Abstract:The optical detection of ultrasound field is to study the acoutooptic effectiveness when the light passes through the sound disturbing field. The information of the sound field can be obtained by analyzing the emergent light. Compared with conventional ultrasound field detection technology, the optical detection is a nonintrusive method, which has no disturbing to the sound field. It has the advantages of fast detection speed, high spatial resolution and wide frequency band. It has been widely utilized in the ultrasound field detection since 1936, when the first accurate model of acoustooptics was introduced. It has made significant progress in both theoretical research and sound field detection. In this study, we summarize the progress in acoustooptic effectiveness, and prospect for the future application of acoustooptic effectiveness.

Abstract:Current intelligent diagnostic models of mechanical systems require massive historical data under different health states and corresponding labels to complete model training. However, the abnormal samples are hard to be acquired in some mechanical systems. In the condition where abnormal samples are absent for training, this paper proposes a novel anomaly detection method of mechanical systems. The new method combines generative adversarial network (GAN) and autoencoder (AE) to establish an encodingdecodingencoding network model. The proposed model is firstly trained by the normal samples of the mechanical system acquired in the early stage, then the model is used to test the online collected real time monitoring samples with unknown health state and outputs the dissimilarity between the latent features obtained in two encoding. Finally, the system is monitored by inspecting the variation of the output dissimilarity. Three groups of experiment analysis results are used to verify the effectiveness of the proposed method. Compared with traditional methods, the proposed method can detect the anomaly earlier, the dissimilarity index has a larger increment when anomaly occurs, and this method can more stably characterize the fault evolutionary process.

Abstract:Steel strand is one of the most important and indispensable stress components in longspan bridges, however, there is still lack of effective method to detect and monitor the stress of the steel strand in existing bridges due to the influence of necessary anticorrosion measures. The ultrasonic guided wave carries obvious stress characteristics in the propagation along the steel strand. Through the wavelet packet decomposition of the guided wave signal in the timefrequency domain, the coefficient matrix of wavelet packet decomposition is extracted under different stress states. Then, the singular value vector of the coefficient matrix is used as the characteristic parameter, the support vector regression model with learning ability is established to detect the stress value of the steel strand. The results show that the singular value vector of the guided wave is an effective characteristic parameter in presenting stress state, and the relationship between the singular value vector distance and steel strand stress value shows a monotonous linear law in stepwise loading process. While adopting the support vector regression model established with the singular value vector to predict the steel strand stress, the determination coefficient of the results reaches to 0973 9, the prediction result obtained using the support vector regression model is more stable than that using neural network method.

Abstract:To improve the accuracy of mediumlong term prediction of satellite clock bias (SCB), a prediction method based on Vondrak filter firstorder differential modified exponential curve model (VDMECM) is proposed. First, the frequent hopping and gross errors phenomenon of SCB before modeling are considers. The median absolute deviation is used to detect and eliminate the clock hopping and gross errors data. Meanwhile, the missing clock data can be recovered by using the Lagrange interpolation method. Secondly, the systematic and random errors of SCB are studied. Vondrak filter smoothing algorithm is used to reduce these errors. Thirdly, the prediction performance of the model is improved by considering effective data bits in SCB. The firstorder difference is used to eliminate the influence of the trend item of the clock bias sequence. And, MECM prediction model is formulated. Finally, the mediumlong term forecast experiments for the next four time periods are implemented based on the postaccuracy precision SCB published in the IGS server. Two typical changing trends are also considered. Experimental results show that the mediumlong term prediction accuracy of this method is better than the quadratic polynomial model (QPM) and the gray model (GM (1, 1)). Compared with these two methods, the average prediction accuracy (RMS) of 60day is increased by 9200% and 8080%, and the average prediction stability (Range) of 60day is increased by 9240% and 8140%.

Abstract:Wear debris in the hydraulic oil contains a lot of important information about the wear of hydraulic components. The detection of wear debris can prevent hydraulic system failures effectively. The inductive sensor has the shortcoming of weak ability to detect nonferromagnetic metal particles. To solve this problem, this paper proposes an integrated wear debris detection device, which consists of a capacitive sensor and an inductive sensor. The capacitive sensor can distinguish the air bubbles and metal particles in hydraulic oil. The inductive sensor can distinguish the ferromagnetic and nonferromagnetic metal particles in hydraulic oil. By combining the detection results of two sensors, air bubbles, ferromagnetic and nonferromagnetic metal particles can be detected with a high sensitivity. The designed wear debris detection device in this study can detect and distinguish 80 μm air bubbles, 30 μm iron particles and 45 μm copper particles. The integration of the capacitive sensor and the inductive sensor compensates for the shortcomings of the two detection methods effectively. The combination of multiple types of sensors is of great significance for improving the detection accuracy of the wear debris detection device.

Abstract:Aiming at the problem that traditional carbon fiber electric field electrode is subject to the capacitive reactance effect and cannot response to low frequency signals, in this paper, a new type of carbon fiber marine electric field electrode is prepared using the modification method of concentrated nitric acid oxidation and silica sol coating. The experiment results show that large number of nitrogenoxygen functional groups are introduced to the modified carbon fiber surface, which not only increases the specific surface area and hydrophilicity of the electrode, but also reduces the effect of the capacitive reactance effect. In terms of electric field response performance, the modified electrodes can correctly respond to the LF signal with frequency as low as 1mHz. Compared with the results before modification, the response accuracy and sensitivity are improved greatly, the linearities of the modified electrodes responding to frequency of 10 mHz and 1 mHz are 33% and 21%, and the sensitivities are 0048 2 and 0050 3, respectively.

Abstract:To meet relevant requirements of the assemble experiment of the orbit telescope, the factors of range, stiffness and sensitivity of sixaxis force sensor are considered. In this way, a sixaxis force sensor with large range and high sensitivity is designed. First, the mathematical model of the classical cross beam six axisforce sensor is formulated. The mathematical expressions of the surface strain of strained beam and the deformation stiffness of elastomer are compared. When each channel acts alone, a scheme for improving the sensitivity of sensor is presented. Then, the sensor structure is designed in detail, and the feasibility of the structure scheme is verified from the finite element analysis. Finally, the sixaxisdimensional force sensor is made and calibrated. Results show that the repeatability error is less than 033%FS, and the measurement sensitivity of the force channel and the moment channel is larger than 083 mV/V and 26 mV/V. The designed sixaxis force sensor meets the requirements of the project and has been applied in the ground assemble experiment of the onorbit telescope.

Abstract:Fiber optic acoustic sensors can be widely used in energy source, security defense and other important fields, however their noise is complex, which affects the measurement accuracy and stability. Aiming at this problem, in this paper, an acoustic sensor system based on linear fiber optic Sagnac interferometer is demonstrated, and an improved wavelet threshold denoising algorithm is proposed. Furthermore, on this bases, combining with the signal characteristics, a synthesis filtering scheme is designed to improve the speech detection quality of the system. Taking the actually measured speech signal as an example, with this algorithm the Allan variance of the signal is reduced to 824×10-14, while the class spacing criterion is increased to 691, which effectively improve the quality of detected speech signal. The proposed algorithm can be widely used in acoustic sensing field, and has great significance for the accuracy of the later stage acoustic source localization for the optical fiber interferometer acoustic sensing system as well.

Abstract:There are great nonlinearity and uncertainty in the Global Positioning System/Inertial Navigation System (GPS/INS) integrated navigation system composed of low cost devices. In order to improve this issue, a filtering method with Sliding Mode Observer (SMO) introduced is proposed in this paper. Firstly, in this method, the integrated navigation system model is established, the calculation process of Extended Kalman Filter (EKF) is introduced and its shortcomings are analysed. Then, the basic principle of SMO is introduced, and the Sliding Mode Observer is constructed according to the system. Finally, the implementation procedure of the EKF integrated navigation algorithm with SMO is explained. The SMO integrates the model error, state estimation and mean variance into EKF algorithm to correct the output of the system. The trajectory simulation experiment and vehicle test prove that the proposed method is superior to the traditional EKF method, and has higher filtering accuracy. In the vehicle test, when the satellite signal is out of lock for 15 s, compared with those of the EKF method, the eastbound and northbound position errors are reduced by 53%, 37%, respectively. The result proves that the proposed method can effectively suppress the error divergence of GPS/INS integrated navigation, and provides certain reference value for future engineering practice.

Abstract:By establishing the error model of heading angle for triaxial magnetometer that widely used in magnetoinertial navigation system of unmanned aerial vehicle (UAV), the heading misalignment error that consists of constant error and half circle compass error is concluded. Therefore, a novel calibration algorithm is proposed, which is based on dual inner products. The principle of proposed method makes use of the invariance of geomagnetic vector and itself inner product, as well as the constant inner product of geomagnetic vector and gravity vector. This method can eliminate the mutual misalignment of triaxial magnetometer that cannot be identified using the invariance of vector norm only, and thus it can achieve complete calibration of triaxial magnetometer. Numerical simulations and experiment results show that the proposed method has better performance compared with scalar checking, dot product invariance and twostep methods. It can effectively reduce the norm error of geomagnetic vector and heading angle error of UAV, and it has preferable robustness against sensor noise.

Abstract:Accurate measurement of oilwater interface in the crude oil tank is a basic requirement of petroleum exploitation process. The development of this technology is very important in the research of petrochemical process system engineering. Firstly, 15 kinds of oilwater interface measurement technologies are analyzed. The measurement principle and application of each measurement technology are introduced. These oilwater interface measurement technologies are compared in three terms, including technical mode, measurement mode and calculation mode. Then, this paper focuses on the oilwater interface calculation mode and compares the research progress of algorithms adopted in this direction. The calculation methods are summarized in four aspects, including direct reading method, keyparameter based method, matrixdata based method and imageanalysis based method. Finally, this paper prospects the future development of oilwater interface measurement technology. The future development and the related difficulties are analyzed in five aspects, including hybridmode technology, high precision measurement processing, noncontact measurement, multidimensional data calculation method and intelligent information system construction platform.

Abstract:In the programmable capacitor, due to the influence of stray capacitance generated by the edge effects, it is difficult to achieve accurate and stable small capacitance standards of subpF and below. In this paper, a method of implementing a small capacitance standards based on Kelvin equipotential protection electrode is proposed, so that the stray capacitance affecting the small capacitance is limited to the stray capacitance generated by the edge effect between the center electrode and the guard electrode. The analytical model of equivalent small capacitance standard is established，and the algorithm for analyzing the variation of stray capacitance is proposed by conformal mapping. It is found that the air gap between the center electrode and the guard electrode is the main influencing factor of stray capacitance when both the electrode thickness and the distance between the electrode plates are determined. Combined with the finite element analysis software Ansoft Maxwell to verify the optimal air gap between the main electrode and the guard electrode, and the small capacitance standards will be realized. It has been tested that in the programmable fused silica capacitor, the actual small capacitor capacitance value of the subpF level is basically the same as the design small capacitor capacitance value.

Abstract:Aiming at the problems that traditional fuzzy Cmeans (FCM) algorithm does not consider the mutual relationship among pixels and requires to obtain the initial cluster center when dealing with image segmentation, the paper proposes a FCM clustering segmentation algorithm considering the relationship among pixels. Firstly, the algorithm adopts the principle of data field, uses the mutual relationship among the pixels to calculate the potential values of the pixels and form the image data field. Then, the initial cluster center of the FCM algorithm is determined with the image data field potential center. Finally, based on the image data field, the FCM algorithm is used to realize the clustering segmentation of the target image. In order to verify the effectiveness of the algorithm, the artificial synthetic image and the workpiece surface defect image were used for experiments. The experiment results show that the algorithm has better segmentation effect. Meanwhile, for different noisy images with streaks, decarburization and hole defects, the segmentation accuracies are above 93%, and has a high mean structural similarity.

Abstract:Aiming at the problem that object tracking is subject to failure in complex scenes such as occlusion and illumination variation, a highaccuracy and robust object tracking algorithm is proposed. Firstly, the target model based on edge information, the filter model based on HOG feature and the color model based on color histogram are merged into a more accurate and strong robust tracking model. Then, the double tracking reliability judgment criterion based on the score of the feature is proposed to detect the reliability of the tracking result. Finally, when the reliability of the tracking result is low, particle filtering, sparse representation and distance constraint positioning are used for redetection to achieve continuous and stable tracking. On the OTB2015 dataset, the average overlap precision of the proposed algorithm is 782%, the average center location error is 231 pixel and the average tracking rate is 308 f/s, which indicates that the accuracy and robustness are better than those of other algorithms. The algorithm was verified on mobile robot and vehicle tracking platform, the average overlap precisions are 975% and 972%, the average center location errors are 68 pixel and 126 pixel, respectively, and the average tracking rates are 291 and 284 f/s, respectively. The proposed algorithm can effectively track the targets in above mentioned complex scenes and meet the realtime requirements.

Abstract:The lateral resolution detection of ophthalmic optical coherence tomography (OCT) equipment is easily affected by the noise and interference fringes in spot image. To improve the accuracy of lateral resolution detection, an ophthalmic OCT equipment lateral resolution detection method based on dynamic weight particle swarm optimization (DWPSO) algorithm is proposed. Firstly, through constructing the dynamic weight factor, the local PSO algorithm is improved and the DWPSO algorithm is established. Then, the beam intensity distribution model is established. The DWPSO algorithm is adopted to identify the beam intensity distribution model parameters, and the beam width is obtained. Finally, the least squares algorithm is used to fit the beam width, and the numerical aperture is obtained. On this basis, the numerical aperture is used to achieve the lateral resolution detection of ophthalmic OCT equipment. The experiment results indicate that the DWPSO algorithm can quickly obtain the global optimal solution of the beam intensity distribution model compared with other algorithms. The lateral resolution of the measured ophthalmic OCT equipment is 1821 μm on close focus, and its lateral resolution is 4991 μm on far focus. The proposed lateral resolution detection method can effectively reduce the influence of noise and interference fringes in spot image, and has good noise robustness.

Abstract:Point cloud registration is a key step for indoor mobile robot pose estimation scenario building. Current point cloud registration methods hardly work in lowtexture scenes. To improve scene adaptability of indoor mobile robot, this paper proposes a novel improved 3DNDT point cloud registration algorithm this paper proposes an improved ORB algorithm to ensure valid feature extraction in lowtexture scenes; In addition, to improve accuracy and effectivity of point cloud registration, this paper proposes improved 3DNDT algorithm for quick and accurate registration matrix solving. Quantitative results on famous TUM datasets show our system performs as good as or better than other popular solutions (lower RMSE value than 002 m), and time consuming decreases 3 time than traditional 3DNDT algorithm; Notably, our algorithm can work in low texture scenes. Therefore, our algorithm can improve scene adaptability of indoor mobile robot.

Abstract:Light series antifriction bearing cages are prone to deformation during the riveting process due to the small diameter of pockets and the relatively large nail hole distance between the two halves, resulting in the defects of riveting skew. Therefore, this paper proposed a pattern recognition method based on image texture features for the accurate identification of cage skew defects. Firstly, a bearing normalization expansion algorithm was improved, which realized the automatic optimization of the starting point of the expansion to avoid missegmentation of the cages, rivets and rolling elements. Secondly, a bearing image cage localization and segmentation algorithm was designed, and 7 cage regions were accurately separated. Finally, the Hu moment and rotation invariant uniform local binary pattern (LBPriu2P,R) were extracted separately as texture features, and the classification model was constructed by combining PCA and SVM. The results showed that the correct recognition rate of the SVM model based on Hu moment and LBPriu2P,R were 85% and 100% respectively. Therefore, the LBPriu2P,R feature combined with the SVM model has a good recognition effect on the bearing cage skew defect. This method was expected to provide a reference for the automatic identification of the defects in the antifriction bearing cage riveting process.

Abstract:The multichannel domescreen projection is an important information visualization technology. Projection images are prone to image distortion and bright bands are generated in the coincident region of the image. To solve these problems, an image fusion processing method based on image correction principle is proposed. Firstly, the geometric position relationship among projection images is analyzed, and the distorted image is corrected by grid division method to realize the grid position splicing and alignment of edge images. Then, the edge attenuation algorithm is used to fuse images. The shading plate device for spherical screen projection fusion is designed, which filters the overlapping light paths generated in the domescreen of multisource projection. In this way, the bright bands on the domescreen image in multisource projection can be eliminated, and the overall brightness of the projection image uniform is reserved. Finally, the edge fusion effect of the projection image is analyzed by experiments. Results show that the average peak signaltonoise ratio of five groups of images selected by the experiment is 33550 dB, which prove that this method has a good edge fusion effectiveness.

Abstract:With the globalization of BeiDou satellite navigation system, BeiDou receivers based on software defined radio (SDR) are used more and more popular in BeiDou navigation and positioning. However, in the high precision measurement applications, such as real time kinematic (RTK) application, the baseband sampling ADC in BeiDou receivers suffers the influence of sampling clock jitter, which is added with the inherent quantization noise and thermal noise, and generates neglectable influence on the ranging error. In this paper, the BeiDou published global signals B1I, B1C, B2I, B2a and B3I are analyzed firstly. Secondly, the math models of the sampling noises including the clock jitter, quantization noise and thermal noise in ADC sampling are described in detail. Thirdly, the synthesized SNR formula is proposed, and the general expression of the ranging error is derived from the carriertonoise ratio (C/N0). An engineering quantitative analysis method for selecting different parameters under high accuracy requirement is given. Furthermore, aiming at the influence of different parameters on the BeiDou signal ranging error, simulation analysis was performed. Simulation results show that under the goal of ranging error of 0.6cm, it is required that the sampling clock jitter should not exceed 45 ps, the central frequency of the receiver should not be larger than 80 MHz, the bandwidth should not be less than 25 MHz, the quantization bit length should not be less than 7 bits, and the C/N0 should not be less than 30 dB. Finally, the test experiment on the BeiDou signal B2I was conducted. Under the conditions of the sampling clock jitter of 35 ps, the receiver central frequency of 61.38 MHz, the bandwidth of 32 MHz, the quantization bit length of 12 bits, and the C/N0 of 67 dB, the solved ranging error is 0.41 cm, which verifies the effectiveness of the proposed method.

Abstract:In this paper, the direct way of designing a stable controller for nonlinear system is studied. A framework of learning controller with Lyapunovbased constraint is proposed, which transforms design and analysis of a controller to straightforward way by solving an optimization item with the Lyapunov constraint. A novel way of the global stability guaranteed controller is realized directly. Firstly, the optimization problem subject to Lyapunovbased constraints is formulated, in which the tracking error is the objective function to be minimized. Secondly, the controller combines with PID and feedforward is given in form of neural networks. Finally, the optimization solution of the controller method is analyzed and solved, in which some deep learning technologies are used to enhance the capability of solution. Test results of two simulations of 2 order linear and nonlinear systems demonstrate that the proposed method has high performance in speed of convergence, tracking error and smoothness and amplitude of control output. Results of comparison simulation with backstepping control to the nonlinear system with disturbance, noise, uncertainty of parameters and the difference of reference output demonstrate that the proposed method has high performance in terms of robustness and generalization. Results of simulated physical experiment of onestage rotary inverted pendulum based on VRep and physical testing of singleaxis controlling for quadrotor prove that the method proposed is capable of high precision control and strong disturbance rejection.

Abstract:In recent years, fastpaced life has led to a gradual increase in the incidence of emotional disorders.In order to alleviate people′s tension and anxiety,this paper builds a relaxation training system in the form of online realtime feedback of EEG. The system extracts the relative power of the brain αwave index as feedback value and feeds it back to the user in the form of visual animation. Finally, 20 healthy subjects were subjected to EEG feedback relaxation training to verify the effectiveness of the system. The results showed that the SAS score of the selfrating anxiety scale of the subjects in the experimental group was reduced obviously, and there was significant difference (t=2411, P<005). Through the analysis of the relative total power of the resting EEG αwave index before and after training, in the experimental group, the relative total power of the αwave index measured after training is significantly higher than that of the previous measurement, and has significant difference (t=-4256, P<0 05); but this phenomenon is not observed in the control group.The above results show that the training system can achieve the effect of making the subject feel relaxed and the emotional reaction tends to be calm.

Abstract:In intelligent manufacturing, the prognostics and health management of the equipment driven by big data has been paid much attention. In recent years, because it can capture more hidden knowledge in the process of feature extraction of hierarchical structure and has good data adaptability across a variety of domains, deep learning has become a hot topic in the field of equipment health management. It has been widely used in equipment fault diagnostics and prognostics. This paper systematically reviews emerging literatures on the application of deep learning in equipment health management. It summarizes, classifies and explains main publications on this trendy topic. Various architectures and related theories are also discussed. As a review, this paper expounds the achievements, challenges and future development trends of the deep learning in the field of equipment fault diagnostics and prognostics. It provides a clear direction for practitioners including the industry to select, design or implement deep learning architecture for the equipment health management.

Abstract:EPFbased UAV reconnaissance moving target localization algorithm needs to use the EKF algorithm to calculate the mean and covariance of all particles in the sampling stage, which results in a large amount of computation. In this paper, an improved adaptive EPF algorithm based on KL divergence is proposed. The method uses the EKF algorithm to update the first half of the particles in the sampling phase. The latter half of the particles is still updated with the prior probability distribution, and then according to the KL divergence between the probability distributions of the two particle sets, the number of the particles at current moment is adaptively updated. Selecting the appropriate number of particles while ensuring accuracy greatly reduces the amount of calculation and improves the speed of operation. Through the verification with actually measured flight data, the average number of particles in each sampling period for this algorithm is 40, and the average calculation time in each sampling period is 8ms. Compared with EPF algorithm, this method can significantly reduce the calculation time while ensuring the positioning accuracy, and has certain engineering application value.

Abstract:In recent years, to reduce the influence of commonmode voltage on voltage source inverters, model predictive commonmode voltage reduction methods are widely studied. However, the conventional methods use only one nonzero voltage vector per control period, in which may bring large current harmonics. Therefore, a hybrid multivectorbased model predictive commonmode voltage reduction method is proposed. First, the principle of the proposed method is presented. Then, the influences of dead time and current ripples on the commonmode voltage are analyzed in detail, and the proposed multivectorbased model predictive control method is further improved. In the improved method, only one nonzero voltage vector is applied at current sector 7. While in other current sectors, multi nonzero voltage vectors are applied. In this way, not only can the commonmode voltage be restricted within ±Vdc/6, but also the current total harmonic distortion is reduced. Simulation and experimental results verify the effectiveness of the proposed method.

Abstract:Combining the advantages of generalized regression neural network (GRNN) in nonlinear fitting and flexible network structure, a prediction model for the anticompression strength of cooked pellets is constructed to determine the proportion of raw materials (Ca, Si, Mg, etc.) and the important parameters characterizing the quality of pellets in the pellet production process. Based on the anticompression strength prediction model and with beetle antennae search (BAS) algorithm, an intelligent recommendation model for optimum pellet ingredient proportion is constructed. In the adjustable range of pellet ingredient, the intelligent recommendation scheme for optimum pellet ingredient proportion is presented. The simulation and experiment results of the recommended scheme show that the prediction model of cooked ball anticompression strength has super interpolation ability and excellent generalization performance. In the pellet ingredient change range of no more than 20%, the intelligent recommended optimal pellet ingredient proportion scheme can increase the cooked ball anticompression strength by more than 16% on average; the system operates steadily and the simulation results are effective. The BAS intelligent recommendation model was applied in the actual pellet manufacturing process, compared with that in the same period of the previous year, the daily average anticompression strength of cooked pellets is improved significantly and the practical application effect is good.