Abstract Multi-threshold image segmentation (MTIS) is a crucial technology in image processing, characterized by simplicity and efficiency, and the key lies in the selection of thresholds. However, the method's time complexity will grow exponentially with the number of thresholds. To solve this problem, an improved arithmetic optimization algorithm (ETAOA) is proposed in this paper, an optimizer for optimizing the process of merging appropriate thresholds. Specifically, two optimization strategies are introduced to optimize the optimal threshold process: elite evolutionary strategy (EES) and elite tracking strategy (ETS). First, to verify the optimization performance of ETAOA, mechanism comparison experiments, scalability tests, and comparison experiments with nine state-of-the-art peers are executed based on the benchmark functions of CEC2014 and CEC2022. After that, to demonstrate the feasibility of ETAOA in the segmentation domain, comparison experiments were performed using 10 advanced segmentation methods based on skin cancer dermatoscopy image datasets under low and high thresholds, respectively. The above experimental results show that the proposed ETAOA performs outstanding optimization compared with benchmark functions. Moreover, the experimental results in the segmentation domain show that ETAOA has superior segmentation performance under low and high threshold conditions.

Abstract The operational optimization of the coal mine integrated energy system (CMIES) is crucial for reducing costs and carbon emissions. However, the system’s multi-objective nature, stringent constraints, and the uncertainty of renewable and mine-derived energy make solving its optimization challenging. Thus, this paper first presents a data-driven uncertainty transformation method to address the uncertainty of renewable energy and mining derived energy output; then, a multi-task multi-objective evolutionary algorithm based on adaptive auxiliary tasks (MMOEA-AS) is proposed, which includes a main task and three auxiliary tasks. Meanwhile, an adaptive update strategy for auxiliary tasks and a matching degree-guided knowledge transfer mechanism are proposed to improve the performance of the algorithm. Finally, taking the energy scheduling problem of a coal mine in Shanxi, China as an example, MMOEA-AS is compared with five advanced evolutionary algorithms. The results show that MMOEA-AS can effectively solve the operation optimization of the CMIES, and obtain the optimal scheduling results.

Abstract Space layout design is a fundamental yet complex problem in architecture. This task’s inherent complexity, arising from the need to balance numerous geometric configurations and topological relations while adhering to specific constraints, poses significant challenges. Recent advancements in deep reinforcement learning have shown promise in addressing similar planning problems, suggesting its potential utility for innovative space layout solutions. However, a critical limitation of deep reinforcement learning is its struggle with generalizing learned strategies to unseen scenarios. In the context of architectural design, this limitation could prevent deep reinforcement learning from being a scalable design method. Pretraining has emerged as a transformative strategy within the field of artificial intelligence, especially in the realm of foundational models, to enhance the generalization capabilities of learning algorithms. While pretraining is being the central focus of this paper, our approach diverges from conventional pretraining methods that focus on pixel-level design of layouts as is in diffusion based model. Instead, we propose an architectural simulation of space layout design that could embody the multifaceted essence of architectural design. To this end, we have developed a space layout simulator called SpaceLayoutGym that serves dual purposes: first, as an environment for the reinforcement learning agent to interact with and learn the intricacies of design, and second, as a tool for generating a dataset of design scenarios and their corresponding design solutions for model pretraining. We then used imitation learning to pretrain the agent on the generated training design scenarios. This process is being followed by a fine-tuning phase by using proximal policy optimization algorithm on new design scenarios. Our results demonstrate that pretraining can enhance the generalization capabilities of deep reinforcement learning in space layout design, paving the way for more adaptable and scalable artificial intelligence-aided architectural design.

Abstract With the increasing complexity of industrial manufacturing tasks, optimizing Multi-Point Manufacturing (MPF) operations, such as drilling and spot welding, has become essential for improving energy efficiency and execution speed. These tasks can be formulated as multi-dimensional Traveling Salesman Problems (TSPs), where robotic manipulators must traverse multiple points optimally while minimizing energy consumption. However, traditional heuristic algorithms, such as Genetic Algorithm (GA) and Ant Colony Optimization (ACO), face significant limitations. Their iterative nature leads to long computation times, and their performance heavily relies on precise parameter tuning to achieve high-quality solutions. To address these challenges, this study proposes a novel minimum energy consumption scheduling strategy for MPF problems, utilizing a Graph Neural Network (GNN)-based approach enhanced by Transfer Learning. The method employs Self-Adjusting Multi-modal Artificial Neural Networks (SAMANN) to reduce the 6D task nodes into a 2D representation, enabling the reuse of pre-trained 2D TSP models. Fine-tuning is then applied to further optimize performance. Compared with traditional heuristic algorithms, the proposed method significantly reduces computation time while maintaining solution quality. It is also found that the new approach surpasses the performance of the current state-of-the-art (SOTA) neural model in terms of both solution quality and training time across multi-dimensional TSPs and MPF problem domains, offering a robust and efficient solution for real-world MPF applications.

Abstract Traditionally, vulnerability assessments for naval ships have relied on empirical calculations to estimate the extent of damage without taking into account the structural capacity of the ship. This study introduces an enhanced damage-extent method that incorporates the structural load-bearing capacity of the hull to assess the vulnerability of naval ships to explosive loads. This vulnerability assessment predicts the area of damage to the hull structure and calculates the probability of onboard equipment experiencing functional losses due to the explosive load, thus allowing various design alternatives to be evaluated. The proposed methodology improves upon traditional damage-volume-based approaches, such as damage-radius and ellipsoid methods, by considering the hull’s structural stiffness and intrinsic damage resistance. It integrates the hull’s structural resistance to the load, enhancing the damage assessment process for both the hull and equipment. This approach facilitates damage prediction for different hull designs by comparing the allowable impulse with the explosive pressure. In assessing the functionality loss and vulnerability of the equipment within the damaged hull, the network of equipment functions is considered. An anti-ship cruise missile with a sea-skimming trajectory is investigated as the explosive charge, with procedures established to simulate its trajectory and impact location on the hull. Hundreds of potential internal and external explosion points are generated, predicting the explosive pressure at each location. The shock wave, including incident overpressure, reflection pressure, and quasi-static gas pressure, is converted into impulses, taking into account the configuration of hull compartments to accurately predict these pressures and equivalent impulse. The resulting impulse is compared with the intrinsic damage capacity of each compartment’s structure to assess potential damage. System network and fault tree analysis evaluate the loss of function and vulnerability of equipment within the damaged hull. Finally, the proposed capacity-based damage extent method demonstrates more accurate damage assessment compared to traditional methods, overcoming the limitations of damage-radius and ellipsoid approaches by considering hull strength.

Abstract With recent advancements in artificial intelligence technology, various studies are being conducted in the shipbuilding industry. Traditionally, hull form variation methods have relied on the intuition and expertise of designers, leading to inconsistent results and unintended changes in the ship’s main dimensions depending on the designer’s competence. Moreover, the iterative process of design variation and analysis to derive the optimal hull form is both costly and time-consuming. To address these issues, this study proposes an optimal hull design technique utilizing reinforcement learning, a type of unsupervised learning in machine learning. Reinforcement learning allows the model to learn from past policies by recording and accumulating the rewards associated with various actions taken by an agent in a specific environment. In this study, after calculating the main parameters of the ship, the agent defines a state representing hull information and performs local transformations of the bow and stern. The reward of reinforcement learning is defined as the change in total resistance due to the hull deformation, constrained by limiting the tolerance of the ship’s prismatic coefficient (CP) and longitudinal center of buoyancy. In this study, the problem is solved by comparing the proximal policy optimization algorithm and the deep deterministic policy gradient algorithm to find the best deep reinforcement learning model for the hull optimization problem. The results were compared with the genetic algorithm and speed-constrained multi-objective particle swarm optimization, and the optimal hull resistance values were less different, but the time of the reinforcement learning model was five times shorter.

Abstract The present study explores the dynamics of a two-dimensional, incompressible nanofluid flow through a stretching curved sheet within a highly porous medium. The mathematical model is formulated by including external forces such as viscous dissipation, thermal radiation, Ohmic heating, chemical reactions, and activation energy by utilizing a curvilinear coordinate system. The viscosity and thermal conductivity of the nanofluids are examined using the Koo–Kleinstreuer–Li model. The choice of $Al_{2}O_{3}$ and $CuO$ nanoparticles in this model stems from their distinct thermal properties and widespread industrial applicability. By non-dimensionalizing the governing partial differential equations, the physical model is simplified into ordinary differential equations. BVP-5C solver in MATLAB is utilized to numerically solve the obtained coupled non-linear ordinary differential equation. Graphical results are presented to investigate the velocity, temperature, and concentration profiles with entropy generation optimization under the influence of several flow parameters. The artificial neural network backpropagated with Levenberg–Marquardt method (ANN-BLMM) used to study the model. The performance is validated using regression analysis, mean square error and error histogram plots. The outcome illustrates that the velocity and temperature profiles increase with increasing the Forchhiemer parameter. Also, the velocity profile increases with increasing curvature parameter, while, reverse effect is observed for temperature profile. This research augments our comprehension of nanofluid dynamics over curved surfaces, which has implications for engineering applications. The insights gained have the potential to significantly contribute to the advancement of energy-efficient and environmentally sustainable cooling systems in industrial processes.

Abstract In optimization, metaheuristic algorithms have received extensive attention and research due to their excellent performance. The slime mould algorithm (SMA) is a newly proposed metaheuristic algorithm. It has the characteristics of fewer parameters and strong optimization ability. However, with the increasing difficulty of optimization problems, SMA has some shortcomings in complex problems. For example, the main concerns are low convergence accuracy and prematurely falling into local optimal solutions. To overcome these problems, this paper has developed a variant of SMA called CCSMA. It is an improved SMA based on horizontal crossover (HC) and covariance matrix adaptive evolutionary strategy (CMAES). First, HC can enhance the exploitation of the algorithm by crossing the information between different individuals to promote communication within the population. Finally, CMAES facilitates algorithm exploration and exploitation to reach a balanced state by dynamically adjusting the size of the search range. This benefits the algorithm by allowing it to go beyond the local space to explore other solutions with better quality. To verify the superiority of the proposed algorithm, we select some new original and improved algorithms as competitors. CCSMA is compared with these competitors in 40 benchmark functions of IEEE CEC2017 and CEC2020. The results demonstrate that our work outperforms the competitors in terms of optimization accuracy and jumping out of the local space. In addition, CCSMA is applied to tackle three typical engineering optimization problems. These three problems include multiple disk clutch brake design, pressure vessel design, and speed reducer design. The results showed that CCSMA achieved the lowest optimization cost. This also proves that it is an effective tool for solving realistic optimization problems.

Abstract Uncertain multi-modal transport path optimization (UMTPO) is a combined optimization non-deterministic polynomial-time hard problem. Its goal is to determine a path with the lowest total transportation cost and carbon emissions from the starting point to the destination. To effectively address this issue, this article proposes a modified dung beetle optimizer (DBO) to address it. DBO is a swarm-based metaheuristic optimization algorithm that has the features of a fast convergence rate and high solution accuracy. Despite this, the disadvantages of weak global exploration capability and falling easily into local optima exist. In this article, we propose a modified DBO called MSHDBO for function optimization and to solve the UMTPO problem. However, for the vast majority of metaheuristic algorithms, they are designed for continuous problems and cannot directly solve discrete problems. Therefore, this article employs a priority based encoding and decoding method to solve the UMTPO problem. To verify the performance and effectiveness of the MSHDBO algorithm, we compared it with other improved versions of the DBO algorithm used in the literature. We confirmed the excellent performance of MSHDBO using 41 benchmark test functions from the IEEE CEC 2017 test suite and IEEE CEC 2022 test suite. Additionally, we compared the MSHDBO algorithm with 10 other state-of-the-art metaheuristic optimization algorithms through a practical UMTPO problem. The experimental results indicated that the MSHDBO algorithm achieved very good performance when solving the UMTPO problem.

Abstract Recently, interest in functional safety has surged because vehicle technology increasingly relies on electronics and automation. Failure of certain system components can endanger driver safety and is costly to address. The detection of abnormal data is crucial for enhancing the reliability, safety, and efficiency. This study introduces a novel anomaly-detection method of designable generative adversarial network anomaly detection (DGANomaly). DGANomaly combines the data augmentation method of a designable generative adversarial network (DGAN) with a generative adversarial network anomaly-detection data classification technique. DGANomaly not only generates virtual data that are challenging to obtain or simulate but also produces a range of statistical design variables for normal and abnormal data. This approach enables the specific identification of normal and abnormal design variables. To demonstrate its effectiveness, the DGANomaly method was applied to an electric power steering (EPS) model when multiple degradations of gear stiffness, gear friction, and rack displacement were considered. An EPS model was constructed and validated using simulation programs such as Prescan, Amesim, and Simulink. Consequently, DGANomaly exhibited a higher classification accuracy than the other methods, allowing for more accurate detection of abnormal data. Additionally, a clearer range of statistical designs can be obtained for normal data. These results indicate that the statistical design variables that are less likely to fail can be obtained using minimal data.