DSpace at UMBB
The institutional repository of the University M'Hamed Bougara Of Boumerdes is a digital archive that contains the scientific output of the University. Dspace manages, preserves and provides access to the academic works of UMBB.
Communities in DSpace
Select a community to browse its collections.
Recent Submissions
Atmospheric Emissions of Volatile Organic Compounds From a Solvent Paint Sludge - Chemical Identification and Quantification .
(Faculty of Sciences M’hamed Bougara University Boumerdes, 2025) Metali, A; Ziati, M; Camselle, C; Gouveia, S
Solvent based paint sludge contains volatile organic compounds (VOCs) that negatively impact both ecosystems and human health. These VOCs contribute to ground-level ozone formation and are key precursors in the creation of secondary organic pollutants in the atmosphere. There are limited studies that quantify the gaseous compounds released from solvent based paint sludge. In this context, we conducted our experimental study on detection of gaseous emission from paint sludge samples taken from an Algerian paint industry, in both short and long period of storage. Results revealed that paint sludge samples emitted a wide range of VOCs; over 36 compounds were identified, quantified and classified into different chemical groups. Among the compounds identified, aliphatic hydrocarbons, aromatic hydrocarbons, esters, ketones and alkanes. The collected sample from the waste storage area(PS2)contains a higher concentration of oxygenated compounds than those collected in the regeneration station (PS1) and in the waste storage chamber (PS3). Therefore, the gas emitted from paint sludge increased depending on the type and duration of storage. Given that many of the identified VOCs are carcinogenic, toxic, or harmful by inhalation, it is essential to regulate gas emissions during waste management and treatment.
EEG Signal Analysis for Biometric Identification Using Machine Learning
(IEEE, 2026) Cherifi, Dalila; Bentayeb, Ghillasse; Saadsaoud, Abdeldjalil; Boubchir, Larbi
Recent research into the uniqueness of an individual's brain activity indicates that it holds great promise as a foundation for biometric recognition systems. However, despite these encouraging findings, several challenges remain before such technology can be fully realized and implemented. This study investigates the potential of EEG-based identification biometric systems using data collected from 158 healthy subjects at the Max Planck Institute in Leipzig. The proposed experimental pipeline involves preprocessing the raw EEG signals, extracting a variety of features in time, frequency and time-frequency domains including statistical measures, power spectrum of the signal, and Wavelet energy. Following feature extraction, several supervised machine learning algorithms are applied to evaluate the discriminative power of these features for individual identification. Additionally, the impact of dimensionality reduction using Principal Component Analysis was assessed on the extracted features to determine its effectiveness in improving classification performance. The results demonstrate that EEG recordings, reflecting unique brain activity patterns, have significant potential for developing robust identification biometric systems. This research contributes to advancing the understanding of EEG-based biometrics and lays groundwork for future studies to address existing challenges and enhance system accuracy and reliability.
Localization-Guided Semantic Segmentation Framework for Breast Ultrasound Lesions Using BUS_UCLM
(2026) Arioua, Faycal; Touazi, Faycal; Gaceb, Djamel; Benzenati, Tayeb; Telkhoukh, Ibtissem; Magaz, Nour Elhouda
Breast cancer tumor detection and segmentation in ultrasound imaging is crucial for improving early diagnosis
and guiding treatment decisions. This paper presents a two-stage deep learning framework that integrates lesion
localization-guided semantic segmentation to enhance tumor detection in breast ultrasound images. In the first
stage, YOLOv11-n-nano, a lightweight detection model with approximately 2 million parameters, is employed
to localize suspicious regions of interest. In the second stage, advanced segmentation architectures—including
U-Net++, U-Net3+, Attention U-Net, and TransUNet—are applied to refine the segmentation of the localized
areas. The proposed framework is evaluated on the publicly available BUS_UCLM dataset. Our approach achieves
state-of-the-art performance, with DeepLabV3 obtaining a Dice score of 94.31% and an Intersection over Union
(IoU) of 92.10%. These results highlight the effectiveness of localization-guided segmentation in reducing false
positives and improving segmentation accuracy in breast ultrasound imaging
Vacancy-type defects in 6H-SiC single crystals under irradiation
(Springer Nature, 2026) Hazem, Rafik; Izerrouken, Mahmoud; Yahsi, Ugur; Tav, Cumali
Vacancy-type defects in 6H-SiC single crystals and their evolution under irradiation were investigated using positron annihilation spectroscopy. Immediately after irradiation, silicon monovacancies (VSi) aggregate into stable (VSi-VC)3 hexavacancy clusters, which persist across the full dose range, indicating early defect saturation and the absence of further clustering. With increasing fluence, additional VSi are primarily eliminated through recombination with interstitials, which becomes the dominant defect-annihilation mechanism. An increase of the irradiation temperature enhances VSi mobility and accelerates the onset of saturation. These results demonstrate that (VSi-VC)3 represents the final, highly stable aggregated form and radiation-resistant vacancy configuration in 6H-SiC under the present irradiation conditions. This remarkable radiation tolerance of 6H-SiC confirms its suitability for next-generation nuclear systems and radiation-hardened electronic devices
Prediction of interfacial stresses in helicoidal FRP-strengthened RC beams under hygrothermal effects
(Springer Nature, 2026) Nebab, Mokhtar; Amara, Rezki; Ait Atmane, Hassen; Bernard, Fabrice; Bennai, Riadh
The structural integrity of reinforced concrete (RC) members retrofitted with externally bonded composite laminates is primarily governed by complex interfacial stress transfer mechanisms, which are frequently compromised by manufacturing imperfections and aggressive environmental exposure. While conventional fiber-reinforced polymer strengthening is well established, the transition toward bio-inspired helicoidal Bouligand architectures offers a transformative approach to enhancing delamination resistance. However, the interfacial mechanics of these advanced configurations under coupled multi-physical loading remain insufficiently characterized. This study presents a robust computational framework leveraging high-precision artificial neural networks to evaluate interfacial shear and normal stress distributions in RC beams featuring stochastic air-void defects. Utilizing an extensive database of 63,975 data points, a multi-layer perceptron architecture was optimized to capture the high-dimensional nonlinear coupling between constitutive material properties, geometric configurations, and tri-stage loading. The model demonstrates exceptional predictive fidelity, yielding an overall correlation coefficient Rall = 0.99915. Principal results reveal that the bio-inspired helicoidal technique significantly homogenizes interfacial stress distribution, effectively mitigating peak concentrations at the plate ends. Parametric investigations demonstrate that optimizing the reinforcement density (12–32 layers) and fiber architecture facilitates a global reduction in peak interfacial shear and normal stresses by up to 40%. Notably, hygrothermal concentrations escalate debonding driving forces by up to 30%, while the helicoidal configuration attenuates the adverse effects of stochastic air-void defects. This research establishes the bio-inspired technique as a superior, computationally efficient alternative for the reliability-based design and optimization of resilient infrastructure