bob手机在线登陆学报自然版

2021, 41(12): .

Abstract(126) PDF(9)

Abstract:

2021, 41(12): .

Abstract(141) PDF(10)

Abstract:

Multi-Source Sensor Fault Diagnosis Method Based on Improved CNN-GRU Network

MA Liling, GUO Jian, WANG Shoukun, WANG Junzheng

2021, 41(12): 1245-1252. doi:10.15918/j.tbit1001-0645.2020.183

Abstract(185) PDF(24)

Abstract:
A fault diagnosis method for multi-source sensors in complex systems was proposed. Based on the correlation between multi-source sensor data,a convolutional neural network (CNN) was used to extract the connections and features between different sensors. In the convolutional neural network,a sensor data calibration module was designed to make the network pay more attention to learning sensor data related to fault signals. Recurrent neural networks were used to model the time series of sensors,and jump connections and auxiliary loss functions were added to the network to reduce the difficulty of network training. Finally,based on the temporal and spatial characteristics,the results of the fault classification and the estimated values of the fault parameters were obtained at one time. The simulation results show that the improved CNN-GRU network can accurately diagnose fixed deviation fault and drift deviation fault of sensors in real time. The sensor data calibration module and the jump connection can effectively improve the accuracy and precision of the diagnosis algorithm.

Global Organization Network for Facial Attribute Editing

DAI Zhongjian, GU Xiaowei

2021, 41(12): 1253-1261. doi:10.15918/j.tbit1001-0645.2020.195

Abstract(160) PDF(14)

Abstract:
In this paper,a novel global organization network for facial attribute editing was proposed based on a generative adversarial network. Facial attribute editing is to generate face images with desired attributes by combining the encoder-decoder structure with GAN. However,the traditional encoder-decoder structure has limited ability to reconstruct face images and edit attributes. Directly combining the encoder features with the attribute label,the method can result in poor attribute editing performance due to the incorporation of the encoder features,while,face restoration degree degrades because of the absence of the encoder features,and the two can not be balanced. Therefore,the global organization units (GOU) andU-shaped transferring method were proposed.U-shaped transferring method was arranged to change the traditional attribute flow mode and generate inverted states. Combining with the inverted states,the global organization unit was used to generate global state,build a bridge between the encoder and decoder,and help the decoder better integrate encoder features and attribute information. Meanwhile,in order to better fit the global organization unit,an encoder down-sampling was redesigned. Experimental results show that the proposed method can improve the ability of face reconstruction and attribute editing simultaneously.

PM_2.5Concentration Prediction Based on PCA-OS-ELM

LI Jihan, LI Xiaoli, WANG Kang, CUI Guimei

2021, 41(12): 1262-1268. doi:10.15918/j.tbit1001-0645.2020.199

Abstract(192) PDF(8)

Abstract:
In order to improve the prediction accuracy of PM_2.5concentration,a method based on the principal component analysis and online sequential extreme learning machine (PCA-OS-ELM) was proposed to predict PM_2.5concentration in this paper. Firstly,principal component analysis (PCA) was used to extract the key variables affecting air quality in high-dimensional atmospheric data,and remove unnecessary redundant variables. Secondly,an online sequential extreme learning machine (OS-ELM) network prediction model was established by using the extracted key variables. Finally,the training data and network parameters were continuously updated to realize the rapid prediction of PM_2.5concentration by combining batch processing with successive iteration. The results show that,taking different batches of training data to update the model,the PCA-OS-ELM prediction method can quickly realize the prediction of atmospheric PM_2.5concentration,proving the effectiveness of the proposed method. Compared with other methods,this method shows little prediction error,higher prediction accuracy and better practical value.

Research on Numerical Simulation Method of Direct Field Acoustic Test Platform

RONG Jili, ZHANG Bohan, CHENG Xiuyan, FAN Bochao, QIN Zhaohong, LI Haibo, WEI Long

2021, 41(12): 1269-1276. doi:10.15918/j.tbit1001-0645.2021.049

Abstract(160) PDF(5)

Abstract:
In order to guide the construction of direct field acoustic test platform and replace high cost reverberation field acoustic test,a method of equivalent loudspeaker with velocity boundary was proposed. Based on this method,a finite element model of direct field acoustic test was established,and the responses of sound pressure level (SPL) in different regions were compared with that of the reverberation field. The results show that the velocity boundary equivalent method can accurately simulate the directivity and SPL response of the loudspeaker,and the response of the direct field model based on this method are consistent with the reverberation field near the center of the field,which can provide a reference for the selection of reliable test area in the direct field acoustic test.

Research on Classification of Commodity Ultra-Short Text Based on Deep Random Forest

NIU Zhendong, SHI Pengfei, ZHU Yifan, ZHANG Sifan

2021, 41(12): 1277-1285. doi:10.15918/j.tbit1001-0645.2019.108

Abstract(138) PDF(6)

Abstract:
In recent years,with the development of mobile communication and information technology,more and more ultra-short text data with a length of no more than 20 words and no auxiliary tag information need to be processed on the network and in actual application scenarios. Because of inherent ambiguity and feature sparseness of ultra-short text,obvious lack of context,and difficulty in distinguishing semantics,an effective classification method is needed in the field of text categorization. To solve the performance problem of those classifiers based on the traditional short text classification method KNN and the decision tree,a new method was proposed based on deep random forest for the classification of commodity short texts. Using a "diversion" strategy and taking an external knowledge base as assistance,the method was arranged to directly determine the commodity name with the clear category in the knowledge base,and to vectorize the description of the incapable extracted commodity name based on a Word2vec tool. And then the vectors in the external knowledge base were classified according to deep random forest. Finally,the classifier was continually optimized until the threshold of training set size was reached. The experimental results show that compared with the traditional classification method KNN and decision tree,the classification method proposed in this paper can improve the average accuracy by 22.78% and 17.22%,and the average recall rate by 22.85% and 15.23% respectively.

DOA Estimation Based on Unknown Mutual Coupling Model

LUO Xue

2021, 41(12): 1286-1292. doi:10.15918/j.tbit1001-0645.2019.160

Abstract(135) PDF(4)

Abstract:
In array signal processing,mutual coupling between sensors has an adverse effect on the estimation of parameters. A new direction of arrival (DOA) estimation algorithm for unknown mutual coupling model was proposed to reduce the effect of mutual coupling. Utilizing the idea of iteration,the algorithm was arranged firstly to preliminary estimate DOA in the whole array. Then the angularly dependent coefficients (ADCs) were estimated based on the well-known subspace theory to further improve the accuracy of DOA estimation. The mutual coupling coefficients were finally determined by solving the least squares problem,and starting the next iteration with the new DOA as the preliminary value. Simulation results show that the proposed method can achieve better performance and more robust than other DOA estimation algorithm for unknown mutual coupling model,in particular at a low signal-to-noise ratio (SNR) or a small-sized array.

Sequence Optimization Classifier Chain Based on Label-Specific Features and Causal Discovery

LUO Senlin, WANG Haizhou, PAN Limin

2021, 41(12): 1293-1299. doi:10.15918/j.tbit1001-0645.2019.224

Abstract(119) PDF(3)

Abstract:
Classifier chain is an important multi-label classification method to mine multi-dimensional label information of specific objects by using the correlation between labels. To solve the problems in the existing classifier chain algorithm,including the redundancy of learning features caused by the base learner training of each label in the complete feature space,and the low efficiency of information utilization among labels caused by the random sequence of label learning and the one-way non-feedback in the training process of classifier chain,a sequence optimization classifier chain based on label-specific features and causal discovery was proposed. In this method,affine propagation clustering was used to construct advanced structured features for each base learner,reducing the difficulty of training single label nodes. At the same time,conditional entropy was used to mine the causal relationship between labels,optimize the learning sequence and improve the utilization density of relevant information between labels. The experimental results on several open datasets show that the sequential optimization classifier chain can effectively enhance the learning effect of single node and the utilization of correlation information between multi-labels,and improve the classification effect of multi-labels,possessing high practical value.

Signal-to-Noise Ratio Estimator with Fast Convergence Based on Empirical Distribution Function

WANG Yongqing, ZHAO Shiqi, SHEN Yuyao, MA Zhifeng

2021, 41(12): 1300-1306. doi:10.15918/j.tbit1001-0645.2021.020

Abstract(184) PDF(8)

Abstract:
Empirical distribution function (EDF)-based estimators are effective for various multilevel constellations in a wide signal-to-noise ratio (SNR) range via the Kolmogorov-Smirnov test. However,there are numerous addition and matching operations between reference cumulative distribution functions (CDFs) and the EDF. A signal-to-noise ratio estimator through continuous iteration with a linear polynomial to accelerate the matching procedure was proposed. On the premise of estimation accuracy,using the idea of "direct substitution curve",the zero point of the maximum distance curve was iteratively approximated by the root of the linear polynomial,and the SNR corresponding to the zero point was used as the estimation value of the received signal. The simulation results show that compared with the original algorithm,the iteration number of the proposed strategy is reduced by more than 90%,which greatly reduces the matching complexity and computational complexity. Compared with the existing reduced-complexity iterative strategy,the proposed strategy exhibited faster convergence and better estimation performance.

Analysis of Ground-Based and Space-Based Optical Observation System Warning Capability of Near-Earth Asteroids

YANG Xu, ZHAO Kexin, GAN Qingbo, LIU Jing, YAO Yongqiang

2021, 41(12): 1307-1313. doi:10.15918/j.tbit1001-0645.2021.033

Abstract(152) PDF(16)

Abstract:
In response to the threat of near-Earth asteroid impacts,the short-term warning capabilities of various typical ground-based and space-based optical observation systems for near-Earth asteroids (NEAs) were analyzed. The concept of short-term warning circle was proposed. Using the visible reflection model of celestial bodies in the solar system,the short-term warning capabilities of large-aperture ground-based telescope LSST,and three typical space-based telescopes deployed in the orbits of Sun-Earth L1 point,Venus-like orbit and distant retrograde orbit were simulated and calculated and the early warning capabilities of three space-based systems for NEAs approaching from the direction of the sun were compared. The results show that the early warning capability of L1 was good but the available warning time was short relatively; the Venus-like orbit constellation had a long warning time and wide coverage area,but the existence of phase gaps would cause a big omission probability; distant retrograde orbit had the advantages of high coverage rate and long warning time to the NEAs from the direction of the sun. The deployment of several space-based telescopes can provide sufficient supplement to the ground-based observation system,solve the problem of coverage omission in the direction of the sun,and provide approaching warning of NEAs in all directions.

Radiation-Hardened by Design Techniques to Mitigate Single- Event Transients in Voltage-Controlled Delay Line

SHI Zhu, WANG Bin, ZHAO Yanpeng, YANG Bo, LU Hongli, GAO Lijun, LIU Wenping

2021, 41(12): 1314-1321. doi:10.15918/j.tbit1001-0645.2100.153

Abstract(165) PDF(5)

Abstract:
The voltage-controlled delay line (VCDL) is one of the most sensitive subcircuits to single event (SE) in delay-locked loops (DLLs),which consists of a bias circuit and voltage-controlled delay cells. The sensitivity of VCDL in a DLL to single-event transient (SET) was analyzed based on double exponential current source and 3-D TCAD mixed-mode simulation. According to the difference in the severity of SET response and circuit structure,the bias circuit was hardened by analog redundancy,while a SET detection circuit was proposed for voltage-controlled delay cells. Simulations,making under the condition of 80 MeV·cm²/mg linear energy transfer (LET) values,1.2 V supply voltage and 1 GHz input reference clock,show the perturbed magnitude of biasing voltages,V_bnandV_bp,can be significantly reduced by 75% and 60%,respectively,completely eliminating missing pulses of output signals compared with the unhardened one. The proposed detection circuit can indicate SET response in voltage-controlled delay cells under different circumstances,improving the reliability of output signals in the DLL.

Preparation of Al/C Composite Particles by Mechanical Milling and Its Effect on Property of HTPB Propellants

LI Guoping, CEN Zhuoqi, REN Xin, LUO Yunjun, TAO Weibin

2021, 41(12): 1322-1330. doi:10.15918/j.tbit1001-0645.2020.193

Abstract(154) PDF(7)

Abstract:
In order to improve the activity and combustion efficiency of aluminum powder,a mechanical ball milling method was used to prepare Al/C composite particles and the milling conditions were optimized. The results show that,when the carbon content is 30%,the ratio of balls mass to powder mass is 14,the rotating speed is 350 rad/s,hexane is used as an inert atmosphere and the milling time is 2 hours,the sizes of Al and C in the composite particles are 33.2 nm and 42.5 nm respectively,the relative effective activity can reach up to 70.33%. The effects of composite particles on the viscosity,mechanics,explosion heat,burning rate and other properties of composite propellant were studied respectively. The results suggest that composite particles can increase the viscosity of HTPB propellant,but the system is a typical shear thinning fluid. The more composite particles are added,the greater the degree of shear thinning will be. Therefore,the propellants containing composite particles can be formed with high quality by selecting appropriate addition amount and processing conditions. And particles will improve the mechanical strength of HTPB propellant. When the particles content increases from 0% to 30%,the average maximum tensile strength and fracture strength enhance from 0.37 MPa and 0.32 MPa to 0.75 MPa and 0.69 MPa respectively. In addition,composite particles can significantly improve the explosive heat and burning rate of the propellant. For example,taking 15% composite particles as the replacement amount,the explosive heat and burning rate can be increased by 9.91% and 48.27%,respectively,compared with the blank.

Submission System

Author Related

Writing Norms

Links

2021 Vol. 41, No. 12

Archive