Abstract Large-scale multi-objective optimization problems are prevalent in practical applications. Consequently, researchers have proposed numerous large-scale multi-objective evolutionary algorithms (LMOEAs) to address these problems. Among them, neural network (NN)-based LMOEAs have garnered significant attention owing to their superior search efficiency. Regrettably, existing NN-based LMOEAs are limited by a single source of training data and pre-fixed network topology, resulting in insufficient generalization and weak adaptability. We propose a heterogeneous operator mechanism and stochastic configuration network-enhanced LMOEA (HMSCN-LMOEA) to overcome these issues. Specifically, a heterogeneous operator mechanism is introduced to address the limited data sources in NN model training, thereby effectively enhancing the quality of the training set. A stochastic configuration network (SCN) model with dynamic structural adaptability is designed to replace the NN model with a pre-fixed network topology, thereby enhancing the algorithm’s adaptability. Finally, to evaluate the algorithm’s performance, four large-scale multi-objective optimization problems—UF, WFG, LSMOP, and ZCAT—are adopted. Experimental results based on IGD, IGD+, Spacing, and HV metrics demonstrate the significant advantages and competitiveness of HMSCN-LMOEA over state-of-the-art LMOEAs. Furthermore, HMSCN-LMOEA is employed to solve the cloud task scheduling problem across different task scales, and the experimental results show that it achieves the top rank in 75% of the evaluated scenarios, further demonstrating its promising potential.

Abstract The intelligent scheduling of autonomous trucks at intersections in open-pit coal mines poses a fundamental challenge due to nonlinear dynamics and complex multi-truck interactions under gradient–load coupling. Unlike existing intersection control studies that are mainly designed for flat urban roads and neglect mining-specific physical constraints, this study proposes a Gradient-Load Aware Multi-Agent Attention-Enhanced DDPG (GLA-MA-MADDPG) approach that explicitly embeds mining-domain physical laws into multi-agent reinforcement learning for intersection scheduling. A gradient–load dynamic coupling model is developed to inject physics fidelity into MARL-based intersection scheduling, capturing nonlinear and load-sensitive kinematics overlooked by flat-road assumptions in prior studies. To handle coordination under these constraints, we design a task-oriented multi-dimensional attention mechanism that jointly interprets physical heterogeneity, asymmetric dynamics, priorities, and collision risks. Additionally, a gradient-aware adaptive priority strategy redefines right-of-way as a physics-grounded, state-dependent process, ensuring safe and preferential passage for loaded and downhill trucks. In the simulated four-branch gradient intersection environment, the proposed GLA-MA-MADDPG method achieves substantial performance gains over traditional MADDPG across 20 independent runs. Specifically, it improves throughput by 23.5%, reduces average transit time by 8.3%, decreases waiting time by 31.2%, and lowers the collision rate from 3.47‰ to 1.32%, achieving efficiency gains of 39.2% in mixed uphill–downhill scenarios. Overall, this study contributes a first-of-its-kind integration of physics-informed gradient–load modeling with attention-driven MARL, providing a generalizable computational framework for intelligent intersection scheduling in open-pit mining and other large-scale, safety-critical engineering systems.

Abstract Global and constraint optimization in engineering structural design problems often involves more complex types, which increases computational complexity. To address this challenge, this paper constructs an exploration mechanism inspired by the hunting behaviors of marine octopuses, along with an exploitation mechanism based on their mating behaviors. These mechanisms aim to balance convergence speed and solution accuracy using a specially designed stochastic regulatory factor. This paper develops a nature-swarm phenomenon-based search strategy and mathematical model, named the Octopus Optimization Algorithm (OOA), by simulating processes of octopuses searching for potential prey, escaping natural predators, attacking prey, and mating behaviors. In addition, inspired by the water-spraying recoil and transient acceleration phenomenon, a recoil motion-based stochastic feedback mechanism is proposed by designing a unique recoil operator to achieve information exchange in different search spaces. To demonstrate the universal applicability of the proposed OOA algorithm, we qualitatively analyzed swarm convergence and swarm search behaviors, population diversity, exploration and exploitation performance on 84 benchmarks covering unimodal, multimodal, fixed-dimensional, and composite functions and quantitatively verified convergence, effectiveness, significance, robustness, population diversity, exploration and exploitation efficiency, progressive scalability, and parameter sensitivity on the CEC2017 suites with 10, 30, 50, and 100 dimensions. Moreover, OOA beats 12 highly cited competitors in terms of computational performance when solving different optimization problems. Based on the pairwise comparisons-based Wilcoxon test and multiple pairwise comparisons-based Friedman test, it indicates that compared to 12 state-of-the-art algorithms, OOA achieved a mean rank of 1.19 across 84 benchmarks. The nonparametric test significance results show OOA contains 981 positive signs out of 1008 comparisons (84 benchmarks), with an optimization efficiency of 97.3%. On the CEC2017 suites, the mean ranks across four dimensions were 1.22 with 10Dim, 1.0 with 30Dim, 1.0 with 50Dim, and 1.0 with 100Dim, respectively, all ranking first. The nonparametric test results indicate OOA contains 1427 positive signs out of 1440 comparisons (120 benchmarks), with a solution efficiency of 99.1%. Thus, the proposed OOA algorithm demonstrates statistically significant advantages in computational performance and scalability. OOA has achieved better results than competitors in eight engineering problems, showing superior computational efficiency and reliability.

Abstract Surface defect detection plays a critical role in quality control of industrial products, particularly in steel manufacturing. However, existing deep learning methods often suffer from high computational costs, limited generalization, and constrained accuracy in complex defect scenarios. To address these challenges, we propose a lightweight detection model, RAGA-YOLO, based on an improved YOLOv11-n. The model incorporates a novel lightweight neck structure, CCFM-PAN-FPN, which combines channel alignment with efficient downsampling to significantly reduce both Params and GFLOPs. Additionally, the Relation-aware Global Attention (RGA) module is introduced to enhance global information modeling, thereby improving the model’s ability to detect cross-regional defects. Moreover, the C3kDC_FFM module, composed of the DCbottleneck and the Feature Fusion Module (FFM), effectively enhances contextual representation and improves the integration of multi-scale features. Furthermore, a dilated convolution layer is incorporated into the detection head to form a new Dilated head, which enlarges the receptive field and enhances the model’s ability to recognize small-scale defects. On the Northeastern University Steel Surface Defect Detection (NEU-DET) dataset, RAGA-YOLO achieves a 2.1% increase in mAP, substantially reduces multiple Toolkit for Identifying Detection and segmentation Errors (TIDE) errors, and decreases the number of parameters by 15.3% while maintaining unchanged GFLOPs. Experiments on the Metallic Surface Defect Detection with 10 classes (GC10-DET) dataset and Anomaly Surface Visual Defect of (ASVD) washer dataset further demonstrate its excellent generalization capability and robustness to scale variations, highlighting its promising potential for practical industrial applications. In addition, experiments conducted on the PCB dataset further verify the superior performance of the Dilated head in recognizing small-scale defects.

Abstract The honey badger algorithm (HBA) has attracted the interest of researchers from various fields for solving optimization problems. However, it is prone to rapid loss of diversity and imbalanced exploration-exploitation behavior, leading to premature convergence in local optima and limiting final solution accuracy. To overcome these drawbacks, this paper proposes a novel dynamic mean-based HBA (DM-HBA) with two enhancements: (1) a mean-based guidance strategy that maintains population diversity and allows the algorithm to escape local optima. (2) A dynamic attraction parameter enables a smooth transition from exploration to exploitation. Finally, 41 benchmark functions from the CEC’17 and CEC’22 test suites, as well as three complex real-world engineering problems, are used. Statistical analyses revealed that DM-HBA outperformed HBA on 72% of the functions, with mean improvements of 21-25% in best-fitness values and approximately 26-34% reductions in standard deviation across 30D, 50D and 100D. For the more challenging CEC’22 suite, at 20D, it outperforms HBA on 75% of the functions, with a mean improvement of 8.7% and an approximately 31% standard deviation reduction, demonstrating that DM-HBA is both robust and scalable, with a well-balanced exploration-exploitation process. The source code is available at GitHub.

Abstract In an air-cooled box ground control station (ABGCS), electronic devices generate heat during operation. Insufficient heat dissipation can degrade performance or cause failures, so an efficient layout design is crucial. This paper proposes a new layout method for devices in the ABGCS. The objective function is established using the energy-grid method. An IPSO-DE (improved particle swarm optimization and differential evolution) algorithm is introduced to enhance search ability. Layouts generated by PSO, IPSO, and IPSO-DE were simulated using computational fluid dynamics (CFD). The results show that the heat dissipation efficiency is proportional to the objective function value. Finally, tests verified the IPSO-DE layout with an average relative error of 1.06%-2.84%. This proves its reliability and provides a reference for similar designs.

Abstract The logic mining model has gained significant attention with the advancement in processing complex and large-scale datasets. This advantage stems from the model’s capability to extract interpretable logical rules. However, many recent logic mining models fail to accurately retrieve optimal logic patterns from datasets because of their inflexible structures, rigid attribute selection, and fixed number of attributes. To address these issues, this paper proposes three innovative strategies through Flexible based G-Type Random 3-Satisfiability Reverse Analysis (FGRA) to enhance the performance of the induced logic in doing knowledge extraction and classification. First, FGRA utilizes G-Type Random 3-Satisfiability (GRAN3SAT) with the Discrete Hopfield Neural Network (DHNN) to diversify the clause of the logical rule. Later, the dataset is utilized in the DHNN training phase that incorporates GRAN3SAT. Second, FGRA adopts a cost function that maximizes classification accuracy and evaluates the effectiveness of the induced logic using multiple evaluation metrics in experiments. This procedure is vital to address the low performance of the potential GRAN3SAT logic before it enters the retrieval phase of the DHNN. Third, FGRA employs a multi-attribute selection ensemble approach, including log-linear selection, random selection, and correlation analysis, to identify key attributes for constructing the GRAN3SAT formulation. These strategies are crucial to ensure the model can adapt to the nature of the datasets based on varying sizes and complexity. The experimental results indicate that the performance of FGRA is promising compared to the other state-of-the-art logic mining models with an average accuracy of 0.873 in extracting the best-induced logic from numerous benchmark datasets. These findings have significant implications for bridging the gap between the logical rule and the data pattern, marking an important contribution to the field.

Abstract Effective airfoil geometry optimization requires exploring a diverse range of designs using as few design variables as possible. This study introduces AirDbM, a Design-by-Morphing (DbM) approach specialized for airfoil optimization that systematically reduces design-space dimensionality. AirDbM selects an optimal set of 12 baseline airfoils from the UIUC airfoil database, which contains over 1,600 shapes, by sequentially adding the baseline that most increases the design capacity. With these baselines, AirDbM reconstructs 99 % of the database with a mean absolute error below 0.005, which matches the performance of a previous DbM approach that used more baselines. In multi-objective aerodynamic optimization, AirDbM demonstrates rapid convergence and achieves a Pareto front with a greater hypervolume than that of the previous larger-baseline study, where new Pareto-optimal solutions are discovered with enhanced lift-to-drag ratios at moderate stall tolerances. Furthermore, AirDbM demonstrates outstanding adaptability for reinforcement learning (RL) agents in generating airfoil geometry when compared to conventional airfoil parameterization methods, implying the broader potential of DbM in machine learning-driven design.

Abstract Enhancing thermal performance in phase change materials (PCMs) is critical for advancing thermal energy storage systems. Passive strategies, such as optimizing geometry and using nanoparticles, offer promising ways to enhance heat transfer and energy efficiency. This study examines a flow of non-Newtonian Casson nanofluid synthesized by sodium sulfate decahydrate PCM, water, borax stabilizer, and aluminum oxide (Al₂O₃) nanoparticles subjected to an external magnetic field in an optimized octagonal cavity with plus-shaped fin. Octagonal cavity is heated from below; the remaining walls of the enclosure are thermally insulated. The governing equations are solved numerically using the Finite Element Method (FEM). Simulations explored the effects of the Casson parameter β, Rayleigh number Ra, and Hartmann number Ha Casson on flow structure, Nusselt number (Nu), and mass Sherwood number (Sh). Results show β and Ha have competing influences. Lower β enhanced convection, raising the mean Nusselt number by ∼55% versus large β, while high Ha suppressed flow and heat transfer. Ra was the dominant factor and increasing Ra shifted the system to convection-dominated regime, strengthening vortices and significantly improving thermal (Nu) and solute (Sh) transfer.

Abstract Fluid-Structure-Contact Interaction (FSCI) phenomena have been studied in recent years for only a limited range of applications, such as the interaction of air with multiple parachute canopies and blood flow through heart valves. Ink transfer in contact-based printing systems, such as flexographic printers, is a significant example of such phenomena and can be adequately understood only when investigated as an FSCI problem. This paper aims to pioneer the study of ink transfer between interacting rollers by proposing a methodology applicable to a wide range of FSCI scenarios. This methodology includes prescribing a small gap where the structures establish physical contact and applying an auxiliary fluid in the gap with a viscosity high enough to prohibit its movement, thereby practically replicating the real closed contact. A Lagrangian conformal mesh approach is employed to maximise accuracy while maintaining reasonable computational cost. Simulations were conducted on a scenario involving substantial deformation of a roller surface and its penetration into the micro-cavities of the other roller, where the ink initially resides. The results demonstrate different phases of ink transfer between surfaces: pre-contact, contact development, isolation, contact opening, and surface separation, during which ink breakup occurs. The pressure was observed to rise significantly during the penetration phase, reaching a peak value of approximately 1 × 10⁶ Pa. The ink transfer rate was calculated to be 27%, consistent with the relatively lower band of reported industrial ranges. These findings provide insights into understanding and controlling ink transfer between rollers, which can help maximise cell evacuation rates by altering the engraved texture shapes on the rollers. The proposed methodology is also applicable to other contact-based printing systems and broader applications involving fluid-structure-contact interactions.