Search Results (22)

The cathode and anode differential pressure control of a proton exchange membrane fuel cell (PEMFC) directly affects its service life and operating efficiency. Existing control methods find it difficult to cope with strong nonlinear perturbations, and fixed differential pressure control is prone to pressure overshoot and threshold exceedance, resulting in unstable pressure regulation. In order to solve the current research problems, a reinforcement learning method based on hybrid experience replay (HP-TD3) is proposed. A CART-based algorithm is first used to classify the states of the test load, and a load-related segmented reward function is designed. In addition, a hindsight experience replay (HER) mechanism is incorporated into the Priority Experience Replay Twin Delayed Deep Deterministic Policy Gradient (PER-TD3) framework to improve sample utilization efficiency and training stability. Finally, the performance of HP-TD3 and its ability to cope with nonlinear disturbances are verified on a fuel cell control unit hardware-in-the-loop (FCU-HIL) platform. In the A test load (frequent switching and high low-load proportion), the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the degradation index of the fuel cell dynamic performance (Δfc) of HP-TD3 are respectively reduced by 17.4%, 20.5%, and 13.3% compared to P-TD3; in the B test load (high-load operation and low switching frequency), these indicators are reduced by 25.7%, 29.4%, and 15.4% respectively. Full article

Background: Water inertia load training using equipment such as water vests provides unstable resistance that enhances trunk muscle activation. However, practical methods for prescribing exercise intensity without expensive electromyography (EMG) equipment remain limited. This pilot study aimed to develop prediction models for estimating trunk muscle activation using rating of perceived exertion (RPE) during water inertia load exercises. Methods: Seventeen healthy adults (20.45 ± 2.02 years) performed lateral trunk flexion exercises wearing a water vest at five progressive loads (8–16 kg in 2 kg increments). Surface EMG was recorded from four trunk muscles (rectus abdominis, external oblique, internal oblique, erector spinae) and normalized to maximal voluntary isometric contraction (%MVIC). Rating of perceived exertion (RPE) was assessed using the Borg CR-10 scale. Load-dependent changes in muscle activation were examined using repeated-measures ANOVA, and relationships between RPE and EMG were analyzed using regression and linear mixed-effects models. Results: All trunk muscles showed significant increases in activation with increasing load (all p < 0.001, ηp² = 0.381). RPE demonstrated significant positive correlations with all abdominal muscles (r = 0.37–0.46, p < 0.001). Simple regression analyses indicated predictive accuracy (R² = 0.267), representing a 29% increase compared with the strongest individual muscle model. Linear mixed-effects modeling confirmed RPE as a significant predictor after accounting for inter-individual variability. Conclusions: This pilot study provides preliminary evidence that RPE can be used to estimate trunk muscle activation during water inertia load exercise. The proposed composite activation index enhances prescription when EMG measurement is not feasible. Full article

Pneumatic separation can exhibit unstable performance when the feed composition fluctuates while operating parameters remain fixed. This work investigates a perception-informed airflow regulation approach, demonstrated on a representative fibrous–granular mixture case study. We propose LGDNet, a lightweight visual ratio estimation network (0.08 M parameters) built with Ghost-based operations and learned grouped channel convolution (LGCC), to estimate mixture composition from dense images. A dedicated 21-class dataset (0–100% in 5% increments) containing approximately 21,000 augmented images was constructed for training and evaluation. LGDNet achieves a Top-1 accuracy of 66.86%, an interval accuracy of 74.10% within a $\pm 5\%$ tolerance, and an MAE of 4.85, with an average inference latency of 28.25 ms per image under the unified benchmark settings. To assess the regulation mechanism, a coupled CFD–DEM simulation model of a zigzag air classifier was built and used to compare a regime-dependent airflow policy with a fixed-velocity baseline under representative prescribed inlet ratios. Under high impurity loading ($r=70\%$), the dynamic policy improves product purity by approximately 1.5 percentage points in simulation. Together, the real-image perception evaluation and the mechanism-level simulation study suggest the feasibility of using visual ratio estimation to inform airflow adjustment; broader generalization and further on-site validation on real equipment will be pursued in future work. Full article

Maintaining lower-limb joint stability is essential for safe and efficient performance during landing and directional changes. This pilot study examined the effects of a 10-week perturbation-based Dynamic Stability Training (DST) program using an inertial water load on lower-limb joint moments during single-leg landing and a 3-s stabilization phase following a 90° cutting maneuver. Fifteen healthy young men completed DST twice weekly. Three-dimensional motion capture and force-plate data were collected at pre-, mid-, and post-training to compute hip, knee, and ankle joint moments. During landing, hip flexion and abduction moments increased, whereas knee abduction moment decreased. During the stabilization phase, hip flexion, hip rotation, and ankle abduction moments decreased, while knee abduction moment increased. These joint-specific changes suggest potential adaptations in frontal- and transverse-plane control when training with unstable inertial water loads; however, interpretations should remain cautious given the exploratory design and absence of a control group. Larger randomized controlled trials are needed to confirm these preliminary findings. Full article

The generation of realistic electrical load profiles is essential for advancing smart grid analytics, demand forecasting, and privacy-preserving data sharing. Traditional approaches often rely on large, high-resolution datasets and complex recurrent neural architectures, which can be unstable or ineffective when training data are limited. This paper proposes a data-driven framework based on a lightweight 1D Convolutional Wasserstein GAN with Gradient Penalty (Conv1D-WGAN-GP) for generating high-fidelity synthetic 24 h load profiles. The model is specifically designed to operate on small- to medium-sized datasets, where recurrent models often fail due to overfitting or training instability. The approach leverages the ability of Conv1D layers to capture localized temporal patterns while remaining compact and stable during training. We benchmark the proposed model against vanilla GAN, WGAN-GP, and Conv1D-GAN across four datasets with varying consumption patterns and sizes, including industrial, agricultural, and residential domains. Quantitative evaluations using statistical divergence measures, Real-vs-Synthetic Distinguishability Score, and visual similarity confirm that Conv1D-WGAN-GP consistently outperforms baselines, particularly in low-data scenarios. This demonstrates its robustness, generalization capability, and suitability for privacy-sensitive energy modeling applications where access to large datasets is constrained. Full article

Stair ambulation requires precise lower extremity control and postural stability. Middle-aged women with degenerative knee arthritis (DKA) are at an increased risk of falls, yet the effects of unstable load training on their postural stability remain underexplored. This study investigated the effects of a 10-week Aqua Vest-based unstable load training program on postural stability and pain during stair ambulation in middle-aged women with DKA. Thirty participants were randomly assigned to an exercise group (EG) or a control group (CG), with 15 participants in each group. The EG completed a 10-week Aqua Vest training program, while the CG received no intervention. Pre- and post-assessments included spatiotemporal parameters, lower extremity kinematics, and center of pressure (CoP) trajectories during stair descent, as well as knee pain evaluated by a visual analog scale (VAS). After training, the EG demonstrated significant improvements in spatiotemporal and kinematic parameters, reduced ML displacement, lower VAS scores, as well as significant changes in AP CoP parameters. These findings suggest that Aqua Vest-based unstable load training may enhance ML postural stability and alleviate pain in DKA patients, potentially contributing to enhanced balance function and improved stair ambulation safety. Full article

Beach variants of popular sports like soccer and handball have grown in participation over the last decade. However, the characterization of the workload demands in beach sports remains limited compared to their indoor equivalents. This systematic review aimed to: (1) characterize internal and external loads during beach invasion sports match-play; (2) identify technologies and metrics used for monitoring; (3) compare the demands of indoor sports; and (4) explore differences by competition level, age, sex, and beach sport. Fifteen studies ultimately met the inclusion criteria. The locomotive volumes averaged 929 ± 269 m (average) and 16.5 ± 3.3 km/h (peak) alongside 368 ± 103 accelerations and 8 ± 4 jumps per session. The impacts approached 700 per session. The heart rates reached 166–192 beats per minute (maximal) eliciting 60–95% intensity. The player load was 12.5 ± 2.9 to 125 ± 30 units. Males showed 10–15% higher external but equivalent internal loads versus females. Earlier studies relied solely on a time–motion analysis, while recent works integrate electronic performance and tracking systems, enabling a more holistic quantification. However, substantial metric intensity zone variability persists. Beach sports entail intermittent high-intensity activity with a lower-intensity recovery. Unstable surface likely explains the heightened internal strain despite moderately lower running volumes than indoor sports. The continued integration of technology together with the standardization of workload intensity zones is needed to inform a beach-specific training prescription. Full article

The present paper investigates the unexpected unstable failure observed commonly on fiber tows tensile-tested under strain-controlled loading, although the force on the fibers should theoretically be relaxed under controlled strain. A model of the reaction of the load train when the fibers break under strain-controlled conditions is proposed. The criterion for instability is based on the comparison of the filament strength gradient and the overstress induced by the reaction of the load train when the fibers fail. The contribution of multiplet filament failures attributed to the fiber inter-friction and stress waves was taken into account. The compliance of the load train for the test results considered in the present paper was measured. It is shown that, depending on the number of filaments sharing the overload, the values of the structural parameters, and the fiber characteristics, the condition of unstable failure may have been fulfilled by the SiC fiber tows that were tested in house, as discussed in the present paper. The critical parameters that were identified and quantified include the load train compliance, gauge length, fiber stiffness, and bonding of the tow ends. This should allow the proper conditions for stable failure. Important implications for the validity and an analysis of the strengths derived from the unstable fracture of the tows are discussed. Full article

Military and law enforcement members’ shooting ability is influenced by their postural balance, which affects their performance and survivability. This study aimed to investigate the effects of a proprioception training program (standing or walking on unstable surfaces) on postural balance and shooting performance. Twenty participants, divided into two groups, completed 60 shots in a shooting simulator while standing, before and after a 4-week proprioception training program. One group (n = 10) followed the training program (EXP), while the other group followed the regular military academy program (CON). The shooting was conducted under four conditions: without load on a stable surface, with load on a stable surface, without load on an unstable surface, and with load on an unstable surface. The findings reveal that the training program had a significant impact on the EXP, improving their balance (p < 0.01). Additionally, only in the EXP, shooting score and the percentage center of gravity increased (p < 0.01) and the stability of the shots, measured by holding time on the target, doubled from 2.2 to 4.5 s (p < 0.01). These improvements were more pronounced when participants had a load and/or were on an unstable surface. In conclusion, a proprioception training program could be beneficial for improving postural balance and shooting performance. Full article

The purpose of the present study was to examine the effects of a back squat exercise with unstable load (UN) and traditional free-weight resistance (FWR) on subsequent countermovement jump (CMJ) performance. After familiarisation, thirteen physically active males with experience in resistance training visited the laboratory on two occasions during either experimental (UN) or control (FWR) conditions separated by at least 72 h. In both sessions, participants completed a task-specific warm-up routine followed by three maximum CMJs (pre-intervention; baseline) and a set of three repetitions of either UN or FWR back squat exercise at 85% 1-RM. During the UN condition, the unstable load was suspended from the bar with elastic bands and accounted for 15% of the total load. Post-intervention, three maximum CMJs were performed at 30 s, 4 min, 8 min and 12 min after the last repetition of the intervention. The highest CMJ for each participant was identified for each timepoint. No significant increases (p > 0.05) in jump height, peak concentric power, or peak rate of force development (RFD) were found after the FWR or UN conditions at any timepoint. The lack of improvements following both FWR and UN conditions may be a consequence of the low percentage of unstable load and the inclusion of a comprehensive task-specific warm-up. Further research is required to explore higher UN load percentages (>15%) and the chronic effects following the implementation of a resistance training programme. Full article

Currently, the theoretical braking force control mode, characterized by actual deceleration as an unstable open-loop output, is the most widely used brake control mode in trains. To overcome the shortcomings of non-deceleration control modes, a deceleration control mode is proposed to realize the closed-loop control of train deceleration. First, a deceleration control algorithm based on parameter estimation was derived. Then, the deceleration control software logic was designed based on the existing braking system to meet the engineering requirements. Finally, the deceleration control algorithm was verified through a ground combination test bench with real brake control equipment and pneumatic brakes. The test results show that the deceleration control can make the actual braking deceleration of the train accurately track the target deceleration in the presence of disturbances, such as uncertain brake pad friction coefficients, line ramps, vehicle loads and braking force feedback errors, as well as their combined effects, and does not affect the original performance of the braking system. The average deceleration in the deceleration control mode is relatively stable, and the control error of instantaneous deceleration is smaller. Full article

(I) The execution of different sports involves a significant number of throws, jumps, or direction changes, so the body must be as stable as possible while performing a specific action. However, there is no classification of unstable devices and their influence on performance variables. Furthermore, the effect on athletes’ experience using instability is unknown. (II) The aim of this study was to analyze the power and speed parameters in bench press with different loads and unstable executions: (1) stable (SB), (2) with asymmetric load (AB), (3) with unstable load (UB), (4) on fitball (FB) and (5) on a Bosu^® (BB). A total of 30 male participants (15 trained and 15 untrained) were evaluated for mean propulsive speed (MPS), maximum speed (MS), and power (PW) with different types of external load: a low load (40% of 1RM), medium load (60% of 1RM), and high load (80% of 1RM) in each condition. Variables were measured with an inertial dynamometer. (III) The best data were evidenced with SB, followed by AB (3–12%), UB (4–11%), FB (7–19%), and BB (14–23%). There were no differences between groups and loads (p > 0.05) except in the case of MS with 60% 1RM, where trained participants obtained 4% better data (p < 0.05). (IV) Executions with implements and equipment such as fitball and Bosu^® do not seem to be the most recommended when the objective is to improve power or execution speed. However, situations where the load is unstable (AB and UB) seem to be a good alternative to improve stabilization work without high performance. Furthermore, experience does not seem to be a determining factor. Full article

Rock image classification is a fundamental and crucial task in the creation of geological surveys. Traditional rock image classification methods mainly rely on manual operation, resulting in high costs and unstable accuracy. While existing methods based on deep learning models have overcome the limitations of traditional methods and achieved intelligent image classification, they still suffer from low accuracy due to suboptimal network structures. In this study, a rock image classification model based on EfficientNet and a triplet attention mechanism is proposed to achieve accurate end-to-end classification. The model was built on EfficientNet, which boasts an efficient network structure thanks to NAS technology and a compound model scaling method, thus achieving high accuracy for rock image classification. Additionally, the triplet attention mechanism was introduced to address the shortcoming of EfficientNet in feature expression and enable the model to fully capture the channel and spatial attention information of rock images, further improving accuracy. During network training, transfer learning was employed by loading pre-trained model parameters into the classification model, which accelerated convergence and reduced training time. The results show that the classification model with transfer learning achieved 92.6% accuracy in the training set and 93.2% Top-1 accuracy in the test set, outperforming other mainstream models and demonstrating strong robustness and generalization ability. Full article

Transient stability (TS) and short-term voltage stability (STVS) assessment are of fundamental importance for the operation security of power systems. Both phenomena can be mutually influenced in weak power systems due to the proliferation of power electronic interface devices and the phase-out of conventional heavy machines (e.g., thermal power plants). There is little research on the assessment of both types of stability together, despite the fact that they develop over the same short-term period, and that they can have a major influence on the overall transient performance driven by large electrical disturbances (e.g., short circuits). This work addresses this open research challenge by proposing a methodology for the joint assessment of TS and STVS. The methodology aims at estimating the resulting short-term stability state (STSS) in stable, or unstable conditions, following critical events, such as the synchronism loss of synchronous generators (SG) or the stalling of induction motors (IM). The estimations capture the mechanisms responsible for the degradations of TS and STVS, respectively. The paper overviews the off-line design of the data-driven STSS classification methodology, which supports the design and training of a hybrid deep neural network RCNN (recurrent convolutional neural network). The RCNN can automatically capture spatial and temporal features from the power system through a time series of selected physical variables, which results in a high estimation degree for STSS in real-time applications. The methodology is tested on the New England 39-bus system, where the results demonstrate the superiority of the proposed methodology over other traditional and deep learning-based methodologies. For reference purposes, the numerical tests also illustrate the classification performance in special situations, when the training is performed by exclusively using measurements from generation and motor load buses, which constitute locations where the investigated stability can be observed. Full article

The development of electric power systems has become more complex. Consequently, electric power systems are operating closer to their limits and are more susceptible to instability when a disturbance occurs. Transient stability problems are especially prevalent. In addition, the identification of transient stability is difficult to achieve in real time using the current measurement data. This research focuses on developing a convolutional neural network—long short-term memory (CNN-LSTM) model using historical data events to detect transient stability considering time-series measurement data. The model was developed by considering noise, delay, and loss in measurement data, line outage and variable renewable energy (VRE) integration scenarios. The model requires PMU measurements to provide high sampling rate time-series information. In addition, the effects of different numbers of PMUs were also simulated. The CNN-LSTM method was trained using a synthetic dataset produced using the DigSILENT PowerFactory simulation to represent the PMU measurement data. The IEEE 39 bus test system was used to simulate the model under different loading conditions. On the basis of the research results, the proposed CNN-LSTM model is able to detect stable and unstable conditions of transient stability only from the magnitude and angle of the bus voltage, without considering system parameter information on the network. The accuracy of transient stability detection reached above 99% in all scenarios. The CNN-LSTM method also required less computation time compared to CNN and conventional LSTM with the average computation times of 190.4, 4001.8 and 229.8 s, respectively. Full article