efficiency of the proposed GA is then demonstrated by drawing a comparison between particle swarm optimization (PSO) and the optimum solution via GAMS considering several small/medium- and large-sized problems.