NTU Journal of Engineering and Technology

Classification of epileptic seizures, using the electroencephalogram (EEG) signal, is a challenging phenomenon in clinical neuroscience due to the high-dimensionality of EEG signals, non-stationarity of EEG signals, and a skewed distribution of classes. This paper illustrates a comparative discussion on the classical and the ensemble machine learning model in binary seizure/non-seizure classification using high-dimensional EEG features that came about after considerable preprocessing on 36,864 channels of information. The use of standardized normalization and variance-based feature screening was used to train several different classifiers, including logistic regression, calibrated support machine, k nearest neighbor (kNN), stochastic gradient descent, random forests, gradient boosting, histogram-based gradient boosting (HistGB), and multi-layer perceptrons. The most effective models evaluated with an accuracy of 98.2, F 1-score 0.89, receiver operating characteristic area under the curve (AUC) of 0.993 were the histogram-Based Gradient Boosting as well as the random forest and gradient boosting models, where the level of discrimination was equally high. The findings give a highly empirical point of reference of classical and ensemble learning in EEG-based seizure-detection and depict how tree-based ensembles perform in modeling complex and non-linear EEG feature space. Upon such validated findings, the contactual spatiotemporal deep learning framework, known as DeepWalk-Transformer Sequence (DeepWalk-TS), is also described in the research that must be adopted in the future to integrate the concept of graph spatial representation with transformer time modeling. The proposed framework is not experimentally tested within the given research and rather is proposed as future work.

Quick Response (QR) codes have become very popular in making payments, marketing/authentication, and are frequently used with malicious intent by cybercriminals using them to inject malicious URLs to steal information or install malware. The current study is concerned with malicious QR codes detection with deep learning. A balanced data set of 316,254 benign and 316,254 malicious URLs were turned into an image of QR code to train a Convolutional Neural Network (CNN). On the validation set, the model attained 99.62% accuracy and on the test set, the model attained 100% accuracy, absolute precision, recall and F1 scores. Average processing speed is 61ms per batch which allows real time scanning. CNN was able to outperform other probabilistic models because of its high performance implying that further comparisons can be carried out in future. This work brings to the fore an efficient, scalable solution devised to identify malicious QR codes and enhance the security in the current contactless digital environment.

Cracks and discontinuities present in the structure have a significant influence on the vibration properties of a component. The research work presented in this paper investigates the effect of a crack present at the fixed edge of the cantilever steel plate on the vibration frequency and mode of the steel plate. The analysis presented above adopts a combination of modeling techniques based on CPT and the numerical analysis technique based on FEM in the ANSYS APDL package and attempts to establish the amount of variation in the vibration properties of a plate with the rise of the a/W ratio of the crack. A detailed finite element analysis has been conducted with special attention to the modeling near the cracked area. The natural frequencies and modes have been analyzed for various values of the crack ratio and validated with theoretical values based on the frequency equations with an account for flexibility caused by the presence of cracks. Both theoretical and FE results show that with an increase in the value of the crack ratio, natural frequencies tend to reduce, especially for mode 1. Mode shapes show localization near the crack area, which indicates the effect of the crack on the system stiffness and vibration. The proposed equations derived by fitting the numerical results into the classical frequency equations offer a reliable means for determining the effect of the crack on the frequency shift without having to go to experimental methods. This research highlights the capabilities of a combined analytical and numerical approach to the prediction of the vibration response of cracked systems with a view to developing the area for complex damage configurations.

In this study, the stability of the slope close to the Khosr River in Mosul, Iraq, is examined. Three different layers make up the slope: sandstone at the bottom, highly malleable clay (CH) in the middle, and gravel, sand, and clay at the top. Concerns regarding possible failure, which may cause large material and human losses, have been raised by the growing construction of multi-story buildings at the top of the slope. The SLOPE/W, SIGMA/W, and SEEP/W modules of the Geo-Studio 2012 program were used to perform a two-dimensional (2D) numerical analysis. The stability of the slope was investigated under a number of circumstances, such as increasing groundwater levels (4 to 10 m), rainfall (10 and 30 mm/day for up to 60 days), and external loads (250, 350, 450, and 500 kPa). The slope's initial factor of safety (FOS) ranges from 3.935 to 4.03, making it stable under typical circumstances. But when exposed to external stresses, more rainfall, and higher groundwater levels, the FOS dramatically drops. When a 500 kPa external load was applied, for example, the FOS decreased by up to 71.9%, and when a 100 kPa load and a 4 m water table were combined, the FOS decreased by more than 70%. The study assessed the efficacy of three reinforcement treatments—anchors, micropiles, and a retaining wall—to reduce the chance of failure. The outcomes demonstrated how well these actions improve stability. When positioned at a 25-degree angle, anchors raised the FOS by 24.1% to 196.8%. Three meters below the surface, micro-piles increased the FOS by 37.4% to 164.2%. The most successful option was a retaining wall, which increased the FOS by 178.5% and restored stability in critical conditions. It was 10 meters high and 6.5 meters wide at the foundation. The results show how important it is to have adequate drainage systems and slope reinforcement in order to protect the growing residential area.

The energy form of the Rayleigh-Ritz Principle was used to investigate the effect of rotating a mass at the free end on the dynamic characteristics of the cantilever beam. The complexity of boundary conditions caused by the end mass was avoided by applying the Rayleigh-Ritz principle. A weak form (energy form) Rayleigh Quotient for the beam with an end mass was derived. The stiffness and mass matrices of the system were obtained using 12 simple fixed-free (without end mass) eigenfunctions as admissible functions. Three different values of the end mass shape factor α (1, 2, and 3×10^(-3)), representing the mass moment of inertia of the end mass about the junction point, were used. The results showed that the influence of α on changing dynamic characteristics was significant in the second mode and above. The observed changes were 1.49%, 3.76% and 7.12% in the second, third, and fourth natural frequencies, respectively. A simple mathematical model was proposed to represent these changes. The end mass and its rotation transform the N+1 mode shape of the cantilever into N-mode shape of a fixed-pinned beam.

Attention Deficit Hyperactivity Disorder (ADHD) is a major challenge in pediatric neurodevelopment and one of the most common conditions in children and adolescents. Clinical diagnosis remains difficult due to symptom overlap with other disorders, increasing the need for objective tools. Electroencephalography (EEG) has emerged as a non-invasive technique that, when combined with artificial intelligence, enables more reliable diagnostic models. This study aims to review pediatric and Adolescent EEG-based ADHD studies published between 2015 and 2025, covering both traditional signal-processing and artificial intelligence approaches.Fifty studies were included in this review study.The originality of this review lies in providing a decade-long, child-focused synthesis that highlights accuracy trends, methodological progress, and challenges related to clinical translation. Reported performance varied widely: traditional spectral analyses with Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) achieved 62–88%, whereas deep learning models including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and attention-based hybrids,often exceeded 99%. The limitations across the literature include small sample sizes, restricted multi-center datasets, and insufficient evaluation of ADHD subtypes.

The fast growth of quantum computing puts widely used public-key cryptosystems like RSA and Elliptic Curve Cryptography (ECC) at risk because Shor's algorithm can quickly factor integers and find discrete logarithms. Grover’s algorithm similarly weakens symmetric ciphers like AES, necessitating larger key sizes. This work proposes the Hyperring RSA–AES Hybrid Encryption Scheme (HRA-HES), a hybrid cryptosystem that achieves post-quantum security for simple ciphers while preserving practical usability. HRA-HES derives session keys via Hyperring Learning with Noise within a Key Encapsulation Mechanism, and AES-256-GCM uses these keys to encrypt large data blocks. The multi-valued hyperaddition in the underlying hyperring structure disrupts the periodicity exploited by quantum period-finding algorithms. Implementation results show an encryption throughput of 850 Mbps and an average key generation time of about 2.1 ms, yielding improvements of up to 44% over prior baselines while maintaining low resource consumption, thus offering a scalable, quantum-aware transition framework.

This review presents a clear comparison of the most popular air quality modeling techniques. The main aims of the review were to compare the generally applied air quality modelling methods regarding assumptions, data demand, and applicability, assess how they can be used in urban areas with limited data, and to support the choice of the Box Model BM to assess air quality in Kirkuk, Iraq. The reviewed modeling approaches were Gaussian Dispersion Models GDM and Lagrangian Models LM of point source emissions in meteorological inputs. Computational Fluid Dynamics CFD and Eulerian Grid Models EGM to perform the high-resolution spatial and temporal analysis. The BM simplified model incorporates homogeneous mixing of pollutants in a given volume. The BM was used in this research to estimate the pollutant levels in Kirkuk, which allows referring to its limitations in combination with the Air Quality Index AQI analysis. The findings on the relevance of the BM in Kirkuk indicate that it can be utilized in initial analyses and rapid scenario testing in cities with limited data. Although the spatial detail is lower, it gives quite satisfactory pollutant estimates in combination with the analysis of AQI. Comparison indicates that BM is a relevant policy and cost-effective tool with minimal data access used to support air quality management in Kirkuk.

As electric vehicles become more widely used, there is a growing demand for more efficient and less complex power conversion technologies. Single-input, multi-output(SIMO) DC-DC converters are an essential component of electric vehicle voltage distribution systems, because it allows multi-level voltage supply from a single DC source such as a battery. Its importance is evident from the needs of the vehicle's various subsystems, including control units and auxiliary devices as well as the traction motor. This study deals with the design and analysis of a non-isolated single-input DC-DC converter (200V input) with three outputs: 400V, 24V, and 12V to provide voltage levels compatible with electric vehicle loads The proposed converter model is built and simulated in MATLAB/Simulink environment using a suitable topology combining one boost stage and two buck stages to provide a stable and highly efficient power supply to both the traction motor and auxiliary subsystems. To compare the dynamic and static performance, two control methods were implemented the first is sliding mode control (SMC) and the second is proportional integral control (PI. According to the obtained results, the SMC outperformed the PI controller in terms of voltage precision, disturbance rejection, reduced oscillations, and high efficiency exceeding 95%. This research work is strengthened by applying both PI and SMC control techniques in the design of a multi-output DC-DC converter within a simulation environment. The findings confirm the superiority of the SMC in maintaining regulated voltage and adjust to changes in load, making it the most suitable choice for electric vehicle.

This paper provides the design, implementation, and experimental assessment of an offline bilingual Arabic-English voice-controlled smart home automation system based on embedded hardware and radio-frequency (RF) communication. The given system uses an Arduino-based master-slave architecture, in which local speech recognition is realized with the help of the Elechouse Voice Recognition Module V3, and control commands being sent via HC-12 RF modules with 433 MHz frequency. In contrast to cloud-based voice assistants, all speech processing goes locally, which means that the user privacy is guaranteed and low-latency response is also provided. The system is capable of Arabic and English voice recognition and operating domestic appliances using optocoupler-isolated relay circuits. The experimental performance of RF communication in indoor settings has shown stable communications up to a range of 60m, end to end response latency of less than 300ms and command recognition accuracy of between 85 and 95 percent per command, and a total average command recognition accuracy of 90 percent under low noise conditions. Under high ambient noise, the performance is degraded, which demonstrates the weakness of template-based speech recognition. The findings suggest that offline bilingual voice interfaces used with low-power RF communication can offer a viable, privacy-sensitive solution to small-scale automation of the smart home as well as a platform upon which noise-resistant on-device learning algorithms will be implemented in future applications.

Solar thermal (ST) collectors are an efficient technology for generating thermal energy without the need for an electrical source, unlike photovoltaic thermal (PVT) collectors, which combine photovoltaic and thermal conversion to produce electricity and low-grade thermal energy. This research aims to present a low-cost strategy for improving the dual performance of PVT and ST systems by connecting them in series, under operating conditions that include different water flow rates and constant air flow in the ST unit. This work evaluates the thermal and electrical performance of a hybrid (PVT-ST) system, focusing on the effect of incorporating porous media and phase change materials (PCMs) into the system design. The results demonstrate that the integrated system enhances the heating efficiency of both water and air, contributing to an increase in the overall energy production efficiency. Experimental data showed that decreasing the water flow rate increased temperatures, while increasing the flow reduced temperature fluctuations and improved thermal production. The highest water temperature difference in the PVT system was recorded at 15.9°C at the lowest water flow, and the highest outlet air temperature was 58.5°C under the same conditions. The PVT-ST system's maximum total thermal power output was 756.13 W, achieving an overall thermal efficiency of 110.6%, an electrical efficiency of 12.22%, and a maximum electrical output of 142.69 W. The integration of thermal energy storage technologies enhances the stability of the system's performance, especially during periods of low solar radiation. The use of porous media improves internal heat transfer and distribution, while phase change materials help store and gradually release thermal energy, reducing temperature fluctuations and enhancing the stability and efficiency of the thermal and electrical systems.