The use of lithium-ion batteries (LIB) is common in portable devices, electric vehicles, and aerospace applications thanks to their considerable amount of energy density, compact size, as well as the low maintenance costs. These advantages make them a key technology in promoting sustain ability and combating climate change. A reliable lithium-ion battery (LIB) model is essential for accurately estimating key parameters such as voltage and state of charge. Developing an efficient and precise battery models are considered vital, as they account for various influencing factors. A battery model was developed in this study, using MATLAB/Simulink, focusing on the effects of temperature and SOC. The model is explained, developed, and validated against experimental data, measuring voltage, SOC, and current under varying temperature conditions. Although this model currently considers only SOC and temperature, it can be extended to include additional factors.
Precise State-of-Charge (SoC) estimation is crucial for accurate electric vehicle range prediction and efficient battery charge management. This paper explores a computationally efficient machine learning approach to address SoC-estimation-related issues. A Long Short-Term Memory (LSTM) network is utilized for predicting the SoC of lithiumion batteries, leveraging real-time data on temperature, voltage and current. The proposed LSTM model demonstrates remarkable performance, achieving a mean absolute error of 1.02%, as well as a 1.35% root mean squared error at 25°C, while ensuring low computational overhead, making it suitable for practical implementation in Battery Management Systems (BMSs). This work lays a solid foundation for future advancements in BMS design, offering significant improvements in SoC prediction accuracy, reliability and efficiency. The findings suggest that LSTM networks can effectively model the complex dynamics of battery behavior, even when having varied operational conditions.
Document categorization continues to be a significant area of research, particularly in the context of automating the indexing of diverse web content such as blogs and forums. However, large document classification poses challenges in relation with both performance and processing time. This paper empirically investigates the impact of text summarization on document classification. For that, we compare classification performance before and after applying summarization techniques, focusing on two key aspects: (1) computational time and (2) classification performance, measured through accuracy, loss, and F1-score metrics. Our findings demonstrate the effectiveness of the proposed method, highlighting its potential to enhance the efficiency and accuracy of document categorization processes.
This paper presents a novel flexible load model tailored for cryptocurrency mining applications, specifically designed to dynamically adjust Application Specific Integrated Circuits-based mining operations based on real-time power availability. The model allows miners to use excess power during peak generation and reduce their usage during low power generation, all in support of maintaining a stable grid. The integration of the System Identification technique and some machine learning methods, such as Linear Regression, Support Vector Machine, and Neural Network Models, are used to identify and verify the proposed model against the genuine load data. Values of R2 ranging from 0.9898 to 0.9953 point toward a very good agreement between the simulated and actual load profiles. The Neural Network Model presents the lowest RMSE, MSE, MAE, and WIE, which means that the model accurately captures the actual behavior of the load. This approach strengthens the efficiency and accuracy of mining processes and is quite compatible with renewable power generation and, therefore, the effective use of energy. This research forms the reference platform for adaptive load management for such intensive applications to provide a viable approach to renewable energy integration into high-power utilization systems.
Crosstalk can reduce the performance of RF, analog, mixed-signal and digital integrated circuits. Measurement methods using appropriate measurement parameters to evaluate substrate crosstalk are discussed. Several layout options and technological parameters are compared. Based on these results the designer can choose and combine a set of options for more isolation of sub circuits to fulfill the requirements for the specific application. The focus is set on the different Silicon on Insulator technologies. Depending on frequency range the influence of isolation measures and their combinations is shown. The S-parameters of several test structures were measured using a vector network analyzer. The main results are differences in crosstalk between 50 MHz and 300 MHz.
To address the problem of illumination intensity resulting from the different usable ages, this study calculated the cost of replacing fluorescent lamps in a dairy factory located in Tripoli, Libya, which operates on a photovoltaic system before and after the end of its useful life. The aim of the study was to save electricity to reduce the load on the solar energy system, as the number of lamps was 6000 lamps. This also helped address the problem of the high cost of replacing damaged lamps individually. Using mathematical formulas as a technique, it was determined that replacing lamps at the end of their useful life costs 42,000 Libyan dinars annually, while replacing lamps after their useful life costs 84,000 Libyan dinars during the same period of time. This means that there are annual savings of 48.21% or 43,500 Libyan dinars. While preserving the transformers and wires, in addition to selling the replaced lamps that can be used again.
Permanent magnet synchronous machines (PMSMs) are of great interest in automation and robotics applications due to their small size, high efficiency, and low maintenance requirements. However, these motors face challenges in controlling their speed and position, such as precise rotor detection. This situation is essential in many Industrial applications. Conventional controllers such as PID controllers are commonly used, but their performance is often affected by changes in system dynamics, necessitating updating of controller parameters based on the accurate mathematical model of the motor. This research aims to investigate several intelligent control strategies that do not rely on the mathematical model of the motor. The performance of the drive system will be evaluated using three intelligent controllers: a neural network controller, a neural fuzzy inference system (ANFIS) controller, and a deep reinforcement learning (DRL) controller. The performance of these smart controllers will be compared with that of the PID controller. The results presented in this paper demonstrate the ability of the proposed smart controllers to regulate the speed of PMSMs compared to a PID controller. Specifically, the ANFIS-based controller shows the best performance in terms of maximum overshoot, while the DRL controller demonstrates the best settling time.
In this work, the effects of electrical polarization on 20 nm FinFET devices are studied at various temperature values. FinFET operation is described using an electrical model based on an enhanced drift-diffusion (D-D) model. However, we introduce a new expression for electron mobility to improve the D-D model. The Finite Element Method is employed to obtain results. Based on experimental and numerical results, our model can effectively interpret the degradation of electrical characteristics in the studied FinFET. Moreover, this research demonstrates that the temperature significantly impacts the IDS-VDS characteristics of the studied FinFET, leading to a decrease in performance.
This paper addresses the estimation problem of sensor and actuator faults for nonlinear systems represented by decoupled multimodel affected by unknown but bounded disturbnces and measurement noise. For this, an interval multiobserver is designed to estimate state and default vectors simultaneously. The existence of the interval observer is formulated using Linear Matrix Inequalities (LMI) to guarantee stability conditions. Finally, the effectiveness of the proposed method is proven by simulation results.
Electrical Impedance Tomography (EIT) is a widely used non-invasive imaging method in a variety of application fields. Most EIT applications require an accurate and high-quality reconstructed inner conductivity distribution image. Several solutions have been developed for reliable EIT image reconstruction, but they still suffer from limitations such as low spatial resolution and inconsistency in real measurement scenarios. In this paper, we propose using physically based, extensive simulation data to train a Convolutional Neural Network (CNN) model with 5-fold cross-validation. We generate a set of 10.000 data samples, which builds a comprehensive basis for its training and ensures reliability and generalization capability. The quantitative evaluation demonstrates good reconstruction performance, achieving a Mean Squared Error (MSE) of 0.0151 and an Image Correlation Coefficient (ICC) of 0.97 in the best-performing fold for the simulated data. The method's effectiveness is particularly validated through experimental measurements, where it significantly outperforms traditional Gauss-Newton reconstruction in terms of image sharpness and boundary definition. The algorithm is robust and easily applicable for EIT imaging.
Timely and accurate detection and classification of lung diseases are critical for effective diagnosis and treatment planning. This paper presents a deep learning approach to distinguish COVID-19 from normal cases using chest X-ray images, employing the VGG-16 convolutional neural network (CNN) model. To enhance interpretability and trust in the model's decisions, we integrate Explainable Artificial Intelligence (XAI) techniques–Grad-CAM, Grad-CAM++, and Score-CAM–to visualize and interpret the features contributing to the classification outcomes. Our model achieved an accuracy of 98.68%, highlighting the effectiveness of the VGG-16 architecture for this task. Each XAI method provided unique insights: Grad-CAM localized general feature regions, Grad-CAM++ offered more refined and detailed heatmaps, and Score-CAM produced clearer, gradient-free visual explanations that enhance transparency and stability. Our findings indicate that Score-CAM outperforms the other techniques by providing stable, precise heatmaps, making it particularly suitable for clinical use. This combination of high accuracy and interpretability supports the applicability of AI-driven diagnostics in real-world healthcare settings, fostering trust and reliability in automated lung disease classification.
Recent years have seen a rapid development in Machine Learning, which has profoundly influenced many areas of science and engineering. Among them, computer vision takes the leading place, where important tasks are image classifications powered by CNNs. Despite the great performance of CNNs in complicated scenarios, they remain sensitive to so-called adversarial attacks, and deliberate perturbations leading them to incorrect predictions. Besides more innocuous consequences, this has serious security implications for critical applications, in-cluding medical diagnostics, where misclassifications might result in disastrous outcomes. This research work discusses adversarial attacks on CNNs and other DNNs in computer vision, studying a full range of the generation and detection methods with details while discussing intrinsic vulnerability and robustness. It also proposes a learning framework that will enhance the robustness and security of DNNs and CNNs against such adversarial perils. The ultimate goal is directed to an improvement in the reliability of such models in absolutely critical scenarios for safe deployment into applications where accuracy is crucial.
Adaptive modulation and coding is a transformative technology that enables the real-time optimization of modulation and coding schemes, dynamically adjusting to varying channel conditions. This capability is crucial for next-generation wireless communication systems, where high performance, efficiency, and reliability are essential. By leveraging real-time feedback from the receiver, adaptive modulation and coding ensures that data transmission is continuously optimized, maximizing spectral efficiency while minimizing errors. The objective of this paper is to provide a comprehensive review of the latest advancements in adaptive modulation and coding techniques, critically evaluating the challenges and opportunities in their practical implementation. The contribution of this work lies in synthesizing key trends in adaptive modulation and coding, identifying the primary obstacles to its adoption, and exploring emerging research directions. Additionally, this paper offers a framework for advancing the efficiency and robustness of future communication systems, providing insights that will drive the next wave of innovation in wireless networks.
The effects of partial shading can be severe, even when only a small fraction of the panel is shaded. This phenomenon can lead to several issues. Reduced Power Output: As mentioned earlier, solar cells are typically connected in series, meaning that the current flowing through the entire string is limited by the lowest-performing cell. A shaded cell acts as a bottleneck, reducing the current in the entire panel and thereby decreasing its output power. In extreme cases, the shaded cell can even reverse current flow, leading to energy loss. Hot Spots and Damage: Partial shading can create “hot spots” on the panel, where the shaded cells overheat due to the backflow of current. These hot spots can cause permanent damage to the solar panel, significantly reducing its lifespan. Prolonged exposure to these conditions may cause the affected cells to fail entirely, which can be costly to replace.
This paper proposes intelligence methods for the prediction of wireless network traffic and the evaluation of its development. This solution analyses the changing patterns of actual business traffic data in cellular networks in the time dimension and uses ensemble learning (EL) to predict the tendency of traffic, providing guidance for operators' network optimization deployment from a short-term perspective. The paper depends on a dataset taken from the Eurecom ElasticMon 5G monitoring platform. The approach applied a multi-dimensional mapping relationship between other operational data and traffic in the network using four ensembles of learning: Extreme Gradient Boosting (XGBoost), ADA boosting, gradient boosting algorithm, and a voting regression algorithm. They applied to find the upper limit of traffic that the cellular cell can carry, optimizing the deployment of the network from a long-term perspective. The results showed that XGBoost outperforms other methods, whose best performance ensures the privacy of traffic data.
In this study the design and simulation of a microstrip antenna array for a radar application using the S-band (2 to 4 GHz) have been examined. To enhance the electrical and radiation properties of the suggested antenna, a slot is incorporated into the radiating element and a new power divider is employed. Two rows, each with four pieces, make up the planned array, which has overall measurements of 11.7 x 11.9 cm in width and length respectively. The findings indicate that at a resonance frequency of 3.76 GHz, the suggested antenna has a reflection coefficient of -34 dB. When the four-element array is assembled, the radiation gain is enhanced, reaching 6.18 dB with an efficiency of 98.66%. Furthermore, the antenna array provides complete 360° coverage, enabling omnidirectional and continuous scanning. This feature is essential for radar applications, as it ensures optimal target detection across the entire perimeter around the radar, with no dead zones or interruptions.
The increasing penetration of distributed energy resources (DERs) poses several challenges for network control. Voltage issues are the main drawback of high DER penetration, especially in photovoltaic (PV) systems. Moreover, ensuring a balance between electricity production and consumption is still a real challenge for the power grid. In response to such constraints, PV inverters should support grid voltage. In this context, inverters play a key role in adjusting the power injected into the grid according to variations in demand and production. Conventionally, inverters were operated with a unit power factor. Still, advanced inverters, called smart inverters, can even supply reactive power to the grid to overcome voltage problems. These advanced functions enable solar PV inverters to monitor and communicate system status, automatically providing auxiliary services to the grid. Thus, advanced inverter control is essential to prevent imbalances that could lead to voltage fluctuations or grid instability. Specifically, Var-priority and Watt-priority modes were designed in the paper for a comparative study. Simulations were conducted to validate the modes' effectiveness in addressing voltage issues under grid disturbances.
This paper discusses the use of image processing to improve object handling in the SCARA robot system. The proposed approach aims to achieve faster and safer movement of the SCARA robot arm for object handling. During the experiment, the optimal performance modification determines the subsystem required to accommodate and compensate for the system's image processing capability. In general, images are processed at lower speeds, as higher speeds could potentially damage the object during system transportation. The results of the proposed approach indicate the development of new object-handling techniques for SCARA robot movements by introducing visual feedback to display and control objects. To improve object handling by the SCARA robot, several parameters in the image are described in the subsequent process. This new application of the SCARA robot system is significant, as the collision-free machine modifications emphasize the safety of the working setup. It was found that the precise manipulation of the SCARA robot when handling objects in open-loop control lacks sufficient technical representation.
This paper proposed a load frequency control (LFC) to ensure power quality of a two-area interconnected hybrid power system (PS) involving integrated electric vehicles (EVs) and communication delay. The control framework consists of three optimization-based proportional-integralderivative (PID) controllers named genetic algorithm (GA-PID), particle swarm optimization (PSO-PID), and grey wolf optimization (GWO-PID) controllers, and fractional order proportional-integral-derivative (FOPID) for improving the system's stability and performance. The performance of the proposed controllers for handling the challenges arising from EV integration, communication delay, and renewable energy sources (RESs) intermittency has been assessed. A thorough performance analysis is subsequently carried out according to the simulation results and various error indices, including integral square error (ISE), integral absolute error (IAE), and integral time absolute error (ITAE). The findings show that the proposed GA-PID controller performs better dynamically than the PSO-PID, GWO-PID, and FOPID controllers in terms of robustness and convergence rate. The results suggest the improvement of LFC in power systems through optimization algorithms.
This paper focuses on the design of interval state observers for uncertain linear im-pulsive systems subject to bounded disturbances and parametric uncertainties. According to positive systems theory, an hybrid interval observer is designed to give upper and lower bounds of the state. The design problem is formulated as Linear Matrix Inequality (LMI) and the efficiencies of the proposed approach are validated through numerical simulations, demonstrating its robustness against bounded uncertainties and disturbances.
