Office Online  
    Office Online
    Manuscript Tracking
    Editor Work
    Office Work
    Editor-in-chief
Journal Online
    Current Issue
    Advanced Search
    Archive
    Read Articles
    Download Articles
    Email Alert
Traffic
Quick Search  
  Adv Search
2022 Vol. 43, No. 10
Published: 2022-10-28
 
1245 100kN·m Deadweight Torque Standard Machine
ZHANG Zhi-min,HU Gang,ZHANG Yue,MENG Feng,ZHANG Wei
DOI: 10.3969/j.issn.1000-1158.2022.10.01
A set of 100kN·m deadweight torque standard machine is newly developed at National Institute of Metrology. The knife edge support mechanism with high precision and large load capacity is adopt to support moment arm at the fulcrum, the moment-arm adopts the single beam structure and is equipped with the arm limit protection mechanism, the weight loading system adopts the combination of inner and outer ring weight group and lifting mechanism. The working principle and mechanical structure of the machine are introduced, the results of performance test and uncertainty assessment are described. The theoretical analysis and performance tests show that the repeatability is smaller than 5×10-5,the uncertainty is less than 1×10-4 (k=3) in the range of 1kN·m to 100kN·m.
2022 Vol. 43 (10): 1245-1249 [Abstract] ( 199 ) HTML (1 KB)  PDF (37453 KB)  ( 102 )
1250 Comparative Analysis of Ultramicro Volume Measurement Methods Based on Static Measurement method and Optical Method
LIU Xin-xing,WANG Si-xian,SUN Bin,ZHANG Jing-yue,WANG Jin-tao
DOI: 10.3969/j.issn.1000-1158.2022.10.02
The static weighing method and optical method are used to measure the ultramicro volume. In the optical method, the measurement systems based on the time-domain and spatial-domain spectroscopic methods are designed and built respectively. The measurement data of the two optical measurement systems at 1 μL and 10 μL are 0.967 μL and 0.929 μL, and 9.976 μL and 9.729 μL, respectively The renaturation is 3.2% and 2.54%, and the spatial spectroscopic measurement system is 3.6% and 2.02%. The same conclusion is obtained with the static measurement method, which proves the feasibility of optical method in the field of ultramicro volume measurement. Compared with the two optical systems, the detector sensitivity of the time-domain spectrometer is better, and more light intensity information can be detected in the ultra-fine volume measurement of low concentration; Compared with the static weighing method,.the optical method for liquid evaporation is simpler, and the design of optical elements can effectively reduce the influence of evaporation.
2022 Vol. 43 (10): 1250-1255 [Abstract] ( 139 ) HTML (1 KB)  PDF (41306 KB)  ( 136 )
1256 Improved YOLOv3 Algorithm for Surface Defect Detection of Aluminum Profile
YAO Bo,WEN Xiu-lan,JIAO Liang-bao,WANG Shu-gang,QIAN Zheng,LI Zi-kang
DOI: 10.3969/j.issn.1000-1158.2022.10.03
Aiming at the problems of low accuracy and low efficiency in the detection of surface defects of aluminum profile, an improved method based on YOLOv3 is proposed. Firstly, the k-means clustering algorithm is used to cluster the collected data sets, and the target candidate box with the optimal size is selected. Then, considering the large surface defects of aluminum profile, the network hierarchical structure of YOLOv3 is adjusted. Six CBL units before the target detection layer are changed into four CBL units, and two residual units are added to improve the reuse of features. Compared with the classical convolution network Faster-RCNN and SSD, a large number of experimental results show that the accuracy of the proposed algorithm can reach 97%, and the detection speed can reach 47 frame/s. The proposed method is obviously better than Faster-RCNN and SSD, which is suitable for the aluminum profile surface defect detection with high accuracy and rapidity.
2022 Vol. 43 (10): 1256-1261 [Abstract] ( 240 ) HTML (1 KB)  PDF (67379 KB)  ( 106 )
1262 A Detection Method for Positioning Accuracy of 3D Phantom System in Situ Based on Refraction Compensation of Laser Tracker
JI Zhen-chao,LI Jia-fu,DU Hua,ZHU Xiao-ping,ZHANG Hui,YU Jing
DOI: 10.3969/j.issn.1000-1158.2022.10.04
A detection method of the positioning accuracy of medical 3D phantom based on laser tracker refraction compensation is proposed, and the model of ADM and IFM ranging error compensation and space position coordinates solution of target ball is established, and theoretical model verification experiment and contrast experiment of 3D phantom positioning accuracy detection are carried out, and the high precision measurement of laser tracker in glass media is realized. The experimental results show that the average deviation of X, Y, and Z coordinates is respectively reduced from 3.410mm, 0.407mm, 1.732mm to 0.022mm, 0.015mm, 0.035mm, and the average deviation of the distance between adjacent points is reduced from 0.266mm to 0.017mm before and after compensation, which is equivalent to the measurement accuracy of the laser tracker in the air. On this basis, taking the suspension detection method without glass shielding as the reference, the positioning errors measured by the two methods are basically consistent. Finally, using Monte Carlo method, the standard uncertainty of the distance error between adjacent points is 0.012mm, which can meet the detection requirement.
2022 Vol. 43 (10): 1262-1270 [Abstract] ( 128 ) HTML (1 KB)  PDF (121172 KB)  ( 75 )
1271 Multi-scale Correlation Filtering Tracking Based on Adaptive Feature Fusion
ZHANG Li-guo,YANG Man,ZHOU Si-en,JIN Mei
DOI: 10.3969/j.issn.1000-1158.2022.10.05
In order to reduce the influence of target deformation, illumination, motion blur and target rotation on the tracking effect in target tracking, a multi-scale correlation filtering tracking algorithm based on adaptive feature fusion is proposed on the basis of correlation filtering KCF. Firstly, the features of conv2-2, conv3-4 and conv5-4 layers and CN features in vgg19 network are extracted, and CN features are added to conv2-2 layer. Then, the three features are used to replace the hog feature for filtering learning, and three pairs of response diagrams are obtained. Then, the weighted fusion of the three response maps is carried out to predict the target position.Finally, multi-scale correlation filter is introduced to determine the scale.Compared with KCF tracking algorithm, the accuracy and success rate of this algorithm are improved by 13.6% and 11.8% respectively.Compared with several existing excellent tracking algorithms, the algorithm has better tracking effect in dealing with motion blur, background clutter, target deformation and plane rotation.
2022 Vol. 43 (10): 1271-1278 [Abstract] ( 189 ) HTML (1 KB)  PDF (146328 KB)  ( 105 )
1279 The Investigation on the Temperature Field Uniformity of a High-precision Vacuum Blackbody
WU Qiang,HAO Xiao-peng,SONG Jian,LI Guo-zhan
DOI: 10.3969/j.issn.1000-1158.2022.10.06
In order to further improve the temperature control accuracy of vacuum black body, the temperature distribution of vacuum black body is studied by means of theoretical analysis, numerical simulation and experiment, the heat transfer formula between black body and environment is derived, and the temperature distribution characteristics of vacuum black body are analyzed.For the specific black body structure, the corresponding temperature field control equation is further simplified, the three-dimensional model is simplified to a one-dimensional model without loss of precision, the calculation complexity is reduced, the simulation time is shortened, and the numerical stability in the process of solving is improved.The numerical results are in good agreement with the experimental data, but the global maximum temperature difference of vacuum black matter obtained by numerical simulation is about 3 times the temperature difference between measuring points, which indicates that the actual temperature control accuracy of black body is much lower than that of experiment results.
2022 Vol. 43 (10): 1279-1284 [Abstract] ( 208 ) HTML (1 KB)  PDF (84848 KB)  ( 71 )
1279 A Principal Component Analysis Method for TDLAS to Remove High Frequency Noise
CHANG Hai-tao,CAI Jing,WEN Yue,ZHU Yu-mei
DOI: 10.3969/j.issn.1000-1158.2022.10.07
Detection of gas temperature with tunable diode laser absorption spectroscopy (TDLAS) is easily affected by the high-frequency noise of detector and hardware circuit. The principal component analysis method commonly used in spectral image denoising is applied to the high frequency noise elimination of absorbance column vector data, and a denoising method based on principal component analysis is proposed. Firstly, the original absorbance column vector is arranged in a matrix. Then it is decomposed into the main component matrix and the score matrix through principal component analysis. The appropriate principal component score is taken, and the intercepted part of the principal component matrix and score matrix are used to reconstruct the original data. The intercepted principal component represents the main information of the original data, while the eliminated part only contains noise information. Experimental results show that this method is used for TDLAS to measure water vapor temperature, and the noise removal rate is 81% and the standard deviation of single temperature calculation is reduced from 8.9 to 5.3.
2022 Vol. 43 (10): 1279-1284 [Abstract] ( 153 ) HTML (1 KB)  PDF (72789 KB)  ( 71 )
1291 Channel Thinning and the Measurement Method of Gas Ultrasonic Flowmeter Based on Contraction Flow
MAO Xian-yi,CUI Li-shui,XIE Dai-liang
DOI: 10.3969/j.issn.1000-1158.2022.10.08
In order to improve the accuracy level of the low-channel number ultrasonic flowmeter, a low-channel number ultrasonic flowmeter with a shrinking flow structure was designed on the basis of the traditional ultrasonic flowmeter. Focusing on the research content, numerical simulation was used to study the flow field characteristics of the shrinking flow in the reducer, and the geometric parameters of the reducer and the installation method of the ultrasonic probe were determined. Through the air flow experiment, the basic characteristics of the flow measurement of mono and dual channels under contracting flow conditions were studied. The scope of flow studied is 27~432m3/h and the range of pipe diameter is 100~150mm. The research results show that the numerical simulation results of shrinkage flow are consistent with the theoretical model. When the shrinkage ratio of the reducer increases from 2 to 6, the proportion of the uniform area of the measured flow field increases significantly. When the shrinkage ratio is fixed, as the flow rate increases, the boundary layer thickness decreases significantly. Compare the measurement results of ultrasonic flowmeters with different structures and configurations through real flow experiments.The accuracy of the actual flow measurement of single-channel gas ultrasonic flowmeter is superior to that of the traditional single-channel gas ultrasonic flowmeter significantly and is equivalent to the traditional dual-channel gas ultrasonic flowmeter.
2022 Vol. 43 (10): 1291-1297 [Abstract] ( 191 ) HTML (1 KB)  PDF (59266 KB)  ( 69 )
1298 Flow collection performance analysis of the electric extruded rubber flow collector for logging
ZHANG Xi,KONG Wei-hang,ZHAO Gong-da,LIU Jia-yu,LI Lei,XING Guang-long
DOI: 10.3969/j.issn.1000-1158.2022.10.09
Aiming at the problems of leakage and scraping of conventional oil logging flow collectors, an overall scheme of electric extruded rubber flow collector with rubber elastomer as the core is proposed. The finite element method and the Mooney-Rivlin hyperelastic constitutive model are used to simulate the rubber elastomer and study the influence of its shape, geometry and material properties on the flow collection performance. Moreover, an electric extruded rubber flow collector is developed and its flow collection performance is tested on a simulated well test rig. Experimental results demonstrate that the structure, hardness, friction coefficient with the pipe and the concave degree of the rubber have a significant influence on the performance of the collector; The rubber flow collector with hardness 58HA, size 80mm × 12mm, 4mm internal concavity and rounded internal concave elastomer has better flow collection performance than conventional umbrella type flow collector.
2022 Vol. 43 (10): 1298-1305 [Abstract] ( 136 ) HTML (1 KB)  PDF (101993 KB)  ( 72 )
1313 Research on Fast Focus Location of Focused Transducer Based on Machine Vision
PENG Yuan,ZHENG Hui-feng,DONG Zhao-cheng,YAO Run-guang
DOI: 10.3969/j.issn.1000-1158.2022.10.11
A method of fast focus location of focused transducer based on machine vision is proposed. According to the feature of the focused transducers geometric center, the arc adjacency matrix-based fast ellipse detection connection algorithm detects the transducers edge contour information, and then gets the focus position, the focus coordinate of the manipulator is calculated by the coordinate transformation relation, and the focus of the transducer can be located quickly by the manipulator. The focus coordinates of the transducer are obtained by scanning the sound field of the focal plane to verify the accuracy of the focus coordinates obtained by the method. The experimental results show that the focus of the method is in the range of -3dB real focus diameter, which proves that the method is feasible.
2022 Vol. 43 (10): 1313-1318 [Abstract] ( 163 ) HTML (1 KB)  PDF (46122 KB)  ( 71 )
1319 Error Analysis and Compensation of Servo Motor Rotor Position Detection
ZHANG Tian-heng,TONG Peng,YANG Ji-sen
DOI: 10.3969/j.issn.1000-1158.2022.10.12
Aiming at the problems of low accuracy, inconvenient installation and high cost in current motor rotor position measurement, a design scheme of rotor position sensor based on the combination of tunnel magnetoresistance effect and time grating technology is proposed. On the premise of not destroying the original structure of the motor, a pair of space orthogonal TMR chips are embedded in the front cover of the motor to realize the embedded position precise detection. According to the principle of TMR sensor rotor position calculation, the error components caused by installation error, electrical error and electromagnetic noise error of TMR sensor are analyzed. an error compensation method based on extreme learning machine is proposed. The model optimal parameters are obtained through training the real sample values and measured values. According to the model parameters, Rotor position error model is established, which is used to realize rotor position error compensation.The experimental results show that under the uniform speed of 2000r/min, the maximum measurement error of rotor position before compensation is 4.64° and the rotor position error is 0.315° after compensation, and the accuracy is improved by 93.2%.The embedded rotor position detection method provides a new method for motor rotor position detection.
2022 Vol. 43 (10): 1319-1325 [Abstract] ( 158 ) HTML (1 KB)  PDF (50126 KB)  ( 77 )
1326 An Improved Feature Extraction Method of Bearing Fault Signal Based on Adaptive Multivariate Variational Mode Decomposition
SHI Pei-ming,ZHANG Hui-chao,YI Si-ying,HAN Dong-ying
DOI: 10.3969/j.issn.1000-1158.2022.10.13
Aiming at the nonlinearity and non-stationarity of bearing signals in practical engineering, an adaptive multi-variable mode decomposition algorithm is proposed. The decomposition effect of multivariate variational modes is mainly related to the number of intrinsic modes k and penalty parameter α. In order to solve the influence of artificial empirical parameter setting on the decomposition results of multivariate signals, an adaptive signal decomposition algorithm is proposed. The specific contents are as follows: Firstly, the hybrid gray wolf algorithm is combined with the multivariate variational mode decomposition algorithm, and the minimum mode overlap component index is proposed, which is used as the fitness function to seek the optimal solution of(k, α). According to the optimal solution, the multivariate signals are decomposed and the fault features are extracted. Simulation signals and actual data are used to verify the effectiveness and accuracy of the proposed method. By comparing with multivariate empirical mode decomposition (MEMD) and cascade variational mode decomposition, the effectiveness and practicability of the proposed method in rolling bearing fault feature extraction are verified.
2022 Vol. 43 (10): 1326-1334 [Abstract] ( 135 ) HTML (1 KB)  PDF (194313 KB)  ( 86 )
1335 UWB Indoor Positioning Algorithm Based on Kalman Filter and Particle Filter Fusion
CHENG Xue-cong,LIU Fu-cai,HUANG Ru-nan
DOI: 10.3969/j.issn.1000-1158.2022.10.14
Indoor positioning technology based on ultra-wideband (UWB) has been widely developed.However, the measurement of UWB in LOS (line-of-sight) and NLOS (non-line-of-sight) environments There are different degrees of error in the distance information, so an improved Kalman filter algorithm is proposed to smooth the UWB original data; then a Kalman filter and particle filter (KPF) particle filter and Kalman filter fusion algorithm is proposed.Particle sampling is carried out through the state quantity and error covariance obtained by Kalman filtering, which overcomes the disadvantage that the kinematic model of traditional particle filtering does not match the actual motion, and greatly reduces the phenomenon of particle degradation.After experiments, the positioning accuracy of the algorithm in LOS and NLOS environments is improved by 20.6% and 15.6%, respectively.
2022 Vol. 43 (10): 1335-1340 [Abstract] ( 197 ) HTML (1 KB)  PDF (79160 KB)  ( 99 )
1341 Motor Imagery EEG Signal Recognition Method Based on GST-ECNN
JIN Hai-long,WU Xia,FAN Feng-jie,WANG Jin-ping
DOI: 10.3969/j.issn.1000-1158.2022.10.15
In the research on the decoding of EEG signals, there are existing time-frequency analysis methods that have limited high-frequency signal processing capabilities, multi-channel signal information redundancy, and the ReLU activation function of the commonly used convolutional neural network classifier is greatly affected by the learning rate. It is difficult to obtain satisfactory results with the same regularization for different layers. To solve the above problems, a method based on the combination of generalized S-transform feature extraction and enhanced convolutional neural network classification is proposed. At the same time, a wrapping method combining Relief algorithm and forward selection search strategy is proposed for channel selection. The results show that the proposed method uses less signal channels and achieves better ability of feature extraction and classification. The highest classification accuracy of 98.44±1.5% is obtained in the fourth BCI dataset I, which is higher than other existing algorithms. The good classification performance of this study not only reduces the calculation consumption, also effectively improves the classification accuracy, which has a certain reference significance for EEG feature extraction and classification.
2022 Vol. 43 (10): 1341-1347 [Abstract] ( 179 ) HTML (1 KB)  PDF (88806 KB)  ( 84 )
1348 The Theory of Time Metrology and its Development
WANG Ruo-pu,DENG Xiang-rui,TONG Shuai,ZHANG Chao
DOI: 10.3969/j.issn.1000-1158.2022.10.16
As a basic physical quantity, time plays an important role in scientific research, national defense construction and economy development and humans life.The development history of time measurement systems is systematically reviewed.The theoretical foundation and methods for establishment of time systems are elaborated.The definition and realization, as well as conversion for every kind of time scale are studied.The advantages and disadvantages of each time system are discussed.The possible development direction of the time measurement systems in the future is also prospected.It is valuable not only in understanding the development history of theory of time metrology but also in application of high precision time measurement techniques.
2022 Vol. 43 (10): 1348-1358 [Abstract] ( 180 ) HTML (1 KB)  PDF (15542 KB)  ( 89 )
1359 Key Comparison of the Air-kerma of the NIM and the BIPM in Low-energy X-rays
ZHAO Rui,WU Jin-jie,FAN Song,WANG Pei-wei,YU Ji-li
DOI: 10.3969/j.issn.1000-1158.2022.10.17
In order to promote the unification of measurement values in the field of low-energy X-ray air kerma measurement in China and achieve international equivalence, National Institute of Metrology, China (NIM) measured the air kerma of low-energy X-ray absolutely with a plate free air ionization chamber which is the primary standard in 2018. Two transfer ionization chambers named Radcal-RC6M were calibrated in low energy X-ray radiation field at NIM and BIPM. The comparison results are equivalent to those of the BIPM within the range of 0.37% of the relative standard uncertainty. The comparison results have been included in the key comparison database (KCDB) of the BIPM.
2022 Vol. 43 (10): 1359-1365 [Abstract] ( 239 ) HTML (1 KB)  PDF (17112 KB)  ( 89 )
1366 Experimental Calibration of CdTe Detection Efficiency Using Single-energy X-ray Radiation Device
YU Tao,GUO Si-ming,ZHOU Jiang-bin,JIANG Zheng,QIE Xiao-yu,GUO Kai-yue,WU Jing-jie
DOI: 10.3969/j.issn.1000-1158.2022.10.18
Before realizing the absolute measurement of the single-energy parallel photon source, it is necessary to calibrate the efficiency of the CdTe detector. The physical model of the CdTe detector was established by using the MCNP5 Monte Carlo simulation program, the intrinsic detection efficiency of the energy point in the 10~260keV energy range was simulated and calculated, the detection efficiency was higher than 75% in the energy range of 10 ~ 60keV. The intrinsic detection efficiency of the CdTe detector was experimentally calibrated with a single energy X-ray device and a HPGe detector. The results show that the simulated efficiency of the CdTe detector is consistent with the experimental efficiency trend in the energy range of 10 ~ 100keV, and the maximum error does not exceed 5.6%. Because the Te element produces escape peaks at 27keV and 32keV, the detection efficiency has a clear downward trend at these two energies. Efficiency calibration of CdTe detector was carried out with 241Am and 133Ba radioactive sources, and the calibration efficiency of radioactive sources with energies of 59.54keV and 81keV was consistent with the measured value of single energy X-ray radiation device.
2022 Vol. 43 (10): 1366-1370 [Abstract] ( 167 ) HTML (1 KB)  PDF (40575 KB)  ( 69 )
1371 Determination of Coal Blending Uniformity in Power Plant Based on Maximum Likelihood Estimation
LIU Fu-guo,ZHAO Zhong-hua,LIU Ke,WANG Shou-en
DOI: 10.3969/j.issn.1000-1158.2022.10.19
It is difficult to evaluate the mixing uniformity of powder or particle mixing material by traditional sampling analysis method. The mixed raw coal is regarded as a mixture with different probability distributions of various components. In the mixed coal formed by mixing raw coal, the probability distribution of a certain component is determined by the composition distribution, mixing ratio and particle mixing state of the raw coal. Taking volatile as the tracer component, the change of the probability distribution of volatile in the mixing process was investigated. At given mine coal volatile distributions and mixed coal sample data set, the maximum likelihood principle was used to estimate the mixing ratio and standard deviation of various raw coal in the mixed coal, and then the mixing state was evaluated. The results show that the prediction results of the mixing model for the actual mixing system are in good agreement with the statistical data of raw coal consumption. The multiple blending conditions simulation experiments also show that the worse particle homogeneity of the mixing system, the higher the prediction accuracy of the model.
2022 Vol. 43 (10): 1371-1377 [Abstract] ( 176 ) HTML (1 KB)  PDF (61383 KB)  ( 69 )
1378 Study on Evaluation Method of Physical Efficiency of Cyclone Microbiological Sampler
LIU Jia-qi,ZHANG Guo-cheng,WU Dan,Tian Ying,SHEN Shang-yi,YANG Zhen-qi,JING Wen-jie,HUO Sheng-wei
DOI: 10.3969/j.issn.1000-1158.2022.10.20
In order to evaluate the collection efficiency of cyclonic microbiological sampler, a static chamber method evaluation device with uniform and stable particle concentration was built to measure the collection physical efficiency of two cyclonic microbiological samplers in China, and the collection efficiency curve was fitted. The measurement results showed that under the fixed sampling flow of Brand 1 sampler, Da50(the aerodynamic diameter when the collection efficiency was 50%) of the physical acquisition efficiency curve was 0.91μm. The Brand 2 sampler had 5 kinds of adjustable sampling flow rates, and Da50 were 1.60μm, 1.36μm, 1.19μm, 1.06μm and 1.05μm, respectively. The acquisition physical efficiency of Brand 2 sampler was compared with that of sample solution with initial capacity of 15mL and 7.5mL, respectively. It was found that the acquisition physical efficiency was lower when the initial capacity of sample solution was lower. The study provided a reference for the performance evaluation of cyclonic microbiological sampler.
2022 Vol. 43 (10): 1378-1381 [Abstract] ( 155 ) HTML (1 KB)  PDF (23291 KB)  ( 72 )
1382 Status and Prospects on the Standardization of Biological Phenotype and Phenome
ZHANG Yong-zhuo,GAO Ying,NIU Chun-yan,FU Bo-qiang,WANG Jing
DOI: 10.3969/j.issn.1000-1158.2022.10.21
With the development of modern analytical techniques such as high-throughput sequencing, the research on exploring the relationship between mechanisms and external representations in life, especially the biological phenotyping, has become more and more in-depth.The precise demands of gene, environment and phenotype association research make the standardization research of biological phenome become crucial.Because phenome covers a wide range, many species, and the phenotypic characteristics differ greatly, there is no standard to follow, which seriously restricts scientific research and industrial development.From the perspective of standardization, the related concepts and development of biological phenotypes, as well as the problems faced by the standardization of biological phenome are analyzed and predicted, so as to provide metrological technology and standard support for the high-quality development in the field of biological phenome with various types and data.
2022 Vol. 43 (10): 1382-1388 [Abstract] ( 148 ) HTML (1 KB)  PDF (5400 KB)  ( 408 )
1306 Flow Pattern Recognition of Gas-liquid Two-phase Flow Based on Convolutional Neural Network and Gated Recurrent Unit
ZHANG Li-feng,WANG Zhi,WU Si-cheng
DOI: 10.3969/j.issn.1000-1158.2022.10.10
The method for identifying the flow pattern of gas-liquid two-phase flow in a vertical pipeline based on convolutional neural network (CNN) and gated recurrent unit (GRU) is presented in this paper. Based on the reconstructed image by the electrical resistance tomography (ERT) system, the discrete cosine transform (DCT) is performed after the filling processing. The difference between the maximum and minimum DCT coefficients is calculated, and a certain frame length data are selected as the network input to identify the flow pattern. The influence of the length of the input sequence on the accuracy of CNN-GRU, CNN and GRU network classification is analyzed, and the optimal input vector dimensions are determined to be 60, 65 and 50. Using experimental data to train and test the CNN-GRU network, and compare it with the GRU and CNN networks, the results show that the CNN-GRU network has the highest classification accuracy, and the average flow pattern recognition accuracy rate can reach 99.40%.
2022 Vol. 43 (10): 1306-1312 [Abstract] ( 143 ) HTML (1 KB)  PDF (103131 KB)  ( 76 )
News
Download
Links
Copyright © Editorial Board of Acta Metrologica Sinica
Supported by:Beijing Magtech