BP Neural Network for Linear Birefringence of Three-Phase Optical Current Transducer

Ye Liu,Yanmin Su,Caitang Wang

DOI: https://doi.org/10.3321/j.issn:0253-987X.2000.06.001

2000-01-01

Abstract:A BP neural network is used to compensate for linear birefringence effect of three-phase optical current transducer. The algorithm and way for compensation are analyzed. The neural network is validated and has been shown successfully for its intended purpose.

What problem does this paper attempt to address?

Evaluation of Chracteristic of Optical Thin Film By Neural Network System

Yuan-sheng Ma,Xu Liu,Pei-fu Gu,Hai-feng Li,Jing-fa Tang

DOI: https://doi.org/10.1364/oic.1995.mb13

1995-01-01

Abstract:Controlling and mastering the optical constant of thin film is a key factor concern for the optical coating. Each optical constant must be controlled very well to make the films meet requires desired. At present, there are many methods to characterize the optical constants of the films, for example: the envelop method, ellipsometry measurement, waveguide prism methode etc.[1]. All of them use numerical inversion technique to obtain the optical constants from the physical parameters measured. These methods are very well for treating their problems, but they have their main shortages of indefinite and very complex. We propose now an artifical neural network (ANN) model to simulate the relationship between the spectral character and optical constant of the thin films. ANN is a simplified mathematical model of biological network, it is extremely strong in dealing fuzzy and indefinite problem, and is widely used for the signal process and recgonization. Neural networks have the ability to learn directly from example data rather than following programmed rules based on a knowledge base, and can treat the problem extremely fast and tolerate data containg experimental error[2].
Improved BP algorithm for neural network and its application in welding process control

裴浩东,苏宏业,褚健

DOI: https://doi.org/10.3785/j.issn.1008-973X.2002.01.008

2002-01-01

Abstract:Multiplayer neural network based on BP algorithm is being applied in many fields, but in spite of BP algorithmic's shortcomings of slow convergence rate and proneness to yield minimal local results. Many simulation tests and analyses, revealed that the adjusted weights of various layers differed much, so a new algorithm of weight balance layer by layer was brought forward to calculate the weights' influences on neural network . The results indicated that the new algorithm converged faster than the standard BP algorithm. And a control system simulation using the new algorithm in welding showed to be a good approach for welding control.
Realization and Application of BP Neural Network in LabVIEW

CAI Qing-hua,YANG Shi-xi,LUO Xi-liang

DOI: https://doi.org/10.3969/j.issn.1002-6886.2006.02.003

2006-01-01

Abstract:LabVIEW is a G programming language,which is widely accepted by industry and laboratory.BP neural network and LabVIEW are combined and applied in inversion algorithms of Multi-layered thickness testing based on eddy current method,which widened the application area,advanced the capability and the developing efficiency of the testing system.
Soft-sensor modeling via neural network PLS approach

梁军,汪小勇,王文庆

DOI: https://doi.org/10.3785/j.issn.1008-973X.2004.06.005

2004-01-01

Abstract:By using the universal approximation property of neural network (NN), a feedforward neural network was embedded into the framework of standard partial least squares (PLS), which results in a nonlinear PLS-NN model. A gradient descent learning algorithm was employed to train the neural network for calculating weights matrix. Based on two sets of real operating data, the proposed PLS-NN model was used as a soft-sensor to estimate the slab temperatures of an industrial reheating furnace with three quality variables by twenty-six process variables. Compared with the linear PLS model and the first principle model, the PLS-NN model gives the smallest estimation error for slab temperatures. The three estimation errors keep a ratio of about 1: 1.7:2.8. Analysis of the relationships between process variables and quality variables based upon PLS-NN model shows that gas flowrates of every burner have crucial influence on the operating condition of reheating furnace.
Temperature Compensation of Elasto-Magneto-Electric (EME) Sensors in Cable Force Monitoring Using BP Neural Network

Ru Zhang,Yuanfeng Duan,Yang Zhao,Xuan He

DOI: https://doi.org/10.3390/s18072176

IF: 3.9

2018-01-01

Sensors

Abstract:Techniques based on the elasto-magnetic (EM) effect have been receiving increasing attention for their significant advantages in cable stress/force monitoring of in-service structures. Variations in ambient temperature affect the magnetic behaviors of steel components, causing errors in the sensor and measurement system results. Therefore, temperature compensation is essential. In this paper, the effect of temperature on the force monitoring of steel cables using smart elasto-magneto-electric (EME) sensors was investigated experimentally. A back propagation (BP) neural network method is proposed to obtain a direct readout of the applied force in the engineering environment, involving less computational complexity. On the basis of the data measured in the experiment, an improved BP neural network model was established. The test result shows that, over a temperature range of approximately −10 °C to 60 °C, the maximum relative error in the force measurement is within ±0.9%. A polynomial fitting method was also implemented for comparison. It is concluded that the method based on a BP neural network can be more reliable, effective and robust, and can be extended to temperature compensation of other similar sensors.
Wavelet Convolutional Neural Network for Robust and Fast Temperature Measurements in Brillouin Optical Time Domain Reflectometry.

Bei Chen,Lianghao Su,Zhaoyang Zhang,Xiaozhi Liu,Tingge Dai,Muping Song,Hui Yu,Yuehai Wang,Jianyi Yang

DOI: https://doi.org/10.1364/oe.451877

IF: 3.8

2022-01-01

Optics Express

Abstract:In this paper, a wavelet convolutional neural network (WNN) consisting of a one-dimensional (1D) convolutional neural network and a self-adaptive wavelet neural network has been proposed and demonstrated experimentally for temperature measurement in a Brillouin optical time domain reflectometry (BOTDR) system. Based on the analysis of the system noise, it follows the Gaussian white noise distribution along the time-related sensing distance. The impact of the noise in time-domain on the measured Brillouin gain spectra (BGSs) could be neglected, so that the BGSs in the fiber can be regarded as a series of 1D input data of the proposed WNN. Different self-adaptive wavelet activation functions connected to each output of the full-connection network are adopted to realize the multi-scaled analysis and the scale translation, which can obtain more local characteristics in frequency-domain. The output extracted by the WNN is Brillouin frequency shift (BFS), which presents linearity correlation to the actual temperature. Considering the multi-parameters including different frequency ranges, signal-to-noise-ratios (SNRs), BFSs and spectral widths (SWs), a general model of the proposed WNN is trained to handle more extreme cases, in which it doesn't require retraining for different single-mode (SM) optical fibers in BOTDR sensing system. The performances of the WNN are compared with other two techniques, the Lorentzian curve fitting based on Levenberg-Marquardt (LM) algorithm and the basic neural network (NN) containing input and output layers together with two hidden layers. Both the simulated and measured results show that the WNN has better robustness and flexibility than the LM and the NN. Besides, the computational accuracy of the WNN is improved and the fluctuation of that is slighter, especially when the SNR is less than 11 dB. Moreover, the WNN takes approximately 0.54 s to measure the temperature from the 18,000 collected BGSs transmitted through the 18 km SM optical fiber. The calculating time of the WNN is greatly reduced by three orders of magnitude in comparison with that of the LM, and is comparable to that of the NN. It proves that the proposed WNN may provide a feasible or even better scheme for the robust and fast temperature measurement in BOTDR system.
Artificial Neural Network Compensation for an Optical Fiber Temperature Sensing System Based on Optical Time Offset

Lu-jun Cui,Yan-long Cao,Hui-chao Shang,Hong-hong Guo,Cheng-yin Wang,Qiang Guo

DOI: https://doi.org/10.23919/iconac.2017.8082075

2017-01-01

Abstract:Through analyzing the optical time offset in optical fiber, the paper presented a new method that using optical fiber sensing technology for monitoring the temperature. Measurement experiments of monitoring temperature were carried out. Through the analysis of experimental data, the results indicated that the temperature variations could be detected by the optical fiber sensor based on optical time offset effectively, and the temperature sensitivity of the optical fiber sensing system is about 2.46mV/°C. Owing to the temperature sensor was affected by the variations and noises seriously, the artificial neural network (ANN) was introduced to the temperature sensor based on optical time offset as a compensation method, experimental results and numerical simulation indicated that the optical fiber temperature sensor with artificial neural network could enhance the nonlinear correction for sensor and decrease the interference effectively. Simultaneously, by way of the contrast test of the different neural networks, experimental results showed that the linearity of full scope could reach 0.11% for temperature sensor based on BRF network, and concluded the BRF network was more suitable for compensation of optical temperature sensor based on optical time offset.
Locally Linear Back-propagation Based Contribution for Nonlinear Process Fault Diagnosis

Jinchuan Qian,Li Jiang,Zhihuan Song

DOI: https://doi.org/10.1109/jas.2020.1003147

2019-01-01

Abstract:This paper proposes a novel locally linear backpropagation based contribution（LLBBC） for nonlinear process fault diagnosis. As a method based on the deep learning model of auto-encoder（AE）, LLBBC can deal with the fault diagnosis problem through extracting nonlinear features. When the on-line fault diagnosis task is in progress, a locally linear model is firstly built at the current fault sample. According to the basic idea of reconstruction based contribution（RBC）, the propagation of fault information is described by using back-propagation（BP） algorithm. Then, a contribution index is established to measure the correlation between the variable and the fault, and the final diagnosis result is obtained by searching variables with large contributions. The smearing effect, which is an important factor affecting the performance of fault diagnosis, can be suppressed as well,and the theoretical analysis reveals that the correct diagnosis can be guaranteed by LLBBC. Finally, the feasibility and effectiveness of the proposed method are verified through a nonlinear numerical example and the Tennessee Eastman benchmark process.
Residual flux density estimation of the three-phase transformer using BP neural network

Yuzhan Ren,Chengcheng Liu,Youhua Wang

DOI: https://doi.org/10.1063/9.0000631

IF: 1.697

2024-01-01

AIP Advances

Abstract:When the off-line transformer is re-energized, the phase-controlled switching strategy can avoid the generation of inrush current by controlling the phase. To determine the closing phase, the residual flux density (Br) in the transformer core needs to be accurately measured. This paper proposes a Br estimation method for three-phase transformers based on the finite element method and BP neural network. Firstly, the direction of Br in each phase core is determined based on the transient current characteristics. Then, the three-phase transformer is simulated and the BP neural network is trained to estimate the Br based on the simulation results. The experimental results on a three-phase transformer show that the proposed method can accurately determine the direction and amplitude of Br in each phase of the three-phase transformer.

materials science, multidisciplinary,physics, applied,nanoscience & nanotechnology
A PSO-BP Model for Nonlinear Fitting of Phyllostachys edulis' Thermal Conductivity

Wu Shuci,Liu Shuai,Li Jianjun,Shen Xuejie,Wang Chuanli

2012-01-01

Abstract:By applying PSO(particle swarm optimization)algorithm to train BP(back propagation)neural network,this article puts forward a PSO-BP model of Phyllostachys edulis' thermal conductivity and regards the process of neural network learning as that of iterative optimization for particle swarm,with the purpose of optimizing neural network weights and thresholds.The research results show that the PSO-BP model of Phyllostachys edulis' thermal conductivity is superior to the standard BP network model in several aspects,such as generalization performance,fitting precision,training,error validation,etc.,which provides a new method for the application of intelligent information processing technology in bamboo material analyses.
Application of BP Neural Net in Fault Diagnosis of Excavator Hydraulic System

ZHANG Bin,YAN Jun,ZHANG Xing

DOI: https://doi.org/10.3969/j.issn.1001-3881.2008.02.065

2008-01-01

Abstract:The structure and algorithm of BP neural net were described,the realization process of the fault diagnosis of hydraulic system based on BP neural net was discussed.According to the experiment,the BP net has better learning function,high net convergence rate and high stability of learning and memory.The diagnosis results indicate that the presented diagnosis method has high reliability and can attain the expected results,which can be applied to fault diagnosis of hydraulic system.
Lnista-Net: Learning Nonlocal Regularized Iterative Shrinkage-Thresholding Network for Noninvasive Cardiac Transmembrane Potential Imaging.

Shuting Xie,Linsheng Cheng,Huafeng Liu

DOI: https://doi.org/10.1109/isbi48211.2021.9433947

2021-01-01

Abstract:ECG inverse problem is a very classic ill-posed problem. Myocardial transmembrane potential (TMP) provide a wealth of physiological information and is often selected as the reconstructed cardiac source signal. Researcher have solved these problems by constructing convex optimization models with constraints or building end-to-end neural network. Based on the nonlocal self-similarity of TMP distribution, we propose a novel method to reconstruct the dynamic distribution of TMP. Combining the physical interpretation and convergence theory provided by the optimization method with the advantages of deep learning to automatically extract features from big data, we propose a learning nonlocal regularized Iterative Shrinkage-Thresholding Network (LNISTANET) for noninvasive cardiac TMP imaging. Experiments verify the superiority of this method over existing methods from the spatial and temporal distribution of TMP sequences.
Nonlinearity Compensation of Magneto-Optic Fiber Current Sensors Based on WOA-BP Neural Network

Zhizhuang Liang,Qun Han,Teng Zhang,Yuliang Tang,Junfeng Jiang,Zhenzhou Cheng

DOI: https://doi.org/10.1109/jsen.2022.3205701

IF: 4.3

2022-10-19

IEEE Sensors Journal

Abstract:Compensating the temperature dependency and the slight nonlinearity of the magneto-optic response of a fiber optic current sensor (FOCS) are crucial to improve its accuracy. In this article, the nonlinearity originations of the FOCS with the static-biasing scheme for temperature compensation are analyzed and characterized. We found that the temperature dependency of the static bias and the varying magnetic domains of the Faraday crystal are the two main sources of the nonlinearity. A back-propagation neural network (BPNN) aided by the whale optimization algorithm (WOA) is developed to compensate for the temperature dependency and the nonlinear response of the fiber current sensor. Experimental results show that in the temperature range of −20 °C–60 °C and measurement range of 0 to 27 mT, the maximum relative error can be reduced to 0.38% with the proposed method, whereas without compensation the relative error is as large as 16.0%. The WOA-BPNN is also compared with the nonoptimized BPNN and the PSO-BPNN, the maximum relative errors of the last two methods are 4.04% and 0.99%, respectively, with the same set of experimental data. The WOA-BPNN enables the sensor to fulfill the accuracy requirement of the metering standard.
A Deep Learning Compensated Back Projection for Image Reconstruction of Electrical Capacitance Tomography

Jin Zheng,Lihui Peng

DOI: https://doi.org/10.1109/jsen.2020.2965731

IF: 4.3

2020-01-01

IEEE Sensors Journal

Abstract:The linear back projection (LBP) algorithm is often used for real-time online image reconstruction of electrical capacitance tomography (ECT) due to its high speed. However, due to the fact that the image reconstruction of ECT is a nonlinear ill-posed inverse problem, reconstructed images obtained by the LBP algorithm that simplifies ECT image reconstruction as a linear problem, tend to have distortion and can only be used for qualitative observation. In this paper, a deep fully-connected neural network, which improves the imaging quality of the LBP algorithm by compensating its imaging results is proposed. Instead of simplifying the ECT image reconstruction as a linear problem, our proposed compensated LBP algorithm uses a deep neural network to map the nonlinear relationship from capacitance to permittivity distribution. Furthermore, the difference between the capacitance regarding the permittivity distribution reconstructed by the LBP algorithm and the actual capacitance is used as the input of the network while the difference between the reconstructed permittivity distribution and the actual permittivity distribution is used as the output of the network. The results of the network can be used to compensate the image reconstruction results of the LBP. This strategy makes the ECT image reconstruction need only to deal with a support interval significantly smaller than that of the original ECT image reconstruction problem and is helpful to suppress the nonlinearity to be trained. Both the training and testing results based on simulation data instances and experimental data show that the proposed compensation network has a great improvement on image reconstruction results of the LBP algorithm. In addition, the computation load is comparable to the original LBP algorithm.
Distributed Brillouin frequency shift extraction via a convolutional neural network

Yiqing Chang,Hao Wu,Can Zhao,Li Shen,Songnian Fu,Ming Tang

DOI: https://doi.org/10.1364/PRJ.389970

IF: 7.6

2020-04-24

Photonics Research

Abstract:Distributed optical fiber Brillouin sensors detect the temperature and strain along a fiber according to the local Brillouin frequency shift (BFS), which is usually calculated by the measured Brillouin spectrum using Lorentzian curve fitting. In addition, cross-correlation, principal component analysis, and machine learning methods have been proposed for the more efficient extraction of BFS. However, existing methods only process the Brillouin spectrum individually, ignoring the correlation in the time domain, indicating that there is still room for improvement. Here, we propose and experimentally demonstrate a BFS extraction convolutional neural network (BFSCNN) to retrieve the distributed BFS directly from the measured two-dimensional data. Simulated ideal Brillouin spectra with various parameters are used to train the BFSCNN. Both the simulation and experimental results show that the extraction accuracy of the BFSCNN is better than that of the traditional curve fitting algorithm with a much shorter processing time. The BFSCNN has good universality and robustness and can effectively improve the performances of existing Brillouin sensors. © 2020 Chinese Laser Press

optics
Brillouin Optical Time Domain Analyser Signal Processing Based on Broad Learning System

Yan Liu,Xiaolong Wang,Yukai Shi,Zhiyun He

DOI: https://doi.org/10.1109/icsp62122.2024.10743492

2024-01-01

Abstract:In this paper, a broad learning system (BLS) is used to process the temperature sensing signals of a Brillouin Optical Time Domain Analyser (BOTDA). The fiber temperature is extracted directly from the local Brillouin gain spectrum (BGS) without the need of Brillouin frequency shift spectrum (BFS) conversion. Since the BGS signal conforms to the Lorentzian line shape, the BLS is trained with the ideal Brillouin gain spectrum and its corresponding temperature in the training phase, and tested with the actual BOTDA monitoring data in the testing phase. The experimental results show that BLS provides a solution with higher accuracy and larger error tolerance for the direct acquisition of BOTDA temperature sensing information. Compared with ANN, the network training speed of BLS is increased by three times, and the weight matrix can be updated quickly using pseudo-inverse ridge regression for the added BGS signals without re-training, which means that its data processing performance is more closely matched to that of BOTDA sensing signals.
3D Image Recognition Technology for Analog Circuit Based on DSP and BP Algorithm

Shiling Huang

DOI: https://doi.org/10.1109/ICCASIT55263.2022.9986999

2022-10-12

Abstract:At present, the "artificial intelligence fever" centers on image recognition and video recognition, driving the computer vision market. BP neural network, as a commonly used processing model in the field of image recognition, has further become a research hotspot of 3D image recognition technology for analog circuits. The purpose of this paper is to design a three-dimensional image recognition technology for analog circuits based on DSP and BP algorithm. The recognition algorithm adopts the fastest version of BP neural network, LMBP, and improves it. On the one hand, the conjugate gradient method is used to speed up the convergence process and improve the performance of the algorithm. The above algorithm is transplanted into DSP platform. Finally, using a breadboard, an analog circuit is built, 200 kinds of faults are artificially created, and the system is tested many times. The results show that using the improved BP algorithm, the correct rate of fault diagnosis is 84%, and the BP algorithm is 74.5%, which are improved compared with the traditional fault dictionary method, so the system has certain use value.

Computer Science,Engineering
Application of BP Neural Network Algorithm in the UV Absorption Spectrum of the Phosphate Solution

WANG Rui,YU Zhen-Hong,YU Ying

DOI: https://doi.org/10.3969/j.issn.1004-8138.2009.04.014

2009-01-01

Abstract:The ultraviolet spectrophotometric non-linear relationship between absorbance and the concentration of solution was treated by the BP neural network with the neural network toolbox in Matlab,which provides many of the neural network design,training and simulation functions,and is convenient and effective that bestowing the BP network to solve the problem. The results show that the network convergence rate of the Levenberg-Marquardt method is the fastest,and the rate of the quantized conjugate gradient method is the slowest.The BP neural network is more suitable to resolve this issue.
A new color blindness cure model based on BP neural network

Yu Ma,Xiao-Dong Gu,Yuan-Yuan Wang

DOI: https://doi.org/10.1007/11760191_109

2006-01-01

Abstract:There is no method of color blindness cure so far although some research on color blindness has been done. In this paper we propose the BP neural network model for Color Blindness Cure (CBC) according to the biological experimental results. This model transforms abnormal output signals of two paths corresponding to two kinds of cones of the dichromats into normal ones of three paths corresponding to three kinds of cones of normal people. Computer simulation results show that this model has good performance and changes the abnormal vision of the dichromats into normal ones. In addition, some improvements of our model are introduced.
BpMC: A Novel Algorithm Retrieving Multilayered Tissue Bio-Optical Properties for Non-Invasive Blood Glucose Measurement

Weijie Liu,Anpeng Huang,Ping Wang

DOI: https://doi.org/10.1109/bibm.2017.8217690

2017-01-01

Abstract:Non-invasive blood glucose measurement is a crucial challenge in both academic and industry communities. Currently, most of non-invasive solutions are developed based on optical signals. However, their accuracy is still far from clinical requirements if these measured optical signals directly used to estimate corresponding glucose levels. To solve this challenge, a novel Back-propagation Monte Carlo (BpMC) algorithm is proposed to retrieve bio-optical properties in human multilayered tissues. Build on BpMC algorithm, two non-invasive blood glucose estimation models, namely BpMC-DEE and BpMC-CNN, are conceived. In contrast to existing black-box solutions, BpMC-DEE is a white-box model that is more reliable in clinical. BpMC-CNN is a gray-box model whose results are more accurate in cost of a larger dataset and higher computing complexity. BpMC-DEE and BpMC-CNN are embedded and implemented into our designed noninvasive device - Earlight, for clinical trials. The clinical trial results demonstrate that correlation coefficients of these two models reach 0.852 and 0.895, respectively, referring to invasive glucometers. In terms of Clarke Error Grids, our proposals account for 90.6% and 93.5% statistic points in regions A and B, respectively. Moreover, the BpMC algorithm can be applied to other components measurement of biological tissues.

BP Neural Network for Linear Birefringence of Three-Phase Optical Current Transducer

Evaluation of Chracteristic of Optical Thin Film By Neural Network System

Improved BP algorithm for neural network and its application in welding process control

Realization and Application of BP Neural Network in LabVIEW

Soft-sensor modeling via neural network PLS approach

Temperature Compensation of Elasto-Magneto-Electric (EME) Sensors in Cable Force Monitoring Using BP Neural Network

Wavelet Convolutional Neural Network for Robust and Fast Temperature Measurements in Brillouin Optical Time Domain Reflectometry.

Artificial Neural Network Compensation for an Optical Fiber Temperature Sensing System Based on Optical Time Offset

Locally Linear Back-propagation Based Contribution for Nonlinear Process Fault Diagnosis

Residual flux density estimation of the three-phase transformer using BP neural network

A PSO-BP Model for Nonlinear Fitting of Phyllostachys edulis' Thermal Conductivity

Application of BP Neural Net in Fault Diagnosis of Excavator Hydraulic System

Lnista-Net: Learning Nonlocal Regularized Iterative Shrinkage-Thresholding Network for Noninvasive Cardiac Transmembrane Potential Imaging.

Nonlinearity Compensation of Magneto-Optic Fiber Current Sensors Based on WOA-BP Neural Network

A Deep Learning Compensated Back Projection for Image Reconstruction of Electrical Capacitance Tomography

Distributed Brillouin frequency shift extraction via a convolutional neural network

Brillouin Optical Time Domain Analyser Signal Processing Based on Broad Learning System

3D Image Recognition Technology for Analog Circuit Based on DSP and BP Algorithm

Application of BP Neural Network Algorithm in the UV Absorption Spectrum of the Phosphate Solution

A new color blindness cure model based on BP neural network

BpMC: A Novel Algorithm Retrieving Multilayered Tissue Bio-Optical Properties for Non-Invasive Blood Glucose Measurement