Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hongbiao Zhou has a total of 53 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Effluent Quality Prediction Using Adaptive Sparse Deep Belief Network with Stochastic Configuration Learning Scheme

    by Le Wang, Chunyang Fan, Rui Yan, Shuke Zhang, Hongbiao Zhou

    The accurate prediction of effluent water quality is essential for monitoring the operational status of wastewater treatment plants and supporting process regulation. To enhance feature representation and prediction accuracy in data-driven effluent quality modeling, this paper proposes an adaptive sparse deep belief network with stochastic configuration learning (ASDBN-SCL). First, an adaptive learning-rate strategy based on gradient-direction consistency is introduced during restricted Boltzmann machine pretraining to dynamically regulate the parameter-update step size. Second, a sparsity constraint based on the average activation probability of hidden neurons is incorporated to suppress redundant feature responses and promote compact representations. Third, stochastic configuration learning is employed as the supervised output module, in which hidden nodes are incrementally generated under a supervisory constraint and the output weights are analytically determined using least squares. The proposed method was evaluated using hourly monitoring data collected from a municipal wastewater treatment plant in Huai’an, China. After data preprocessing, 4500 valid samples were retained and chronologically partitioned into training, validation, and independent test sets. For effluent chemical oxygen demand prediction, the proposed method achieved a test RMSE of 0.3868±0.0174, an MAE of 0.3037±0.0142, and an R2 of 0.9954±0.0004. For effluent total nitrogen prediction, the corresponding values were 0.2239±0.0094, 0.1756±0.0082, and 0.9849±0.0013, respectively. These results indicate that the proposed framework provides accurate and reliable effluent water quality predictions for the investigated wastewater treatment process.

  • Open Access

    A GMM-Based Spatio-Temporal Distribution Knowledge Transfer MOEA/D for Dynamic Multi-UAV Cooperative Path Planning

    by Shuke Zhang, Hongbiao Zhou, Tengfei Ma, Le Wang

    Multi-UAV cooperative path planning in time-varying environments requires balancing flight efficiency, threat avoidance, and coordination consistency while responding rapidly to environmental changes. To address slow population recovery, insufficient inter-task cooperation, and unreliable reuse of historical search information, this paper proposes a Gaussian mixture model (GMM)-based spatio-temporal knowledge transfer multi-objective evolutionary algorithm within the MOEA/D framework, termed STKTM-MOEA/D. The proposed method represents mission requirements and flight constraints using objective functions and feasibility constraints. Elite decision vectors selected through non-dominated sorting and crowding-distance ranking are used to construct GMMs that probabilistically describe promising search regions. Each GMM captures the locations, dispersions, and relative importance of multiple high-quality regions, providing a unified distribution-level knowledge representation for spatial and temporal transfer. Spatial knowledge transfer jointly considers task similarity and distribution complementarity between the main and auxiliary tasks to improve collaborative search. Temporal knowledge transfer retrieves reliable historical GMMs from a knowledge pool according to environmental similarity. After an environmental change, an adaptive reconstruction strategy combines current elites, temporally transferred individuals, spatially transferred individuals, and randomly generated exploratory individuals. The reconstructed population then continues MOEA/D evolution, environmental selection, GMM updating, and knowledge-pool updating under the new environment, improving convergence, diversity, and adaptability. Experiments on 14 DF and 5 FDA problems show that STKTM-MOEA/D achieves the best mean MIGD on 10 problems. In multi-UAV path planning, it attains a 100% success rate and outperforms KTM-DMOEA in HV, PD, and runtime, demonstrating strong effectiveness and efficiency in dynamic environments.

  • Short-term traffic flow prediction method based on ATT-CNN-BiLSTM hybrid network

    by Shiqi Hu, Ningyi Sun, Congguo Ma, Hongbiao Zhou, Wen Shi, Le Wang

    Short-term traffic flow forecasting is a fundamental component of intelligent transportation systems (ITS), enabling proactive traffic management and congestion mitigation. However, traditional statistical methods often struggle to capture the nonlinear and time-varying characteristics of traffic flow, while single deep learning models may fail to learn complementary temporal patterns across multiple scales. To address these limitations, this paper proposes a hybrid forecasting framework, attention mechanism-convolutional neural network-bidirectional long short-term memory (ATT-CNN-BiLSTM), which integrates one-dimensional convolutional neural networks for local temporal feature extraction, a BiLSTM network for modeling long-range bidirectional dependencies, and a squeeze-and-excitation attention mechanism for adaptive feature recalibration. Extensive experiments conducted on real-world traffic datasets demonstrate that the proposed model achieves stable and competitive prediction performance across datasets collected from different years. In addition, ablation studies verify the effectiveness of the attention mechanism and network configuration in improving overall forecasting accuracy. The proposed framework therefore provides an effective and practical solution for short-term traffic flow forecasting in ITS applications.


  • Manifold transfer-based dynamic multi-objective optimal control for wastewater treatment processes

    by Weifeng Ding, Hongbiao Zhou, Feng Li, Shuke Zhang

    To address the dynamic complexity in wastewater treatment processes and the failure of traditional prediction-based dynamic multi-objective optimization algorithms to meet the i.i.d. (IID) assumption, this study proposes a dynamic multi-objective optimization control strategy called MTDI-OC, which is based on manifold transfer learning and elite differential immigration for optimization, and employs a fuzzy neural network (FNN) for control. First, the self-organizing fuzzy neural network (SO-FNN) model is employed to capture the dynamic nonlinear relationship between energy consumption (EC) and effluent quality (EQ) in real-time. By leveraging manifold transfer learning, historical optimization experience is adapted to the new dynamic environment, enabling rapid generation of Pareto-optimal setpoints for dissolved oxygen ( S O ) and nitrate nitrogen ( S NO ) concentrations. Simultaneously, an elite differential immigration strategy is adopted to maintain population diversity. Then, an improved Chebyshev decision-making method is applied to balance the conflict between energy consumption and water quality. Finally, an FNN controller tracks the optimized setpoints to establish a dynamic disturbance-resistant control system. Experiments on S O and S NO tracking control under different weather conditions demonstrate the effectiveness of the proposed strategy. Comparative studies under varying weather scenarios show that the strategy excels in reducing energy consumption while maintaining relatively stable water quality. In dry weather, the proposed MTDI-OC can reduce EC to 3633.14 kWh d −1 , achieving a 15.24% reduction. Compared to other optimal controller, the proposed strategy ensures full compliance with water quality standards, significantly lowers energy consumption, and improves the removal rates of critical indicators such as ammonia nitrogen ( S NH ) and total nitrogen ( N tot ).


  • Open Access

    RFAM-YOLO: Reinforced Feature Aggregation and Multi-Attention YOLOv8s for High-Precision Traffic Sign Detection

    by Le Wang, Ziqi Jiang, Chuang Chen, Hongbiao Zhou

    In the field of intelligent driving, including autonomous and assisted driving, traffic sign detection is critical, as its performance directly affects vehicle safety. To address the challenges posed by complex backgrounds and the small size of traffic signs, this paper proposes an improved small-object traffic sign detection algorithm based on YOLOv8s, named RFAM-YOLO. First, since the shared parameter mechanism in standard convolution kernels can limit performance, RFCAConv and C2f_RFCAConv are integrated into the backbone network to enhance convolution operations, thereby improving multi-scale feature extraction. Second, a C2f_MLCA attention module is incorporated into the Neck component to enhance the representation of key object features while maintaining computational efficiency, thereby improving feature fusion. Third, to enhance the detection of multiple small traffic signs, an additional small-object detection head module is introduced to better capture their characteristics and contextual information. Finally, to account for differences in aspect ratios between predicted and ground truth bounding boxes, the original CIoU loss function is replaced with Focaler-ShapeIoU, improving bounding box regression accuracy and detection robustness. Experiments on two benchmark traffic sign datasets, TT100K and CCTSDB2021, demonstrate that RFAM-YOLO consistently improves detection performance over the YOLOv8s baseline while maintaining a lightweight model size and real-time inference speed. On the TT100K dataset, RFAM-YOLO achieves a 4.38 percentage point increase in mean Average Precision (mAP@50), from 86.09% to 90.47%, with simultaneous gains in precision and recall. On the CCTSDB2021 dataset, it further boosts mAP@50 from 96.54% to 98.20%. These results show that the proposed architecture provides a favorable accuracy-efficiency trade-off and improved robustness to small and cluttered traffic signs in realistic driving scenarios.


  • Kernel-space oversampling and adaptive ensemble learning for robust imbalanced data classification

    by Le Wang, Ningyi Sun, Shiqi Hu, Hongbiao Zhou, Han Chen

    This paper proposes a robust fuzzy kernel-based possibilistic clustering-means adaptive weighted random forest (FKAW-RF) framework for imbalanced data classification that integrates kernel-space oversampling with adaptive ensemble learning. The method first maps the original dataset into a high-dimensional kernel space, where a fuzzy kernel-based possibilistic clustering-means (FKPCM) algorithm is employed to identify and remove boundary samples, thereby reducing the influence of noise and class overlap. Within this kernel space, a kernel-conventional synthetic minority oversampling technique (KSMOTE) strategy is applied to perform high-quality oversampling of minority class instances, and the generated samples are projected back into the input space through preimage estimation. To further enhance classification robustness, an adaptive weighted random forest (AW-RF) is constructed as the ensemble classifier, in which underperforming or redundant decision trees are pruned based on feature correlation analysis, and the remaining trees are adaptively weighted using dual penalty factors reflecting both local and global error rates. Extensive experiments on UCI, KEEL, and Tennessee-eastman (TE) benchmark datasets demonstrate that the proposed kernel-space oversampling and adaptive ensemble learning approach consistently outperforms conventional synthetic minority oversampling technique (SMOTE) variants and state-of-the-art classifiers across multiple evaluation metrics, including accuracy, precision, recall, and G-mean, confirming its effectiveness and robustness for challenging imbalanced classification tasks.


  • Prediction of remaining useful life of lithium-ion battery using convolutional neural network-bidirectional long short-term memory with attention mechanism

    by Hengrui Zhou, Feng Li, Shuke Zhang, Le Wang, Hongbiao Zhou

    Accurately predicting the remaining useful life (RUL) of lithium-ion batteries is crucial for ensuring the safe and reliable operation of battery systems. However, due to the highly nonlinear degradation behavior of lithium-ion batteries under complex working conditions, traditional prediction methods often fail to achieve satisfactory accuracy. This study proposes a hybrid prediction model that integrates an attention (ATT) mechanism, a convolutional neural network (CNN), and a bidirectional long short-term memory (BiLSTM) network, termed ATT-CNN-BiLSTM. First, a one-dimensional CNN is employed to extract local degradation patterns and inter-channel correlated features from multi-dimensional sensor time-series data. Then, the BiLSTM network captures long-term dependencies within the degradation process. Furthermore, an attention mechanism is introduced to dynamically adjust the contribution of key time-step features and suppress noise interference. Experimental results based on the NASA battery aging datasets demonstrate that the proposed ATT-CNN-BiLSTM model reduces RMSE and MAE by 49.08% and 54.67%, respectively, compared with the LSTM model, and by 10.89% and 17.85%, respectively, compared with the ACNN-LSTM-MMD model. Ablation experiments also confirm that further confirm that the ATT-CNN-BiLSTM model decreases RMSE and MAE by 26.45% and 33.33%, respectively, compared with CNN-BiLSTM, and by 21.85% and 23.36%, respectively, compared with ATT-BiLSTM.


  • Non‐Destructive Measurement of Sugar Content in Litchis Using Visible and Near‐Infrared Spectroscopy and Fuzzy Stochastic Configuration Network

    by Hongbiao Zhou, Feng Li, Ningyi Sun, Shuke Zhang

    Non‐destructive measurement of sugar content in litchis is essential for achieving precise and efficient sorting. This study developed a predictive model for sugar content in litchis based on visible‐near‐infrared (Vis‐NIR) spectral data and intelligent algorithms. To overcome the limitations of existing models, such as poor fitting performance and time‐consuming training processes, an efficient fuzzy stochastic configuration network (FSCN) was proposed by integrating the strengths of fuzzy neural networks and stochastic configuration networks. The FSCN incorporates fuzzy membership functions to quantify uncertainties in input features and leverages the incremental learning mechanism of its stochastic configuration layer to dynamically optimize network parameters. This approach enhances the model's capacity to capture the complex nonlinear relationship between sugar content and spectral wavelengths. Furthermore, the effects of 12 spectral preprocessing methods and 4 feature extraction techniques were systematically evaluated. Using 656 spectral data samples from litchis, experimental results demonstrated that the combination of multiplicative scatter correction (MSC) with the successive projections algorithm (SPA) yielded the most effective outcome. The proposed FSCN model exhibited superior predictive performance, achieving an R 2 of 0.9676, a root mean square error (RMSE) of 0.5487, and a mean absolute error (MAE) of 0.4441. These values surpassed those obtained by partial least squares regression (PLSR), artificial neural networks (ANNs), and support vector regression (SVR). These findings confirm that the FSCN model is well‐suited for accurate sugar content detection in practical litchi sorting systems.


  • Self-organizing fuzzy neural network using a fuzzy self-tuning multi-objective particle swarm optimization

    by Le Wang, Xiang Ma, Liyan Wen, Hongbiao Zhou

    This paper presents a novel fuzzy self-tuning multi-objective particle swarm optimization (FSTMOPSO) algorithm for self-organizing fuzzy neural networks (SOFNNs). The proposed approach addresses the multi-objective optimization problem by simultaneously optimizing three key factors: training accuracy, network complexity, and the smoothness of connection weights. FSTMOPSO employs a fuzzy self-tuning strategy that adaptively adjusts the parameters of each particle throughout the optimization process, thereby enhancing both convergence speed and solution quality. Additionally, a crowding operator based on Mahalanobis distance ensures diversity among the archived solutions, preventing clustering and enhancing coverage of the Pareto front. The FSTMOPSO-based SOFNN (FSTMOPSO-SORBF) can optimize the parameters and determine the number of fuzzy rules, achieving an optimal trade-off between network complexity and prediction accuracy. Furthermore, a detailed convergence analysis of FSTMOPSO-SORBF is provided to support its applicability in real-world scenarios. To verify the effectiveness of the proposed FSTMOPSO-SOFNN, four test cases are used: multi-objective optimization problems, the identification of the Mackey–Glass time series, effluent total phosphorus prediction, and short-term traffic flow forecasting. The simulation results demonstrate that the proposed FSTMOPSO-SOFNN not only achieves high approximation accuracy but also does so with minimal structural complexity.


  • Pipeline leak diagnosis using multisource multiscale attention entropy and enhanced least square twin support vector machine

    by Hongbiao Zhou, Shilin Zhang, Longfeng Li, Le Wang, Congguo Ma

    To achieve rapid and precise identification of water supply pipeline leakage faults, this study introduces a diagnostic framework that integrates multisource multiscale attention entropy (MMATE), an enhanced least-squares twin support vector machine (ELSTSVM), and an adaptive boosting (AdaBoost) algorithm. The MMATE-ELSTSVM-AdaBoost architecture follows a three-stage workflow. First, an improved wavelet threshold denoising (IWTD) technique featuring a novel threshold-processing function is developed to suppress signal noise. Second, MATE values are extracted from the denoised signals, and a multidimensional feature vector is constructed by integrating MATE characteristics across multisource sensors. Third, the ELSTSVM’s objective function is enhanced through a reformulated distance metric, augmented with regularization and compactness terms to optimize model robustness and generalizability. The refined MMATE feature vectors are then fed into the ELSTSVM-AdaBoost ensemble for fault classification. Experimental evaluations demonstrate that the proposed framework significantly outperforms conventional single-feature approaches, standalone models, and state-of-the-art SVM variants, achieving a peak classification accuracy of 99.10% in leakage detection tasks.