Search

This study introduces the Self-adaptive Emperor Penguin Optimizer (SA-EPO), a new variant that addresses the exploration–exploitation balance limitations of the original EPO due to its statics control parameters. SA-EPO integrates multiple parameter adaptation strategies with unique features and selection probabilities, enabling dynamic modification of control parameters based on individual solution performance. An intelligent selection mechanism within SA-EPO’s framework periodically updates the selection probabilities of these parameter adaptation strategies based on their historical effectiveness in enhancing solution quality, ensuring the optimal strategy is consistently employed. SA-EPO's efficacy is validated against 15 leading optimization algorithms through tests on 41 benchmark functions from the CEC2017 and CEC2022. Furthermore, SA-EPO's capability are demonstrated on seven real-world engineering challenges. Comprehensive non-parametric statistical analyses, including Friedman test and Wilcoxon signed rank test, confirm the superior accuracy and convergence speed of SA-EPO across a range of optimization scenarios. The SA-EPO demonstrates substantial performance enhancements compared with the EPO, with improvements of 47.9 % and 52.4 % in Freidman rank for CEC2017 and CEC2022, respectively. Additionally, the Wilcoxon signed rank test reveals a 100 % improvement, indicating a complete advantage over the EPO in all tested scenarios. These findings highlight its potential to drive industrial and process innovation in diverse optimization tasks.

Feature selection is a pivotal preprocessing step in deploying machine learning solutions, aimed at removing redundant features from datasets while preserving predictive accuracy. Despite the popular use of wrapper-based feature selection techniques with metaheuristic search algorithms (MSAs) such as Teaching Learning Based Optimization (TLBO), balancing exploration and exploitation in identifying optimal feature subsets remains a fundamental challenge. This paper introduces an advanced wrapper-based feature selection method utilizing an enhanced TLBO variant, termed Modified TLBO with Hierarchical Learning Scheme (MTLBO-HLS). MTLBO-HLS incorporates significant enhancements in both teacher and learner phases to better balance exploration and exploitation, thus improving robustness in solving complex problems like feature selection. Specifically, the Multi-Hierarchical Teacher Phase (MH-TP) divides the MTLBO-HLS population into multiple group tiers based on fitness, where learners are guided by superior peers from better tiers. The Dynamic Peers Interaction Based Learner Phase (DPI-LP) enriches the learning process by enabling interactions with single and multiple elite peers and retaining valuable knowledge across dimensions. Extensive simulations demonstrate MTLBO-HLS’s superior performance in feature selection, consistently identifying the smallest feature subsets and achieving the highest classification accuracies across various datasets. These results highlight its potential to drive industrial innovation through enhanced machine learning model efficiency and accuracy.

Summary: Achieving lightweight real-time object detection necessitates balancing model compression with detection accuracy, a difficulty exacerbated by low redundancy and uneven contributions from convolutional layers. As an alternative to traditional methods, we propose Rigorous Gradation Pruning (RGP), which uses a desensitized first-order Taylor approximation to assess filter importance, enabling precise pruning of redundant kernels. This approach includes the iterative reassessment of layer significance to protect essential layers, ensuring effective detection performance. We applied RGP to YOLOv8 detectors and tested it on GTSDB, Seaships, and COCO datasets. On GTSDB, RGP achieved 80% compression of YOLOv8n with only a 0.11% drop in mAP0.5, while increasing frames per second (FPS) by 43.84%. For YOLOv8x, RGP achieved 90% compression, a 1.26% mAP0.5:0.95 increase, and a 112.66% FPS boost. Significant compression was also achieved on Seaships and COCO datasets, demonstrating RGP’s robustness across diverse object detection tasks and its potential for advancing efficient, high-speed detection models.

Abstract Under grid voltage sags, over current protection and exploiting the maximum capacity of the inverter are the two main goals of grid-connected PV inverters. To facilitate low-voltage ride-through (LVRT), it is imperative to ensure that inverter currents are sinusoidal and remain within permissible limits throughout the inverter operation. An improved LVRT control strategy for a two-stage three-phase grid-connected PV system is presented here to address these challenges. To provide over current limitation as well as to ensure maximum exploitation of the inverter capacity, a control strategy is proposed, and performance the strategy is evaluated based on the three generation scenarios on a 2-kW grid connected PV system. An active power curtailment (APC) loop is activated only in high power generation scenario to limit the current’s amplitude below the inverter’s rated current. The superior performance of the proposed strategy is established by comparison with two recent LVRT control strategies. The proposed method not only injects necessary active and reactive power but also minimizes overcurrent with increased exploitation of the inverter’s capacity under unbalanced grid voltage sag.

There is a growing need for a more streamlined and automated method of refining biofuels, as there are currently no universally applicable process inspection instruments on the market. All process variables in bio-oil upgrading operations are maintained according to the offline specifications of the products and intermediates. Failure of the process and loss of resources can result from batch-wise monitoring not having real-time product standards. Consequently, in order to cut down on waste and lessen the chances of process failure, a quick and accurate tool for specifying intermediates and products is required. To resolve this issue, we created a model using response surface methodology and an artificial neural network that can increase the bio-oil yield involving parameters, i.e., biomass particle size (mm), temperature (°C), and residence time (min). The maximum bio-oil production (47.0883%) was achieved at 3 mm particle size, 523°C temperature, and 20 min residence time. All results are “better” for root mean squared error (∼1), and the highest coefficient of regression for bio-oil production is in the range of 0.97–0.99. Temperature is the most significant factor in bio-oil yield, followed by particle size and residence time. Based on physicochemical properties, bio-oil has the maximum kinematic viscosity (11.3 Cst) and water content (18.7%). Making bio-oil precious compounds allows it to be used as boiler feedstock and steam generation fuel.

Conventional perturb and observe (P&O) algorithm fails to track global maximum power point (GMPP) under complex partial shading conditions (PSC) in photovoltaic (PV) system. While many of the latest maximum power point tracking (MPPT) algorithms are designed to handle simpler PSCs with fewer peaks, their capability to handle highly complex PSCs remains uncertain. This study presented more practical, challenging, and complex PSCs that have over five peaks and extremely close peak values. A new deterministic peak hopping (PH)-based MPPT algorithm with simple mechanisms is proposed to address these complex PSCs. An agent is utilized to scan and hop between the lower and higher duty cycle regions of P-V curve with optimum step size, thereby effectively narrowing down the tracking region, moving towards the GMPP. Additionally, the proposed algorithm utilizes an adjustable sampling time during scanning and hopping process to accelerate the convergence. Extensive simulation studies have revealed the effectiveness of the proposed algorithm in tracking GMPP. Moreover, the proposed algorithm outperforms five of the latest MPPT algorithms. In experimental setup, the proposed algorithm is successfully implemented into real-time TI C2000 microcontroller and performed robustly using ITECH IT6012C-800-40 PV simulator, achieving tracking time shorter than 0.83s and tracking accuracy over 98.70%.

The increasing complexity of modern mechanical systems, especially rotating machinery, demands effective condition monitoring techniques, particularly deep learning, to predict potential failures in a timely manner and enable preventative maintenance strategies. Health monitoring data analysis, a widely used approach, faces challenges due to data randomness and interpretation difficulties, highlighting the importance of robust data quality analysis for reliable monitoring. This paper presents a two-part approach to address these challenges. The first part focuses on comprehensive data preprocessing using only feature scaling and selection via random forest (RF) algorithm, streamlining the process by minimizing human intervention while managing data complexity. The second part introduces a Recurrent Expansion Network (RexNet) composed of multiple layers built on recursive expansion theories from multi-model deep learning. Unlike traditional Rex architectures, this unified framework allows fine tuning of RexNet hyperparameters, simplifying their application. By combining data quality analysis with RexNet, this methodology explores multi-model behaviors and deeper interactions between dependent (e.g., health and condition indicators) and independent variables (e.g., Remaining Useful Life (RUL)), offering richer insights than conventional methods. Both RF and RexNet undergo hyperparameter optimization using Bayesian methods under variability reduction (i.e., standard deviation) of residuals, allowing the algorithms to reach optimal solutions and enabling fair comparisons with state-of-the-art approaches. Applied to high-speed bearings using a large wind turbine dataset, this approach achieves a coefficient of determination of 0.9504, enhancing RUL prediction. This allows for more precise maintenance scheduling from imperfect predictions, reducing downtime and operational costs while improving system reliability under varying conditions.

Computed tomography (CT) is used in a wide range of medical imaging diagnoses. However, the reconstruction of CT images from raw projection data is inherently complex and is subject to artifacts and noise, which compromises image quality and accuracy. In order to address these challenges, deep learning developments have the potential to improve the reconstruction of computed tomography images. In this regard, our research aim is to determine the techniques that are used for 3D deep learning in CT reconstruction and to identify the training and validation datasets that are accessible. This research was performed on five databases. After a careful assessment of each record based on the objective and scope of the study, we selected 60 research articles for this review. This systematic literature review revealed that convolutional neural networks (CNNs), 3D convolutional neural networks (3D CNNs), and deep learning reconstruction (DLR) were the most suitable deep learning algorithms for CT reconstruction. Additionally, two major datasets appropriate for training and developing deep learning systems were identified: 2016 NIH-AAPM-Mayo and MSCT. These datasets are important resources for the creation and assessment of CT reconstruction models. According to the results, 3D deep learning may increase the effectiveness of CT image reconstruction, boost image quality, and lower radiation exposure. By using these deep learning approaches, CT image reconstruction may be made more precise and effective, improving patient outcomes, diagnostic accuracy, and healthcare system productivity.

The performance evaluation of a Photovoltaic (PV) system heavily relies on accurately estimating the parameters based on its current—voltage relationships. However, due to the PV model’s inherent complexity, obtaining these parameters with precision and efficiency is a challenging task. In this study, a new variant known as teaching learning-based optimization with unique exemplar generation schemes (TLBO-UEGS) is proposed to address PV module parameter estimation problems with robustness and effectiveness. To enhance the performance of TLBO-UEGS, a modified initialization scheme that leverages the strengths of chaotic maps and dynamic oppositional based learning is introduced. This scheme ensures the generation of an initial population with improved solution quality. Furthermore, both the modified teacher phase and modified learner phase are integrated within the TLBO-UEGS optimization framework. This integration allows for different learning strategies to be employed based on the fitness values of each learner, effectively updating their search trajectories. Within the modified teacher phase, two unique exemplar generation schemes are designed to facilitate more effective guidance for learners in the first half of the population while maintaining population diversity. Meanwhile, the modified learner phase emulates a realistic knowledge acquisition process by enabling learners in the second half of the population to engage in collaborative learning with multiple peer learners or retain valuable knowledge from previous learning processes. Extensive simulations demonstrate that TLBO-UEGS achieves superior results, with the minimum root mean square error (RMSE) values of 3.5644 × 10−04 ± 0.0014, 1.3237 × 10−04 ± 0.0043, and 6.6016 × 10−06 ± 0.00011 obtained for Photowatt-PWP201, Leibold Solar (LSM 20), and Leybold Solar (STE 4/100) PV modules, respectively.

Abstract One of the greatest challenges for widespread utilization of solar energy is the low conversion efficiency, motivating the needs of developing more innovative approaches to improve the design of solar energy conversion equipment. Solar cell is the fundamental component of a photovoltaic (PV) system. Solar cell’s precise modelling and estimation of its parameters are of paramount importance for the simulation, design, and control of PV system to achieve optimal performances. It is nontrivial to estimate the unknown parameters of solar cell due to the nonlinearity and multimodality of search space. Conventional optimization methods tend to suffer from numerous drawbacks such as a tendency to be trapped in some local optima when solving this challenging problem. This paper aims to investigate the performance of eight state-of-the-art metaheuristic algorithms (MAs) to solve the solar cell parameter estimation problem on four case studies constituting of four different types of PV systems: R.T.C. France solar cell, LSM20 PV module, Solarex MSX-60 PV module, and SS2018P PV module. These four cell/modules are built using different technologies. The simulation results clearly indicate that the Coot-Bird Optimization technique obtains the minimum RMSE values of 1.0264E-05 and 1.8694E−03 for the R.T.C. France solar cell and the LSM20 PV module, respectively, while the wild horse optimizer outperforms in the case of the Solarex MSX-60 and SS2018 PV modules and gives the lowest value of RMSE as 2.6961E−03 and 4.7571E−05, respectively. Furthermore, the performances of all eight selected MAs are assessed by employing two non-parametric tests known as Friedman ranking and Wilcoxon rank-sum test. A full description is also provided, enabling the readers to understand the capability of each selected MA in improving the solar cell modelling that can enhance its energy conversion efficiency. Referring to the results obtained, some thoughts and suggestions for further improvements are provided in the conclusion section.

The imbalance between parking availability and demand has led to a rise in traffic challenges in many cities. The adoption of technologies like the Internet of Things and deep learning algorithms has been extensively explored to build automated smart parking systems in urban environments. Non-human-mediated, scalable smart parking systems that are built on decentralized blockchain systems will further enhance transparency and trust in this domain. The presented work, PARKTag, is an integration of a blockchain-based system and computer vision models to detect on-field free parking slots, efficiently navigate vehicles to those slots, and automate the computation of parking fees. This innovative approach aims to enhance the efficiency, scalability, and convenience of parking management by leveraging and integrating advanced technologies for real-time slot detection, navigation, and secure, transparent fee calculation with blockchain smart contracts. PARKTag was evaluated through implementation and emulation in selected areas of the MIT Art Design Technology University campus, with a customized built-in dataset of over 2000 images collected on-field in different conditions. The fine-tuned parking slot detection model leverages pre-trained algorithms and achieves significant performance metrics with a validation accuracy of 92.9% in free slot detection. With the Solidity smart contract deployed on the Ethereum test network, PARKTag achieved a significant throughput of 10 user requests per second in peak traffic hours. PARKTag is implemented as a mobile application and deployed in the mobile application store. Its beta version has undergone user validation for feedback and acceptance, marking a significant step toward the development of the final product.

Titanium alloys has significance in engineering applications owing to their enhanced properties and its ability to retain the shape at elevated temperatures. A new Teaching Learning Based Optimization (TLBO) variant was developed with multiple search features to mitigate cutting force and surface irregularities in Titanium samples. This assists to achieve the best quality of the product at minimal cutting energy. Experiments were conducted and the significance of machining parameters on cutting force and surface finish were analyzed. It is ascertained that the best surface is attained at a lower tool feed rate with higher cutting speed. The increase of nose radius has more influence on the surface quality. Chaotic multiobjective TLBO with multiple effective guidance was applied in both single objective and multiobjective optimization, where useful information of other non-fittest learners is leveraged for effective more population search. The performance of the new algorithm was evaluated and comprehensively discussed. The minimum cutting force Fz=65.06N and Ra=1.41μm can be achieved with v =130m/min, f=0.051mm/rev, nr=0.4mm and ap=0.5mm. The predicted results were validated experimentally and verified with other existing optimizers. It is concluded that this new algorithm can be applied in machining and production wastage can be greatly minimized.

Introduction: In pattern recognition and data mining, feature selection is one of the most crucial tasks. To increase the efficacy of classification algorithms, it is necessary to identify the most relevant subset of features in a given domain. This means that the feature selection challenge can be seen as an optimization problem, and thus meta-heuristic techniques can be utilized to find a solution. Methodology: In this work, we propose a novel hybrid binary meta-heuristic algorithm to solve the feature selection problem by combining two algorithms: Dipper Throated Optimization (DTO) and Sine Cosine (SC) algorithm. The new algorithm is referred to as bSCWDTO. We employed the sine cosine algorithm to improve the exploration process and ensure the optimization algorithm converges quickly and accurately. Thirty datasets from the University of California Irvine (UCI) machine learning repository are used to evaluate the robustness and stability of the proposed bSCWDTO algorithm. In addition, the K-Nearest Neighbor (KNN) classifier is used to measure the selected features’ effectiveness in classification problems. Results: The achieved results demonstrate the algorithm’s superiority over ten state-of-the-art optimization methods, including the original DTO and SC, Particle Swarm Optimization (PSO), Whale Optimization Algorithm (WOA), Grey Wolf Optimization (GWO), Multiverse Optimization (MVO), Satin Bowerbird Optimizer (SBO), Genetic Algorithm (GA), the hybrid of GWO and GA, and Firefly Algorithm (FA). Moreover, Wilcoxon’s rank-sum test was performed at the 0.05 significance level to study the statistical difference between the proposed method and the alternative feature selection methods. Conclusion: These results emphasized the proposed feature selection method’s significance, superiority, and statistical difference.

Differential evolution (DE) is a popular optimization algorithm with easy implementation and fast convergence rate. For evolutionary algorithms such as DE, the initialization process of solution members is crucial because the distribution of initial population can govern the overall quality of final solution obtained in terms of accuracy and convergence speed. This study leverages the strengths of both chaotic maps and oppositional-based learning strategy to design a new DE variant with modified initialization scheme, namely chaotic oppositional DE (CODE) in order to generate the initial population with good quality of mean fitness and diversity of the solutions. The effectiveness of CODE variants incorporated with seven different chaotic maps are investigated using CEC 2014 benchmark functions and the chaotic circle oppositional DE (CCODE) is revealed as the best performing CODE variants. The optimization performance of CCODE is further compared with other existing optimization algorithms in terms of solution accuracy and convergence speed. Extensive simulation studies prove that the proposed algorithm is able to outperform its peers by achieving better trade-off between two contradicting requirements of fast convergence speed and population diversity preservation.

Differential evolution (DE) is a popular evolutionary algorithm inspired by Darwin’s theory of evolution and has been studied extensively to solve different areas of optimisation and engineering applications since its introduction by Storn in 1997. This study aims to review the massive progress of DE in the research community by analysing the 192 articles published on this subject from 1997 to 2021, particularly studies in the past five years. The methodology used to search for relevant DE papers and an overview of the original DE are firstly explained. Recent advances in the modifications proposed to enhance the effectiveness and efficiency of the original DE are reviewed by analysing the strengths and weaknesses of each published work, followed by the potential applications of these DE variants in solving different real-world engineering problems. In contrast to most existing DE review papers, additional analyses are performed in this survey by investigating the impacts of various parameter settings on given DE variants to identify their optimal values required for solving certain problem classes. The qualities of modifications incorporated into selected DE variants are also evaluated by measuring the performance gains achieved in terms of search accuracy and/or efficiency against the original DE. The additional surveys conducted in this study are anticipated to provide more insightful perspectives for both beginners and experts of DE research, enabling their better understanding about current research trends and new motivations to outline appropriate strategic planning for future development works.

Effective fault detection and classification play essential roles in reducing the hazards such as electric shocks and fire in photovoltaic (PV) systems. However, the issues of interest in fault detection and classification for PV systems remain an open-ended challenge due to manual and time-consuming processes that require the relevant domain knowledge and experience of fault diagnoses. This paper proposes a hybrid deep-learning (DL) model-based combined architectures as the novel DL approaches to achieve the real-time automatic fault detection and classification of a PV system. This research employed the wavelet packet transform (WPT) as a data preprocessing technique to handle the PV voltage signal collected and feeding them as the inputs for combined DL architectures that consist of the equilibrium optimizer algorithm (EOA) and long short-term memory (LSTM-SAE) approaches. The combined DL architectures are able to extract the fault features automatically from the preprocessed data without requiring any previous knowledge, therefore can override the traditional shortages of manual feature extraction and manual selection of optimal features from the extracted fault features. These desirable features are anticipated to speed up the fault detection and classification capability of the proposed DL model with higher accuracy. In order to determine the performance of the proposed fault model, we carried out a comprehensive evaluation study on a 250-kW grid-connected PV system. In this paper, symmetrical and asymmetrical faults have been studied involving all the phases and ground faults such as single phase to ground, phases to phase, phase to phase to ground, and three-phase to ground. The simulation results validate the efficacy of the proposed model in terms of computation time, accuracy of fault detection, and noise robustness. Comprehensive comparisons between the simulation results and previous studies demonstrate the multidisciplinary applications of the present study.

Teaching learning-based optimization (TLBO) is a popular algorithm used to solve various optimization problems. Nevertheless, conventional TLBO and some improved variants tends to suffer with premature convergence due to rapid loss of population diversity, especially when handling the challenging optimization problems. Furthermore, it is not practical to tackle real-world multiobjective problems using prior approach given the frequent changes of customers’ requirements. Motivated by these challenges, an improved variant known as Modified Multi-objective Teaching Learning Based Optimization-Refined Learning Scheme (MMTLBO-RLS) was proposed as a posterior approach to solve challenging multiobjective optimization problems, including the prediction of optimum turning parameters to machine Polyether ether ketone material (PEEK). Substantial modifications were introduced for teacher and learner phases of MMTLBO-RLS to achieve better balancing of exploration and exploitation searches without incurring excessive computational cost. For modified teacher phase of MMTLBO-RLS, each learner was guided by a unique teacher solution and unique mean position to perform searching with better diversity. Meanwhile, two new learning strategies are incorporated into the modified learner phase of MMTLBO-RLS, enabling all learners to enhance their knowledge more efficiently based on their learning preferences. A systematic approach was followed to develop modelling equations required for optimization. The developed algorithm was then employed in single objective optimization as well as multiobjective optimization to cater its performances in any real-world environment. The prediction model reports that surface roughness of $1.1042~\mu m$ and material removal rate of 22.8991 cm3/minute can be achieved. The predicted results differ from validation results by less than 2.69% in any case of optimization. A benchmarking on the performance of MMTLBO-RLS in solving CEC 2009 multiobjective benchmark functions was further carried out with other seven meta-heuristic algorithms. The superior performance of MMTLBO-RLS proves that it is not only suitable to be used in industries to produce the parts of PEEK with supportive quality and quantity, but it is also able to solve other multiobjective optimization problems with competitive performances.

The deep learning diagnosis of aircraft engine-bearing faults enables cost-effective predictive maintenance while playing an important role in increasing the safety, reliability, and efficiency of aircraft operations. Because of highly dynamic and harsh operating conditions of this system, such modeling is challenging due to data complexity and drift, making it difficult to reveal failure patterns. As a result, the objective of this study is dual. To begin, a highly structured data preprocessing strategy ranging from extraction, denoising, outlier removal, scaling, and balancing is provided to solve data complexity that resides specifically in outliers, noise, and data imbalance problems. Gap statistics under k-means clustering are used to evaluate preprocessing results, providing a quantitative estimate of the ideal number of clusters and thereby enhancing data representations. This is the first time, to the best of authors’ knowledge, that such a criterion has been employed for an important step in a preliminary ground truth validation in supervised learning. Furthermore, to tackle data drift issues, long-short term memory (LSTM) adaptive learning features are used and subjected to a learning parameter improvement method utilizing recursive weights initialization (RWI) across several rounds. The strength of such methodology can be seen by application to realistic, extremely new, complex, and dynamic data collected from a real test-bench. Cross validation of a single LSTM layer model with only 10 neurons shows its ability to enhance classification performance by 7.7508% over state-of-the-art results, obtaining a classification accuracy of 92.03 ± 0.0849%, which is an exceptional performance in such a benchmark.

This study introduces ETLBOCBL-CNN, an automated approach for optimizing convolutional neural network (CNN) architectures to address classification tasks of varying complexities. ETLBOCBL-CNN employs an effective encoding scheme to optimize network and learning hyperparameters, enabling the discovery of innovative CNN structures. To enhance the search process, it incorporates a competency-based learning concept inspired by mixed-ability classrooms during the teacher phase. This categorizes learners into competency-based groups, guiding each learner’s search process by utilizing the knowledge of the predominant peers, the teacher solution, and the population mean. This approach fosters diversity within the population and promotes the discovery of innovative network architectures. During the learner phase, ETLBOCBL-CNN integrates a stochastic peer interaction scheme that encourages collaborative learning among learners, enhancing the optimization of CNN architectures. To preserve valuable network information and promote long-term population quality improvement, ETLBOCBL-CNN introduces a tri-criterion selection scheme that considers fitness, diversity, and learners’ improvement rates. The performance of ETLBOCBL-CNN is evaluated on nine different image datasets and compared to state-of-the-art methods. Notably, ELTLBOCBL-CNN achieves outstanding accuracies on various datasets, including MNIST (99.72%), MNIST-RD (96.67%), MNIST-RB (98.28%), MNIST-BI (97.22%), MNST-RD + BI (83.45%), Rectangles (99.99%), Rectangles-I (97.41%), Convex (98.35%), and MNIST-Fashion (93.70%). These results highlight the remarkable classification accuracy of ETLBOCBL-CNN, underscoring its potential for advancing smart device infrastructure development.

In powder metallurgy, the sintering process is a high-power consuming and critical process for better mechanical properties of composites due to proper diffusion of atoms. In this context, different sintering processes were investigated along with their sintering condition. The present work focused on optimizing conventional sintering process parameters for carbon nanotubes (CNTs) reinforced aluminum composites using Taguchi optimization methods. The Taguchi L9 orthogonal array (OA) experiment was considered for the investigation. CNT’s wt.%, sintering temperature, and time were chosen as process parameters in the sintering process, while macro-hardness and relative density were evaluated as performance evaluation characteristics. The signal-to-noise ratio (S/N) and ANOVA statistical procedures were utilized to evaluate the effect of sintering parameters/levels on the micro-hardness and relative density of the Al/CNTs composite sintered. ANOVA statistical analyses revealed that the CNTs wt.% significantly influences relative density (83.58%), followed by temperature (14.58%), whereas CNTs wt.% significantly influenced micro-hardness (77.75%), followed by temperature (13.64%). The sintering of Al/CNTs composites using these optimum conditions is recommended to reduce power consumption and enhance the quality of the sintered composite.

Convolutional neural networks (CNNs) have excelled in artificial intelligence, particularly in image-related tasks such as classification and object recognition. However, manually designing CNN architectures demands significant domain expertise and involves time-consuming trial-and-error processes, along with substantial computational resources. To overcome this challenge, an automated network design method known as Modified Teaching-Learning-Based Optimization with Refined Knowledge Sharing (MTLBORKS-CNN) is introduced. It autonomously searches for optimal CNN architectures, achieving high classification performance on specific datasets without human intervention. MTLBORKS-CNN incorporates four key features. It employs an effective encoding scheme for various network hyperparameters, facilitating the search for innovative and valid network architectures. During the modified teacher phase, it leverages a social learning concept to calculate unique exemplars that effectively guide learners while preserving diversity. In the modified learner phase, self-learning and adaptive peer learning are incorporated to enhance knowledge acquisition of learners during CNN architecture optimization. Finally, MTLBORKS-CNN employs a dual-criterion selection scheme, considering both fitness and diversity, to determine the survival of learners in subsequent generations. MTLBORKS-CNN is rigorously evaluated across nine image datasets and compared with state-of-the-art methods. The results consistently demonstrate MTLBORKS-CNN’s superiority in terms of classification accuracy and network complexity, suggesting its potential for infrastructural development of smart devices.

The rapid development of information and wireless communication technologies together with the large increase in the number of end-users have made the radio spectrum more crowded than ever. Besides, providing a stable and reliable service is challenging, as electromagnetic environments are evolving and becoming more sophisticated. Accordingly, there is an urgent need for more reliable and intelligent communication systems that can improve the spectrum efficiency and the quality of service to provide agile management of network resources, so as to better meet the needs of future wireless users. Specifically, Automatic Modulation Recognition (AMR) plays an essential role in most intelligent communication systems especially with the emergence of Software Defined Radio (SDR). AMR is an indispensable task while performing spectrum sensing in Cognitive Radio (CR). Thanks to the significant advancements in Deep Learning (DL) applications, new and powerful tools have been provided which can tackle problems in this space. Thus, today, integrating DL models into AMR has gained the attention of many researchers. This work aims to provide a comprehensive state-of-the-art review of the most recent Machine Learning (ML) based AMR methods for Single-Input Single-Output (SISO) and Multiple-Input Multiple-Output (MIMO) systems. Furthermore, the architecture of each model will be identified along with a detailed comparison in terms of specifications and performance. Finally, an outline of the open problems, challenges, and potential research directions is provided along with discussion and conclusion.

The electrical generator is the key part of the electrical generation system for the oil and gas industry, and it is easy to fail, which disturbs the availability and reliability of the electrical generation in the power industry. Therefore, extracting and diagnosing the fault features from the process signals are useful to diagnose the status of the machine. Though, a common challenge in many applied applications is the practical knowledge about the risk of failure or historical records, which is totally unlabeled and difficult to be identified by traditional fault approaches. Hence, in the present study, a novel deep learning (DL) framework is proposed to fill the gap by balancing the three stages of fault feature extraction, fault detection, and parameter optimization based on the long short term memory- recurrent neural networks (RNN- LSTM), stacked autoencoders (SAE), and particle swarm optimization (PSO) techniques. The suggested framework focuses on failure detection through a sequence of numerous features for the unlabeled historical data and unknown anomaly. To validate the effectiveness of the proposed DL framework, an experiment for failure detection of the electrical generator was conducted for the data of risky environment at Yemen oil and gas plant. The experimental results compared with the earlier studies validate that, the DL framework can address the faults for vibration signals of the electrical generator in a well- diagnosis performance effectively.

PSO is a simple and yet powerful metaheuristic search algorithm widely used to solve various optimization problems. Nevertheless, conventional PSO tends to lose its population diversity drastically and suffer with compromised performance when encountering the optimization problems with complex fitness landscapes. Extensive studies suggest the needs of preserving high population diversity for PSO to escape from the local optima in order to solve complex optimization problems effectively. Inspired by these ideas, a hovering swarm PSO (HSPSO) is proposed in this paper, where a computationally efficient diversity preservation scheme is first introduced to divide the population of HSPSO into a main swarm and a hovering swarm. An exemplar construction scheme is subsequently proposed in the main swarm of HSPSO to generate a universal exemplar by considering the promising directional information contributed by the other non-fittest particles. The proposed universal exemplar is envisioned to suppress the negative impacts of global best particle, while remain effective to guide all particles of main swarm converging towards the promising solution regions. While hovering around the main swarm, an intelligent scheme is introduced to dynamically adjust inertia weights of all hovering swarm members to achieve proper balancing of exploration and exploitation searches at swarm levels. Extensive performance analyses are conducted by using the proposed HSPSO to solve 30 benchmark functions of CEC 2014 and five real-world engineering applications. Simulation results reveal that the HSPSO is able outperform the state-of-art optimizers when solving most tested functions due to its excellent diversity preservation capability.

Hybrid aluminum metal matrix composites have prominence than conventional metals. Aluminum is used as base metal in many metal matrix composites, as it is widely used in industries and researches. The low density that is obtained after alloying is the major advantages of hybrid composite. The present study involves synthesis of Al6061/B4C/Mica hybrid metal matrix composite using stir casting method by mica particles as 3, 4, 5, 6wt.%, while maintaining 10wt.% of boron carbide. The best mechanical properties are reported by inclusion of 3wt.% of mica, while further addition of mica more than 3wt.% pulls down the mechanical properties and hence not sensible. The early-stage of strain hardening, and high tensile strength-yield strength ratio is observed, which concludes the fabricated hybrid aluminum composite is well suitable for automotive-related sheet forming operations.

Automatic modulation recognition (AMR) has become increasingly important in the field of signal processing, especially with the advancements of intelligent communication systems. Deep Learning (DL) technologies have been incorporated into the AMR field and they have shown outstanding performances against conventional AMR methods. The robustness of DL-based AMR methods under varying noise regimes is one of major concerns for the widespread utilization of this technology. Furthermore, most existing works have neglected the contributions of hand-crafted features (HCFs) in boosting the classification performances of DL-based AMR methods. In order to address the aforementioned technical challenges, a novel and robust DL-AMR method is proposed by leveraging the benefits of both contextual features (CFs) and HCFs for a specific range of signal-to-noise ratio (SNR). A novel feature selection algorithm is also proposed to search for the optimal sets of HCFs in order to reduce the dimensions of feature vectors without losing any important and relevant features. Simulation studies are performed to investigate the feasibility of proposed method in classifying 11 types of modulation schemes. Extensive performance analyses revealed the superiority of proposed method over baseline method in terms of the classification performance as well as the excellent capability of proposed feature selection algorithm in determining an optimal subset of HCFs.

Swept Friction Stir Spot Welding is provoked and fascinated for its captive performance in building space frame structures, ship buildings and panels of automotive vehicles. The aim of the current research is to avoid the wastage in welding of 6061-T6 aluminium alloy and to improve the sustainabilty of the weld process. Al2O3 nanoparticles were added into the guide hole for improving the weld characteristics. The percentage of reinforcement was varied by changing the guide hole diameter as 1.5, 2.0, 2.5 and 3.0 mm. An improved NSTLBO algorithm was proposed and applied. The proposed algorithm predicted the optimum welding parameters as D = 2.8 mm, N = 1387 rpm and F = 17 mm/minute with the predicted wear, lap shear load and microhardness of 197 microns, 6134 N and 94.6 HV, respectively. The validation experiments were further done and the error within 0.5% was found between the actual and predicted results. The results of proposed algorithm was evaluated with other popular meta-heuristic algorithms for the effectiveness. The microscopic examinations on weld zones and fracturerd samples were also presented. Industries can use these best process parameters predicted by the proposed algorithm for earning the best weld characteristics from eco-friendly welding.

Wire cut EDM is a quite regularly used machining process in mechanical and electronic industries. This research has attempted to machine aluminum alloy for which experimental design was prepared using Box-Behnken design. Different combinational options of pulse-on time (P1), pulse-off time (P2), servo wire feed (WF), and current (I) were investigated and surface roughness after machining was observed. Collected 27 datasets were further used in Adaptive Neuro Fuzzy Inference System (ANFIS) to produce about 500 datasets. These 500 datasets are approximated data derived from experimental datasets, known as synthetic data. Data model was further developed and used in Gorilla Troops Optimizer (GTO) to locate the optimum machining parameters. With the excellent three search operators: move towards other gorillas, migrate towards unknown places, and migrate towards known places, GTO has produced the lowest surface roughness value of 0.500953 μm when the machining parameters of pulse-on time, pulse-off time, wire feed, and current values were set as 121 μs, 52 μs, 3 m/min, and 166A, respectively. To ensure the accuracy of the synthetic data-based model and optimality, verification and validation were conducted. Wilcoxon signed rank test was conducted for the pairwise comparison of GTO with each of its competing algorithms at the significance level of σ = 0.05. Friedman test was conducted to calculate the average ranking of each algorithm and to detect the global differences between all compared algorithms. Outperforming performance by GTO algorithm in machining of the selected material is found.

Despite of its simplicity, the conventional learning strategy of canonical particle swarm optimization (PSO) is inefficient to handle complex optimization problems due to its tendency of overemphasizing the fitness information of global best position without considering the diversity information of swarm. In this article, a modified particle swarm optimization with effective guides (MPSOEG) is proposed, aiming to improve the algorithm's search performances in handling the optimization problems with different characteristics. Depending on the search performance of algorithm, two types of exemplars can be generated by an optimal guide creation (OGC) module incorporated into MPSOEG by referring to the particles with valuable directional information. Particularly, a global exemplar is generated by OCG module to guide the swarm converging towards the promising solution regions of search space, whereas a unique local exemplar can be customized for each particle to enable it escaping from local or non-optimal solution regions. In contrary to global best particle, the exemplars generated by OGC module are able to guide all MPSOEG particles more effectively by considering both fitness and diversity information of swarm, hence can achieve better balancing of algorithm's exploration and exploitation searches. Another notable contribution of MPSOEG is the simplicity of its learning framework through the elimination of both inertia weight and acceleration coefficients parameters. Comprehensive simulation studies are conducted with 25 benchmark functions and the proposed MPSOEG is reported to outperform its six peer algorithms in terms of search accuracy, search reliability and search efficiency in most tested problems.

One of the relevant factors in smart energy management is the ability to predict the consumption of energy in smart households and use the resulting data for planning and operating energy generation. For the utility to save money on energy generation, it must be able to forecast electrical demands and schedule generation resources to meet the demand. In this paper, we propose an optimized deep network model for predicting future consumption of energy in smart households based on the Dipper Throated Optimization (DTO) algorithm and Long Short-Term Memory (LSTM). The proposed deep network consists of three parts, the first part contains a single layer of bidirectional LSTM, the second part contains a set of stacked unidirectional LSTM, and the third part contains a single layer of fully connected neurons. The design of the proposed deep network targets represents the temporal dependencies of energy consumption for boosting prediction accuracy. The parameters of the proposed deep network are optimized using the DTO algorithm. The proposed model is validated using the publicly available UCI household energy dataset. In comparison to the other competing machine learning models, such as Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Multi-Layer Perceptron (MLP), Sequence-to-Sequence (Seq2Seq), and standard LSTM, the performance of the proposed model shows promising effectiveness and superiority when evaluated using eight evaluation criteria including Root Mean Square Error (RMSE) and R2. Experimental results show that the proposed optimized deep model achieved an RMSE of (0.0047) and R2 of (0.998), which outperform those values achieved by the other models. In addition, a sensitivity analysis is performed to study the stability and significance of the proposed approach. The recorded results confirm the effectiveness, superiority, and stability of the proposed approach in predicting the future consumption of energy in smart households.

Human skin diseases have become increasingly prevalent in recent decades, with millions of individuals in developed countries experiencing monkeypox. Such conditions often carry less obvious but no less devastating risks, including increased vulnerability to monkeypox, cancer, and low self-esteem. Due to the low visual resolution of monkeypox disease images, medical specialists with high-level tools are typically required for a proper diagnosis. The manual diagnosis of monkeypox disease is subjective, time-consuming, and labor-intensive. Therefore, it is necessary to create a computer-aided approach for the automated diagnosis of monkeypox disease. Most research articles on monkeypox disease relied on convolutional neural networks (CNNs) and using classical loss functions, allowing them to pick up discriminative elements in monkeypox images. To enhance this, a novel framework using Al-Biruni Earth radius (BER) optimization-based stochastic fractal search (BERSFS) is proposed to fine-tune the deep CNN layers for classifying monkeypox disease from images. As a first step in the proposed approach, we use deep CNN-based models to learn the embedding of input images in Euclidean space. In the second step, we use an optimized classification model based on the triplet loss function to calculate the distance between pairs of images in Euclidean space and learn features that may be used to distinguish between different cases, including monkeypox cases. The proposed approach uses images of human skin diseases obtained from an African hospital. The experimental results of the study demonstrate the proposed framework’s efficacy, as it outperforms numerous examples of prior research on skin disease problems. On the other hand, statistical experiments with Wilcoxon and analysis of variance (ANOVA) tests are conducted to evaluate the proposed approach in terms of effectiveness and stability. The recorded results confirm the superiority of the proposed method when compared with other optimization algorithms and machine learning models.

The research was attempted to mimic the locomotion of the salamander, which is found to be one of the main animals from an evolutionary point of view. The design of the limb and body was started with the parametric studies of pneumatic network (Pneu-Net). Pneu-Net is a pneumatically operated soft actuator that bends when compressed fluid is passed inside the chamber. Finite Element Analysis software, ANSYS, was used to evaluate the height of the chamber, number of chambers and the gap between chambers for both limb and body of the soft mechanism. The parameters were decided based on the force generated by the soft actuators. The assembly of the salamander robot was then exported to MATLAB for simulating the locomotion of the robot in a physical environment. Sine-based controller was used to simulate the robot model and the fastest locomotion of the salamander robot was identified at 1 Hz frequency, 0.3 second of signal delay for limb actuator and negative π phase difference for every contralateral side of the limbs. Shin-Etsu KE-1603, a hyper elastic material, was used to build the salamander robot and a series of experiments were conducted to record the bending angle, the respective generated force in soft actuators and the gait speed of the robot. The developed salamander robot was able to walk at 0.06774 m/s, following an almost identical pattern to the simulation.

This paper deals with the optimization of machining parameters of speed, feed rate, and depth of cut that aim to simultaneously achieve the low surface roughness (SR) and high material removal rate (MRR) of a version of ACETAL homopolymer material known as Delrin. First, an L27 orthogonal array with three-level of cutting speed (Vc), feed rate (f ), and depth of cut (ap) is formulated, and the experiments are conducted accordingly in a CNC turning machine using cemented carbide tool with insert angle of 80°. A response surface model is rendered from these experimental results, and two objective functions representing the SR and MRR of Delrin are derived. An enhanced multi-objective teaching-learning-based optimization (EMOTLBO) is then proposed to solve the multi-objective machining problem, aiming to minimize the SR and maximize the MRR of Delrin simultaneously. A fuzzy decision maker is also integrated to softly select the preferred solution from Pareto-front based on the importance level of both objective functions. Extensive simulation studies prove that EMOTLBO is more competitive than other existing algorithms for being able to produce a more uniformly distributed Pareto-front. Simulation results are further validated Experimentally, and the difference of lower than 5% is observed that imply to good agreement between the simulation and experimental results.

Most existing particle swarm optimization (PSO) variants use a single learning strategy and a fixed neighborhood structure for all particles during the search process. The adoption of rigid learning pattern and communication topology may restrict the intelligence level of each particle, hence degrading the performance of PSO in solving the optimization problems with complicated fitness landscapes. Recent studies suggested that the employment of self-adaptive mechanism in adjusting the search strategy and topology connectivity of each particle along the search process may serve as a potential remedy to improve the performance of PSO, especially when dealing with complex problems. For this reason, a self-adaptive topologically connected (SATC)-based PSO equipped with an SATC module and an improved learning framework is proposed. The SATC module is envisioned to facilitate each particle to perform searching with different exploration and exploitation strengths by adaptively modifying its topology connectivity in different searching stages. A modified velocity update scheme and an alternative search operator are also introduced to formulate an improved learning framework to enhance the performance of proposed work further. Substantial numbers of benchmark functions and two real-world optimization problems are used to compare SATC-based PSO (SATCPSO) with several well-established PSO variants. Extensive studies have verified that SATCPSO is more competitive than its peers in most of the tested problems.

Demand response (DR) is an effective solution used to maintain the reliability of power systems. Although numerous demand bidding models were designed to balance the demand and supply of electricity, these works focused on optimizing the DR supply curve of aggregator and the associated clearing prices. Limited researches were done to investigate the interaction between each aggregator and its customers to ensure the delivery of promised load curtailments. In this paper, a closed demand bidding model is envisioned to bridge the aforementioned gap by facilitating the internal DR trading between the aggregator and its large contract customers. The customers can submit their own bid as a pairs of bidding price and quantity of load curtailment in hourly basis when demand bidding is needed. A purchase optimization scheme is then designed to minimize the total bidding purchase cost. Given the presence of various load curtailment constraints, the demand bidding model considered is highly nonlinear. A modified genetic algorithm incorporated with efficient encoding scheme and adaptive bid declination strategy is therefore proposed to solve this problem effectively. Extensive simulation shows that the proposed purchase optimization scheme can minimize the total cost of demand bidding and it is computationally feasible for real applications.

An innovative solution to provide a demand response for power system is proposed in this paper by considering the feasibility of two-way direct load control (TWDLC) of central air conditioning chiller system for wide-area in real-time manner. Particularly, the proposed TWDLC scheme is designed to tackle the load shedding ratio optimization problem for all under-controlled customers, aiming to satisfy the target load curtailment defined in each scheduling step. Another notable contribution of this work is the utilization of constraint loosening concept on actual load, curtailed to overcome the uncertainties of load reduction during TWDLC. Given the presence of fuzzy constraints, the proposed load shedding ratio optimization problem can be tackled using fuzzy linear programming. A delicate strategy is then formulated to transform the proposed fuzzy linear programming problem into a regular linear programming problem. A selection scheme used to obtain the feasible candidates set for load shedding at every sampling interval of TWDLC is also designed along with the fuzzy linear programming.