Ying Zhang’s research while affiliated with Guizhou University and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (23)


Flowchart of NB-LTO preparation.
SEM and EDS mapping images of Na and Br during the heating of NB-LTO from 500 to 900 °C: a–c NB-LTO-500 °C, d–f NB-LTO-700 °C, and g–i NB-LTO-900 °C.
XRD spectrum during heating at 900 °C: a partial enlargement of LTO, b partial enlargement of NB-LTO, c relative content curves of various phases in LTO, and d relative content curves of various phases in NB-LTO.
XPS spectra of LTO and NB-LTO during 900 °C heating: a full XPS spectra of LTO and NB-LTO, b Ti2p, c O1s, d Li1s, e Na1s, and f Br3d.
Diffusion path of Li⁺ in the TiO2 unit cell. The gradient green sphere represents the diffusion process of Li⁺. Numbers 1, 2, 3, and 4 are the four sites that Li⁺ finally reaches through diffusion after structure optimization.

+2

Exploring the evolution behavior of Li, Na, and Br during the continuous phase transition of lithium titanate
  • Article
  • Publisher preview available

January 2025

Journal of Materials Science

Ying Zhang

·

Hang Xiong

·

Bo Zhu

·

[...]

·

Benjun Xu

This study examines the continuous phase transition of NaBr-doped Li4Ti5O12 during calcination, focusing on the migration and evolution of Li⁺, Na⁺, and Br⁻ and their impact on the transition. Results from high-temperature in situ X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and density functional theory (DFT) calculations indicate that NaBr incorporation increased the TiO2–Li2TiO3 transition temperature and reduced the Li2TiO3–Li4Ti5O12 transition temperature. An appropriate amount of Li⁺ was doped into TiO2 at low temperatures. Notably, with increasing temperature, Na⁺ was doped into the TiO2 cell through the gap, while Br⁻ was adsorbed on the surface of TiO2 without entering the TiO2 cell. With the continuous increase in the temperature and doping of Li⁺, TiO2 transforms into Li2TiO3. Na⁺ and Br⁻ gain more energy, Na⁺ enters the Li2TiO3 unit cell for gap doping, and Br⁻ enters the Li2TiO3 unit cell to replace O²⁻, promoting the transformation of Li2TiO3 to Li4Ti5O12. Overall, this research provides an intrinsic connection between the microscopic properties of anions and cations during NaBr-doped Li4Ti5O12 phase transition, clarifies the states of Li⁺, Na⁺, and Br⁻ in this transition, and offers a theoretical basis for the states of anions and cations during continuous Li4Ti5O12 phase transition.

View access options


Time-Varying Formation-Surrounding Control for Multi-Quadrotors Pursuit-Evasion Games with Disturbances and Collision Avoidance

January 2024

·

1 Read

IEEE Transactions on Aerospace and Electronic Systems

This article addresses the time-varying formation-surrounding control (FSC) problem for the multiple quadrotors pursuit-evasion (MQPE) games subject to time-varying disturbances and collision constraints. Firstly, two appointed-fixed-time observers are presented to estimate the position and velocity of the evader leader within the user-appointed time. Then, based on the H robust control theory and artificial potential function (APF), the novel cost functions are designed for quadrotors to deal with disturbances and collision problems. Subsequently, the adaptive dynamic programming (ADP)-based time-varying FSC schemes are proposed to solve the MQPE games, which employ critic-only neural networks (NNs) to approximate the optimal cost functions. Finally, rigorous theory analyses and numerical simulation examples are provided to demonstrate the Nash equilibrium of MQPE games and the effectiveness of the proposed control schemes.



FedSiM: A Similarity Metric Federal Learning Mechanism based on Stimulus Response Method with Non-IID Data

September 2023

·

38 Reads

·

2 Citations

FL (Federal Learning) based on parameter sharing under the assumption that the data obey IID (Independent Identical Distribution) has already achieved good results in areas such as fault diagnosis. Data collected by the decentralized devices often do not obey IID. However, when faced with the scenario of client data obeying Non-IID distribution, its diagnostic accuracy is usually weak. Based on this, we did an investigation on the mechanism causing this phenomenon and found that it was attributed to the weight shift of the network. Therefore, based on the stimulus response principle, we investigated the network similarity of federal clients under different data distributions and explain the reasons for the weight shift. Firstly, it was pointed out that there are differences in the regions where the network is activated when performing different classification tasks. Then, FedSiM (Similarity Metric Federal Learning) was proposed based on the principle that there are differences between the activated regions. Finally, experiments were designed on the Case Western Reserve University bearing failure dataset for different degrees of independent identical distribution cases. The results show that FedSiM can improve the diagnostic accuracy by 15.8 percentage points in the case of Non-IID, and a few shared FedSiM methods to further improve the accuracy were also given.




Figure 4. Structure of the adaptive multilayer filter algorithm. The pseudo code for the adaptive multilayer filtering algorithm is shown as Algorithm 2: Algorithm 2 Adaptive Multilayer Filter Input: n: number of nodes LGP: The link generation probability matrix L i : The node maximum load matrix Output: Fm: The filtered matrix 1: The First Layer: Connectivity-based filtering 2: Max X : Top 2 of highest probability in each node based on LGP 3: th 1 ←The minimum value of Max X 4: for i = 0 → n do 5: for j = 0 → n do 6: if i = j and LGP[i, j] ≥ th 1 then 7: Fm[i, j] ← 1 8: else 9: Fm[i, j] ← 0 10: end if 11: end for 12: end for 13: 14: The Second Layer: Power-law-based filtering 15: k: Node initial level 16: Max X : Top k of highest probability in each node based on LGP 17: th 2 ←The minimum value of Max X 18: for i = 0 → n do 19: for j = 0 → n do 20: if i = j and LGP[i, j] ≥ th 1 then 21: Fm[i, j] ← 1 22: else 23: Fm[i, j] ← 0 24: end if 25: end for 26: end for
Figure 6. Network topology before and after algorithm optimization: (a) initial network; (b) optimized network.
Figure 8. Effect of different values α on algorithm optimization: (a)
Figure 9. Effect of different values c on algorithm optimization: (a) c = 0.1; (b) α = 0.3; (c) α = 0.5; (d) c = 0.7; (e) c = 0.8; (f) c = 0.9.
Network invulnerability entropy measures and robustness metric before and after algo- rithm optimization.
VGAE-AMF: A Novel Topology Reconstruction Algorithm for Invulnerability of Ocean Wireless Sensor Networks Based on Graph Neural Network

April 2023

·

38 Reads

·

6 Citations

Journal of Marine Science and Engineering

Ocean wireless sensor networks (OWSNs) play an important role in marine environment monitoring, underwater target tracking, and marine defense. OWSNs not only monitor the surface information in real time but also act as an important relay layer for underwater sensor networks to establish data communication between underwater sensors and ship-based base stations, land-based base stations, and satellites. The destructive resistance of OWSNs is closely related to the marine environment where they are located. Affected by the dynamics of seawater, the location of nodes is extremely easy to shift, resulting in the deterioration of the connectivity of the OWSNs and the instability of the network topology. In this paper, a novel topology optimization model of OWSNs based on the idea of link prediction by cascading variational graph auto-encoders and adaptive multilayer filter (VGAE-AMF) was proposed, which attenuates the extent of damage after the network is attacked, extracts the global features of OWSNs by graph convolutional network (GCN) to obtain the graph embedding vector of the network so as to decode and generate a new topology, and finally, an adaptive multilayer filter (AMF) is used to achieve topology control at the node level. Simulation experiment results show that the robustness index of the optimized network is improved by 39.65% and has good invulnerability to both random and deliberate attacks.



On the Sparse Gradient Denoising Optimization of Neural Network Models for Rolling Bearing Fault Diagnosis Illustrated by a Ship Propulsion System

September 2022

·

38 Reads

·

7 Citations

Journal of Marine Science and Engineering

The drive rolling bearing is an important part of a ship’s system; the detection of the drive rolling bearing is an important component in ship-fault diagnosis, and machine learning methods are now widely used in the fault diagnosis of rolling bearings. However, training methods based on small batches have a disadvantage in that the samples which best represent the gradient descent direction can be disturbed by either other samples in the opposite direction or anomalies. Aiming at this problem, a sparse denoising gradient descent (SDGD) optimization algorithm, based on the impact values of network nodes, was proposed to improve the updating method of the batch gradient. First, the network is made sparse by using the node weight method based on the mean impact value. Second, the batch gradients are clustered via a distribution-density-based clustering method. Finally, the network parameters are updated using the gradient values after clustering. The experimental results show the efficiency and feasibility of the proposed method. The SDGD model can achieve up to a 2.35% improvement in diagnostic accuracy compared to the traditional network diagnosis model. The training convergence speed of the SDGD model improves by 2.16%, up to 17.68%. The SDGD model can effectively solve the problem of falling into the local optimum point while training a network.


Citations (15)


... The paper by Wang and Zhang [16] introduced Faddism, a similarity metric federal learning mechanism designed to improve diagnostic accuracy in federated learning (FL) with Non-IID data. Faddism uses a stimulus response distance (SR-distance) to measure the similarity of network parameters among clients and incorporates a similarity metric (Sims) loss function to reduce the impact of weight shift caused by different data distributions. ...

Reference:

Novel practical predictive maintenance measurement solutions for industrial systems
FedSiM: A Similarity Metric Federal Learning Mechanism based on Stimulus Response Method with Non-IID Data

... During the operation of the ship's hydraulic system, some contaminants will be mixed into the hydraulic oil, and when their content exceeds a certain limit, thus affects the normal operation of the whole system. Therefore, in order to avoid the failure of the ship's hydraulic equipment, it is necessary to detect the solid metal particle contaminants in the hydraulic oil [3,4]. ...

On the Sparse Gradient Denoising Optimization of Neural Network Models for Rolling Bearing Fault Diagnosis Illustrated by a Ship Propulsion System

Journal of Marine Science and Engineering

... Among the papers that reported that the use of an explanation method impacts the FL training, a salient subset uses XAI to defend the FL process from the negative effects of defaulting nodes [33], malicious behavior from certain FL nodes [52], [53], and instances of the GAN attack in FL [54]. Other positive effects reported include improved accuracy in [55], [36], [56] and enhanced learning efficiency [57]. It must be noted that such benefits stem from incorporating XAI as a component of the FL algorithm design. ...

Multi-Level Federated Network Based on Interpretable Indicators for Ship Rolling Bearing Fault Diagnosis

Journal of Marine Science and Engineering

... In recent years, few-shot learning has emerged as a research hotspot in deep learning [9]. Techniques such as meta-learning, metric learning, and data augmentation have been proposed to address the poor performance of deep learning methods in few-shot environments, and these methods are equally applicable in the field of cybersecurity [10]. By simulating few-shot scenarios to optimize the learning process of models, their ability to recognize novel and rare threats has been enhanced, with relevant algorithms showing preliminary success in the cybersecurity domain. ...

On IoT intrusion detection based on data augmentation for enhancing learning on unbalanced samples
  • Citing Article
  • March 2022

Future Generation Computer Systems

... Recently, models based on stochastic differential equations (SDEs) have played an important role across various fields of science and industry [1][2][3][4]. In particular, they are extensively applied in biology and neural networks, which has led to considerable attention being given to the stability analysis of differential equations [5][6][7][8][9][10][11][12][13][14]. Additionally, the process of synaptic transmission in nervous systems involves noise and can be regarded as stochastic perturbation [15]. ...

Photovoltaic power prediction based on hybrid modeling of neural network and stochastic differential equation
  • Citing Article
  • November 2021

ISA Transactions

... Further, Routing schemes based in RL have been applied to diverse networks such as mobile ad-hoc network (MANET) [13,14], vehicular adhoc networks (VANET) [15], wireless sensor networks (WSN) [16][17][18], wireless mesh networks (WMN) [19] and delay tolerant networks (DTN) [20] for performance improvement. RL-based routing has also been employed in other contexts, such as underwater sensor networks [21,22], software-defined networks [23,24], and information-centric networks [25]. ...

Reinforcement Learning-Based Opportunistic Routing Protocol for Underwater Acoustic Sensor Networks
  • Citing Article
  • February 2021

IEEE Transactions on Vehicular Technology

... In the DCRRP protocol, during the clustering process, a cluster head is selected from the high-energy and short-distance clusters of the mobile receiver. The members are [23] Clusters also join the cluster head based on distance. Still, because of sink mobility, the sink may exist in another area and collect data, forcing the sensor node containing the information to send it to the node that replaces it. ...

FW-PSO Algorithm to Enhance the Invulnerability of Industrial Wireless Sensor Networks Topology

Sensors

... The studies on resilient algorithms originated in wireless sensor networks (WSNs) under static scenarios [7]- [10]. Subsequently, the mobility-based resilient algorithms were developed for mobile WSNs [11]- [20], and were later extended for USNETs [21]- [24]. The mobility-based algorithms can be divided into three categories. ...

A Novel Hybrid Optimization Scheme on Connectivity Restoration Processes for Large Scale Industrial Wireless Sensor and Actuator Networks

Processes

... The dilemma is that improving tracking accuracy requires more nodes to track the target and consumes more energy [4,5]. Therefore, it is important to investigate how to balance tracking accuracy and energy consumption to prolong the network lifetime while obtaining favorable tracking performance [6]. ...

Sensor-Networked Underwater Target Tracking Based on Grubbs Criterion and Improved Particle Filter Algorithm

IEEE Access

... This section offers background on ML techniques, namely FL and FSL, alongside the Hyperband algorithm used for hyperparameter optimization in this study. [18] 2019 FANETs FANET Dataset, CRAWDAD VANET dataset [22] Distributed Training [17], 2020 FANETs FANET Dataset, CRAWDAD VANET dataset [22] Distributed Training [8] 2023 UAV swarm UAV Attack Dataset [29] Distributed Training [6] 2023 FANETs FANET Dataset Distributed Training FSL [9] 2021 IoT UNSW-NB15 [16], Bot-IoT [11] Central Training [15] 2023 IoT UNSW-NB15 [16], TON IoT [3], a IoT dataset Central Training [13] 2023 IoT CIC-DDoS2019 [33], CIC-IDS2017 [21], Central Training CSE-CIC-IDS2018 [25], NSL-KDD [26], and UNSW-NB15 [16] FSL + FL Our study 2024 FANETs FANET Dataset Distributed Training ...

Intrusion Detection for IoT Based on Improved Genetic Algorithm and Deep Belief Network

IEEE Access