Table of Content

20 December 2022 Volume 49 Issue 6

  

Information and Communications Engineering

Precoding scheme for massive MIMO with one-bit DACs based on cross entropy

ZHANG Hangyu, ZHANG Rui, LIAO Fangyuan, LI Yongzhao

Journal of Xidian University. 2022, 49(6):  1-8.  doi:10.19665/j.issn1001-2400.2022.06.001

Abstract ( 329 )   HTML ( 592 )   PDF (953KB) ( 222 )   Save

Figures and Tables | References | Related Articles | Metrics

Massive multiple-input multiple-output (MIMO) systems have the advantages of high spatial resolution,high spectral efficiency,and wide coverage.Utilizing one-bit digital-to-analog converters (DACs) can significantly reduce the power consumption,hardware complexity,and deployment costs in massive MIMO systems.However,due to the quantized noise introduced by the one-bit DACs,the symbols after linear precoding suffer from inevitable distortions.The system bit error rate (BER) performance will significantly fall and reach saturation prematurely with the increase of the signal-to-noise rate (SNR).Unlike linear precoding,nonlinear precoding considers the effect of one-bit quantization,and directly designs the quantized symbols,which can significantly improve the BER.Considering that the symbols after one-bit quantization belong to a finite set with a few elements,we modeled the nonlinear precoding problem from the perspective of combinatorial optimization,and a cross-entropy based algorithm was proposed.The proposed algorithm adaptively updates the probability distribution of each element in the precoding vector at each iteration by minimizing the cross entropy,and can quickly converge to obtain the precoding vector.Meanwhile,the proposed algorithm can be readily extended to the systems with multi-bit DACs.Simulation results show that the proposed algorithm outperforms existing algorithms based on convex optimization in terms of BER under a high SNR,and is robust to channel estimation errors.Besides,the applicability of the proposed scheme to multi-bit DACs is also verified by simulation.

Block Markov superposition transmission of 5G LDPC codes

GUO Kongjing, WANG Qianfan, MA Xiao

Journal of Xidian University. 2022, 49(6):  9-14.  doi:10.19665/j.issn1001-2400.2022.06.002

Abstract ( 212 )   HTML ( 534 )   PDF (866KB) ( 119 )   Save

Figures and Tables | References | Related Articles | Metrics

With the growing demand for high reliability in the enhanced mobile broadband (eMBB) scenario,we present a new class of block Markov superposition transmission codes using 5G low-density parity-check (LDPC) codes as the basic codes.At the transmitter,multiple code blocks (CBs) within a transport block (TB) are encoded separately by 5G LDPC codes,and then superimposed onto the following codewords of the next TB via a row/column interleaver.At the receiver,we present an iterative sliding-window decoding algorithm on a normal graph.Based on the cyclic redundancy check(CRC) in 5G protocols,we also employ a CRC-and-parity-check stopping criterion in the decoding algorithm,with which we can stop the updating procedure of specific codewords which has passed the check.Simulation results show that the proposed scheme using the 5G LDPC code[1 056,528] as the basic code can yield a coding gain of up to 1.2 dB over the conventional 5G LDPC code at the bit-error-rate of 10^-6,and that increasing the size of the decoding window has a limited effect on lowering the error floor.It is also revealed that the complexity of the proposed code is similar to the conventional 5G LDPC codes in the working SNR region.

Robust node placement in TDOA-based multiple sources localization

ZHAO Yue,LI Zan,LI Bing,LU Xiaoju,HAO Benjian

Journal of Xidian University. 2022, 49(6):  15-22.  doi:10.19665/j.issn1001-2400.2022.06.003

Abstract ( 196 )   HTML ( 250 )   PDF (2078KB) ( 96 )   Save

Figures and Tables | References | Related Articles | Metrics

This paper focuses on the scenario consisting of nodes with fixed positions and nodes with flexible positions,and studies the robust placement method of the flexible nodes in the presence of the uncertainty area of sources and uncertainty interval of TDOA noise strength.First,a TDOA-based multiple source localization scenario with fixed nodes and flexible nodes is constructed.Then,the weighted average worst-case CRLB (WAW-CRLB) is proposed to robustly measure the source localization accuracy in the presence of various uncertainties.Second,an optimization problem is formulated with the objective function as the WAW-CRLB and the decision variable as the position vectors of flexible nodes.The problem is then resolved by the genetic algorithm.Simulation results validate the performance advantage of the proposed robust node placement algorithm with two baselines;and verify accuracy improvement with the number of flexible nodes.

Parameter estimation method for LFM signals suppressing impulse noise

ZHANG Yuhong, ZHANG Yixin, ZHANG Chao, BAO Junmin

Journal of Xidian University. 2022, 49(6):  23-31.  doi:10.19665/j.issn1001-2400.2022.06.004

Abstract ( 220 )   HTML ( 361 )   PDF (2604KB) ( 83 )   Save

Figures and Tables | References | Related Articles | Metrics

To solve the problems of a low parameter accuracy,noise-dependent prior knowledge and a limited scope of application in the parameter estimation method of Linear Frequency Modulation under impulse noise,a new parameter estimation method of Linear Frequency Modulation signal is proposed.First,the principle of LV’s Distribution is introduced,and it is pointed out that the performance of the algorithm will be seriously degraded under the impulse noise,and that the parameter estimation ability will fail completely under the strong impulse noise.Then,the property of the nonlinear function Sigmoid is quoted and derived,leading to the result that the phase information of the Linear Frequency Modulation signal before and after the nonlinear function transformation remains unchanged.By combining the LV’s Distribution and the properties of the nonlinear function,using the nonlinear function to process the original signal,and then being divided by its conjugate,the two-dimensional Fourier transform of its symmetric parameter instantaneous autocorrelation function is carried out according to the definition of LV’s Distribution,and the signal is transformed into the Centroid Frequency Chirp Rate domain.Finally,the parameter estimation of the Linear Frequency Modulation signal can be realized according to the peak coordinates.Simulation results show that for the single-component and two-component Linear Frequency Modulation signal,the proposed algorithm can effectively suppress impulse noise and does not depend on the prior knowledge of noise,and can accurately estimate the parameter of the Linear Frequency Modulation signals under strong impulse noise and an extremely low signal-to-noise ratio.The method is simple to implement and has good robustness.Moreover,the parameter estimation of complex signals can be realized,which is beneficial to the application in signal processing.

Stable relay selection method under an uncertain preference ordinal for UAV in post-disaster

XU Zimeng,WANG Bowen,YUN Xiao,WANG Xiaolin

Journal of Xidian University. 2022, 49(6):  32-41.  doi:10.19665/j.issn1001-2400.2022.06.005

Abstract ( 162 )   HTML ( 162 )   PDF (1579KB) ( 78 )   Save

Figures and Tables | References | Related Articles | Metrics

Natural disasters,accident disasters,public health events and social security events are the main types of emergency,which are frequent in China in recent years,so it is urgent to build an all-dimensional and stereoscopic emergency communication support system.To overcome the communication interruption caused by the damage of the ground infrastructure after the disaster,the relay UAVs can be used to realize the interrupted transmission.Aiming at the problem that D2D communication cannot obtain accurate two-side information in the complex environment after the disaster,this paper studies the UAV-assisted relay selection problem of D2D users based on the uncertain preference ordinal from the perspective of matching theory.First,by considering the real-time characteristics of the network,the optimization problem is abstracted as the maximization of the average transmission success rate of D2D users,and the optimization problem is solved by using the stable matching in dynamic scenarios.Second,according to the transmission rate within the predicted flight ranges,the uncertain preference ordinals of both the D2D pairs and the UAVs are obtained.Further the uncertain preference ordinal is comprehensively evaluated to generate corresponding preference lists,and a many-to-one two-side matching model is established on this basis.Third,to avoid the influence of peer effects among the D2D user clusters that reuse the same UAV relay,a two-side exchange-stable matching relay selection algorithm is proposed to ensure the stability of the matching.Finally,simulation results demonstrate that,considering the uncertainty of the post-disaster emergency scenario,the proposed algorithm can effectively improve the transmission success rate of D2D users compared with existing algorithms.

Multi-data mixed FFT processing optimization method

HONG Qinzhi,WANG Zhijun,GUO Yifan,LIANG Liping

Journal of Xidian University. 2022, 49(6):  42-50.  doi:10.19665/j.issn1001-2400.2022.06.006

Abstract ( 218 )   HTML ( 374 )   PDF (1764KB) ( 77 )   Save

Figures and Tables | References | Related Articles | Metrics

A new computational method is proposed to support multiple FFT processing simultaneously.The method can effectively solve the performance loss of memory-based architecture FFT processors due to the computational path pipeline bubbles and the unbalanced throughput of different FFT points.A deeply pipelined WFTA algorithm-based configurable butterfly unit and a new multiple block floating-point processing unit are designed,which can support high precision computing and include one radix-9/two radix-8/three radix-5/four radix-4/five radix-3 in parallel.Based on the proposed method,a multi-mode high performance FFT processor for 4G/5G is implemented,which can support 60 modes including 64~4 096-point FFT/iFFT and 12~3 240-point DFT/iDFT computing.The processor is implemented based on 55 nm CMOS technology,with a 1.46 mm² layout,supports 1.5GS/s in a single mode and 2.2 GS/s in a mixed mode at 500MHz.It is shown that the proposed processor can support more points and has a better performance than previous designs.

Design of a SiGe BiCMOS broadband low noise amplifier

GUO Fei,LIANG Yu,ZHANG Wei,YANG Xue

Journal of Xidian University. 2022, 49(6):  51-57.  doi:10.19665/j.issn1001-2400.2022.06.007

Abstract ( 279 )   HTML ( 161 )   PDF (1522KB) ( 103 )   Save

Figures and Tables | References | Related Articles | Metrics

To meet the design requirements of a low noise amplifier in the receiving front-end of broadband radio frequency communication,a new broadband matching structure of a miller capacitor based on the common emitter stage is proposed.The structure uses the miller capacitance of the heterojunction bipolar transistor,with the load incorporated into the input matching network for design,so as to realize broadband input matching.The structure achieves a sufficiently good low-noise performance.At the same time,it can effectively expand the working frequency band of the amplifier.A broadband low noise amplifier using a 0.13 μm SiGe BiCMOS process is designed.The circuit of the amplifier consists of three amplifiers,with the first stage adopting the Miller capacitor broadband matching structure to reduce noise and realize broadband matching and the latter two stages adopting the common base cascode structure to compensate for the gain.Simulation results show that in the 6~30 GHz band,the gain of the low noise amplifier is 16.5~19.1 dB,the noise figure is 1.43~2.66 dB,the input reflection coefficient S11 is less than -11.9 dB,and the output reflection coefficient S22 is less than -13.7 dB,and that the amplifier is unconditionally stable in the whole frequency band.The DC power consumption of the circuit is 38.7 mW at 1.8 V supply voltage.The overall chip area is 0.88 mm2.The amplifier has a desired comprehensive performance and can be used in a broadband receiving system.

Software PUF with multiple entropy sources based on path sensitization

WANG Pengjun,CHEN Jia,ZHANG Yuejun,ZHUANG Youyi,LI Lewei,NI Li

Journal of Xidian University. 2022, 49(6):  58-66.  doi:10.19665/j.issn1001-2400.2022.06.008

Abstract ( 171 )   HTML ( 384 )   PDF (3616KB) ( 48 )   Save

Figures and Tables | References | Related Articles | Metrics

The Physical Unclonable Function (PUF),as a chip fingerprint,has been widely used in the field of information security.However,the current mainstream PUF designs need to add additional hardware to obtain feature information,and the application in extremely cost-constrained systems faces huge challenges.In this paper,with path sensitization taken as the research object,a software PUF scheme with multiple entropy sources for extracting deviation data from the existing hardware structure is proposed by combining the characteristic of device delay deviation and the uncertainty of register sampling.First,several sets of test patterns are selected to sensitize the target paths and establish the mapping relationship between PUF response and chip feature.Second,the scan chain structure is inserted into the circuit netlist,and different overclocking clock signals are applied in the sampling stage of the trigger to extract the chip abnormal data.Finally,the data is compared with the standard output to count the number of error paths at different clock frequencies,and the PUF response is obtained by random combination of the numbers.Experimental results show that the uniqueness of the proposed PUF is 47.58%,that the randomness is 49.7%,and that the PUF can resist machine learning attacks.

Computer Science and Technology & Artificial Intelligence

Algorithm for segmentation of remote sensing imagery using the improved Unet

LI Jiaojiao, LIU Zhiqiang, SONG Rui, LI Yunsong

Journal of Xidian University. 2022, 49(6):  67-75.  doi:10.19665/j.issn1001-2400.2022.06.009

Abstract ( 534 )   HTML ( 167 )   PDF (3496KB) ( 127 )   Save

Figures and Tables | References | Related Articles | Metrics

Existing remote sensing image segmentation algorithms simply combine edge information with semantic information,they often fail to ensure an overall improvement in semantic modelling.To solve such problems,an improved remote sensing image segmentation algorithm using Unet networks is proposed.The improved algorithm adds an edge extraction module to the base encoder module.The module fuses the semantic feature information of the backbone network and the boundary feature information obtained from the input image by Canny operator and dilated mathematical morphological operations to learn the edges of remote sensing image.To further acquire global information of remote sensing images for improving segmentation accuracy,an edge-guided context aggregation module is proposed.This module enhances the intra-class consistency by capturing the long-distance dependencies between pixels in the boundary region and pixels inside the object,and then aggregates the contextual information.Under the test of the "Tianzhi Cup" AI Challenge dataset,the overall accuracy of the improved model reached about 84.5% and the average intersection ratio reached about 68.6%,with an accuracy improvement of 5.3% and 9.2% respectively compared with the Unet model.The improved model achieved an overall accuracy of 91.2% and 91.6% on the ISPRS Vaihingen and Potsdam benchmark datasets respectively,making it more suitable for accurate remote sensing image segmentation.

Computer Science and Technology & Artificial Intelligence

Online classification jointed RGBT tracking based on the dual attention Siamese network

ZHANG Zhaoyu,TIAN Chunna,ZHOU Heng,TIAN Xilan

Journal of Xidian University. 2022, 49(6):  76-85.  doi:10.19665/j.issn1001-2400.2022.06.010

Abstract ( 406 )   HTML ( 293 )   PDF (2846KB) ( 78 )   Save

Figures and Tables | References | Related Articles | Metrics

The imaging mechanism of visible light and that of thermal infrared are different.Visible light and thermal infrared images reflect different information on the object.A dual-modal visual tracker based on visible light and thermal infrared sequences can comprehensively utilize the inherent correlation and complementarity of two modals,which reduces limitations and uncertainties of single-modal information,and improves the robustness of the visual tracking system.We propose an end-to-end dual-modal tracking algorithm with the Siamese network based on infrared and visible light sequences.The network learns the depth features from the visible light and thermal infrared frames at the same time,and then adaptively fuses the two-model features through intra-modal and cross-modal dual attention mechanisms,which leads to more robust tracking.At the same time,in view of the insufficiency of the Siamese network in distinguishing the target and semantic background,we incorporate the online classification module into the tracking framework.The online learned classifier reduces the interference and adapts to the target changes during tracking.According to experimental results,the proposed algorithm effectively improves the performance of the tracker.Its precision rate and success rate are 90.6% and 73.8% on the RGBT benchmark dataset GTOT,which are 5.5% and 4.3% higher than those of the baseline algorithm.The overall performance is better than that of other advanced tracking algorithms.

Computer Science and Technology & Artificial Intelligence

Bilevel optimization approach for annealing parameter estimation in the image denoising problem

FENG Xiangchu, WEI Lili

Journal of Xidian University. 2022, 49(6):  86-94.  doi:10.19665/j.issn1001-2400.2022.06.011

Abstract ( 177 )   HTML ( 160 )   PDF (2065KB) ( 56 )   Save

Figures and Tables | References | Related Articles | Metrics

One of the important issues in variational image denoising is to select reasonable regularization parameters and use regularization parameters efficiently.The simulated annealing algorithm uses the iterative method to gradually approximate the minimum solution to the energy generalization function.The monotonically increasing regularization parameters are set in its iterative process.In general,the determination of the regularization parameters/annealing parameters and the monotonic increase pattern in the simulated annealing model are empirically based.In this paper,it is desired to learn the optimal annealing parameters adaptively from the data.The new bilevel model for estimating annealing parameters is proposed by combining the bilevel optimization structure with the simulated annealing algorithm.The lower level of the model is the iterative method containing annealing parameters,where the Laplace regularization term is added to ensure good properties of the lower level problem.The upper level problem is the loss function based on the L₂ norm.Meanwhile,this paper proposes an accurate solution algorithm for estimating the annealing parameters by utilizing the back propagation algorithm.A simple interpolation generalization method is given for adapting the proposed model to noise removal problems of different intensities.Experimental results show that the annealing parameters learned adaptively from the data by the algorithm proposed satisfy the assumption of increasing a priori monotonicity of the simulated annealing algorithm.Compared with common regularization parameter selection methods,the proposed algorithm not only ensures the computational efficiency but also improves the denoising effect.Experimental results also verify that the proposed algorithm has a good generalization ability.As shown in the paper,the reasonableness of the monotonic increase of the regularization parameters during the iterative process is demonstrated from the perspective of data learning.Further,it is shown that the proposed algorithm can be used to obtain numerically optimal annealing parameters and variation trends.

Computer Science and Technology & Artificial Intelligence

Implementation of EEG emotion analysis via feature fusion

YANG Liying,MENG Tianhao,ZHANG Qingyang,CHAO Si

Journal of Xidian University. 2022, 49(6):  95-102.  doi:10.19665/j.issn1001-2400.2022.06.012

Abstract ( 319 )   HTML ( 366 )   PDF (1150KB) ( 86 )   Save

Figures and Tables | References | Related Articles | Metrics

Since the EEG signal has the characteristics of non-stationary,weak,and large frequency difference,it is difficult to obtain a higher recognition accuracy.In order to improve the performance of EEG sentiment analysis,this paper conducts research from two aspects:feature extraction and feature selection.In terms of feature extraction,based on the power spectrum intensity,the balanced power spectrum intensity feature (BPSI) is adopted.For feature selection,a feature fusion algorithm FFS is proposed,which combines the Relief and mRMR to reduce the feature dimension and improve the recognition performance.This paper uses the SVM classification algorithm,and carries out experiments on DEAP data.Experimental results show that,compared with the power spectrum intensity,the classification accuracy of the BPSI feature is increased by 6.22% on average.The performance is increased by 3.9 points with features selected by the FFS compared with the baseline,by 1.84 points compared with the Relief,and by 2.05 points compared with the mRMR.The average accuracy of the emotion recognition algorithm based on the BPSI and FFS reaches 88.89% in Valence dimension and 87.73% in Arousal dimension,and meanwhile the average feature dimension is reduced from 160 to 67.

Computer Science and Technology & Artificial Intelligence

High efficient framework for large-scale zero-shot image recognition

ZHANG Zehuan, LIU Qiang, GUO Difei

Journal of Xidian University. 2022, 49(6):  103-110.  doi:10.19665/j.issn1001-2400.2022.06.013

Abstract ( 190 )   HTML ( 170 )   PDF (1497KB) ( 73 )   Save

Figures and Tables | References | Related Articles | Metrics

For large-scale zero-shot image recognition tasks,because of a large number of classes,model training is difficult and training costs of the model are high.In order to solve those problems,this paper designs a high-efficient zero-shot learning framework,which improves the accuracy and generalization ability at low training costs.This framework designs the joint space,uses the image branch network and the semantic branch network to map different modal vectors to the joint space to complete model training and inference.In the image branch network,in order to change the distribution of image feature vectors,this paper uses the perceptron network to map image feature vectors to the joint space.In the semantic branch network,graph convolutional networks are used to map semantic vectors to the joint space.In addition,the loss function is designed to constrain the joint space,so that the discrimination of different classes in the joint space is increased,which is conducive to model training.Experimental results on the ImageNet show that on the “2-HOPS” test set,compared with existing methods without fine-tuning,the accuracy of our algorithm increases by 1.1%,and the training time decreases by 57.8%;compared with existing algorithms after fine-tuning,the accuracy of our algorithm saves 98.4% of training time without any loss of accuracy.Experimental results show that the method improves the model performance with low training costs.

Computer Science and Technology & Artificial Intelligence

Multi-scale fire detection algorithm with an anchor free structure

QIN Rui,ZHANG Wei

Journal of Xidian University. 2022, 49(6):  111-119.  doi:10.19665/j.issn1001-2400.2022.06.014

Abstract ( 169 )   HTML ( 229 )   PDF (2604KB) ( 60 )   Save

Figures and Tables | References | Related Articles | Metrics

In view of the low detection accuracy of multi-scale flames and false alarms in complex backgrounds,a new fire detection algorithm with an Anchor Free structure is proposed.The algorithm cancels the anchor and adopts a point-by-point prediction method to reduce the hyperparameters of the network,thus effectively reducing the influence of artificial prior knowledge.The BFP module is introduced to optimize feature fusion,and the integration of inter-layer information effectively utilizes the global information on features and enhances the expression ability of multi-scale features.The fusion factor is set to control the information transfer between layers,which ensures the fusion of feature information while reducing the influence of high-level features,and improves the learning ability of shallow features for small targets.A dynamic sampling method is designed to adjust the training process and strengthen the network’s learning effect on flame characteristics by adopting the principle of central sampling and confidence to improve the quality of the sample points.The algorithm not only achieves 96.9% accuracy on the self-built dataset,but also has a good performance on the public fire dataset.Experimental results show that the proposed algorithm has a high detection accuracy and a strong anti-interference ability.The algorithm has a good detection effect for multi-scale flames in complex backgrounds,can better suppress the occurrence of false alarms,and meets the needs of actual fire detection tasks.

Mural inpainting algorithm for group sparse based on multi-scale contourlet transform decomposition

CHEN Yong,ZHAO Mengxue,TAO Meifeng

Journal of Xidian University. 2022, 49(6):  120-128.  doi:10.19665/j.issn1001-2400.2022.06.015

Abstract ( 132 )   HTML ( 293 )   PDF (4032KB) ( 48 )   Save

Figures and Tables | References | Related Articles | Metrics

Mural image restoration is a process of recovering the original image from the damaged mural with missing pixels by using the prior information on the original mural image.In view of the problem that sparse representation does not consider the difference between mural structure information and texture information in mural image restoration,resulting in texture blur and structural line fracture,a group sparse mural restoration algorithm based on multi-scale contour wave decomposition is proposed.First,the nonsubsampled contourlet transform is used to decompose the mural image to be repaired into the low-frequency texture component and high-frequency structure component,which overcomes the deficiency that the difference of mural structure and texture information is not considered in the existing sparse representation of mural repair.Then,the proposed improved group sparse algorithm is used to construct the similar group set of sample blocks for the low-frequency components of texture,and the adaptive group dictionary and sparse coefficients are obtained through the iterative optimization of singular value decomposition and the split Bregman iteration algorithm,so as to complete the repair of low-frequency components.Second,the cubic convolution interpolation algorithm is used to realize the interpolation repair of the high-frequency components of the mural structure.Finally,the restored scale components are fused and reconstructed by the inverse transform of the non down sampled contour wave.Through the restoration experiment of real Dunhuang murals,the results show that the proposed method achieves a better subjective and objective restoration effect and evaluation than the comparison algorithm.

Computer Science and Technology & Artificial Intelligence

VAE-Fuse:an unsupervised multi-focus fusion model

WU Kaijun, MEI Yuan

Journal of Xidian University. 2022, 49(6):  129-138.  doi:10.19665/j.issn1001-2400.2022.06.016

Abstract ( 237 )   HTML ( 235 )   PDF (2820KB) ( 74 )   Save

Figures and Tables | References | Related Articles | Metrics

In the multi-focus image fusion problem,in order to preserve as much original image information as possible and improve the quality of image fusion,a two-stage image fusion network based on unsupervised learning is designed by combining the variational autoencoder structure and the gray variance product function in the no-reference image clarity evaluation index.In the training phase,the multi-scale structural similarity is proposed as the loss function and the total deviation loss is introduced to suppress the noise in the image.An encoder-decoder network based on the variational autoencoder structure is constructed to train the original image reconstruction task.In the fusion stage,after using the trained encoder to encode the features of the fused image,the improved gray variance product function method is used to distinguish the clear pixels.The final decision map is generated by mathematical morphology optimization.Finally,the weighted fusion strategy is used to complete the final fusion of the image.Experimental results show that although this method uses fewer model parameters,it retains more original image information in the encoding and decoding process,and is superior to the traditional spatial frequency-based discrimination method in the pixel discrimination process.Compared with a variety of representative image fusion methods,the proposed method has achieved a superior fusion performance in both subjective and objective evaluation.

Computer Science and Technology & Artificial Intelligence

Cyclic wiener filtering for compound fault diagnosis of an aero-engine rolling element bearing

ZHANG Weitao,JI Xiaofan,HUANG Ju,LOU Shuntian

Journal of Xidian University. 2022, 49(6):  139-151.  doi:10.19665/j.issn1001-2400.2022.06.017

Abstract ( 180 )   HTML ( 300 )   PDF (6190KB) ( 53 )   Save

Figures and Tables | References | Related Articles | Metrics

Fault diagnosis of an aero-engine spindle bearing is an important part of engine prognostics and health management.As is known,the diagnosis of the compound fault of an aero-engine spindle bearing is very difficult and easily affected by other vibration interference signals.We present a compound fault diagnosis method of an aero-engine spindle bearing based on blind signal extraction of canonical correlation analysis (CCA) and cyclic Wiener filtering.First,an adaptive conjugate gradient algorithm is proposed for extracting the blind signal by optimizing CCA criterion.Then,combined with the fault feature frequency,the blind signal extraction algorithm is used to extract the fault feature signal from the observed signal.The extracted fault fe12ature signal is regarded as the expected response of the cyclic Wiener filter.Finally,the cyclic Wiener filter is designed to recover the fault signal,and the envelope spectrum of the filtered signal is analyzed to complete the diagnosis of the bearing composite fault.The proposed algorithm overcomes the problem that the existing methods rely on the bearing parameters too much,and that the expected signal obtained from the fixed mathematical model is too ideal to be applied to practical engineering.Both simulated data and experimental data are used to verify the effectiveness of the algorithm in compound fault diagnosis.

Computer Science and Technology & Artificial Intelligence

Edge-cloud collaborative transfer of process knowledge for digital manufacturing monitoring

CAO Xincheng, YAO Bin, HE Wangpeng, CHEN Binqiang, QING Tao

Journal of Xidian University. 2022, 49(6):  152-163.  doi:10.19665/j.issn1001-2400.2022.06.018

Abstract ( 308 )   HTML ( 231 )   PDF (3649KB) ( 63 )   Save

Figures and Tables | References | Related Articles | Metrics

Intelligent manufacturing is the inevitable trend in the development of high-end equipment,and intelligent operation and maintenance are of great significance for ensuring the quality and reliability of digital processing.In the discrete intelligent manufacturing process,process diversity is the key bottleneck restricting the construction and implementation of digital models for intelligent operation and maintenance.This paper proposes an edge-cloud collaborative process knowledge migration scheme,which integrates edge computing and cloud computing to realize the rapid evolution of intelligent operation and maintenance models.First,a parallel multi-scale convolutional network (PMsCNN) in the cloud is trained to abstractly model the degradation process of the equipment under the historical process plan.Then,the unlabeled data samples under the new process plan is used to carry out transfer learning,so that PMsCNN can adapt to the new process plan.For this reason,an improved maximum mean difference loss function is proposed to overcome the problem of data imbalance.Finally,the evolved PMsCNN is applied to edge devices,and intelligent device operation and maintenance are implemented online.By taking the performance operation and maintenance of the equipment core and basic parts as a research case,the advanced nature of the proposed process knowledge migration scheme is verified.Compared with the existing monitoring method based on deep learning,the test accuracy rate under the new process scheme is improved by more than 20%,which is better than that of the existing migration diagnosis method.

Computer Science and Technology & Artificial Intelligence

Quasi-min-max optimization of dynamic output feedback robust MPC

PING Xubin,LIU Siwei,WU Zongyuan,LIU Ding,LI Zhiwu

Journal of Xidian University. 2022, 49(6):  164-176.  doi:10.19665/j.issn1001-2400.2022.06.019

Abstract ( 278 )   HTML ( 394 )   PDF (1135KB) ( 71 )   Save

Figures and Tables | References | Related Articles | Metrics

For unknown system states in constrained linear parameter varying systems with bounded disturbances,a dynamic output feedback robust model predictive control approach via quasi-min-max robust optimization is designed.In the optimization problem,the dynamic output feedback controller takes a parameter-dependent form,and the optimization control problem can be formulated as convex optimization by the techniques of linear matrix inequalities.In the quasi-min-max robust optimization control problem,by constraining the current and predicted closed-loop system states to be within different robust positively invariant sets,and considering the exactly known model parameters at the current sampling time,the conservativeness of the designed dynamic output feedback controller parameters is reduced.Furthermore,the updates on real-time estimation error sets are performed by considering the invariance of the predicted closed-loop system states in the robust positively invariant set,which avoids the requirement of an auxiliary optimization to update estimation error sets in common output feedback robust model predictive control algorithms.The proposed algorithm not only improves the control performance and guarantees recursive feasibility of the optimization control problem,but also reduces the online computational burden on solving the optimization control problem.When the nominal closed-loop system is steered to the origin,the closed-loop system with bounded disturbances is stabilized within a region in the neighborhood of the origin.A simulation example is given to verify the effectiveness of the algorithm.