Abstract:When the imaging system records image information by using sensor, it loses important phase information due to only record intensity information. Traditional interferometric phase retrieval techniques are restricted in their use due to the strict interference conditions. With the significant development of computational imaging techniques, the non-interference phase retrieval techniques represented by transport of intensity equation, coherent diffraction imaging, coherent modulation imaging and ptychography which are both based on CDI, have attracted great attention. They have been utilized in the field of optical measurement. This kind of techniques does not require reference light with a simple system structure, which can retrieve phase information directly from diffracted intensity images. They have great potential for applications in the measurement field. On this base, the current researches and recent progress of several typical non-interference phase retrieval techniques are presented, and the main technical features of each method are discussed and analyzed.

Abstract:The UWB ranging technology is widely used in the industrial field because of its low power consumption and high security. Due to the attenuation of penetration and multi-channel reflection of UWB pulse signal in industrial non line of sight (NLOS) scene, the direct path signal attenuation and multi-path delay will reduce the positioning accuracy of the system. To solve this problem, a disambiguation and reconstruction algorithm is proposed, which is based on UWB channel impulse response by modeling UWB signal transmission and response. This algorithm analyzes the channel impulse cluster signals by Poisson distribution and enhances the characteristics of the head path signals representing the linear propagation distance. Therefore, it realizes the extraction and separation of the direct path signals under complex multipath interference, which improves the identification accuracy of signal arrival time and reduces the UWB ranging error. Furthermore, we simulate the UWB channel response according to the IEEE802015. 4a standard and establish the experimental research system under different sheltered environments. Simulation and experimental results show that the proposed disambiguation and reconstruction algorithm can significantly improve the positioning accuracy in the multi-interference environment. Especially in the industrial NLOS scene with multiple reflectors and multiple obstructions, the error can be reduced by about 79. 1% compared with threshold method.

Abstract:To address the issue that it is difficult to detect the internal cavity of the small-size ballscrew nut and evaluate its machined surface quality due to its complex internal geometrical configuration and small aperture, a multi-scale measurement method of ballscrew nut based on the spectral confocal principle is proposed. A non-contact optical displacement sensor is used to perform indexed sectional scanning axially for the internal cavity, and three-dimensional point cloud data of the cavity surface expressed in the cylindrical coordinate system are obtained, in which the abnormal points at the interface between the raceway and inner wall are removed by using uniform interpolation filtering. Then, the denoise process is implemented by the wavelet threshold method. Based on the multi-resolution analysis characteristics of the two-dimensional wavelet decomposition, its number of decomposition layers is adaptively determined by means of the energy conservation property of the wavelet decomposition. The inner surface of the screw nut transformed into the Cartesian coordinate system unwrapped along the generatrix is dissociated at different scales, in which the low-frequency components are employed to calculate the profile structure parameters and the high-frequency components for assessing the surface roughness. The inspection results show that the proposed measurement method can evaluate not only the dimensional accuracy of the inner cavity structure but pitch error and lead angle error of the surface quality are 2 μm and 0. 016°, respectively.

Abstract:The prismatic Li-ion battery is a heterogeneous structure consisted of jelly-roll and aluminum shell. The thermal conductivity of the jelly-roll and the heat transfer coefficient between the jelly-roll and the shell are key parameters affecting heat dissipation. At present, there is no direct test method. It is proposed to use battery heat storage to construct a heat source, trigger unsteady heat transfer through the temperature change of the cooling surface. The thermal imaging is used to record the spatial distribution of the shell temperature along the heat transfer direction and its time evolution, and substitute it into the three-dimensional unsteady heat transfer inversion model. Meanwhile, core facing, longitudinal thermal conductivity, core and shell bottom, and maximum surface heat transfer coefficient are calculated. Experiment platform is constructed, tests carried out for two battery cells and three standard specimens. The relative standard deviation of the thermal conductivity test results of the two batteries cells is between 5% ~ 10% . The relative deviation of the results for three standard specimens are all below 5% . The method provides a new tool for experimental evaluation of thermal parameters for various non-uniform structures, for example the cylindrical battery cells.

Abstract:The whole-angle hemispherical resonator gyro ( HRG) is a typical rate integrating gyroscope with the advantages of high precision, high reliability, and long life. The system bandwidth of the whole-angle HRG closed-loop control circuit is studied. The working principle of the whole-angle HRG is analyzed. The input-output control circuit model and the disturbance-output control circuit model are established respectively. Two models are formulated and simulated by Simulink program simulation. The dynamic characteristics of two control circuits are discussed. Finally, the experiment of the whole-angle HRG is carried out. Simulation results show that when the input angular velocity exceeds 532. 8°/ s, the parameter output fluctuation caused by the disturbance signal reaches the maximum and remains constant, which corresponding to the bandwidth of the control circuit is 1. 48 Hz. The simulation and experiment results indicate that the low bandwidth of the whole-angle HRG control circuit can lead to the fluctuation of the ellipse parameters caused by the disturbance signal and deteriorates the control effect. In addition, the frequency of the disturbance signal is proportional to the input angular velocity. The fluctuation tends to be stable when the input angular velocity exceeds the angular velocity with respect to the cut-off frequency of the closed-loop control circuit. The research results in this paper provide a theoretical basis for the dynamic performance analysis of the whole-angle HRG.

Abstract:To achieve low-frequency, multi-directional energy harvesting and high output performance, a piezoelectric energy harvesting system with arc helix structure is proposed in this article. The arc helical piezoelectric energy harvester cannot only reduce the resonance frequency and the volume of the system, but also the arc type cantilever beam has asymmetry. Therefore, the multi-directional collection can be performed. The stress distribution, the resonant frequency and the output performance of the energy harvesting system with different radians are analyzed theoretically. The 2π, 3π, and 4π arc spiral piezoelectric vibration energy harvesting system are fabricated, and their performance are measured and compared. Results show that the 4π arc spiral piezoelectric energy harvesting system has better output performance. The resonant frequency is 47 Hz, the output voltage is 23 V, and the output power is 353 μW. The arc spiral piezoelectric vibration energy harvesting system can be applied to human health detection, environmental control systems, embedded systems, military security and other application fields.

Abstract:The ultrasonic guided wave temperature sensor is currently an emerging temperature measuring instrument in the industrial field, and the selection of its guided wave has important impact on the performance of the sensor. Magnetostrictive torsional wave is suitable for the application, as the basis for temperature measurement with its wave speed proportional to the temperature as well as the characteristics of low decay and easily pick-up. In this article, the magnetostrictive material Fe83Ga17 wire is selected to generate and conduct magnetostrictive torsional waves. The relationship between the flight time and temperature of the torsional wave at a fixed distance is calculated to achieve the fitting relationship between the wave velocity and temperature. According to the fitting relationship and the measured torsional wave velocity, the temperature measurement is realized, and an output voltage model based on a magnetostrictive torsional wave temperature sensor is proposed. In experiments, the output voltage of the magnetostrictive temperature sensor reaches 215. 7~ 465. 2 mV under the room temperature up to 500℃ . The fitting relationship between the measured temperature and the wave velocity can be used as a basis for temperature measurement. To increase the upper temperature measurement limit of the temperature sensor, Fe83Ga17 wire is coupled with the heat sensitive material Ni 20Cr80 wire into a new waveguide wire, and the temperature sensor after structural optimization is obtained. The unamplified output voltage of the optimized sensor reaches 44. 9~85. 6 mV under the temperature from room temperature up to 1 200℃ . The wave speed of the torsion wave remains highly proportional to the temperature in the temperature region, which is capable for reliable temperature measurement under 1 200℃ .

Abstract:In this article, a portable tactile sensor for texture recognition is designed. The sensor uses fiber Bragg grating ( FBG) to recognize and detect different textures and sliding contact speed, which is convenient for robot system integration. Meanwhile, it has low requirement on hardware and software configuration, which is less affected by the environment. Based on the three-dimensional modeling, the sensor structure is statically analyzed and optimized to improve the sensitivity of FBG to force sensing information. A special experimental platform is designed and established to carry out static calibration experiments and complex multi-texture surface detection experiments on the sensor. Through time-frequency analysis of experimental data, it shows that the sensor can recognize different sliding contact speed and different texture. Among the sensors, FBG3 has the highest sensitivity, with an average sensitivity of about 51. 1 pm/ N and a linearity of 0. 998 under loading. When it is unloaded, the average sensitivity is about 50. 8 pm/ N, and the linearity is 0. 998. The repeatability error and hysteresis error of FBG2 are the largest, which are 2. 35% and 2. 23% , respectively.

Abstract:To solve the problem of complex echo signal and low signal-to-noise ratio caused by internal and external radiation of guided wave in the traditional electromagnetic ultrasonic omnidirectional guided wave transducer, a novel electromagnetic ultrasonic omnidirectional guided wave transducer (OUT-EMAT) is designed. The control method of double coil parameters and double excitation source is studied. The intensity and direction of guided wave radiation are controlled. According to the theory of the electromagnetic ultrasonic omnidirectional transducer, the mathematical model of wave displacement superposition is formulated, and the performance of OUT-EMAT is simulated and tested. Compared with the traditional omnidirectional EMAT, Simulation results show that the OUT-EMAT makes the ratio of internal and external radiation guided wave amplitudes change from 1 ∶ 1 to 2 ∶ 9. The external radiation is enhanced about 100% and the internal radiation is suppressed about 55. 6% . The new OUT-EMAT enhances the signal to noise ratio of detection echo, reduces the complexity of signal and improves the accuracy of crack location. It provides a new method for identifying crack defect effectively, which has guiding significance for practical engineering application.

Abstract:To overcome the problem that traditional ice sensors cannot detect multiple icing areas in a wide range of airfoils, a multiple icing area localization method based on the grayscale centroid method is proposed. The vibration spectrum of piezoelectric array before and after icing is collected, and the attenuation rate of spectrum amplitude is extracted as the positioning eigenvalue. Combined with the reconstruction algorithm based on multi-point icing probability detection (RAMIPD) and the grayscale centroid method, the center coordinates of the multiple icing areas are calculated to get the positions of the icing regions on a plane or a curved surface. When the sweep frequency range is the full band of 0. 5~ 5 kHz and the diameter of icing area is 70 mm, the average positioning error of single icing area on the aluminum plate is 27. 4 mm. The average positioning error of two icing areas is 29. 0 mm. The average lateral positioning error of icing areas on the airfoil is 22. 6 mm. To improve the positioning accuracy, a feature extraction method for sensitive frequency band selection based on wavelet packet decomposition (WPD) is proposed. Experimental results show that the single icing area positioning accuracy on the aluminum plate is increased by 34. 59% , the fluctuation of the positioning error is reduced by 45. 67% , the lateral positioning accuracy on the airfoil is increased by 54. 87% , and the fluctuation of the lateral positioning error is reduced by 46. 63% . The proposed icing area localization method can detect the multiple icing areas with high precision and provide a new idea for icing area detection.

Abstract:The traditional eddy current probes may neglect short cracks in special orientation. To address this issue, an absolute Koch planar eddy current probe is proposed. Both the excitation coil and the pickup coil of the probe adopt the Koch snowflake curve based on fractal self-similarity theory. Firstly, the finite element analysis for the Koch probe and the circular probe with the same dimension is conducted. Then, for cracks with different lengths, directions, widths and depths, the difference of response signals of two probes is compared. Finally, an experimental system is established to evaluate the finite element analysis results. Results show that for the inspection of cracks with different lengths, directions and widths, the finite element analysis results are qualitatively consistent with the experimental results. For the inspection of the crack with 3 and 5 mm in lengths, the variations of signals output from the Koch probe are at least 40% higher than those of the circular probe. In the detection of crack in different orientations, the variation of signal output from the Koch probe is at least 49% higher than that of the circular probe for the inspection of the 90° crack which is difficult to be detected. In the detection of crack in different widths, the variation of signal output from the Koch probe is at least 29% higher than that of the circular probe. In the detection of crack in different depths, the variation of signal output from the Koch probe is at least 6% higher than that of the circular probe. The Koch probe has more significant edge for the inspection of short cracks relative to the circular probe. The research result has important reference significance for the design of the electromagnetic structure of planar eddy current probe.

Abstract:The long-term positioning accuracy of shearer independent positioning device is the key to intelligent coal mining. However, the state-of-the-art positioning accuracy of shearer is difficult to meet the demand of automatic mining. This article puts forward the coal mine working face end automatic calibration of inertial navigation positioning system, constructs the underground local positioning system based on UWB system, and calibrates the shearer position which can achieve the long-term autonomous cycle positioning of shearer. Considering the low positioning accuracy of the UWB system in the non-line of sight (NLOS) environment, the constrained square root unscented Kalman filter (CSRUKF) positioning algorithm is proposed. To further reduce the interference of NLOS error, the positioning results are optimized based on the velocity robust Taylor series (VRTS) algorithm with consideration of the motion speed to improve the final positioning accuracy. Based on the UWB positioning system and mobile platform, the simulated terminal positioning is conducted. Experimental results show that the CSRUKF-VRTS method is able to reduce the localization error and improve the positioning accuracy in the NLOS environment. The mean error of the x, y and z-axis is reduced from 0. 332, 0. 404 and 0. 306 m to 0. 266, 0. 212 and 0. 159 m, respectively. The corresponding average accuracy is improved by 17. 4% , 47. 5% , and 48. 1% , respectively. The proposed cyclic positioning strategy of coalmine working face end provides a novel idea for the long-term independent positioning and a reference for the positioning of NLOS environment.

Abstract:In this article, a single mode-D-type-multimode fiber structure is designed and fabricated. The interference among the guided modes and the asymmetrical structure of the D-type fiber can be utilized to obtain the specklegram, which is extremely sensitive to external changes, and the eigenvalues of the specklegram are demodulated based on the texture feature algorithm. The refractive index (RI) measurement could be achieved via the relationship between the eigenvalues and RIs. In the study of the RI sensing characteristics, the effects of different polishing lengths and polishing depths on the sensing characteristics are discussed from the point of view of simulation and experiment. Both the simulation and experimental results show that the D-type optical fiber structure with the parameters of polishing length 2 mm and polishing depth 18 μm has the best refractive index sensing performance. The consistence between the simulation and experiment demonstrates the correction of the research method and the feasibility of the D-type fiber specklegram sensor for RI sensing. The maximum RI sensitivity of the EN reaches -1. 86 / RIU, and the fitting degree reaches 0. 964. The maximum RI sensitivity of the COR reaches -0. 23 / RIU, and the fitting degree reaches 0. 979.

Abstract:To improve the localization accuracy of the LIO-SAM algorithm, the LiDAR-IMU external parameter calibration is studied in this article. To address the low calibration accuracy of existing sensor calibration algorithms in vehicle-mounted conditions, a new joint calibration algorithm is proposed for vehicle sensors. Due to the low degree of freedom under vehicle conditions, the constraints of pitch and roll direction are not established sufficiently. To solve this problem, we first eliminate the influence of translation parameters by using a wide range of vehicle trajectories. Then, the normal distributions transform and iterative closest point algorithm are used to quickly obtain the initial values of rotation parameters. Furthermore, the calibration accuracy of pitch angle and roll angle is improved. In the coarse calibration process, the LiDAR odometer drifts and translation external parameters are not calibrated. Therefore, we further implement the full parameter calibration scheme based on the point cloud optimization method and make some enhancements. In this scheme, the turning region is utilized to construct constraints on the translation external parameters. Then, we combine the statistical error average effect and the displacement constraint to construct a new objective function. Finally, the full parameter calibration results are obtained by iterative optimization. Compared with the original LIO-SAM algorithm, experimental results show that the localization accuracy of LIO-SAM algorithm with external parameter calibration module is improved by 1. 74% ~ 5. 92% .

Abstract:To improve the quality of images reconstructed by planar electromagnetic tomography (EMT), a planar EMT image fusion algorithm is proposed, which is based on array rotation and improved evidence theory. For the first time, the sensor array rotation is applied to the field of planar EMT to increase the number of independent measurement. The measurement information of different rotation angles are used to achieve EMT image fusion. To address the problem that Dempster′s rule is insufficient in dealing with conflicting evidence, the data are clustered for all evidences to be fused to obtain an adaptive threshold. According to this threshold, all evidences are divided into two types, which are conflicting and non-conflicting. The different fusion rules are applied separately. The relationship between the number of rotation and the imaging effect is studied by simulation. Based on the array rotation and an improved D-S evidence theory, the artifacts are removed. The planar electromagnetic sensor with eight coils is developed. The aluminum plate fake detection shows the algorithm is effective. The relative image error of new method is reduced by 17. 0% and the correlation coefficient is increased by 22. 7% compared with traditional method.

Abstract:The process of target detection of FMCW mmWave vehicle mounted radar can be easily affected by clutter. To address this issue, a detection method that can effectively suppress clutter is proposed. Firstly, the companding algorithm is used to improve the three-frame difference method. In this way, it can better separate the clutter and target signal in the echo waves. Then, the CFAR algorithm is introduced to filter the clutter in the echoes, and the density peak clustering algorithm is used to find the target clusters. Finally, the power value of the target point in the target cluster is converted into the corresponding weight to improve the classical center of gravity method. The accuracy of target positioning is effectively improved. Experimental results show that the detection accuracy of this method reaches 96. 5% , 95. 9% , 95. 6% , and 94. 4% in four scenarios, including parking lot, straight road, urban street and campus road. Meanwhile, in terms of detection speed, the processing cycle of single frame radar echo signal can reach about 0. 3 s, which can meet the requirements of practical applications.

Abstract:Structural displacement response is the basic data for structural health monitoring (SHM) and safety condition evaluation. To make full use of the advantages of vision-based sensors, overcome its shortcomings, and improve the accuracy of structural displacement, an accurate structural displacement monitoring method by fusing tilt photogrammetry-based and accelerometer measurement is proposed in this article. On the one hand, the dynamic displacement at higher frequencies is reconstructed by the measured acceleration. On the other hand, the acceleration reconstructed and vision-based displacements in the same frequency band can be used to calculate the timedomain variable scaling factor, which can reduce the error caused by the conversion from pixel displacement to real displacement. Experimental tests on a self-anchored suspension model bridge and field tests on a simply supported beam bridge are carried out to explore the efficiency of the proposed method. The results of experimental tests show that, compared with the linear variable displacement transducer results, the maximum normalized root mean square error of the proposed method is 2. 70% which is about 60% lower than the traditional vision-based approaches. The proposed method can promote the further application of computer vision in the field of bridge structural health monitoring.

Abstract:The gradation and fineness modulus of manufactured sand are important quality indicators in industrial sand production. To solve the problem that the traditional method of grading detection of manufactured sand cannot be implemented online under the actual working condition. This article combines experimental research to propose an online detection method for manufactured sand grading based on deep learning. Firstly, the images of the stacked manufactured sand on the conveyor belt are collected. Then, the manufactured sand images are segmented by the convolution neural network ( CNN) . Finally, the gradation and fineness modulus of manufactured sand is computerized online by the image processing technology. Comparative experimental results show that the mask R-CNN instance segmentation model can effectively segment intact particles in the manufactured sand stacking scenario. The equivalent elliptical Feret short diameter is used as the equivalent particle size parameter and the area gradation is used as the gradation characterization parameter. The maximum repeatability error values of online detection of two groups of fineness modulus manufactured sand are 0. 03 and 0. 05. The maximum repeatability error values of particle size interval are 2. 97% and 3. 43% . Compared with traditional detection methods, this method is feasible and can meet the requirements of online detection in industrial sand production.

Abstract:The change of vehicle view may affect the accuracy of the re-identification algorithm. To solve the influence of changing viewpoints on the accuracy of re-identification, we propose a vehicle re-identification method based on local features and viewpoint perception. First, a parsing module is trained to parse a vehicle into four different views, front, back, side, and top. In this way, the fine-grained representation of the vehicle is improved. Then, we intrduce a vehicle viewpoint-aware network. The output of the network is the predicted probability information of the viewpoint, and the vehicle viewpoint perception effect is dynamically and smoothly represented according to the probability information. Finally, the viewpoint-aware effect is used to assign different weights to each local area of the vehicle to shorten the intra-class distance, expand the inter-class distance, and reduce the impact of viewpoint changes on vehicle re-identification. This method is evaluated on public datasets, including VeRi776 and VehicleID. The accuracy of mAP on VeRi776 dataset has achieved 80. 9% . Experimental results show that the proposed method can effectively improve the accuracy of vehicle re-identification. Ablation experiments demonstrate the effectiveness of the viewpoint-aware smooth representation for vehicle reidentification from multiple viewpoints.

Abstract:The measurement system consisted of vision and 3D laser ranging is the main sensing device for motion estimation and environment modeling. To realize the unification of the sensing data coordinate system of the measurement system, a method for calibrating the structural parameters of the vision and 3D laser ranging system based on the self-combination of space vectors is proposed in this article. It mainly includes three aspects: 1) The plane calibration method is used to solve the internal parameters of the camera and the normal vector of the plane target in the camera coordinate system and the plane iterative fitting algorithm is utilized to solve the normal vector of the plane target in the LiDAR coordinate system. The plane target normal vector set is established under two coordinate systems. 2) According to the angle between the vectors, the cross vector is independently selected in the plane target normal vector combination, and the structural parameter calibration parameters are established to solve the objective function, and the calibration objective function of structure parameters is established. 3) The nonlinear optimization algorithm is used to solve the least-squares problem and obtain the optimal estimation of external parameters. The effectiveness and accuracy of the method are evaluated by simulation and actual calibration experiments. Results show that the error between the image object-side projection and the 3D point cloud of this method is less than 30 mm (3σ), which not only satisfies high-precision 3D measurement but also has a high calibration efficiency. It meets the requirements of accurate measurement of sensor fusion measurement.

Abstract:To address the problem of multi-object tracking under the unknown motion of the intelligent vehicle, a visual multi-object tracking method is proposed, which is based on the coherent point drift. First, the unknown motion model of the intelligent vehicle is formulated by the coherent point drift algorithm. The local object state transformation relationship is achieved. Secondly, an adaptive feature fusion function is constructed, which is based on appearance similarity and motion consistency. Therefore, the Hungarian algorithm is utilized to solve the correspondence between the track and the detection. Finally, the robust data association for the intelligent vehicle is realized. Compared with the current five mainstream multi-object tracking methods, results show that the proposed algorithm has better results in multiple indicators. Compared with the SCEA algorithm, the multi-object tracking accuracy of the proposed method is increased by 6. 3% in the large motion scene of the KITTI dataset. Under the real-shot experimental data, the multi-object tracking accuracy of the proposed method is increased by 7. 3% , which can effectively perform multi-object tracking under the unknown motion of the intelligent vehicle.

Abstract:The accurate quality detection of hardwood logs can realize efficient utilization and profit maximization of wood. However, due to the difference in extraction principle of acoustic parameters and interaction mechanism between parameters and wood properties, the evaluation results in log quality are different to some extent. On this basis, a method for acoustic feature extraction and defect detection is proposed, which is based on the improved particle swarm optimization-variational modal decomposition ( IPSO-VMD). By analyzing the sparse characteristics of defect signals, the minimum average envelope entropy is determined as the fitness function of PSO optimized VMD to search for the optimal combination parameters (K, α). And the search of PSO is accelerated and the global optimal solution is achieved through improving the inertia weight and learning factor of PSO. Then, the effective sub-modes from the IPSO-VMD are selected based on marginal spectrum and energy ratio of sub-band components, and the frequency band distribution and energy ratio of each effective mode are used as the characteristic parameters characterized the defect signal to realize the accurate quality detection of hardwood logs. The actual sawing results show that the major defect types and priorities of logs are detected with accuracy of 88. 6% and 72. 7% , respectively, and which not identifiable by global parameters could be examined effectively based on the IPSO-VMD method. The effectiveness of the new feature parameters can provide a reliable basis for the accurate detection in log quality through fusing the multi-parameter features and constructing the artificial intelligence recognition system.

Abstract:To improve the defect detection ability of heavy thick wall components, a method that combines the compounding diverging wave imaging (CDWI) and circular statistics vector (CSV) is proposed in this article. Based on the virtual source technology, the method realizes a wide range of deflection wavefront emission of full aperture array elements by transmitting delay to increase the acoustic energy of wavefront in depth and width propagation direction. To effectively improve the detection ability outside the effective coverage area of aperture, the quality index of defect echo image is further increased by combining CSV in coherent combination process. For the Φ2 side drill holes with the lateral distribution which exceeds the coverage range of the probe aperture by 20 mm, and the longitudinal depth that is between 2~ 3 times size of the probe aperture in the aluminum block, the amplitudes of the SDH echoes of the proposed method are about 17. 5~ 18. 5 dB and 10. 6~ 14. 8 dB higher than them of the plane wave imaging, respectively

Abstract:Aiming at the non-contact detection needs of nonlinear effects of metal fatigue damage and its low signal-to-noise ratio, an electromagnetic ultrasonic scheme is proposed to collect signals for nonlinear effect. The Duffing chaotic system is used to quantitatively assess the fatigue degree of metal. Then, the nonlinear characteristics of material fatigue evolution are characterized according to the phase trajectory diagram and the Lyapunov index. In this article, the evolution process of fatigue damage of aluminum alloy is analyzed by the finite element method, and the relative nonlinear coefficient change law during the evolution of fatigue damage is studied based on the material Murnaghan model and the micro-crack equivalent spring model. In addition, the noise immunity of the Duffing chaotic system for the extraction of nonlinear effect features is investigated. When the signal-to-noise ratio is 20 dB, the relative nonlinearity coefficient error is 132. 12% , while the Lyapunov index error is 8. 82% . Therefore, the Lyapunov index has significant noise immunity compared with the nonlinear coefficient. In addition, based on the experimental study of fatigue detection of aluminum alloy, the feasibility and accuracy of the Lyapunov index characterization and analysis method of nonlinear effect of electromagnetic ultrasound are evaluated. Results show that the Lyapunov index can effectively address the problem of low signal-to-noise ratio in the process of electromagnetic ultrasonic nonlinear detection. In this way, the sensitivity and repeatability of nonlinear feature picking are improved, and the contribution of non-contact ultrasonic detection methods is further enhanced, such as electromagnetic ultrasonic in the evolution of fatigue online detection.

Abstract:In the phase sensitive time domain reflectometer (φ-OTDR) system, the disturbance signal is usually hidden in the noise, which makes it difficult to locate the disturbance. To improve the signal-to-noise ratio (SNR) of disturbed signals, this article proposes a moving variance averaging algorithm, which combines the characteristics of variance algorithm with outstanding data discreteness. Compared with traditional algorithms, such as cumulative average, moving average, separation average, wavelet denoising algorithm, moving average and moving differential, simulation and experimental results show that the moving variance averaging algorithm has better denoising effect and higher SNR. Therefore, the moving variance averaging algorithm can use fewer cumulative acquisition times to locate the disturbance signal, which can further improve the response frequency of the φ-OTDR system and improve the real-time performance of the system.

Abstract:A siamese network model is proposed for fault diagnosis of rolling bearings under small samples and strong noise. First, a series of time-frequency images are obtained from fault signals by the continuous wavelet transform, and the convolutional neural network is introduced to realize the pattern recognition. Secondly, the small samples are recombined with each other to form new sample pairs through cross matching. Thus, the number of fault samples are increased dramatically. Thirdly, a siamese network model including two sub-models is formulated, which uses the new sample pairs. Finally, a new classifier is designed for the siamese network model to realize fault classification with small samples under strong noise. The proposed faulty diagnose method is evaluated by using fault samples from both noise database and experimental measurement. The accuracy vaules are 96. 25% and 97. 08% , respectively. Results show that one fault can be identified by the proposed siamese network model using only 20 samples, which is less than the samples required by CNN model to reach a similar accuracy.

Abstract:The bearing fault signatures are difficult to be extracted effectively under strong background noises. To address this issue, this article proposes a rolling bearing fault diagnosis method based on the parameter adaptive feature mode decomposition. Firstly, to overcome the shortcoming that the original characteristic mode decomposition (FMD) needs to rely on human experience to set its key parameters without adaptability, the grid search method based on feature energy ratio of squared envelope spectrum ( FER-SES) is presented to automatically determine the mode number n and the filter length L of FMD. Then, the original bearing vibration signals are divided into n mode components by parameter optimized FMD. The mode component with the maximum FER-SES is selected as the sensitive mode component. Finally, the fault characteristic frequency is extracted by calculating the squared envelope spectrum of sensitive mode component to distinguish bearing fault types. The effectiveness of the proposed method is evaluated by simulation signal and engineering case analysis. Compared with variational mode decomposition and spectral kurtosis, the proposed method has better fault feature extraction performance.

Abstract:To address the vibration coupling modulation problem caused by the multi-stage transmission of the wind turbine gearbox, a new ring gear fault model considering the inter-stage amplitude modulation and frequency modulation is proposed. It is applied to the monitor ring gear by the parameter identification technology. The fault characteristic frequency of a certain stage of ring gear will modulate the meshing frequency of different gear trains in a way of amplitude modulation and frequency modulation, which shows the phenomenon of coupling modulation. In the view of this special modulation law, this article establishes the coupling modulation model of vibration signal under two-stage ring gear fault. On this basis, the parameter identification technology based on local mean decomposition and Levinberg-Marquart algorithm is proposed to determine the amplitude modulation coefficient of the fault model. Then, the condition indicator can be easily constructed to achieve the purpose of ring gear fault monitoring. The field data are tutilized to evaluate the content of this paper. Results show that the new model describes the vibration signal more comprehensively than the traditional model. The index constructed by parameter identification technology can locate the fault gear.

Abstract:The existing deep transfer learning-based diagnosis methods usually require that the same fault class space is shared by training and test data, which fail to effectively identify new faults. Thus, a multi-domain emerging fault identification method based on selective weighted adaptive network is proposed. Firstly, a one-dimensional convolutional neural network is adopted to extract depth discriminative features across domains. Then, a domain discriminator and multi-classifier structures are integrated to construct weight functions of source and target domains to adaptively measure the similarity across different categories. The adversarial learning strategy is utilized to effectively reduce the distribution differences of shared classes across domains. Finally, the Gaussian distribution-based fitting method is adopted to automatically discriminate weight thresholds to realize effective fault diagnosis of known faults and emerging faults in the target domain. Experiments are conducted on a gearbox transmission test rig, where the transfer diagnosis tasks under variable operation conditions are designed. The proposed method obtains 0. 8 E-score in various tasks. The effectiveness and the superiority of the proposed method are fully validated in comparison with other existing methods.