Bilad Alrafidain Journal for Engineering Science and Technology (BAJEST)

Microgrids serve as primary links for diverse loads, while numerous locally situated power supply systems are concurrently established in various locations. Grid-integrated hybrid renewable energy power conversion systems have gained significant popularity in recent years. However, a novel and effective control methodology must be implemented to supply high-quality power at the local load bus. The 3rd Order Degree of Freedom – PID (3DOF-PID) controllers stand out among numerous controllers, demonstrating exceptional performance across a wide range of operating conditions. In this paper, a hybrid system comprising a solar power plant and a wind power plant is examined, with a focus on its benefits to the power grid. Furthermore, Sliding Mode Control (SMC) techniques are applied to all photovoltaic (PV) and wind power conversion systems to facilitate the operation of the boost converter as Maximum Power Point Tracking (MPPT) devices. A shared DC bus is created by combining all converters in the solar power facility and the wind energy conversion plant. An inverter is integrated between the DC bus and the utility grid's point of common coupling (PCC). The study presents a robust control strategy utilizing 3DOF-PID controllers to ensure high-quality power delivery at the PCC while accommodating a diverse range of load combinations, including both linear and nonlinear loads, as well as single-phase and three-phase reactive power loads. 3DOF-PID controllers have been created for each converter and incorporated using OPAL-RT technology and modules within a Real-Time Simulator (RTS). The results are presented through RTS to evaluate the suggested control methodologies for the power supply system that integrates grid-connected hybrid renewable energy sources (RESs).

Cybersecurity has seen preparatory AI technologies developing quickly and aiming to fill the gap that arises among human defenders. Interdisciplinary areas of cooperation are emphasized in sponsored research programs in the government, academia, and industry. Practice also develops assets for cybersecurity using AI technologies, including datasets and challenges. Strategies for privacy and structure are also relevant, particularly the fact that systems holding sensitive cybersecurity data that themselves require protection will benefit most from AI. The purpose of this study is to strengthen familiarity with the potential gaps in guarding various types of cybersecurity data. This research identifies several key findings: (1) AI-based systems can enhance threat detection by 78.5% compared to traditional methods; (2) machine learning algorithms demonstrate 93.7% accuracy in identifying zero-day attacks; and (3) natural language processing techniques significantly improve phishing detection with 94.3% accuracy rates. Each type of computer security-related data may have its vulnerable attributes and be susceptible to attack, loss, and disruption in its methods of data gathering, study, and knowledge, and in its development and use. Security awareness should help stakeholders consider how to use AI to leverage and preserve the data types that they need to help monitor, protect, and rebuild current, future, and evolutionary cyber-physical systems. Building on data responsibilities described in different properties and standard operating procedures of several computationally challenging and defense-related data lifecycle scenarios, this study also identifies specific original questions related to data destruction, the immediate manipulation of AI labels, and other cybersecurity risks that AI teams may wish to consider.

Effective prediction of electric power consumption is vital for accurate energy management systems, particularly in growing urban regions. Conventional statistical techniques usually fail to capture the complicated patterns and non-linear tendencies in time-series energy data. The development of deep learning models has demonstrated encouraging findings in a variety of prediction tasks. Therefore, this paper concentrates on the application of diverse deep learning models, comprising One-dimensional Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Bidirectional LSTM, and a hybrid model for power consumption prediction based on a dataset gathered from January 1, 2017, until January 1, 2018, in Tetouan, Morocco. These applied models were assessed utilizing diverse robust assessment metrics, and the findings demonstrated that the hybrid model (One-dimensional CNN with LSTM) reached superior performance with a Mean Squared Error (MSE) of 0.42548, Root-MSE of 0.65228, Mean Absolute Error (MAE) of 0.54395, and Median-AE of 0.51401, surpassing the other standalone and relevant models. The attained findings emphasize the merits of incorporating recurrent and convolutional structures to obtain more effective and accurate predictions of energy consumption time series.

Modern medical devices use MRI images as a file format to explain the components of the human body, such as the human brain. Successful treatment and clinical diagnosis of brain tumors rely on accurate tumor classification. A method used to categorize brain tumors in this paper, based on machine learning classifiers and deep features, is proposed. The suggested framework makes use of multiple pre-trained deep convolutional neural networks (CNN) and the transfer learning concept to extract deep features from brain MRI images. Numerous machine learning classifiers assess the retrieved deep features and the best three models of deep features are selected. Many machine learning classifiers are defined and sequenced as a set of deep features, which are then input into several ML classifiers for use in making output predictions. Using three publicly available brain MRI datasets, we compare the performance of various pre-trained models for brain tumor classification, including deep feature extractors, a deep feature ensemble, and machine learning classifiers. In most instances, a CNN is utilized, and the output of the experiments demonstrates that a collection of deep features can greatly enhance tumor detection performance. The efficiency of the proposed method's is tested by measuring accuracy, which averages of about 99.2%, denoting the ability to be used in medical applications. The proposed method could be used in all other human body components for anomaly detection.

The most common kind of cybercrime now is phishing, it attempts to deceive users into divulging sensitive information like passwords, bank details, and account numbers. Such cyberattacks frequently take advantage of electronic communication channels such as emails, instant messaging, and phone calls. Nowadays, there has been a dramatic uptick in the building of computer networks. Looking at the present pattern among people who use computers worldwide, it is evident that they are required to establish a connection between their PCs and the web. These results highlight the critical nature of Internet connectivity, whether for personal or professional reasons. However, users' privacy is at risk due to the widespread usage of this network, particularly for those users who do not activate security software on their computers. Once this vulnerability is exploited, hackers would be able to breach networks and launch attacks. Hackers may steal sensitive information, including login credentials to online accounts like banks and social media, making this a major concern for anybody using the internet. Some of the assaults that can be launched include phishing attempts. Reviewing the many forms of phishing attempts and the solutions now employed to avoid them is this research set out to do. Deep and machine learning has shown to be an effective tool in the fight against phishing, according to the study. It is possible to use a variety of approaches in the deep and machine learning approach to ward off such assaults.

The current work used the laser cladding method to prepare a composite coating of titanium nitride (TiN) combined with boron nitride (BN) on 420 stainless steels. The microstructure, micro-hardness, wear resistance, and bio-corrosion resistance of the coatings were analyzed using scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS), a micro-hardness tester, a wear tester, a bio-corrosion resistance test, and electrochemical techniques. The results indicated that a strong metallurgical bond was successfully formed between the composite coating and the substrate. SEM analysis revealed that the sample produced at low power demonstrated superior performance compared to the high-power sample with respect to the microstructure. The path coefficients related to adhesion factors, such as microhardness, were higher in the low-power sample, suggesting stronger and more consistent interactions between the coating and the substrate. EDS results showed that the high-power sample had a lower iron (Fe) signal, indicating superior performance in that regard; however, the low-power sample displayed strong titanium (Ti) and BN components. Micro-hardness decreased with high power and increased with low power. The high-power sample exhibited a higher wear rate compared to the low-power counterpart. In the bio-corrosion test, the uncoated martensitic steel displayed a more negative and unstable open circuit potential (OCP) profile. In contrast, the TiN + BN-coated sample showed a less negative and more stable OCP response, indicating enhanced corrosion resistance. Electrochemical techniques demonstrated that the "Mix" sample had superior electrochemical performance compared to the "Base" sample. It exhibited a broader current density range, a higher current response, and clearer separation of anodic and cathodic reactions, making it more effective for applications in medical instruments such as orthopedic implants, dental implants (abutments and dental screws), surgical tools (scalpels, forceps, bone drills), and spinal implants and plates. Electrochemical techniques demonstrated that the "Mix" sample had superior electrochemical performance compared to the "Base" sample. It exhibited a broader current density range, a higher current response, and clearer separation of anodic and cathodic reactions, making it more effective for applications in medical instruments

Advanced models of deep learning have been transformational tools for discovering drugs and solving problems related to cost, time, and complexity. Using complex network frameworks such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), generative adversarial networks (GANs), and graph neural networks (GNNs), researchers have made major improvements in predicting drug-target interactions, performing virtual screens, and developing novel drugs. Those frameworks effectively occupy elaborate biochemical relations and precisely imitate complicated molecular reciprocities. Nevertheless, significant issues remain, like data shortages arising from restricted access to (high-quality) datasets, model predictions' interpretability, and the scalability to include considerable and assorted datasets. To efficiently address these issues, innovative strategies, including diverse techniques of data augmentation, like molecular graph transformations, have been applied to improve datasets. Reinforcement learning has helped improve molecular structures to accomplish desired characteristics, while ‎ensemble learning, which integrates various model structures, has proven effective in improving prediction ‎reliability. Incorporating multi-modal datasets, like pharmacophores properties, 3D molecular ‎representations, and molecular graphs, increases the accuracy of prediction by occupying spatial and ‎even functional molecular properties. Despite these advances, issues remain in multi-drug modeling, drug resistance management, and accurate toxicity prediction. Future works focus on the importance of explainable AI in strengthening model interpretability, with hybrid structures that incorporate machine learning and experimental feedback to simplify the therapeutic scheme. By addressing these challenges and adopting innovative approaches, deep learning is set to revolutionize drug discovery, enabling a more efficient, accurate, and reliable development pipeline for novel therapeutics. This study highlights how model interpretability and confidence can be enhanced by integrating multigene data and leveraging explainable AI techniques. By focusing on these cutting-edge developments, this study aims to provide practical insights for researchers and practitioners to accelerate the development of safe, effective, and personalized therapeutics

X-ray diffraction (XRD) data analysis is a very important technique in materials science. However, traditional methods are time-consuming, require human expertise, and are prone to errors due to the large and complex datasets. This study explores the potential of machine learning (ML) and deep learning (DL) techniques to automate the classification of large X-ray diffraction datasets. This work provides a historical overview of X-ray diffraction, how it occurs, and the fields it involves. It also presents the contributions of artificial intelligence (AI) fields, including machine learning and deep learning, to improve classification and prediction material. The study attempts to present the best of what researchers have achieved in this field, providing a scalable solution for high-throughput materials characterization and its potential for advancing automated XRD data interpretation in both research and industrial applications.