Wyniki wyszukiwania - Biblioteka Nauki

1

Design of quasi-resonant flyback secondary intrinsically safe power supply

100%

Lu S. , Wang B.

Archives of Electrical Engineering

|

2019

|

tom Vol. 68, nr 1

5--13

EN

In order to improve the efficiency and ensure the security of power supply used in a mine, this paper mainly studies the quasi-resonant flyback secondary power supply and analyzes its operational principles based on the requirements of soft-switching technology. In accordance with the maximum energy of a short-circuit and the request of maximum output voltage ripple, this paper calculates the spectrum value of the output filter capacitor and provides its design and procedures to determine the parameters of the main circuit of power supply. The correctness and availability of this theory are eventually validated by experiments.

2

A robust inversion of logging-while-drilling responses based on deep neural network

80%

Zhu G. , Gao M. , Wang B.

Acta Geophysica

|

2024

|

tom Vol. 72, no. 1

129--139

EN

Resistivity inversion plays a significant role in recent geological exploration, which can obtain formation information through logging data. However, resistivity inversion faces various challenges in practice. Conventional inversion approaches are always time-consuming, nonlinear, non-uniqueness, and ill-posed, which can result in an inaccurate and inefficient description of subsurface structure in terms of resistivity estimation and boundary location. In this paper, a robust inversion approach is proposed to improve the efficiency of resistivity inversion. Specifically, inspired by deep neural networks (DNN) remarkable nonlinear mapping ability, the proposed inversion scheme adopts DNN architecture. Besides, the batch normalization algorithm is utilized to solve the problem of gradient disappearing in the training process, as well as the k-fold cross-validation approach is utilized to suppress overfitting. Several groups of experiments are considered to demonstrate the feasibility and efficiency of the proposed inversion scheme. In addition, the robustness of the DNN-based inversion scheme is validated by adding different levels of noise to the synthetic measurements. Experimental results show that the proposed scheme can achieve faster convergence and higher resolution than the conventional inversion approach in the same scenario. It is very significant for geological exploration in layered formations.

3

Normal Mode Solutions of Target Strengths of Solid-filled Spherical Shells and Discussion of Influence Parameters

70%

Jia B. , Fan J. , Li G.-J. , Wang B. , Chen Y.-F.

Archives of Acoustics

|

2024

|

tom Vol. 49, nr 1

83--93

EN

The normal mode solution for the form function and target strength (TS) of a solid-filled spherical shell is derived. The calculation results of the spherical shell’s acoustic TS are in good agreement with the results of the finite element method (FEM). Based on these normal mode solutions, the influences of parameters such as the material, radius, and thickness of the inner and outer shells on the TS of a solid-filled spherical shell are analyzed. An underwater spherical shell scatterer is designed, which uses room temperature vulcanized (RTV) silicone rubber as a solid filling material and does not contain a suspension structure inside. The scatterer has a good TS enhancement effect.

4

Research on the Performance Optimization of Turbulent Self-Noise Suppression and Sound Transmission of Acoustic Windows Made from Functionally Graded Material

70%

Li B. , Zhou F.-L. , Fan J. , Wang B. , Tan L.

Archives of Acoustics

|

2023

|

tom Vol. 48, No. 4

475--495

EN

For a simplified sonar dome model, an optimization method for internal gradients of functionally graded material (FGM) acoustic windows is proposed in this paper. This method can be used to design optimized FGM acoustic windows with better turbulent self-noise suppression and sound transmission performances. A theoretical model of FGM acoustic windows to evaluate the reduction of self-noise caused by the turbulent boundary layer (TBL) pulsating pressure and the sound transmission loss (STL) is derived through the double Fourier transform and the wavenumber frequency spectrum analysis, respectively, based on the transfer matrix idea and the classical elastic theory. The accuracy of the theory is verified by the finite element results of COMSOL Multiphysics. Utilizing the genetic algorithm (GA) and taking the monotonic gradient as the constraint condition, the internal gradient optimization method of FGM acoustic windows obtains the optimization variables in the Bernstein polynomial when the optimization objective is minimized by iterating the optimization variables in the deviation function represented by the Bernstein polynomial that is introduced in the gradient function. The STL, the turbulent self-noise reduction or a weighting function of the STL and turbulent self-noise reduction of FGM acoustic windows is chosen as the optimization objective. The optimization calculation of the sound transmission or turbulent self-noise suppression performances is carried out for the FRP-rubber FGM (FGM with fiber reinforced plastic (FRP) as the substrate material and rubber as the top material) acoustic window. The optimized results show that both the sound transmission and turbulent self-noise suppression performance are effectively improved, which verifies the effectiveness of the optimization method. Finally, the mechanism of the sound transmission and self-noise suppression characteristics before and after optimization are explained and analyzed based on the equivalent model of graded materials. The research results of this paper provide a reference value for the future design of FGM acoustic windows for sonar domes.

5

Observation and Modelling on the Shipping Noise in Shallow Waters with Complex Islands and Reefs of the East China Sea

70%

Peng Z. , Zhou F. , Fan J. , Wang B. , Wen H.

Archives of Acoustics

|

2021

|

tom Vol. 46, No. 2

301--311

EN

The impact of the noise radiated from merchant ships on marine life has become an active area of research. In this paper, a methodology integrating observation at a single location and modelling the whole noise field in shallow waters is presented. Specifically, underwater radiated noise data of opportunistic merchant ships in the waters of Zhoushan Archipelago were collected at least one day in each month from January 2015 to November 2016. The noise data were analyzed and a modified empirical spectral source level (SSL) model of merchant ships was proposed inspired by the RANDI-3 model (Research Ambient Noise Directionality) methodology. Then combining the modified model with the realistic geoacoustic parameters and AIS data of observed merchant ships, the noise mappings in this area were performed with N × 2D of Normal Mode calculations, in which the SSL of each ship was estimated using the modified model. The sound propagation at different receiving positions is different due to the shielding effect of islands and bottom topography. The methodology proposed in this paper may provide a reference for modelling shipping noise in shallow waters with islands and reefs.

6

Comparisons between 2D and 3D MPFEM Simulations in Modeling Uniaxial High Velocity Compaction Behaviors of Ti-6Al-4V Powder

70%

Zhou J. , Xu H. , Zhu C. , Wang B. , Liu K.

Archives of Metallurgy and Materials

|

2022

|

tom Vol. 67, iss. 1

57--65

EN

Multi-particle finite element method (MPFEM) simulation has been proven an efficient approach to study the densification behaviors during powder compaction. However, comprehensive comparisons between 2D and 3D MPFEM models should be made, in order to clarify which dimensional model produces more accurate prediction on the densification behaviors. In this paper, uniaxial high velocity compaction experiments using Ti-6Al-4V powder were performed under different impact energy per unit mass notated as Em. Both 2D and 3D MPFEM simulations on the powder compaction process were implemented under displacement control mode, in order to distinguish the differences. First, the experimental final green density of the compacts increased from 0.839 to 0.951 when Em was increased from 73.5 J/g to 171.5 J/g. Then detailed comparisons between two models were made with respect to the typical densification behaviors, such as the density-strain and density-pressure relations. It was revealed that densification of 2D MPFEM model could be relatively easier than 3D model for our case. Finally, validated by the experimental results, 3D MPFEM model generated more realistic predictions than 2D model, in terms of the final green density’s dependence on both the true strain and Em. The reasons were briefly explained by the discrepancies in both the particles’ degrees of freedom and the initial packing density.

7

Developing Hessian-free second-order adversarial examples for adversarial training

61%

Qian Y. , Zhang L. , Wang Y. , Ji B. , Yao T. , Wang B.

International Journal of Applied Mathematics and Computer Science

|

2024

|

tom Vol. 34, no. 3

425--438

EN

Recent studies show that deep neural networks (DNNs) are extremely vulnerable to elaborately designed adversarial examples. Adversarial training, which uses adversarial examples as training data, has been proven to be one of the most effective methods of defense against adversarial attacks. However, most existing adversarial training methods use adversarial examples relying on first-order gradients, which perform poorly against second-order adversarial attacks and make it difficult to further improve the robustness of the model. In contrast to first-order gradients, second-order gradients provide a more accurate approximation of the loss landscape relative to natural examples. Therefore, our work focuses on constructing second-order adversarial examples and utilizing them for training DNNs. However, second-order optimization involves computing the Hessian inverse, which typically consumes considerable time. To address this issue, we propose an approximation method that transforms the problem into optimization within the Krylov subspace. Compared with the Euclidean space, the Krylov subspace method typically does not require storing the entire matrix. It only needs to store vectors and intermediate results, avoiding explicitly calculating the complete Hessian matrix. We approximate the adversarial direction by a linear combination of Hessian-vector products in the Krylov subspace to reduce the computation cost. Because of the non-symmetrical Hessian matrix, we use the generalized minimum residual to search for an approximate polynomial solution of the matrix. Our method efficiently reduces computational complexity and accelerates the training process. Extensive experiments conducted on the MNIST, CIFAR-10, and ImageNet-100 datasets demonstrate that our adversarial learning using second-order adversarial samples outperforms other first-order methods, leading to improved model robustness against various attacks.