Advance Journal of Science Engineering and Technology

Machine learning (ML), a key branch of artificial intelligence, enables computers to learn from data, identify patterns, and make predictions without explicit programming. With rapid growth in data availability and computational power, ML has become widely used in areas such as healthcare, finance, transportation, and natural language processing. This review provides an overview of fundamental machine learning concepts, major learning paradigms, and their practical applications. It focuses on three main approaches: supervised learning, unsupervised learning, and reinforcement learning. An experimental comparison of representative algorithms—Support Vector Machine (SVM), Decision Tree, Linear Regression, K-means clustering, and Q-learning—was conducted using standard datasets and evaluated through accuracy, precision, recall, and F1-score. Results indicate that supervised learning algorithms performed better for prediction tasks with labeled data. SVM achieved the highest performance with 90% accuracy, followed by Linear Regression (87%) and Decision Tree (85%). K-means clustering showed moderate performance, while Q-learning demonstrated lower accuracy in static prediction tasks. The study concludes that algorithm selection should depend on data characteristics, problem requirements, and computational constraints. While supervised learning is most effective for labeled datasets, unsupervised and reinforcement learning remain valuable for pattern discovery and sequential decision-making.

Jordan’s electricity sector faces major challenges due to its heavy reliance on imported fossil fuels, rising electricity demand, and infrastructure inefficiencies. Approximately 94% of the country’s energy needs are met through imports, making the sector vulnerable to geopolitical instability and price fluctuations. At the same time, population growth and economic development have increased electricity demand, highlighting the need for sustainable energy solutions. This review analyzes the key challenges, infrastructure conditions, policy frameworks, and future opportunities in Jordan’s electricity sector. Data from national institutions such as the Ministry of Energy and Mineral Resources (MEMR), National Electric Power Company (NEPCO), and international sources were examined to assess consumption trends, system losses, renewable energy potential, and regional interconnection projects. Electricity consumption in Jordan increased from 6.1 billion kWh in 2000 to 17.0 billion kWh in 2021. Natural gas currently accounts for about 80% of electricity generation, while renewable energy contributes around 20%. Solar photovoltaic capacity has reached 2,063 MW and wind power 370 MW. However, transmission and distribution losses remain significant, and electricity tariffs remain below actual system costs, creating financial pressures. The study concludes that expanding renewable energy, modernizing grid infrastructure, improving policy implementation, and strengthening regional energy cooperation are essential for enhancing energy security and sustainability. Jordan’s strong solar and wind potential provides a promising pathway for reducing import dependence and achieving a more resilient electricity system.

The rapid growth of network traffic and the increasing use of encrypted applications have made traditional traffic classification methods such as port-based techniques and deep packet inspection less effective. Software-Defined Networking (SDN) provides centralized control and programmability, enabling advanced traffic management when combined with machine learning (ML). This study reviews and evaluates the performance of ML algorithms for network traffic classification in SDN environments. Experiments were conducted using the UNB ISCX Network Traffic Dataset containing multiple traffic categories such as CHAT, FILE, STREAMING, VIDEO, AUDIO, and MAIL. Three machine learning models—Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR)—were implemented. Recursive Feature Elimination with Cross-Validation (RFECV) was applied for optimal feature selection and to reduce overfitting. Model performance was evaluated using accuracy, precision, recall, and F1-score. Results show that Random Forest achieved the best performance with 97.49% accuracy and a macro F1-score of 0.97, followed by SVM with 95.55% accuracy and Logistic Regression with 92.87%. Feature selection significantly improved classification performance and model generalization, particularly in handling imbalanced traffic classes. The study concludes that Random Forest combined with RFECV-based feature selection provides an effective solution for accurate and efficient traffic classification in SDN. The integration of machine learning with SDN can improve network security, quality of service, and resource management.

Self-Compacting Concrete (SCC) offers superior workability and can be placed without mechanical vibration, making it suitable for heavily reinforced structures. However, SCC is more vulnerable to high temperatures due to its dense microstructure, which can trap vapor and cause explosive spalling during fire exposure. This study investigates the combined effect of basalt fibers (0.4% by volume) and water-absorbent polymer beads (WAPB) on the fresh properties, mechanical performance, thermal conductivity, and residual strength of SCC exposed to temperatures up to 700 °C. Four SCC mixtures were prepared: a reference mix with basalt fibers only and three mixes containing 3%, 4%, and 5% WAPB by weight of cementitious materials. Fresh properties were evaluated using EFNARC tests, while compressive, splitting tensile, and flexural strengths were measured at 7, 28, and 56 days. Specimens were then exposed to 300 °C, 500 °C, and 700 °C to determine residual strengths. Results showed that WAPB improved workability and segregation resistance, with slump flow increasing up to 9.92%. Although compressive strength decreased initially at 28 days, WAPB mixes exhibited greater strength gain by 56 days due to internal curing. After exposure to 700 °C, residual compressive strengths increased from 32.91% in the reference mix to 46.84% in the mix with 5% WAPB. Similar improvements were observed in tensile and flexural strengths. The findings indicate that incorporating WAPB in basalt fiber-reinforced SCC enhances fire resistance and structural stability. The internal curing effect and formation of pressure-relief voids help reduce thermal damage and spalling. This approach shows strong potential for developing fire-resistant SCC for structures such as high-rise buildings, tunnels, and industrial facilities.

Background: The global energy landscape is undergoing a profound transformation driven by the dual imperatives of sustainable development and climate change mitigation. With conventional fossil fuel reserves depleting rapidly and environmental concerns escalating, renewable energy sources have emerged as viable alternatives for meeting growing energy demands. Among these, solar photovoltaic and wind power systems represent the most mature and widely deployed technologies. However, the inherent intermittency and variability of these sources necessitate sophisticated control strategies to ensure reliable power delivery and maximum energy harvesting. Hybrid renewable energy systems combining solar and wind sources offer enhanced reliability through complementary generation patterns. In this context, the present study focuses on the design, implementation, and experimental validation of an optimal fuzzy logic controller (FLC) for hybrid solar–wind energy systems implemented on a field-programmable gate array (FPGA) platform to achieve improved tracking efficiency, faster response times, and enhanced system stability. Methods: The proposed hybrid system integrates photovoltaic (PV) modules and wind turbine generators with permanent magnet synchronous generators (PMSG). Mathematical modeling of both subsystems was developed using MATLAB/Simulink, incorporating detailed PV array characteristics and wind turbine aerodynamics. A Mamdani-type fuzzy logic controller was designed using error (E) and change in error (dE) as inputs with five linguistic variables (NB, NS, ZE, PS, PB). The controller determines optimal duty cycles for DC–DC converters to achieve maximum power point tracking (MPPT). For hardware implementation, the Xilinx System Generator (XSG) platform was utilized to convert Simulink models into FPGA-compatible designs. The controller was implemented on a Virtex-6 XC6VLX315T FPGA using Xilinx ISE for synthesis and bitstream generation, and system performance was evaluated under varying atmospheric conditions. Results and Conclusion: Experimental results demonstrated excellent performance of the proposed FPGA-based fuzzy logic controller. The system achieved maximum power point tracking efficiency of 99.7%, significantly outperforming conventional approaches. The duty cycle stabilized at 0.38 within approximately 10 ms, indicating rapid convergence to the optimal operating point. Comparative analysis showed that the XSG-based implementation achieved 5% faster power stabilization than conventional MATLAB/Simulink models. The inverter output produced clean sinusoidal waveforms with minimal harmonic distortion and accurate 120° phase separation among the three-phase voltages. Under varying wind speeds up to 15 m/s, the controller maintained stable operation with duty cycle stabilization around 0.41. These results confirm that FPGA-based fuzzy logic controllers provide superior parallel processing capability and enhanced control efficiency. The proposed system offers a promising solution for efficient integration of hybrid renewable energy sources, particularly for addressing growing electricity demands in developing countries. Future work may explore integration of deep learning and reinforcement learning techniques for improved adaptability and grid integration.