Search Results (294)

In the domain of password recovery, deep learning has emerged as a pivotal technology for enhancing recovery efficiency. Despite its effectiveness, the inherent computation complexity of deep learning-based password generation algorithms poses substantial challenges, particularly in achieving synergistic acceleration between deep learning inference, and plaintext encryption process. In this paper, we introduce PassRecover, a multi-FPGA-based computing system that can simultaneously accelerate deep learning-driven password generation and plaintext encryption in an end-to-end manner. The system architecture incorporates a neural processing unit (NPU) and an encryption array configured to operate under a streaming dataflow paradigm for parallel processing. It is the first approach to explore the benefit of end-to-end offline password recovery. For comprehensive evaluation, PassRecover is benchmarked against PassGAN and five industry-standard encryption algorithms (Office2010, Office2013, PDF1.7, Winzip, and RAR5). Experimental results demonstrate excellent performance: Compared to the latest work that only accelerate encryption algorithms, PassRecover achieves an average 101.5% speedup across all tested encryption algorithms. When compared to graphics processing unit (GPU)-based end-to-end implementations, this work delivers 93.01% faster processing speeds and $3.73\times$ superior energy efficiency. These results establish PassRecover as a promising solution for resource-constrained password recovery scenarios requiring high throughput and energy efficiency. Full article

Surface classification is critical for ground robots operating in diverse environments, as it improves mobility, stability, and adaptability. This study introduces IMU-based deep learning models for surface classification as a low-cost alternative to computer vision systems. Two feature fusion models were introduced to classify the surface type using time-series data from an IMU sensor mounted on a ground robot. The first model, a cascaded fusion model, employs a 1-D Convolutional Neural Network (CNN) followed by a Long Short-Term Memory (LSTM) network and then a multi-head attention mechanism. The second model is a parallel fusion model, which processes sensor data through both a CNN and an LSTM simultaneously before concatenating the resulting feature vectors and then passing them to a multi-head attention mechanism. Both models utilize a multi-head attention mechanism to enhance focus on relevant segments of the time-sequence data. The models were trained on a normalized Internal Measurement Unit (IMU) dataset, with hyperparameter tuning achieved via grid search for optimal performance. Results showed that the cascaded model achieved higher accuracy metrics, including a mean Average Precision (mAP) of 0.721 compared to 0.693 for the parallel model. However, the cascaded model incurred a 44.37% increase in processing time, which makes the parallel fusion model more suitable for real-time applications. The multi-head attention mechanism contributed significantly to accuracy improvements, particularly in the cascaded model. Full article

This study explores the integration of the JADE (Java Agent Development Framework) platform with deep reinforcement learning (DRL) to enhance governmental interoperability and optimize administrative workflows in municipal settings. The proposed approach combines the JADE’s robust multi-agent system (MAS) capabilities with the adaptive decision-making power of DRL to address prevalent challenges faced by government agencies, such as fragmented operations, incompatible data formats, and rigid communication protocols. By enabling seamless communication between agents across departments such as the Treasury, the Event Management department, and the Public Safety department, the hybrid system fosters real-time collaboration and supports efficient, data-driven decision making. Agents leverage historical and real-time data to adapt to environmental changes and make optimized decisions that align with overarching governmental objectives, such as resource allocation and emergency response. The result is a system capable of managing intricate administrative duties using structured agent communication and the integration of DRL-driven learning models, improving governmental interoperability. Key performance indicators highlight the system’s effectiveness, achieving a task completion rate of 95%, decision accuracy of 96%, and a communication latency of just 120 ms. Additionally, the framework’s flexibility ensures seamless scalability, accommodating complex and large-scale tasks across multiple governmental units. This research presents a scalable, automated, and resilient framework for optimizing governmental processes, offering a pathway to more efficient, transparent, and adaptive public sector operations. Full article

With the large-scale integration of wind power, photovoltaic, and other renewable energy sources into the power grid, their inherent randomness and variability present significant challenges to the frequency stability of power systems. Conventional thermal power units with limited frequency regulation capabilities face further strain, as frequent power fluctuations accelerate wear and tear, thereby shortening their operational lifespans. This makes it increasingly difficult to meet the demands for frequency regulation. Electrode boilers, as flexible electrical loads, can be retrofitted to enhance their flexibility and participate in grid frequency regulation alongside renewable energy units. This not only improves frequency stability but also reduces wear on generating units. However, the frequency regulation process involves balancing multiple objectives, such as maintaining system frequency stability, ensuring economic efficiency, and optimizing operational effectiveness. Traditional control strategies often struggle to address these competing objectives effectively. To address these challenges, this paper proposes a multi-objective collaborative optimization control decision model for electrode boilers to assist in grid frequency regulation. The model not only meets the frequency regulation requirements but also considers additional constraints, including the operational efficiency of electrode boilers, economic benefits, and equipment degradation. A genetic algorithm is employed to solve the model, and simulation analysis is conducted using the IEEE 14-node system. The results demonstrate that this strategy significantly enhances frequency stability, improves boiler operational efficiency, and boosts economic benefits, offering a viable solution for integrating electrode boilers into grid frequency regulation. Full article

Lithium-ion power battery pack life, capacity and safety depend primarily on consistency between battery cells. However, inconsistencies between battery cells are inevitable due to the inherent variability in production processes and operational environments. In parallel circuits, battery management systems can usually only monitor the total module current and terminal voltage, which results in limitations that lead to inter-unit inconsistency, reducing overall safety and energy efficiency. The conventional method of battery sorting involves analyzing static parameters such as capacity, internal resistance and voltage to ensure static consistency between cells. Nonetheless, cell-to-cell variations are more pronounced during dynamic and complex operations. The direct integration of static and dynamic features may result in data scale discrepancies and redundant information. Thus, the present study proposes a two-stage multi-parameter clustering method based on static and dynamic features. Initially, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) was applied to sort abnormal batteries and identify the number of subsequent clusters, using discharge capacity, internal resistance and open-circuit voltage (OCV) as inputs. Then, a Principal Component Analysis (PCA) was used to downscale and extract features from the discharge voltage profile. The principal component data were used as inputs to the Self-Organizing Map (SOM) clustering algorithm, which uses its self-organized and unsupervised learning characteristics to mine more dynamic time-series features and complete the final clustering and sorting. Finally, the effectiveness of the two-stage sorting method in parallel circuits was verified by determining clustering evaluation indexes, as well as the cycle life and discharge curves of batteries reassembled in parallel. Full article

In order to enable oat ears to be quickly and accurately identified in the natural environment, this paper proposes an oat ears detection and counting model based on an improved Faster R-CNN. In the backbone network, the commonly used single convolutional neural network is replaced by a parallel convolutional neural network to realize the feature extraction of oat ears, and a feature pyramid network (FPN) is incorporated to improve the small target-missed detection problem and the multi-scale problem of oat ears. Then, the anchor box configuration is optimized according to the size and distribution of the labeled boxes in the dataset, which improves the efficiency of the model to detect oat ears. Finally, progressive non-maximum suppression (Progressive-NMS) was used to replace non-maximum suppression (NMS) to optimize the screening process of prediction boxes. According to the data from different experiments designed, the optimized model can effectively detect oat ears in the natural environment and complete the counting of oat ears per unit area. Compared with the traditional Faster R-CNN detection model, the mean average precision (mAP) of the improved model is increased by 13.01%, which could provide reference for oat yield prediction and intelligent operation. Full article

This paper introduces a novel approach to the design of multi-element planar solar concentrators, aimed at optimizing solar energy harvesting systems. The proposed methodology is based on the integration of identical unit cells, strategically arranged to enhance solar radiation capture efficiency and achieve high angular selectivity. Mathematical modeling of the operational principles of the unit cells forms the foundation for determining production parameters and streamlining the concentrator assembly process. Particular emphasis is placed on analyzing key performance metrics, such as solar radiation concentration and optical efficiency, thereby advancing the understanding of the relationship between design parameters and energy output. The study employs MATLAB R2022b and ZemaxOpticStudio 13 software to model the solar concentrator, identifying the optimal cell configuration to achieve a geometric concentration ratio of 3.45, with angular selectivity ranging from 23° to 90°. This research contributes significantly to the field of solar concentrator technology, offering a pathway for more efficient utilization of renewable energy sources and improved adaptability to diverse operating conditions. Full article

The rapid expansion of data has made global access easier, but it also demands increasing amounts of energy for data storage and processing. In response, neuromorphic electronics, inspired by the functionality of biological neurons and synapses, have emerged as a growing area of research. These devices enable in-memory computing, helping to overcome the “von Neumann bottleneck”, a limitation caused by the separation of memory and processing units in traditional von Neumann architecture. By leveraging multi-bit non-volatility, biologically inspired features, and Ohm’s law, synaptic devices show great potential for reducing energy consumption in multiplication and accumulation operations. Within the various non-volatile memory technologies available, flash memory stands out as a highly competitive option for storing large volumes of data. This review highlights recent advancements in neuromorphic computing that utilize NOR, AND, and NAND flash memory. This review also delves into the array architecture, operational methods, and electrical properties of NOR, AND, and NAND flash memory, emphasizing its application in different neural network designs. By providing a detailed overview of flash memory-based neuromorphic computing, this review offers valuable insights into optimizing its use across diverse applications. Full article

The current method for dividing slope units primarily relies on hydrological analysis methods, which consider only geomorphological factors and fail to reveal the geological boundaries during landslides. Consequently, this approach does not fully satisfy the requirements for detailed landslide susceptibility assessments at the township scale. To address this limitation, we propose a new landslide susceptibility evaluation model based on geomorphological and geological characteristics. The key challenges addressed include: (i) Optimization of the slope unit division method. This is accomplished by integrating geomorphological features, such as slope gradient and aspect, with geological features, including lithology, slope structure types, and disaster categories, to develop a process for extracting slope units based on both geomorphological and geological characteristics. The results indicate that the proposed slope units outperform the hydrological analysis methods in three key indicators: overlap, shape regularity, and spatial distribution uniformity. (ii) Development and validation of the evaluation model. A landslide susceptibility index system is developed using multi-source data, with susceptibility prediction conducted via the XGBoost model optimized by Bayesian methods. The model’s accuracy is validated using the Receiver Operating Characteristic (ROC) curve. The results show that the proposed slope units achieve an AUC value of 0.973, surpassing the hydrological method. (iii) Analysis of landslide susceptibility variations. The susceptibility of the two types of slope units is analyzed through landslide case studies. The consistency between the proposed slope units and field verification results is explained using engineering geological characteristics. The SHAP model is then used to examine the influence of key disaster-inducing and individual factors on landslide occurrence. Full article

The implementation of predictive maintenance (PM) in aviation presents unique challenges due to strict safety requirements, complex operational environments, and regulatory constraints. This paper develops a comprehensive decision-making framework for evaluating the feasibility of implementing PM for aircraft components, addressing the critical need for systematic integration of technical, economic, and regulatory considerations. Through expert surveys involving 78 aviation maintenance professionals and the application of multi-criteria decision analysis, this study identifies and validates 14 key criteria across four categories: technical and operational, economic and feasibility, regulatory and compliance, and organizational and human factors. The analytic hierarchy process is employed to establish criteria weights, with flight safety impact, reliability predictability, and data sufficiency emerging as primary drivers. The framework’s effectiveness is demonstrated through case studies comparing turbofan engines and avionics units, validating its ability to discriminate between components suitable for PM implementation. Results indicate that successful PM implementation requires not only technological readiness but also organizational alignment and regulatory compliance. This study contributes to aviation maintenance practice by providing a structured, evidence-based approach to PM implementation decisions, while establishing a foundation for future innovations in maintenance strategies. The framework’s practical applicability is enhanced through a detailed implementation roadmap and validation methods, ensuring its relevance for maintenance decision-makers while maintaining alignment with aviation safety standards. Full article

This study investigates the operational efficiency of the lab-scale oxidation ditch (OD) functioning in simultaneous nitrification and denitrification modes, focusing on forecasting biochemical oxygen demand (BOD₅) concentrations over a five-day horizon. This forecasting capability aims to optimize the operational regime of aeration tanks by adjusting the specific load on organic pollutants through active sludge dosage modulation. A comprehensive statistical analysis was conducted to identify trends and seasonality alongside significant correlations between the forecasted values and various time lags. A total of 20 time lags and the “month” feature were selected as significant predictors. These models employed include Multi-head Attention Gated Recurrent Unit (MAGRU), long short-term memory (LSTM), Autoregressive Integrated Moving Average–Long Short-Term Memory (ARIMA–LSTM), and Prophet and gradient boosting models: CatBoost and XGBoost. Evaluation metrics (Mean Squared Error (MSE), Mean Absolute Error (MAE), Symmetric Mean Absolute Percentage Error (SMAPE), and Coefficient of Determination (R²)) indicated similar performance across models, with ARIMA–LSTM yielding the best results. This architecture effectively captures short-term trends associated with the variability of incoming wastewater. The SMAPE score of 1.052% on test data demonstrates the model’s accuracy and highlights the potential of integrating artificial neural networks (ANN) and machine learning (ML) with mechanistic models for optimizing wastewater treatment processes. However, residual analysis revealed systematic overestimation, necessitating further exploration of significant predictors across various datasets to enhance forecasting quality. Full article

This paper presents an enhanced fault prediction framework for synchronous condensers in UHVDC transmission systems, integrating Large Language Models (LLMs) with optimized Wavelet Packet Transform (WPT) for improved diagnostic accuracy. The framework innovatively employs LLMs to automatically optimize WPT parameters, addressing the limitations of traditional manual parameter selection methods. By incorporating a Multi-Head Attention Gated Recurrent Unit (MHA-GRU) network, the system achieves superior temporal feature learning and fault pattern recognition. Through intelligent parameter optimization and advanced feature extraction, the LLM component intelligently selects optimal wavelet decomposition levels and frequency bands, while the MHA-GRU network processes the extracted features for accurate fault classification. Experimental results on a high-capacity synchronous condenser demonstrate the framework’s effectiveness in detecting rotor, air-gap, and stator faults across diverse operational conditions. The system maintains efficient real-time processing capabilities while significantly reducing false alarm rates compared to conventional methods. This comprehensive approach to fault prediction and diagnosis represents a significant advancement in synchronous condenser fault prediction, offering improved accuracy, reduced processing time, and enhanced reliability for UHVDC transmission system maintenance. Full article

AmeriFlux is a network of hundreds of sites across the contiguous United States providing tower-based ecosystem-scale carbon dioxide flux measurements at 30 min temporal resolution. While geographically wide-ranging, over its existence the network has suffered from multiple issues including towers regularly ceasing operation for extended periods and a lack of standardization of measurements between sites. In this study, we use machine learning algorithms to predict CO₂ flux measurements at NEON sites (a subset of Ameriflux sites), creating a model to gap-fill measurements when sites are down or replace measurements when they are incorrect. Machine learning algorithms also have the ability to generalize to new sites, potentially even those without a flux tower. We compared the performance of seven machine learning algorithms using 35 environmental drivers and site-specific variables as predictors. We found that Extreme Gradient Boosting (XGBoost) consistently produced the most accurate predictions (Root Mean Squared Error of 1.81 μmolm⁻²s⁻¹, R² of 0.86). The model showed excellent performance testing on sites that are ecologically similar to other sites (the Mid Atlantic, New England, and the Rocky Mountains), but poorer performance at sites with fewer ecological similarities to other sites in the data (Pacific Northwest, Florida, and Puerto Rico). The results show strong potential for machine learning-based models to make more skillful predictions than state-of-the-art process-based models, being able to estimate the multi-year mean carbon balance to within an error ±50 gCm⁻²y⁻¹ for 29 of our 44 test sites. These results have significant implications for being able to accurately predict the carbon flux or gap-fill an extended outage at any AmeriFlux site, and for being able to quantify carbon flux in support of natural climate solutions. Full article

Flash memory, as the core unit of a compute-in-memory (CIM) array, requires multiple positive and negative (PN) high voltages (HVs) for word lines (WLs) to operate during storage and computation. A traditional WL driver generates these voltages using several charge pumps (CPs), leading to significant area overhead. This paper presents a novel multi-mode CP (MMCP) that generates all required HVs for a CIM array under a single CP, supporting CIM unit operation in programming, readout, and erase modes. Unlike traditional voltage multipliers, the MMCP eliminates the need for multiple CPs, reducing area and pump capacitor usage. Compared to a PN CP that drives a common load, the MMCP can provide multiple PN HVs by using level shifters (LSs) and switches. The MMCP is designed in a 55 nm standard CMOS process with an area of only 0.021 mm². Additionally, this paper proposes global PN HV switches, which can correctly deliver the PN HVs generated by the MMCP from the same port (at different times) to the upper and lower power rails of WL driver circuits. Simulation results show that with a 2.5 V supply, 100 pF load, and 50 $\mathsf{\mu }$A current, the maximum error due to ripple is only 0.28%. Full article

Graph-based neural networks have proven to be useful in molecular property prediction, a critical component of computer-aided drug discovery. In this application, in response to the growing demand for improved computational efficiency and localized edge processing, this paper introduces a novel approach that leverages specialized accelerators on a heterogeneous edge computing platform. Our focus is on graph convolutional networks, a leading graph-based neural network variant that integrates graph convolution layers with multi-layer perceptrons. Molecular graphs are typically characterized by a low number of nodes, leading to low-dimensional dense matrix multiplications within multi-layer perceptrons—conditions that are particularly well-suited for Edge TPUs. These TPUs feature a systolic array of multiply–accumulate units optimized for dense matrix operations. Furthermore, the inherent sparsity in molecular graph adjacency matrices offers additional opportunities for computational optimization. To capitalize on this, we developed an FPGA GFADES accelerator, using high-level synthesis, specifically tailored to efficiently manage the sparsity in both the graph structure and node features. Our hardware/software co-designed GCN+MLP architecture delivers performance improvements, achieving up to $58\times$ increased speed compared to conventional software implementations. This architecture is implemented using the Pynq framework and TensorFlow Lite Runtime, running on a multi-core ARM CPU within an AMD/Xilinx Zynq Ultrascale+ device, in combination with the Edge TPU and programmable logic. Full article