Search Results (8,998)

Addressing the critical challenge of signal degradation in biosensor cardiac monitoring, this paper introduces an efficient hybrid dilated–long short-term memory (LSTM) with residual learning (HDLR), which is a novel architecture engineered for a high-fidelity electrocardiogram (ECG) denoising signal. The proposed HDLR model synergistically integrates with dilated convolutions to expand the receptive field for multi-scale feature extraction. This is an LSTM-based backbone used to resolve long term temporal dependencies with residual learning paths to stabilize gradient flow and accelerate convergence. The proposed HDLR architecture integrates three core functional components with dilated convolutional layers utilized for local temporal feature extraction, where varying dilation rates expand the receptive field to capture both local waveform patterns and broader morphological structures without increasing computational complexity. Experimental results demonstrate a significant leap in performance, with the HDLR model achieving a mean squared error (MSE) of 0.002176, a signal-to-noise ratio (SNR) of 14.4420 dB, and a Matthews correlation coefficient (MCC) of 0.9822. Beyond quantitative metrics, the proposed HDLR architecture exhibits exceptional robustness in preserving cardiac morphology, specifically the P-wave and QRS complex of the ECG signal under stochastic noise conditions. These findings underscore the HDLR model’s potential as a backbone for next generation, real time diagnostic systems in intelligent healthcare. Full article

Dynamic target tracking is a challenging task for autonomous ships due to the continuous variation of relative motion states and the requirement for coordinated control of position, heading, and speed. This paper proposes a task-oriented continuous decision-making framework based on Soft Actor-Critic (SAC) reinforcement learning for the autonomous approach and stable following of dynamic target vessels. A finite-history-enhanced SAC framework is developed by incorporating LSTM-based sequence encoding into the policy and value networks to capture recent evolution patterns of target motion and own-ship maneuvering responses. Furthermore, a sector-annular tracking region defined by distance and relative bearing constraints is constructed and a multi-component reward function is designed to integrate distance convergence, heading adjustment, region maintenance, speed matching, and control smoothness into policy learning. Simulation experiments under straight-line motion, curved motion, and randomized initial conditions demonstrate that the proposed SAC-LSTM method achieves improved task completion capability and control quality compared with SAC, PPO, and DDPG under the same task settings. Compared with standard SAC, SAC-LSTM improves the average success rate by 7.5 percentage points, reduces the average episode length by approximately 22.0%, and decreases the average heading error by approximately 35.5%. Additional sequence-length analysis, reward-component ablation, and multi-level disturbance tests further validate the effectiveness of the proposed design. The results indicate that the proposed method provides an effective solution for continuous decision-making in dynamic target tracking tasks. Full article

A fractional-order equivalent circuit model (FOECM) provides a compact and physically interpretable representation of the memory-dependent polarization behavior of lithium-ion batteries. Leveraging this property, a fractional order model-guided residual-correction framework is proposed for state-of-charge (SOC) estimation, in which the FOECM, unscented Kalman filter (UKF), and long short-term memory (LSTM) residual learner are integrated into a unified estimation chain rather than treated as separate modules. In this framework, the FOECM is parameterized using Dynamic Stress Test (DST) data and incorporated into the UKF to construct the FOECM + UKF estimator. The LSTM learns the history-dependent SOC residual from sequences of measured operating signals and FOECM + UKF SOC estimates, and its output is added to the UKF estimate without replacing the fractional-order physical model. The proposed hybrid estimator is trained and configured using the available DST, Supplemental Federal Test Procedure (US06), and Federal Urban Driving Schedule (FUDS) data, and is independently evaluated on the US06 and FUDS profiles of Cell 008. Compared with the FOECM + UKF estimator, the proposed hybrid estimator reduces the SOC root mean square error (RMSE) from 2.70% to 0.90% on US06 and from 2.93% to 1.20% on FUDS, with mean absolute error (MAE) values of 0.66% and 1.03%, respectively. These results demonstrate the effectiveness of coupling fractional-order memory modeling with sequence-based residual correction under the tested dynamic operating profiles. Full article

With the global shift toward green energy, solar photovoltaic (PV) power has expanded rapidly. However, the unpredictable nature of PV generation, caused by changing weather conditions like irradiance and temperature, challenges grid stability and power scheduling. Therefore, developing smart forecasting models for high-precision PV power prediction is essential for modern grid management. This paper introduces an optimized forecasting framework using Long Short-Term Memory (LSTM) networks. By integrating historical power generation data with localized meteorological factors, a multivariate predictive model was developed and validated using empirical data from a 100 MW PV plant. Based on Pearson correlation analysis, four key features—temperature, direct normal irradiance (DNI), relative humidity, and cloud cover—were chosen as the main drivers of PV output. A multivariate LSTM model was then trained and carefully tested using time series cross-validation. Results show the multi-feature architecture consistently outperforms and surpasses single-feature benchmarks. Specifically, the model achieved a peak R² of 0.9864 and minimum MAE of 1.4057 kW. Across four validation sets, R² remained stable (0.9755–0.9836), with most errors tightly bounded within ±5 kW. These findings confirm the model’s excellent accuracy and its value for improving power system dispatch and resource planning. Full article

Multivariate structural-response classification supports bridge monitoring by integrating synchronized measurements from multiple sensors and assigning an observation window to a predefined structural or loading condition. The main technical challenge is to compress temporally and spatially distributed sensor information into a fixed-length representation without losing transient or localized condition-sensitive features. However, previous studies have mainly focused on backbone architecture design, whereas the influence of sequence-level aggregation has rarely been isolated under a controlled backbone and training configuration. In this study, xLSTM-m is proposed as a modified variant of the extended Long Short-Term Memory (xLSTM) network for multivariate bridge-response classification. The conventional last-hidden-state readout is replaced with masked mean pooling over all valid hidden states, while the hybrid sLSTM/mLSTM backbone and the remaining training configuration are kept unchanged. This controlled design isolates the sequence-level aggregation strategy as the only architectural variable. A comparative evaluation is conducted using eight neural models. These models include xLSTM-m, the baseline xLSTM, and six Transformer-based architectures. Three bridge-monitoring datasets are considered: an FBG strain-response dataset and a PCB Piezotronics acceleration-response dataset acquired from a laboratory-scale cable-stayed bridge model, together with the field-scale Z24-9Setup acceleration benchmark. A strict five-fold cross-validation protocol is adopted. The proposed xLSTM-m ranks first on all three datasets. It achieves a mean accuracy of 95.29%, a mean F1-score of 95.57%, and the lowest inter-fold standard deviation of 0.84. Relative to the strongest Transformer baseline for each dataset, the corresponding accuracy gains are 6.30 percentage points for FBG, 3.66 percentage points for PCB, and 31.76 percentage points for Z24-9Setup. These results indicate that, for the evaluated bridge strain and acceleration datasets, sequence-level aggregation substantially affects both classification accuracy and inter-fold stability. Further validation using quasi-static responses, environmental variables, other structural systems, and unseen operating conditions is required before broader applicability across the SHM domain can be established. Full article

Accurate streamflow prediction is an important part of disaster management, as forecasting flow rates is an indispensable element of early warning systems, yet it remains challenging due to the complexity and nonlinearity of climatic inputs. This study evaluated the performances of Long Short-Term Memory (LSTM) and random forest (RF) machine learning algorithms for daily streamflow prediction across sixteen hydroclimatically diverse basins in the contiguous United States using the CAMELS dataset. The models were trained on five climatic features augmented with antecedent streamflow lag features at 7-, 14-, and 30-day intervals. The random forest algorithm demonstrated a better accuracy, achieving an average Nash–Sutcliffe efficiency (NSE) of 0.755 compared to 0.632 for the LSTM model. Both models performed well in the snowmelt-dominated basins and were least effective in flashy humid regimes. Additionally, both models exhibited high precision for flood detection, with accuracy rates exceeding 88% for distinguishing flood events and F1 scores of 0.734 and 0.797 for LSTM and RF, respectively. These results recommend RF for operational streamflow forecasting across hydroclimatically diverse settings and LSTM for perennial snowmelt- and groundwater-influenced catchments where long-range temporal dependencies govern runoff generation. Full article

Rapid urban population growth has intensified traffic congestion in smart cities. This has resulted in longer travel times, higher fuel consumption, increased environmental pollution, greater operational costs, and slower emergency response services. Existing traffic congestion prediction models primarily rely on traffic-flow and temporal features; the effects of road accidents and peak-hour conditions are not adequately addressed. This limitation is particularly significant in smart cities where both recurrent congestion (peak-hour demand) and non-recurrent congestion (road accidents) influence traffic conditions they have a significant impact on the performance of the road network. This study introduces a novel Historical Accident-Aware Peak-Hour GAN-GRU (APG-GRU) framework. The proposed framework employs a data processing pipeline integrating traffic data with historical accident-related features to predict traffic congestion using these features. Extensive experiments are conducted on a novel integrated dataset consist on Automatic Number Plate Recognition (ANPR) traffic data and ANPR traffic observations with historical accident features. The results demonstrate that the APG-GRU framework achieved superior performance on the integrated features dataset, attaining an accuracy of 97.50%, a congested precision of 91.86%, a congested recall of 97.31%, and a congested F1-score of 94.51%, outperforming both the ANPR traffic-only dataset and all baseline models. The APG-GRU framework significantly outperforms a suite of benchmark models, including XGBoost, Long Short-Term Memory (LSTM), and Random Forest as baselines, which achieved accuracies between 84% and 95.5% with correspondingly lower precision, recall, and F1-scores. External validation using a traffic dataset collected from Lahore, Pakistan, further demonstrated the robustness and generalizability of the proposed APG-GRU framework. A web-based interface developed for the APG-GRU framework to visualize accident hotspots and route-level traffic conditions. Routes with smooth traffic flow are highlighted in green, whereas congested routes are highlighted in red, demonstrating the practical applicability of the proposed framework for smart city traffic management systems. Full article

Parameter optimization of injection and production is an important method used to enhance recovery rate and reduce water cut, mainly aiming to determine the injection and production strategies during the development process to maximize the economic benefits throughout the reservoir development process. However, current optimization approaches for injection and production face challenges such as complex and inefficient optimization models and high-dimensional discrete variables, making it difficult to improve the global optimization ability of the algorithm (avoiding local optimal solutions during the optimization process) and the controllability of the time for completing the optimization of injection and production parameters under real and limited numerical simulations. This paper proposes a high-dimensional multi-discrete injection and production parameter non-gradient optimization method (MNOM), which combines the upper confidence bound (UCB) algorithm and effectively explores better injection and production systems and small-layer water-allocation schemes, achieving the maximization of net present value (NPV) over the entire development period. Specifically, this method models the injection and production parameter optimization problem as a Monte Carlo tree search process (Monte Carlo tree search, MCTS), and implements the optimization of injection and production cycles and small-layer water allocation through a genetic algorithm (CLGA) proxy optimized by a convolutional long short-term memory network (ConvLSTM). This method effectively overcomes the spatial and temporal limitations of the search process, maps production dynamics to the random strategies of injection and production parameters, and estimates the expected return of each policy. The CLGA proxy rapidly identifies suitable well-control schedules and water-allocation schemes in real time based on the production status at different development stages, thereby improving overall production performance. The proposed method has two innovative points. Firstly, MCTS can explore the large-scale discrete space of injection and production parameter optimization variables through tree decomposition, combined with the UCB incentive mechanism, to improve the global optimization ability. Secondly, the model training process is completely based on existing physical laws, with good temporal evolution, and the trained strategy can quickly adapt to the production status of the target layer without the need for a complete re-training from the beginning, enabling offline application and having good real-time controllability. In order to verify the effectiveness of the method proposed in the article, tests were conducted on a 3D reservoir actual model. Compared with gradient-based approaches, classical evolutionary algorithms, and proximal policy optimization (PPO), MNOM not only achieves stronger global search performance and requires 55–78% fewer iterations, but also improves the effective sweep volume by 1.1% to 5.5% compared to other optimization methods; furthermore, when compared with the PPO method, it is found that in offline optimization, if the production regime changes, the training strategy has better real-time controllability. The MNOM method can still maintain the original optimization effect compared to the PPO method when the production regime changes, demonstrating better engineering adaptability. The research results show that the proposed multi-strategy fusion high-dimensional multi-discrete injection and production parameter non-gradient optimization method can effectively improve recovery rate, expand effective sweep volume, and balance global optimization ability, optimization efficiency, and real-time controllability under the constraint of limited numerical simulations, and it has good engineering adaptability and application prospects. Full article

Satellites are critical components of modern space information systems. During long-term on-orbit operation, satellite telemetry often exhibits multi-scale temporal dynamics, heterogeneous channel behavior, and time-varying inter-variable dependencies, which pose substantial challenges to anomaly detection. Existing methods remain limited in adaptively representing anomaly patterns across temporal scales, jointly modeling temporal evolution and dynamic asymmetric channel dependencies, and preventing over-generalized reconstruction of anomalous inputs. To address these limitations, this paper proposes MSGMamba, a multi-scale graph state space model for satellite telemetry anomaly detection. First, a multi-scale temporal patch decomposition and gated fusion mechanism partitions telemetry sequences into patches of different granularities and adaptively integrates their representations at each temporal position, enabling the joint modeling of short-term transients and relatively slow-varying patterns. Second, a graph–sequence alternating propagation mechanism couples selective state space updates with dynamic graph interaction. At each temporal patch, a directed and asymmetric dependency graph with self-connection priors is generated from the temporally encoded features, allowing temporal evolution and time-varying cross-channel dependencies to be modeled within a unified framework. Third, an orthogonal memory-augmented anomaly discrimination mechanism introduces an orthogonality-constrained memory bank to reduce redundancy among nominal prototypes and constrain the reconstruction space. A dual-pathway anomaly score further combines signal-space reconstruction error with encoder–memory discrepancy to improve the separability of nominal and anomalous samples. Experiments on the SMAP, MSL, and EIRSAT-1 datasets show that MSGMamba outperforms representative baseline methods in terms of average PA-F1 and AFF-F1. Full article

Background: Cardiovascular diseases (CVDs) are among the leading causes of death globally. In order to treat CVDs successfully in the early stages, it is crucial to diagnose them in time. The ECG is one of the most common and non-invasive methods to detect heart rhythms and to diagnose arrhythmias. However, the analysis of ECG recordings manually requires a lot of time and experience because the morphology of ECG signals and the characteristics of their waveforms are very complex and show large overlaps between different types of arrhythmias. So far, various approaches for automated analysis of ECG signals have been developed, mostly based on deep learning (DL). In general, these methods are able to analyze ECG signals automatically and to detect different types of arrhythmias. Most approaches, however, are based on a purely data-driven feature learning and do not pay attention to the morphology-sensitive temporal structure of ECG signals, which is important for a discriminative diagnosis of arrhythmias. Methods: In this paper, we propose an Attention-Based Morphology-Aware (ABMA) framework to leverage multilead ECG signals in conjunction with automatically computed physiological features using a hybrid deep learning architecture. ABMA leverages multi-scale convolutional neural networks to learn local morphology features, and bidirectional long short-term memory (BiLSTM) networks to model temporal rhythms in ECG signals. We designed an ABMA module that incorporates a morphology scoring network (MSN) in order to (1) estimate the morphology-aware importance of different ECG segments and (2) learn the temporal importance of ECG features. The learned attention weights enable learning to focus on key sections of ECG signals without predefined boundaries or manual annotation of fiducial points. To understand the contribution of each individual component of the framework, we performed an extensive ablation study, where we removed the handcrafted feature branch, the ABMA module, the MSN, and the multi-head attention mechanism, one at a time, and compared the results against a fixed set of experimental configurations. Results: To assess the performance of the proposed framework in three-class classification between sinus rhythm (SA), atrial fibrillation (AFIB), and ventricular tachycardia (VT), we employed a stratified 10-fold cross-validation protocol. Our approach achieved a mean accuracy of 95.18 ± 1.18%, followed by a corresponding weighted F1-score of 95.19 ± 1.18% and a macro F1-score of 94.66 ± 1.35%. Notably, the performance of the proposed complete ABMA framework considerably outperformed the baseline CNN–BiLSTM architecture. Furthermore, in the primary evaluation metrics (i.e., accuracy, F1-score), the complete framework showed statistically significant improvements against the baseline through paired two-sided t-tests (p < 0.001). The ablation study indicated that each architectural component contributed positively to the overall classification performance, with the complete ABMA framework outperforming all reduced variants. Conclusions: The framework was evaluated by stratified cross-validation on a publicly available dataset. Our framework outperformed the baseline CNN–BiLSTM model as well as the respective ablation models in terms of classification performance. The findings from the current study are based on a retrospective analysis and therefore future studies using an independent external dataset, from multiple centers, or as part of a prospective clinical study are necessary in order to establish the generalizability and clinical utility of the proposed framework. The ABMA framework is currently viewed as a very promising research framework for intelligent ECG analysis, but it is not yet a clinically validated diagnostic tool. Full article

Accurate estimation of the state of health (SOH) of lithium-ion batteries is crucial for battery management systems (BMS), as it directly affects system safety, operational reliability, and remaining useful life prediction. Traditional SOH estimation methods generally rely on complete charge–discharge cycle data, which limits their practicality in real-time and online applications. To address this limitation, this paper proposes a novel voltage-charge increment curve-based dual-branch bidirectional long short-term memory network (VCIC-DB-BiLSTM) for battery SOH estimation under incomplete data. First, non-uniformly sampled battery current-voltage data are processed into standardized sequences with equal voltage intervals via voltage-charge increment curves based on ampere-hour integration and cubic spline interpolation. Subsequently, sliding window segmentation is applied to extract fixed-length curve segments from continuous voltage intervals as input features, while the corresponding complete voltage interval curves are used as labels. Finally, the VCIC-DB-BiLSTM network is designed, which uses a dual-branch structure to integrate feature extraction from both patch-processed and raw data, combined with bidirectional sequential modeling. Experimental validation on four benchmark datasets, CALCE, Oxford, XJTU, and TJU, demonstrates that the proposed method achieves competitive performance in SOH estimation under incomplete discharge data conditions, confirming its effectiveness and practical applicability. Full article

Accurate streamflow forecasting is critical for water resource management and disaster mitigation, yet conceptual models often suffer from systematic biases while machine learning lacks physical interpretability. This study proposes a coupled XAJ-LSTM model to integrate the mechanistic strengths of the Xin’anjiang (XAJ) model with the predictive capability of Long Short-Term Memory (LSTM) networks. Applied in the Woluo River basin using an 11-year daily record from 2012 to 2022, the framework was evaluated with leave-one-year-out validation. Results indicate that XAJ-LSTM achieved the highest full-period NSE of 0.885 and the lowest full-period RMSE of 19.48 m³/s, compared with NSE values of 0.867 for XAJ and 0.724 for LSTM. XAJ-LSTM also performed best during the flood period, with NSE of 0.820 and RMSE of 28.81 m³/s. The SHAP framework indicated that XAJ-simulated flow and precipitation remained the dominant predictors, suggesting that the LSTM relied strongly on the conceptual-model baseline and rainfall information. These findings indicate that XAJ-LSTM can improve selected full-period and flood-period diagnostics for daily streamflow simulation in the Woluo River Basin, while further tests with longer records and additional catchments are needed before broader application. Full article

Ensuring the reliability and operational efficiency of critical scientific equipment is a central challenge in high-stakes research environments such as nuclear physics laboratories and particle accelerator facilities. Unexpected failures entail significant financial cost and prolonged interruptions to experimental programmes. We present a predictive maintenance (PdM) framework built around a two-layer Vanilla Long Short-Term Memory (LSTM) network trained on multivariate sensor streams collected at NRF-iThemba LABS between January 2021 and December 2023. Four channels, namely supply voltage, vibration velocity, differential pressure, and rotational speed, were recorded at 5 $\min$ intervals using a suite of industrial-grade transducers (power quality analyser, IEPE accelerometers, differential pressure transmitters, and proximity encoders) feeding a multi-channel data-acquisition chassis via OPC-UA, yielding a time-synchronised dataset of 315,360 observations. A normalised failure score converts the binary classifier output into a continuous, interpretable health indicator that supports tiered scheduling of maintenance. The Vanilla LSTM achieved a test-set F1-score of $75\%$ and an area under the receiver-operating-characteristic curve (AUC) of $0.856$, outperforming five competing architectures (PCA/$T^2$, Random Forest, Deep Neural Network, LSTM Autoencoder, and Bidirectional LSTM Autoencoder), and delivered a mean failure lead time of $(42.3\pm 7.2)$$\mathrm{h}$, exceeding the 36 $\mathrm{h}$ engineering requirement for proactive maintenance scheduling. Full article

Precise prediction of voltage degradation is critical for the prognostics and health management of proton exchange membrane fuel cells (PEMFCs). The performance degradation of PEMFCs is governed by a complex coupling of multiple physicochemical mechanisms, including catalyst layer and proton exchange membrane degradation. Crucially, these internal degradation processes evolve across highly heterogeneous time scales, ranging from transient high-frequency fluctuations to low-frequency and long-term irreversible performance fade. Conventional predictive models, which typically rely on single-scale architectures or fixed receptive fields, are inherently ill-equipped to simultaneously decouple and capture these cross-scale temporal dynamics. To tackle this challenge, this paper innovatively proposes a multi-scale deep learning framework that integrates a multi-scale degradation trend perception module, a long short-term memory (LSTM)-based encoder–decoder architecture, and a multi-head attention mechanism. One-dimensional convolutional layers with different kernel sizes are employed to simultaneously extract local temporal features at multiple granularities, followed by the LSTM encoder–decoder to model long-range temporal dependencies, while the cross-attention mechanism dynamically allocates attention across the encoded context at each autoregressive decoding step. Experimental outcomes indicate that the proposed model realizes excellent predictive accuracy across five evaluation indices in comparison with standard baselines. In particular, the mean absolute percentage error (MAPE) reaches 1.6696%, and the maximum absolute percentage error (Max-APE) is strictly bounded within 5%, substantiating the reliability of the proposed framework for high precision and long-horizon health prognostics for PEMFCs. Full article

Surrounding rock deformation in soft rock tunnels is controlled by complex nonlinear interactions among geological conditions, construction parameters, and support measures, making accurate prediction challenging. In this study, a project-scale deformation prediction dataset was constructed using field monitoring data from a sandy shale tunnel project. Eight engineering factors were selected as input variables, including excavation method, initial support strength, closure time, tunnel burial depth, lithology, rock integrity, groundwater condition, and the relative orientation between major structural planes and the tunnel. A hybrid prediction framework integrating a novel whale optimization algorithm (NWOA) and a long short-term memory (LSTM) network was developed. The proposed NWOA improves the standard whale optimization algorithm by introducing a nonlinear convergence strategy, an adaptive weight coefficient, and a dynamic spiral position updating mechanism to enhance the hyperparameter search process of the LSTM model. Model performance and stability were further assessed using repeated and nested cross-validation. The corresponding RMSEs were 0.2294 ± 0.0734 and 0.2219 ± 0.0751 percentage points, the MAEs were 0.1427 ± 0.0350 and 0.1505 ± 0.0585 percentage points, and the R² values were 0.8660 ± 0.0548 and 0.8732 ± 0.0599, respectively. These comparable results support project-specific predictive performance for the investigated tunnel sections. Full article