Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mohcin Mekhfioui has a total of 52 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    LLM-Based Supervisory Framework for Context-Aware MPPT Algorithm Selection in Photovoltaic Systems

    by Amal Satif, Abdelhadi Zineddine, Mohcin Mekhfioui, Rachid Elgouri

    Maximum power point tracking (MPPT) is essential for maximising energy extraction from photovoltaic (PV) systems. Classical algorithms such as Perturb & Observe (P&O) and Fuzzy Logic Control (FLC-P&O) perform effectively under uniform and transient irradiance conditions, respectively, but fail under partial shading due to local maximum power point trapping. Metaheuristic approaches such as the Grey Wolf Optimizer hybrid (GWO-P&O) can escape local optima under partial shading but exhibit increased convergence time and unnecessary search overhead under uniform conditions. Since no single MPPT algorithm performs optimally under all operating conditions, this paper proposes an LLM-based adaptive MPPT selector that dynamically dispatches the most suitable algorithm according to real-time PV operating conditions. Using structured irradiance descriptors and contextual system information, the LLM performs zero-shot classification of the operating scenario and selects the corresponding MPPT strategy. In contrast to fixed threshold-based switching approaches, the proposed selector offers flexible and extensible supervisory decision-making without requiring redesign when new algorithms or operating conditions are introduced. Simulation results obtained on a 400 W, 4S×2P PV array connected to a boost converter under different irradiance scenarios demonstrate that the proposed selector achieves an average tracking efficiency of 98.7%, outperforming fixed P&O (86.8%), fixed FLC-P&O (91.1%), and fixed GWO-P&O (90.5%) deployments.

  • Open Access

    Machine Learning Operations on ZYNQ FPGA Board for Real-Time Face Recognition

    by Bouchra Kouach, Mohcin Mekhfioui, Rachid El Gouri

    Nowadays, MLOps approaches are gaining popularity thanks to their ability to apply DevOps best practices to machine learning models. They enable the automation and optimization of model training, deployment, and monitoring in various environments, while ensuring effective Continuous Integration/Continuous Deployment (CI/CD). These approaches thus promote real-time applications that can react quickly and improve continuously. This paper examines the feasibility of implementing MLOps practices in embedded systems, specifically on the Zynq-7000 FPGA board. We present a comprehensive MLOps architecture that enables the automated deployment and monitoring of a convolutional neural network model for face recognition on an embedded hardware platform for datacenter physical access control scenarios. This architecture integrates GitLab CI/CD for version control and pipeline automation, MLflow for experiment tracking and model lifecycles management, Prometheus and Grafana for monitoring, and data storage in an S3 Bucket cloud connected to DVC for dataset versioning. The results demonstrate that the proposed pipeline can be effectively deployed on a Zynq-7000 FPGA board enabling automated model retraining, redeployment, and performance monitoring. This approach reduces operational complexity and supports faster adaptation to dataset changes.

  • Open Access

    Real-time fetal arrhythmia detection using deep learning on an embedded fECG monitoring system

    by Mohcin Mekhfioui, Maroua Bouksaim, Nabil El Bazi, Oussama Laayati, Amal Satif, Chaïmaâ Kissi, Tarik Boujiha, Ahmed Chebak


  • Open Access

    A Comprehensive Decade-Long Review of Advanced MPPT Algorithms for Enhanced Photovoltaic Efficiency

    by Maroua Bouksaim, Mohcin Mekhfioui, Mohamed Nabil Srifi

    Photovoltaic energy has become a key pillar in the transition to sustainable energy systems, driven by the need for efficient energy conversion and the reduction of dependency on fossil fuels. Maximum Power Point Tracking (MPPT) is central to optimizing the performance of photovoltaic systems by ensuring the maximum extraction of solar energy, even under fluctuating environmental conditions. This review provides a comprehensive analysis of MPPT algorithms developed and refined over the past decade (2015–2025), highlighting major breakthroughs in algorithmic approaches, from conventional methods such as Perturb and Observe (P&O) and Incremental Conductance (IncCond) to more advanced techniques incorporating artificial intelligence, fuzzy logic, and hybrid systems. The paper evaluates the evolution of MPPT techniques, focusing on their effectiveness in real-world applications, particularly in optimizing photovoltaic output under diverse operating conditions such as partial shading, temperature variations, and rapid irradiance changes. Furthermore, it discusses the ongoing challenges in the field and the promising directions for future research, aiming to further enhance the reliability and efficiency of solar power systems worldwide.


  • Open Access

    FPGA-Based Implementation of Convolutional Neural Networks for Enhanced Physical Security in Data Center Door Access Systems

    by Bouchra Kouach, Mohcin Mekhfioui, Azzedine El Mrabet, Rachid El Gouri

    Digital data security has become a critical topic, whereas the physical security of the hardware on which digital data is stored remains equally critical. The loss of a server is equivalent to the loss of data. Access to the data center rooms is provided using badges assigned to each individual. However, these badges are at risk of being lost or stolen, which can lead to unauthorized access. This represents a risk of the loss of physical equipment and, consequently, of the loss of confidential data. This study proposes an intelligent security system for access doors in data center rooms to ensure optimal security measures, based on CNN image classification applied on a Xilinx Zynq FPGA board using the PYNQ framework. Python was used to develop and train CNN models for image classification, exploiting frameworks such as TensorFlow and Keras. The results demonstrate that Deep Learning (DL) models can be applied on a ZYNQ FPGA board to optimize inference time and highlight Faster R-CNN as the most effective model for image classification, contributing to strengthening the physical security of building access.


  • Open Access

    Integrating A Six-Axis Robotic Arm And Fused Deposition Modeling Extruder For An Enhanced Robotic Additive Manufacturing Process

    by Fatima Ezzahra Hammouch, Mohcin Mekhfioui, Rania Majdoubi, Omar Mouhib, Marouan Bouchouirbat, Ahmed Chebak


  • Implementation of Deep Learning models on an FPGA board for voice classification

    by Kouach Bouchra, Mekhfioui Mohcin, El Gouri Rachid

    This paper presents an implementation of an intelligent door security system, based on voice classification of people requesting access to datacenter rooms, by applying deep learning models (CNN and RNN) implemented on an FPGA board. Our approach is to use the ZYBO Xilinx-Zynq7000 FPGA board, which integrates the PYNQ framework and enables the use of the PetaLinux operating system, to facilitate the deployment of our deep learning models on the FPGA board. We also use deep learning frameworks such as TensorFlow and Keras for model development and training. We add the MFCC (Mel Frequency Cepstrum Coefficient) layer to extract parameters features to describe the voice in form of data. The results of this work demonstrate that the ZYNQ FPGA board is well suited to parallel computation of deep learning models (CNN and RNN) for real-time applications, with the highest accuracy.


  • Open Access

    Optimized Digital Watermarking for Robust Information Security in Embedded Systems

    by Mohcin Mekhfioui, Nabil El Bazi, Oussama Laayati, Amal Satif, Marouan Bouchouirbat, Chaïmaâ Kissi, Tarik Boujiha, Ahmed Chebak

    With the exponential growth in transactions and exchanges carried out via the Internet, the risks of the falsification and distortion of information are multiplying, encouraged by widespread access to the virtual world. In this context, digital image watermarking has emerged as an essential solution for protecting digital content by enhancing its durability and resistance to manipulation. However, no current digital watermarking technology offers complete protection against all forms of attack, with each method often limited to specific applications. This field has recently benefited from the integration of deep learning techniques, which have brought significant advances in information security. This article explores the implementation of digital watermarking in embedded systems, addressing the challenges posed by resource constraints such as memory, computing power, and energy consumption. We propose optimization techniques, including frequency domain methods and the use of lightweight deep learning models, to enhance the robustness and resilience of embedded systems. The experimental results validate the effectiveness of these approaches for enhanced image protection, opening new prospects for the development of information security technologies adapted to embedded environments.