National Polytechnic School (Algeria)

In smart manufacturing, efficient job shop scheduling (JSS) and resource management remain critical challenges, especially in dynamic production environments. Traditional methods often struggle to adapt to real-time changes and unexpected events. To address these limitations, this paper proposes a novel Generative Adversarial Network (GAN)-based generative AI framework that augments scheduling data with realistic synthetic scenarios and integrates Local Outlier Factor (LOF)-enhanced Q-learning-based reinforcement learning (QRL) for adaptive JSS optimization in Industry 5.0 environments. The GAN is trained to generate realistic synthetic scheduling scenarios, which are combined with real-world data from a state-of-the-art Festo Didactics Cyber Physical Lab to augment the diversity and coverage of training samples. The LOF algorithm enables real-time bottleneck detection, while the QRL agent learns robust scheduling policies that minimize makespan and prioritize bottleneck mitigation. Experimental results demonstrate that the proposed GAN-LOF-QRL approach achieves an average makespan reduction of 70.8% across varying production volumes (12, 15, and 18 orders), significantly improving scheduling efficiency and resource utilization compared to traditional RL and heuristic methods. This research advances smart manufacturing initiatives and Industry 5.0 goals by providing a scalable, adaptive scheduling solution that leverages generative AI to address the complexities of modern supply networks.

Accurate parameter estimation is essential for reliable modeling and performance evaluation of Proton Exchange Membrane Fuel Cells (PEMFCs). This paper introduces the application of Tianji’s Horse Racing Optimization (THRO), a recently developed metaheuristic algorithm, to precisely identify the unknown parameters of a widely used PEMFC model. The proposed THRO-based framework is evaluated on six commercial PEMFC stacks, namely NedStack PS6, Horizon 500W, BCS 500W, 250W, Avista SR-12, and Ballard Mark V, and its performance is benchmarked against five recent metaheuristic algorithms: Flood Algorithm (FLA), Educational Competition Optimizer (ECO), Kepler Optimization Algorithm (KOA), Fata Morgana Algorithm (FATA), and Spider Wasp Optimizer (SWO). Comprehensive comparative and statistical analyses demonstrate that THRO consistently achieves superior parameter identification accuracy, robustness, and solution stability across all tested PEMFC models. In particular, THRO achieves the lowest sum of squared errors (SSE), with values of 2.06, $$1.12 \times 10^{-2}$$, $$1.16 \times 10^{-2}$$, 5.25, 1.056, and 0.813 for the NedStack PS6, Horizon 500W, BCS 500W, 250W, Avista SR-12 , and Ballard Mark V PEMFC stacks, respectively. Additionally, THRO attains an extremely low standard deviation levels, indicating strong convergence reliability and resistance to premature stagnation. The obtained results confirm the effectiveness, robustness, and generalization capability of THRO for PEMFC parameter extraction, highlighting its potential as a reliable optimization tool for PEMFC parameter extraction and energy system applications.

Geothermal energy is a promising technology that can harness an abundant and sustainable resource for large-scale energy generation. This study investigates the thermal behavior of a deep borehole heat exchanger (DBHE) with a depth of 3030 m, located in Algeria. The analysis is based on a numerical model developed using ANSYS-CFX to simulate fluid–soil heat transfer over long operating periods. The borehole is modeled as a U-tube heat exchanger inserted into a stratified geological formation with distinct thermophysical properties. Simulation results reveal that the soil temperature decreases gradually during the first years of operation, and the system reaches quasi-steady-state conditions after approximately five years, as the rate of temperature decline becomes negligible. The influence of flow velocity and inlet temperature on the thermal performance of the system is also investigated. Results demonstrate that increasing the flow velocity enhances heat extraction, while lower inlet temperatures significantly improve the long-term stability of the exchanger. A simulation with a 3030 m depth, 0.01 m diameter, 303.15 K inlet temperature, and 0.5 m/s inlet velocity predicted a 366 K outlet temperature after five years. This work highlights the potential of deep abandoned oil wells in Algeria to be repurposed for geothermal applications, contributing to the diversification of sustainable energy resources.

Efficient path planning is a fundamental requirement for autonomous mobile robots to navigate in complex environments. Particle Swarm Optimization (PSO) provides an efficient search mechanism for path planning. However, its reliance on random population initialization often leads to a suboptimal path and slow convergence. This paper proposes a hybrid PSO-Q learning approach that uses Q-learning path as an initial population for the PSO. By combining the global search capabilities of PSO with the adaptive decision-making of Q learning, the proposed method produces smoother and shorter trajectories while reducing computation time and iterations. Simulation results show that PSO-Q learning performs better than classical PSO with 11% decrease in path length, 31% decrease in computation time and reduction on the iterations. These results show the great potential of joining learning-based guidance into the metaheuristic optimization to achieve efficient and robust autonomous path planning.

The variability in estimating the noise variance can considerably diminish the effectiveness of the energy detection (ED). This study analyzes the performance of a newly introduced goodness-of-fit test called the modified Anderson-Darling (MAD) test, which shows improved statistical power when noise uncertainty is present. We derive and empirically validate the analytical formulations for the theoretical performance of the MAD regarding false alarm and detection probabilities. Additionally, we compare our developed method with existing techniques to assess its performance, including ED, generalized ED (GED), and a two-sample likelihood ratio statistic test. The MAD surpasses the investigated methods without the need for prior knowledge of a particular set of noise samples. Our findings indicate that the proposed spectrum sensing technique also results in reduced computational complexity. Moreover, we propose the idea of spectrum sensing based on channel bandwidth rather than detecting by frequency bin, which is more appropriate for enhancing the efficiency of tactical radio band detection in tactical radio communications.