In academia, research on gear units is increasingly focused on electrification, higher rotational speeds, lightweight architectures and stricter efficiency and acoustic constraints. Advanced multiphysics modelling now incorporates contact mechanics, structural flexibility, lubrication, thermal behaviour and electromechanical coupling. Significant advances have been made in the fields of dynamics, vibroacoustic, Rolling Contact Fatigue modelling and coupled power losses and thermal analyses. Furthermore, numerical tools facilitate analyses at the system-level as opposed to the component level. Within the industrial sector, transmissions are required to demonstrate an increase in power density, durability and efficiency with reduced noise, a requirement that is particularly pertinent in the context of electric transmissions. At the same time, manufacturing trends include enhanced precision, surface superfinishing, dry machining and full digitalisation with digital twins. Health monitoring is evolving towards hybrid strategies. Sustainability considerations now influence materials, lubrication strategies, eco-design, repairability and life cycle optimisation. In the field of education, transmission engineering remains fundamental. It is evolving towards systemic, multiphysical and digital approaches. Study programmes are increasingly incorporating tribology, Noise, Vibration & Harshness, thermal modelling, durability prediction, optimisation and data-based monitoring, alongside close links between research laboratories and industry. Digital skills, simulation tools and interdisciplinary training are becoming essential. Overall, far from being a mature or declining field, gear transmission engineering is undergoing a profound transformation. Its future is driven by high-speed electrified systems, integrated modelling, predictive maintenance, sustainable design and enhanced collaboration between research, industry and education.
Determining the characteristics of impact forces is a key component of structural health monitoring systems, as it allows for continuous assessment of the structural integrity of engineered systems. This study addresses two complementary inverse problems related to impact force identification. The first focuses on the reconstruction of the impact force when the impact location is known, while the second deals with the localization of the impact when the position of excitation is unknown. An analytical model was developed to compute the transfer matrix, taking into account different boundary conditions, represented by both rigid and elastic supports, to evaluate their influence on reconstruction accuracy. The impact force was reconstructed by using the regularization methods and comparing two techniques: the TGSVD method and the Tikhonov method. For the latter, two optimization strategies for the regularization parameter λ were considered. The results demonstrate that the S-curve approach provides a more accurate and stable estimation of the impact force. For the localization problem, an optimization-based approach was implemented to estimate the impact position by minimizing the discrepancy between measured and predicted responses. The findings highlight the significant influence of support stiffness on both the accuracy of force reconstruction and the precision of impact localization, emphasizing the importance of appropriate boundary modeling in inverse dynamic analysis.
Phase-sensitive optical time-domain reflectometry (Φ-OTDR) offers advantages such as a simple structure, multi-point vibration localization, and long-distance disturbance detection in optical fiber networks. However, accurately distinguishing diverse environmental vibration events remains challenging. In this study, experiments were conducted using a distributed optical fiber vibration sensing system. Seven common types of environmental vibration events were simulated, and the corresponding signals were collected. Ten time-domain and frequency-domain features were extracted to construct a 10-dimensional feature vector for model training and classification. An improved decision tree optimal ensemble (IDTOE) method was introduced to optimize the random forest model, resulting in the proposed IDTOE-RF classifier. Model performance was evaluated using confusion matrices, accuracy, recall, and F1-score. The IDTOE-RF model outperformed the conventional random forest and support vector machine (SVM) models, achieving an average recognition accuracy of 93.46%, which was 3.84 percentage points higher than that of the conventional random forest model. The proposed method demonstrates good statistical stability and practical applicability for perimeter security monitoring.
Many current prediction methods used in power system load forecasting are adversely affected by an imbalance of feature distributions at different time scales, yielding constrained prediction accuracy and elevated error rates. This paper introduces a multiscale prediction method integrating a multiscale attention convolutional network with a long short-term memory (LSTM) network. First, historical charging-load data from charging stations are obtained and structured into distinct temporal segments for multiscale convolutional feature learning. The multiscale attention convolutional network employs parallel convolutional branches with varying dilation depths to process the input sequence, integrating the outputs via feature concatenation and channel weighting to form unified conceptual representations. An LSTM network models the multilevel time-series data derived from the various convolutional filters, enabling predictions across multiple time scales. The attention mechanism assigns weighted importance scores to the fused features based on temporal relevance prior to sequence modeling, improving prediction stability across various load periods for electric vehicle charging loads. The multi-time-scale prediction experiment demonstrates consistent accuracy at the medium and long time scales, where the mean absolute percentage error across all periods does not exceed 5%, and the coefficient of determination remains stable above 0.9. Through ablation experiments, this paper validates that the integrated multiscale attention convolutional network and LSTM framework yield a mean absolute error of 8.8 kW and a root mean square error of 15.5 kW during peak periods, outperforming single-scale architectures. Accordingly, the model provides an accurate, quantifiable foundation for power system load forecasting.
To better understand the application of computational fluid dynamics (CFD) and genetic algorithms (GAs) in indoor ventilation systems, the author proposes a study based on digital twin simulation and optimization: CFD and GAs for indoor ventilation systems. The author first analyzes the basic concepts, application prospects, technological connotations, and development trends of digital twin technology in the fields of complex industrial systems and complex equipment. Second, a classroom model is established through CFD, and relevant data are obtained. The BP neural network possesses strong nonlinear mapping capabilities and robustness, thereby meeting the requirements for fitting complex fluid simulation data. A substitute model for the CFD model is established using the BP neural network, and a GA objective function is formulated based on predicted mean vote indicators and air-age data. Different weights are set to optimize the model and then compared with the original CFD model. The results show that the CFD-coupled GA model takes only 1–3 h. The combination of CFD and GA can effectively replace the model optimized by directly calling the CFD program within the GA, reducing computation time and improving indoor air quality.
Ballistic protection devices aim to intercept projectiles and mitigate their kinetic energy in order to limit bodily harm. However, even with this mitigation, a shock wave can propagate through the protection to the human body. The emission of shock waves can cause rapid expansion followed by cavitation in biological fluids, resulting in bubble nucleation and growth. Consequently, the unstable implosion of cavitation bubbles, which can be potentially devastating, remains partially misunderstood regarding its effects on human tissues. The objective of this research is to better understand the mechanical loads related to this phenomenon. The main objective is to identify the factors that influence cavitation thresholds, bubble dynamics, and their implosion in order to assess their impact on human tissues. Laboratory tests are performed using a gas launcher that shoots steel balls at intermediate velocity (100–200 m/s) onto a steel plate associated with a liquid. This experiment offers the possibility to monitor impact conditions while observing the creation and collapse of cavitation bubbles through a high-speed camera. Pressure sensors simultaneously record dynamic pressure changes in the fluid near the cavitation region. Initial trials validated the existence of post-impact cavitation, with bubbles visualized by fast imaging. Pressure analysis enabled the identification of factors that influence bubble creation, including impact velocity, initial pressure, and the pressure peak produced during implosion. The analysis also enabled the determination of bubble lifetime as well as the temporal evolution of their radius, providing essential data to improve understanding of the phenomenon.
Architectured materials–also referred to as architected materials in the literature–have experienced significant growth over the past two decades, largely driven by rapid advances in fabrication techniques at the mesoscopic scale, ranging from a few hundred microns to several centimeters. These technological developments have stimulated parallel progress in numerical and theoretical modeling, considerably expanding the accessible design space. Beyond the classical stiffness-to-weight paradigm, attention has extended to properties such as yield strength, buckling behavior, and multifunctional responses, including thermo-electro-magneto-mechanical couplings. The concept of mesoscopic architecture has further been combined with composite materials whose microstructure lies at the micron scale, resulting in systems governed by three intrinsic length scales: micro-, meso-, and macroscales. This multiscale interplay offers an even broader and more versatile design landscape. Although research activity in architectured materials is thriving at the laboratory scale, their large-scale industrial deployment and integration into everyday applications remain in early stages. Despite their remarkable ability to achieve properties rarely found in natural materials, challenges related to design methodologies, manufacturability, scalability, and reliable numerical analysis still need to be addressed. This article provides a broad perspective on the field, with particular emphasis on the contributions o f the French research community over the past 15 years.
In this study, a hybrid structure with an embedded rhombicuboctahedral core was fabricated using a material extrusion process, and its compressive behavior was investigated. Polylactic acid (PLA) was used as a stiff internal core, while thermoplastic copolyester (TPC) served as a compliant matrix. The hybrid structure exhibited a specific compressive strength and mass-normalized energy absorption approximately 1.46 times and 1.74–1.86 times higher than that of the TPC-based cubic structure. While the TPC structure showed a typical plateau behavior associated with progressive collapse, the hybrid structure demonstrated a higher initial stiffness and a continuously increasing stress response with noticeable stress fluctuations due to the sequential yielding of the PLA core. This behavior reflects a brittle–ductile synergistic effect, where the TPC matrix enables energy absorption and the PLA core enhances load-bearing capacity. These findings highlight the potential of multi-material additive manufacturing for designing mechanically efficient hybrid structures.
Vehicle restraint systems (VRS) are components installed on roadsides or freeways to ensure the safety of road users. The validation of a VRS is conditioned by crash tests in accordance with standards EN 1317-1 and EN 1317-2. Due to the high cost of crash testing, manufacturers rely on their experience, analytical calculations, and upstream numerical simulations to design VRS that meet the standards before experimental validation. However, these simulations can have a high computational cost. Therefore, the aim of this study is to propose a simplified model of a wooden VRS, close to a reference numerical model, enabling a significant reduction in computational time. The proposed design comprises rails and posts connected by joints, with a solid wood rail made up of 2-m-long elements. Numerical simulations show that this design satisfies the TB32 test, corresponding to a 1500 kg vehicle launched at 110 km/h with an impact angle of 20°. However, as the computational time is strongly influenced by the assemblies, a reference model with a continuous rail has been proposed, constituting a first simplification. The results indicate that only the parts of the VRS close to the impact zone are subject to high levels of deformation. Consequently, it is possible to simplify the parts remote from this zone in order to further optimize the computational time. Three approaches are proposed to represent these parts: the LS model replaces the remote zones with longitudinal springs; the LTS model uses longitudinal and transverse springs; and the beam (B) model uses the remote parts using beam-type finite elements. All three simplified models are validated and are able to reproduce the reference model with a continuous rail. The EN 1317-1 and EN 1317-2 criteria and the computational time show that the B model is closer to the reference model than the other two. In particular, the computational time is reduced by a factor of approximately 1.8 for the simplified models, while maintaining good agreement with the reference results. The B model, therefore, provides a good representation of the reference model at a lower computational cost.
The dynamic simulation of traditional electromagnetic gear transmission systems often depends on complex multi-physics modeling and intensive computational resources, which limits efficient prediction and rapid design iteration in engineering practice. To address this challenge, this paper proposes a multiscale dynamic simulation method based on a conditional generative adversarial network to efficiently predict the response behavior of electromagnetic gear transmission systems under diverse operating conditions. First, a high-dimensional dataset containing key parameters of electromagnetic gears (such as air gap spacing and permanent magnet arrangement) and their corresponding dynamic responses is generated through finite element simulation. Then, a deep convolutional generator with a fused residual structure is designed, and a physical consistency loss function based on the conservation of energy is introduced to enhance the physical interpretability of the generated results. A multiscale discriminator architecture is adopted to improve the discrimination capability of dynamic features across different temporal and spatial resolutions. Experimental results demonstrate that the proposed cGAN model achieves an average relative error of 5.2% in dynamic response prediction under typical working conditions, with a single prediction time of less than 0.08 s. The average relative error is measured relative to high-fidelity finite element simulation results, underscoring the model’s capability to achieve significant computational speedup while retaining high predictive accuracy. The method thus satisfies the dual requirements of high precision and efficiency, providing a feasible technical solution for intelligent modeling and rapid simulation of electromagnetic gear transmission systems.
Electricity generation relies on a set of complex facilities where mechanical engineering plays an essential role in ensuring the performance, reliability, safety, and sustainability of systems. This document presents a non-exhaustive overview of the contribution of mechanics in the nuclear, hydraulic, and wind power sectors, from a fundamental understanding of the phenomena to advanced methods of design, experimentation, and numerical modeling. We show how fluid mechanics, structural mechanics, and materials science complement each other in order to design, dimension, control, and optimize the components and structures necessary for electricity generation. This complementarity of mechanical sciences is made possible by advanced scientific work providing detailed knowledge of flows, heat transfer, mechanical stresses, vibration phenomena, and damage mechanisms, which occur in various contexts such as reactor cooling, dam resistance, wind turbine stability, and nuclear fuel performance. This summary highlights an important feature of mechanical sciences, namely the parallel evolution of experimental and numerical approaches, which complement each other in understanding the complex phenomena affecting energy facilities. Scale model testing, real-world measurements, and modern imaging techniques provide data that is essential for validating three-dimensional simulations. Numerical models, on the other hand, make it possible to explore extreme conditions that are difficult to reproduce in the laboratory and to test multiple design variants. These models are becoming bigger thanks to high performance computing and allow today chaining or coupling different physics and scales. The decarbonization of energy and the resulting increase in electricity production are associated with several scientific challenges that must be addressed: realistic consideration of dynamic phenomena, improvement of physical models, control of material aging, management of fluid-structure interactions, simulation of two-phase flows, and evaluation of uncertainties in calculations. These challenges will be met more easily if industry and research actively collaborate to maintain a high level of innovation and thus guarantee the safety of production facilities.
Tailless blended wing body (BWB) mini-UAVs offer important aerodynamic and structural advantages, but the absence of conventional tail surfaces makes inherent static stability difficult to achieve during conceptual design. This study proposes an integrated framework for early-stage static stability assessment of tailless BWB mini-UAVs by combining XFLR5 aerodynamic analysis, Latin Hypercube Sampling (LHS), and Gaussian Process Regression (GPR). Seven geometric variables describing chord distribution, sweep, twist, and dihedral are sampled within prescribed ranges, and XFLR5 is used to compute the key longitudinal and lateral-directional stability derivatives Cmα, Cm0, Clβ, and Cnβ. GPR surrogate models are then trained to predict these derivatives and enable rapid exploration of the design space. Cross validation shows that an LHS size of 50 provides the best compromise between predictive accuracy and computational cost among the tested sampling levels. Sensitivity analysis indicates that root chord and root-tip twist difference dominate longitudinal stability, outward dihedral mainly influences roll stability, and kink chord strongly affects directional stability. The framework also identifies feasible tailless BWB configurations satisfying the static stability criteria. The main contribution of this work is a unified and computationally efficient methodology for multi-axis static stability screening of tailless BWB mini-UAVs at the conceptual design stage.
Physics-Informed Neural Networks (PINNs) have recently emerged as a powerful framework for solving forward and inverse problems involving partial differential equations, by embedding physical laws directly into the training process of neural networks. A key advantage of PINNs lies in their ability to infer hidden or hard-to-measure physical quantities from limited data. In this work, we explore the capability of PINNs to investigate dry granular flows, complex systems that are challenging to study due to their non linear rheology and the high computational cost of traditional simulations. To describe the flow dynamics, we adopt the μ(I) rheology, a well-established constitutive model for granular materials. We propose an approach based on a simple and accessible experiment: the collapse of a granular column. Using synthetic data generated from this setup, we demonstrate that PINNs can accurately reconstruct the pressure field and reliably identify key rheological parameters such as μs and μ2. This study highlights the potential of PINNs as a low-cost, flexible tool for rheological characterization of granular materials, potentially replacing more complex and expensive experimental protocols, like rheometers.
The aim of this study was to develop a 3D-printed plantar orthosis composed of a single, recyclable material without a covering layer. This study had two objectives: the protection of the podiatrist's health during the creation of the orthosis and the reduction of the orthosis's environmental impact. Indeed, currently podiatrist realize thermoformed orthotics using many manual and arduous operations, as well as the handling of numerous glues and toxic substances. In this paper, the developed orthosis mimics the geometrical and mechanical properties of a thermoformed orthosis including specific pores in the orthosis. Firstly, the materials composing the thermoformed orthosis were characterized from a mechanical point of view. Secondly, specific isotropic pores were applied to a solid TPU block. The size of these pores was calculated using finite element analysis to reproduce the behavior of each material of the thermoformed orthosis. Thereafter, a 3D-printed orthosis was produced and an initial trial with one patient yielded promising outcomes; however, further research involving a larger group of patients is required to confirm these findings. Finally, a fatigue test was conducted on a representative sample, showing that the defined solution appears to withstand one year of use. In the future, the goal is to no longer use thermoformed orthosis but to directly design the orthoses using computer-aided design software. This research opens new possibilities for designing personalized plantar orthosis, enabling the adjustments of local mechanical characteristics to address specific pathological requirements.
The present paper revisits recent challenges in computational mechanics where data-driven modeling offers unexpected possibilities. For that purpose, the main concepts related to data and learning are first introduced. Then, physics-based, data-driven, and hybrid modeling approaches in the different domains of mechanics: solids and structures, fluids and flow, and processing and manufacturing will be addressed. Finally, technology needs, recent advances, and remaining challenges in the industrial sector will be highlighted.
Rolling bearings operating at high rotational speeds are subject to frictional losses that generate significant self-heating and may affect performance and durability. Predicting temperature fields in sealed wheel bearing assemblies is therefore essential for design and thermal management. In this work, an experimentally calibrated thermo-mechanical finite element framework is developed to simulate the thermal behavior of complete third-generation wheel bearings, including rolling contacts, sealing interfaces, and the surrounding test bench environment. Heat generation is introduced from measured friction torque and distributed at the ball–raceway and seal–hub contacts, while conduction and convection mechanisms are implemented through dedicated thermal boundary conditions. Convective parameters are identified from controlled cooling tests. The model is validated using a dedicated bench equipped with dynamic torque and infrared thermography measurements. Simulations are assessed on two-wheel bearings with significantly different geometries, without any additional parameter recalibration. Predicted steady-state temperatures show good agreement with experiments, with deviations remaining below 3%–5%. The proposed FEM methodology provides a transferable tool for analyzing internal temperature distributions in bearing components and supports predictive thermal design of wheel bearing systems.
This study proposes fractional models to describe the linear viscoelastic behavior of polymethyl methacrylate over a wide range of frequencies and temperatures. The objective of this work is to investigate the efficiency of the fractional model in the identification of the viscoelasticity of the polymer. The experimental data were obtained based on the dynamic mechanical analysis over a limited frequency range at different temperatures. The time–temperature superposition principle was utilized to extend the experimental data across an expanded frequency range. The validity range of this approach was confirmed through the Cole–Cole plot. A master curve for polymethyl methacrylate at a reference temperature was constructed by shifting the dynamic mechanical analysis curves acquired at different temperatures along the frequency axis. The horizontal shift factors were efficiently fitted using the Williams–Landel–Ferry equation. The experimental data, which failed to affirm the time–temperature superposition principle, were precisely characterized using the fractional element model. The master curve was accurately characterized by employing the fractional Zener model. A good agreement between the numerical models and experimental data was achieved. The efficiency of these models was validated by error estimation. The superiority of the fractional models was substantiated through comparative analysis with the integer models. The fractional model was confirmed to be accurate for the prediction of the viscoelastic behavior of polymethyl methacrylate. The Williams–Landel–Ferry equation can be incorporated into the fractional models to address the temperature-dependent viscoelastic properties of the material.
The results presented here concern studies related to air treatment and ventilation flows in habitable enclosures. These studies fall within a broader context involving human health and well-being, comfort, indoor environmental quality, and energy savings. In certain configurations, impinging jets used in such enclosures can generate whistling noises, which are perceived as acoustic nuisances. To address this issue, the characteristics of the vortex structures generated by jets interacting with ventilation openings and the noise produced by a jet impinging on a slotted surface were studied experimentally. Experiments were conducted for an impingement distance of 4 cm and two Reynolds numbers, 4700 and 4800. A dedicated experimental setup was designed for this study to enable a quantitative analysis of the correlations between vortex dynamics upstream and downstream of the slotted plate and the radiated acoustic field, using a dual stereo-PIV configuration combined with microphones. The results show that, when transitioning from Re = 4700 to Re = 4800, the acoustic pressure level drops by 8 dB. Spectral analysis reveals a shift from a self-sustained tonal feedback loop at a single frequency (204 Hz) to a dual-frequency loop (129 and 270 Hz). The analysis of vortex dynamics, using the Lambda-2 criterion, indicates that this drop is associated with a transition from a symmetric to an antisymmetric vortex organization. The spectral analysis of velocity signals extracted from the S-PIV measurements makes it possible to interpret all acoustic frequencies and identify the sources of the acoustic noise. The use of correlation functions between acoustic and velocity signals confirms the presence of aeroacoustic coupling.
Peridynamics is a nonlocal extension of classical continuum mechanics and is increasingly used to solve fracture mechanics problems. However, some issues remain, such as its dispersion characteristics and the use of the constant micromodulus. The introduction of the weighted or kernel functions can effectively address these issues. In this work, several micromodulus functions in the bond-based peridynamics approach are used to explore the influence of the kernel functions on wave dispersion, as well as on the evaluation of dynamic stress intensity factors (DSIFs) and crack propagation. First, a wave dispersion analysis for a 1D problem is performed for different kernel functions. Then, Mode-I and Mode-II DSIFs are computed. The DSIFs are calculated from the displacement field in the vicinity of the crack tip using the displacement extrapolation method. Finally, the Kalthoff–Winkler benchmark is simulated to assess the effect of the kernel functions on dynamic crack propagation.
Natural gas pipeline network emergencies are frequent. Existing monitoring and dispatching systems are mostly static and rule-driven, making them difficult to adapt to changing environments and coupled risks and lacking self-learning and adaptive optimization capabilities. To address this, this paper constructs a dynamic control system for natural gas pipeline emergencies based on reinforcement learning. First, SCADA (Supervisory Control and Data Acquisition) and sensor data fusion are used to achieve multi-source state perception and establish a temporal state space suitable for training. A double deep Q-network intelligent decision engine is then introduced to learn stable policies through experience replay and a target network. A state-action adaptive mapping mechanism is designed to achieve intelligent adjustment of valves, pressure, and flow under different emergency levels. A multi-objective reward function is constructed by combining safety, timeliness, and energy consumption to achieve dynamic system balance. Finally, a visual control platform based on Python and TensorFlow is developed to complete the closed-loop optimization from data perception to policy execution. Experiments show that the average response time of the proposed method is only 0.96 s, a significant improvement over traditional method. After training, the valve control stability index reaches 0.98, and the adjustment time is shortened to 2.1 s. Under level 6 emergency conditions, the safety retention rate still reaches 91.7%, energy consumption is reduced by 13.2%, and the average reward under complex disturbances is 75.8, verifying its high efficiency and robustness in dynamic regulation.
