Abstract:Class A and B butt welds of pressure vessels are important stress-bearing parts, and the measurement of their quality parameters is an important part of welding quality evaluation. This article studies the detection method of weld quality parameters of pressure vessels based on deep learning active vision. A calculation method for weld parameters is proposed under the coexistence of multiple defects, which breaks through the problem that the weld quality parameters are difficult or impossible to calculate under the coexistence of weld defect parameters. We carry out the structural design of the encoding-decoding image feature point extraction network (EDE-net), which can better realize the one-time and accurate extraction of weld surface parameter feature points. We study the deep network structured channel pruning method to effectively improve the real-time performance of pressure vessel weld detection. Taking the welds of pressure vessels of different sizes as the experimental objects, the results show that the EDE-net network with the backbone of Resnet50 has CR= 0. 5 as the overall compression rate of the model, and the extraction time of a single image is reduced from the original 0. 31 s to 0. 19 s, a reduction of 38. 7% . The test report is given by the third-party testing agency, and the device simultaneously measures 5 parameters of the butt weld (Class A, B) weld, which takes less than 0. 63 s, and the allowable error of the measurement error is ≤0. 08 mm.

Abstract:To address the problems existed in the process of tacho-less order tracking (TLOT) based fault diagnosis, such as the error accumulation effect of harmonic components extraction and the difficulty of accurate instantaneous phase estimation, a speed-varying fault diagnosis method for the accessory transmission system based on adaptive harmonic components extraction is proposed. Firstly, the vibration signal is low-pass filtered and down-sampled for computation speed acceleration. Harmonic components are subsequently extracted by the adaptive symplectic geometry mode decomposition (ASGMD) based on the autocorrelation average periods. Secondly, surrogate data test is applied for pseudo components identification. Therefore, interferences induced by background noise can be filtered out adaptively. Finally, Hilbert transform is applied to obtain the instantaneous phase of the decomposed fundamental harmonic component. The TLOT is conducted to realize fault diagnosis for rotating machinery equipment under speed-varying conditions. The simulation analysis and an experiment from a Safran aero-engine during the frequency sweep test exhibit that the proposed method has a relative error of 0. 059% in TLOT. This result is better than those of conventional approaches. The bright prospect is evaluated in industrial application.

Abstract:Deep neural network compression methods with a single and fixed pattern are difficult to compress the model sufficiently due to the limitation of accuracy loss. As a result, the compressed model still needs to consume costly and limited storage resources when it is deployed, which is a significant barrier to its use in edge devices. To address this problem, this article proposes an adaptive joint compression method, which optimizes model structure and weight bit-width in parallel. Compared with the majority of existing combined compression methods, adequate fusion of sparsity and quantization methods is performed for joint compression training to reduce model parameter redundancy comprehensively. Meanwhile, the layer-wise adaptive sparse ratio and weight bit-width are designed to solve the sub-optimization problem of model accuracy and improve model accuracy loss due to the fixed compression ratio. Experimental results of VGG, ResNet, and MobileNet using the CIFAR-10 dataset show that the proposed method achieves 143. 0 ×, 151. 6 ×, and 19. 7 × parameter compression ratios. The corresponding accuracy loss values are 1. 3% , 2. 4% , and 0. 9% , respectively. In addition, compared with 12 typical compression methods, the proposed method reduces the consumption of hardware memory resources by 15. 3×~ 148. 5×. In addition, the proposed method achieves maximum compression ratio of 284. 2× whilemaintaining accuracy loss within limited range of 1. 2% on the self-built remote sensing optical image dataset.

Abstract:Thermal error modeling and compensation is an important tool to improve the machining accuracy of machine tools. It is important to apply the obtained thermal error models to similar tasks to reduce the cost of model construction and data collection. In this article, an easy transfer learning (EasyTL) with intra-domain alignment method for spindle thermal error modeling is proposed to realize the transfer reuse of error models under different working conditions. Further, the respective effects of different types of intra-domain alignment and distance matrices on model migration performance are analyzed. Finally, the EasyTL model is compared and validated with machine learning kNN and deep learning CNN to predict the thermal errors of the Z-direction and Y-direction of spindle under different working conditions, respectively. This method provides a new idea for modeling and compensating the thermal errors of machine spindles. In addition, a workpiece compensation machining experiment is carried out according to the thermal error of the spindle established by the thermal error prediction. The average error of the workpiece after compensation is reduced. This method provides a new idea for the thermal error modeling and compensation.

Abstract:To obtain characteristic parameters and fault influence law required for fault diagnosis of the train air supply system, and address the problems that some fault modes cannot be injected or the injection is destructive, the structure composition, working principle and fault modes are explained. The equivalent fault model design method and the equivalent fault injection process are proposed. Based on the equivalent fault injection test, the influence laws of five typical fault modes at different failure levels were studied and the causes of the faults are analyzed. The research results show that the equivalent fault injection test could reproduce the three typical fault phenomena of insufficient air supply, high oil temperature and compressor not loading without causing damage to the air supply system. The influence laws and characteristic parameters of different fault modes are different. When the air intake filter is severely blocked, the total air pressure is less than 740 kPa, resulting in insufficient air supply. When the temperature control valve is severely stuck, the maximum fuel injection temperature and the compressor head exhaust temperature are increased by about 20℃ and 17℃ compared with the normal values, and the maximum cooling oil flow is decreased by about 14 L/ min. The cooling effect is reduced due to the accumulation of dust in the cooler, and the maximum drop is about 3℃ when the ambient temperature values are 20℃ , 35℃ , and 50℃ , respectively. When the oil is seriously leaking, the fuel injection flow is about 19 L/ min higher than the normal value, and the compressor head exhaust temperature could reach up to 105℃ , resulting in high oil temperature. When the opening degree of the unloading solenoid valve is from 25% ~ 100% , the total air pressure is only 690 kPa, resulting in insufficient air supply. When the opening degree is 0, the air compressor could not restart after the first normal operation and shutdown, causing the compressor not loading. The research results provide a reference for the design of fault diagnosis algorithm based on multi - sensor data such as temperature and pressure and sensor layout of intelligent air supply systems.

Abstract:To provide better subjective auditory perception of speech enhancement for different listeners in different environments, a controllable speech enhancement model based on the perceptual conditional network is proposed. First, a quantile loss function is designed to balance the overestimation and underestimation of speech, which is used to guide the training of network. In this way, the output characteristics of model are controlled by adjusting the level of noise residual and speech distortion in the output of the network. Then, to make a single speech enhancement network has variable output characteristics, the conditional network is introduced. The conditional information is generated by the quantile value related to auditory perception in the quantile loss function to modulate the noisy speech features, and a controllable speech enhancement model is established. The experimental results show that, the designed quantile loss function can effectively adjust the level of residual noise and speech distortion in the enhanced speech, and the proposed controllable speech enhancement model based on the perceptual conditional network can provide variable characteristics of enhanced speech that can be actively controlled by the listener. The listener can get a better speech enhancement experience.

Abstract:The existing data-driven methods in the early detection of rolling bearing degradation have problems of low sensitivity and high false alarms. To address these issues, a dynamically adjustment grey incidence analysis ( DAGIA) method for transient mechanical equipment health monitoring is proposed. First, the Hilbert transform is applied to demodulate the vibration data of the rolling bearing to obtain the envelope signal. To weaken the influence of the value of the resolution coefficient to highlight the degree of discrimination of the correlation value, the feature-to-noise energy ratio (FNER) method is introduced into the traditional grey incidence analysis (TGIA) to dynamically adjust the resolution coefficient, which can characterize the strength of bearing faults. Then, the first set of data is extracted at the initial stage of bearing operation as reference data. The dynamic grey incidence analysis is calculated between the remaining data and the reference data and the bearing performance degradation index is established. Finally, according to the normal samples and combined with Chebyshev′s inequality, the control line is set to identify the starting position of the early degradation of the rolling bearing. The IMS and XJTU-SY databases are used to complete the early degradation recognition of rolling bearings. The results show that the proposed method can accurately recognize the starting position of early degradation and the false alarm is close to 0. It has both sensitivity and robustness, which is beneficial for equipment maintenance personnel to better grasp the operating status of rolling bearings.

Abstract:The ultrasonic guided-wave technique is widely used in the inspection and evaluation of pipeline defects. The defects in pipeline are mainly corrosion defects, and the actual form of pipeline corrosion is diverse and complex. A large number of researches for pipeline corrosion waveguide detection are conducted by means of simulation. The common simplified defect models cannot fully present the complexity of the actual corrosions and the mistakes may be made when analyzing the signals with them. The work in this article proposes a corrosion defect simulation model based on W-M fractal function, and further studies the finite element modeling of the pipeline with corrosion for guided waves inspection. The model is evaluated by simulation results of inspecting different corrosion defects. Results show that the obtained reflections can provide more defect information, which are conducive to reveal the quantitative relationship between the characteristics of pipeline corrosion and the guided-waves signal. Research results of this work are expected to provide theoretical support for further research on the detection and evaluation of corrosion defects in pipeline.

Abstract:To improve the end precision of fiber Bragg grating ( FBG) flexible structure by using the orthogonal curvature 3D reconstruction method, a mapping relationship is established in the reconstructed curvature end coordinates and actual spatial coordinates through neural network. Firstly, the model of polyurethane glue rod is established by COMSOL simulation software. Two fiber Bragg grating strings are orthogonal arranged with 8 gratings, and a dynamic coordinate system is established by the recursive angle algorithm for three-dimensional reconstruction. The reconstructed end point coordinates are trained by back propagation (BP) neural network and extreme learning machine (ELM) neural network. The results show that the average training errors of BP neural network and ELM neural network are 0. 443 6 and 0. 008 2, respectively. Finally, an experimental platform is established to reconstruct the shape of the polyurethane glue stick under stress, and it is substituted into the ELM model for training. The correlation coefficient R 2 of the training results is 0. 985 8, and the root mean square error is 1. 363 0, which effectively improve the precision of the end coordinates of the shape reconstruction compared with the BP neural network.

Abstract:When a wheeled robot moves inside a tapered thin-walled deep cavity cylinder, how to reduce the coaxial deviation between the robot and the cylinder parts is crucial for improving machining accuracy. However, the wheeled robot system using the distributed cylinders as the variable radius mechanism has nonlinear, time-varying, and complex frictional characteristics, which make precise control of coaxial motion deviation extremely difficult. Therefore, this article proposes an adaptive motion control method based on the variable domain fuzzy control theory to improve the precision of robot motion radial displacement and yaw and pitch angle control precision. Firstly, the tree-like kinematic model of the wheeled robot walking mechanism is formulated, and a pose calculation method is proposed. Then, the adaptive motion control method is proposed, and a joint simulation system based on Simulink and Adams is established to evaluate the effectiveness of the method. Finally, the cylinder internal motion control experiments are implemented by using a robot prototype. The results show that the proposed adaptive motion control method can reduce the internal motion deviation of the robot and ensure that the radial displacement deviation of the robot is ≤±1 mm and the yaw and pitch angle deviation are ≤±1°.

Abstract:To realize the precise puncture of prostate brachytherapy robot, this article conducts research on needle deflection deformation prediction model and puncture control strategy in view of the existing deficiency and development trend of robot-assisted prostate brachytherapy surgery. The parameter acquisition problem of needle deflection deformation prediction model is studied, and a Young′s modulus recognition method based on corrective force is proposed to solve the problem of difficulty in obtaining intraoperative tissue Young′s modulus. Then, the puncture control strategy based on corrective force is established, which is divided into two stages. The first is preoperative needle tip trajectory planning stage, and the other is the intraoperative puncture strategy adjustment stage. In the preoperative needle tip trajectory planning stage, a cost function is established, which is based on the needle deflection prediction model to obtain the optimal needle tip trajectory and puncture parameters. In the adjustment stage of the intraoperative puncture strategy, a reverse needle tip deflection prediction model is formulated to compensate the value of the correction force during the operation. The adaptive PID method based on reinforcement learning (RL) is also used to design the controller to achieve the application of correction force and achieve accurate puncture. An experimental platform of prostate seed implantation robot is established independently to evaluate the effectiveness of the proposed puncture control strategy. The average error of seed implantation is 1. 96 mm and the standard error is 0. 56 mm. Experimental results show that the puncture control strategy based on correction force can effectively reduce the tip deflection value and improve the accuracy of seeds implantation.

Abstract:Although blood cell analyzers have been widely used in hospitals, the manual microscopy is still the “ gold standard” for leukocyte detection. In this article, T-DETR, a DETR-based deep learning model with Transformer architecture is proposed for the detection of peripheral blood leukocytes, which aims to relieve the pressure of manual microscopy. First, PVTv2 is used as the backbone of DETR to extract multi-scale feature maps to improve detection accuracy. Then, the deformable attention module is introduced into the DETR model to reduce the computational complexity to accelerate the model convergence. Finally, to obtain the optimal weights, the training method of transfer learning is used on the filtered public leukocyte dataset. Experimental results show that T-DETR has an mAP of 0. 476 on the COCO dataset and 0. 954 on the leukocyte dataset, which is better than DETR and the classical CNN model. Results verify the feasibility of the Transformer structured model for medical image detection applications.

Abstract:When a user holds a rigid tool to slide on the material surface, the texture features of the material surface through vibration of the tool is felt. These vibration acceleration data contain rich texture category information, which provides a basis for texture classification. Texture classification based on tactile sense is of great significance for applications such as haptic human-computer interaction and fine manipulation of robots. At present, the methods of manually designing features related to texture and simple feature extraction using convolutional neural network have been applied to tactile texture classification. However, these methods fail to pay attention to the selection of time scale and the time dependence between tactile serial data, and there are still problems such as insufficient feature extraction of tactile data and poor classification accuracy. To solve the above problems, this article proposes a fusion model which combines multi-scale convolutional network and bidirectional long short memory network to capture multi-scale geometric local spatial features and time dependent features of tactile signals at the same time. The proposed model learns the tactile features of material surface texture from an open tactile data set, and trains them on the open texture vibration acceleration database. The experimental results show that the proposed model achieves the highest texture classification accuracy of 92. 1% robustly and efficiently.

Abstract:Channel equalization is used to suppress the inter symbol interference caused by atmospheric turbulence in the broadband optical wireless coherent communication system. In this article, intermediate frequency signals in turbulent environment are used as training samples. Back propagation (BP) neural network and long short-term memory (LSTM) neural network are utilized for training. The trained stable network model is used as a channel equalizer, and the output intermediate frequency signal by equalizer is used as the evaluation index of system performance and compared with the adaptive optics wavefront distortion correction algorithm. The simulation results show that by using BP neural network channel equalization technology, LSTM neural network channel equalization technology, and wavefront distortion correction technology, the peak values of intermediate frequency signal histogram are located at 0. 49, 0. 38, and 0. 38 V, and the corresponding system bit error rate is 3. 79×10 -5 、1. 64×10 -4 and 8. 48×10 -2 , respectively. Compared with wavefront distortion correction technology, intelligent channel equalization technology has significantly improved on the random fluctuation of intermediate frequency signal amplitude and bit error rate.

Abstract:Electroencephalography (EEG) has been considered as one of the best physiological signals to detect drivers′ mental fatigue. However, EEG signals vary significantly among different subjects and recording sessions, and it is still challenging to design a calibration-free system for EEG fatigue detection. In recent years, many deep learning-based methods have been developed to address this issue and achieve significant progress. However, the “ black-box” nature of deep learning models makes their decisions mistrust. Therefore, an interpretable deep learning model is proposed to recognize cross-subject fatigue states from single-channel EEG signals in this article. The model has a compact network structure. Firstly, a shallow CNN is designed to extract the EEG features. Then, the adaptive feature recalibration mechanism is introduced to enhance the features extraction ability. Finally, the time series of extracted features are linked to classification with LSTM. The interpretable information of the classified decision is achieved through a visualization technique that is taking advantage of hidden states output by the LSTM layer. Extensive cross-subject experiments are implemented on an open EEG dataset with a sustained-attention driving task, and the proposed model achieve the highest average accuracy of 76. 26% . In addition, compared with the advanced compact deep learning models, the parameters and computation are effectively reduced. Visualization results indicate that the proposed model has discovered neuro-physiologically reliable interpretation.

Abstract:The problem of laser short-range omnidirectional detection is considered. Based on the research of laser short-range dynamic circumferential scanning detection mechanism, a single-pulse laser short-range static circumferential detection method based on light cone beam expansion mechanism is proposed. Based on the theory of laser near-field detection and the spatial geometric relationship of static detection field, the echo equation of single pulse laser short-range static detection using the mechanism of light cone expansion is derived. The probability distribution model of single pulse laser short-range ranging is formulated and the laboratory static detection experiment is established. The influence mechanism of pulse laser emission power, inverted reflection cone angle, pulse laser beam divergence angle and target size projection area on the probability distribution of laser short-range circumferential detection is studied. The results show that as the transmit power and the target projection size increase from 10 W and 0. 01 m 2 to 30 W and 0. 25 m 2 , the echo signal amplitude also increases from 0. 16 and 0. 43 μV to 4. 22 and 5. 95 μV. As the inverted reflective cone angle and beam divergence angle increase from 30° and 10 mrad to 120° and 30 mrad, the echo signal amplitude decreases from 3. 18 and 2. 52 μV to 0. 88 and 1. 92 μV. The circumferential detection probability distribution decreases with the increase of the transmission power and the target projection size, and the peak increases and deviates to the left. With the increase of the inverted reflected light cone angle and the beam divergence angle, the half width increases and the peak decreases and deviates to the right. The symmetry of detection distribution is not affected by the above 4 factors.

Abstract:A calibration method of the included angle measuring system based on the autocollimation contrast system is proposed. This method is used to improve the accuracy of the included angle measuring system. Therefore, it can accurately measure the deformation of the spacecraft structure in the space environment. The calibration coefficient is achieved by calibrating the included angle measuring system, and the coefficient is evaluated by simulation data. Results shows that the measuring range of the included angle measuring system can be up to ± 25′ and the measuring resolution can be up to 0. 1″. Experimental results show that the calibration method is simple and convenient with high accuracy. It proves that the measurement accuracy of the angle measuring instrument reaches ± 0. 2″ by computing the absolute value of the difference between the measured results and certified value. In general, the accuracy of the angle measuring instrument after calibration meets the requirements of satellite utilization. This method can provide technical support for the study of micro deformation of satellite structures in orbit.

Abstract:There are the problems of limited working frequency range and serious output vibration waveform distortion of a tri-axial standard vibrator, which are affected by the low natural frequency and nonlinear characteristics of the leaf-spring decoupling structure. Firstly, the natural frequency and output waveform distortion characteristics of the tri-axial standard vibrator under corresponding changing parameters are obtained through ANSYS modal and transient dynamic analyses based on the determined key parameter simplified combinations of the decoupling structure. In further, a multi-element nonlinear regression method is used to establish the regression equations that can accurately represent the nonlinear changing relationships between the natural frequency, distortion degree and key structural parameters. A multi-parameter optimization of the leaf-spring decoupling structure is implemented based on the NSGAII genetic algorithm. The optimal leaf-spring structural parameters are obtained by increasing the first 3 natural frequencies and bending mode natural frequencies by 123. 95% and 166. 85% , respectively. The distortion is reduced by 36. 83% . Finally, the experimental test shows that the performance of the tri-axial standard vibrator corresponding to the optimal leaf-spring decoupling structure has been improved in both natural frequency and output waveform distortion, which evaluate the effectiveness of the multi-parameter optimization design of the leaf-spring decoupling structure based on the proposed genetic algorithm method. It provides a reference for the optimization design of other multi-dimensional flexible structures.

Abstract:During the past decades, the structured illumination microscopy (SIM) has been drawing great attention for both the technique development and applications. However, the conventional SIM that uses a spatial light modulator for fringe projection often has limited imaging field of view. Therefore, we propose a large-field SIM method that combines a 2D grating for fringe pattern projection and an SLM for selecting fringe orientation and performing phase shifting digitally. The proposed SIM technique breaks the bottleneck of fringe number limited by the digital projection devices, while maintaining the advantage of high-speed (digital) phase shifting of conventional SIM. An algorithm of spatial reconstruction is developed to improve the super-resolution imaging speed. Finally, a SIM fluorescence microscopy system is developed. A large field view of 1 380 μm × 1 035 μm with a 20 × / NA 0. 75 objective is experimentally demonstrated, and an enhancement of 1. 8 times higher resolution is realized. The spatial reconstruction algorithm can greatly reduce the computational time of super-resolution reconstruction.

Abstract:The fracture and shedding of turbine blades in aero-engine systems are intimately linked to blade temperature. The precise measurement of turbine blade temperature is significant for ensuring the safe operation of aero-engines. However, traditional infrared radiation temperature measurement methods often face challenges in maintaining accuracy due to factors such as the nonuniform emissivity of turbine blade materials, pronounced reflections from high-temperature objects in the surroundings, and variations in detection angles. To address these issues, this article leverages the principles of radiation transfer theory and infrared thermal imaging temperature measurement to analyze the key factors that influence temperature measurement results. A modified model for infrared radiation temperature measurement of three-dimensional curved surfaces under complex backgrounds is formulated. To obtain the necessary parameters for the temperature correction model, the experimental measurement is designed to determine crucial factors such as emissivity, bidirectional reflection distribution function, and angle coefficient for the curved surface. By applying the temperature correction model proposed in this study, the infrared-corrected temperature values for the curved surface are derived. Comparative analysis of these results with temperature measurements obtained from thermocouples positioned at representative locations demonstrates a reduction in temperature measurement error from approximately 4% prior to correction to less than 1% . This result substantiates the high accuracy and adaptability of the proposed correction model, underscoring its potential to provide valuable support for enhancing aerodynamic heat transfer test technology and facilitating the development of major aero-engine equipment, particularly turbine blades.

Abstract:The traditional centralized metering method for oil development is difficult to get the parameters linked to the oil-gas-water three-phase flow in a single oil well. To address this issue, the volume of fluid and finite element analysis are used in this study to optimize the structural parameters and impacts of the gas-liquid separation on the basis of the numerical simulation model of the measuring device. Therefore, the optimal structural parameters of the monitoring device are determined. Based on the afore-mentioned research, a permanent multi-well group single-well patrol three-phase flow multi-parameter measuring device is developed for long-term, stable and reliable utilization in the existing centralized metering environment. In addition, the experimental studies are conducted on a permanent multi-well group single-well patrol three-phase flow multi-parameter measuring platform built for oil production. The experimental results show that the developed device has less than 10% error in water holdup and gas holdup measurement and less than 4% error in flow rate measurement under mixed fluids such as gas and liquid phase flow rate range (5~ 70 m 3 / d) and liquid phase water holdup range ( 50% ~ 90% ). Both simulations and experiments demonstrate that measuring device performs well in water holdup measurement.

Abstract:The magnetic particle imaging (MPI) is a new dynamic targeting imaging method in the human body. However, the current MPI system mainly utilizes a closed magnetic field scanning structure, which severely limits its clinical application. In this article, an open electrical scanning narrowband MPI system is designed. On the basis of analyzing the harmonic magnetization response of superparamagnetic particles, an open spatial positioning magnetic field based on field free line is formed by using eight gradient coils through the calculation of the coupling magnetic field on the coil surface and the current control. The imaging area is 30 mm×30 mm. A single side excitation coil is used to generate a 20. 7 kHz excitation magnetic field. The superparamagnetic nanoparticles tracer in the high-frequency excitation magnetic field and the positioning magnetic field generates a locatable superparamagnetic magnetization signal with rich harmonic components. A high signal to noise ratio Gradiometer coil is used to detect its third harmonic signal to form a voltage cloud image. Furthermore, the non-negative least square method is used to reconstruct the voltage nephogram through the pre-measured system function matrix to form the tracer concentration distribution nephogram. The imaging experiment results show that the detection sensitivity of the system in the open imaging area is 20 μg Fe, the spatial resolution of image reconstruction is 2 mm, and the imaging speed is 1 fps. The good open imaging effectiveness is achieved. This system is also the first open magnetic particle imaging system independently developed in China.

Abstract:To meet the industrial needs of miniaturization and maintain favorable precision and resolution, a circular time-grating angular displacement sensor is proposed by using multi-layer structure and re-modulation scheme. The sensor is a three-layer structure. Its inner and outer rings complete the re-modulation of the signal through the multi-layer structure by using the axial space. The inner ring is used as the rough measurement part to realize the absolute positioning and the outer ring through re-modulation is used as the fine measurement part to improve the resolution of the sensor. A prototype sensor with an outer diameter of Φ= 100 mm and an inner diameter of Φ = 50 mm is manufactured by PCB technology. Preliminary experiments show that when the inner and outer rings are excited at the same time, the inner and outer rings are affected by the crosstalk generated by each other. Therefore, a time division method is further proposed to avoid the influence of signal crosstalk to improve the measurement accuracy. The final experimental results show that the sensor can achieve absolute positioning and the resolution of the fine measurement part is doubled, and the original measurement accuracy reaches ±4. 1″.

Abstract:There are rich sensory tactile receptors in human skin. Through touch, the human body can recognize materials with different thermal properties. Flexible thermosensation sensors can simulate the function and structure of human skin. They can be applied to intelligent robots and smart sensing. This article presents a flexible thermosensation sensor based on the constant temperature difference (CTD) method, which is consisted of metal electrodes and flexible polymer films. Combined with the application for machine finger touch and recognition, a “heating-contact” heat transfer model is formulated. The geometric characteristics of the sensor, object, and robot fingers are considered. The numerical Laplace inverse transformation algorithm is used to solve the model. The model is evaluated by finite element simulation and thermal perception experiment. A conditioning circuit for the CTD operating mode is designed. Closedloop temperature difference control is implemented by using the PID algorithm, and disturbance is suppressed by introducing the Kalman filter. The relative error of temperature control is within 3. 0% , and the system can recognize 11 materials with different thermal properties. This flexible system has potential application in fields such as robot perception, wearable electronics, and virtual reality.

Abstract:In recent years, with the continuous improvement of industrial robot technology, the application of combining camera and laser rangefinder for measurement is increasing. To facilitate the coordination between the binocular camera and the single point laser rangefinder, calibration between them is essential. This article proposes a calibration method of binocular camera and single point laser rangefinder. The light spot of the single point laser rangefinder is hit on the calibration board, and the light spot is formed on the calibration board. The pixel point at the center of the light spot is extracted by the gray center of the gravity method, and the threedimensional coordinate value of the light spot is extracted by using the internal parameter matrix and depth data of the binocular camera. Then, two three-dimensional coordinate points are used to establish formulas to obtain the rough solution of the external parameter parameters. Meanwhile, the constraint equations are established through the three-dimensional coordinates of multiple light spots in the camera coordinate system, and optimization algorithms are used to optimize the constraint equations to obtain accurate external parameter values. This method is simple and has no special requirements for calibrating devices. The external parameters obtained through this method have high accuracy and good robustness. The actual measurement shows that the average relative error of the external parameter parameters obtained through this algorithm after re-projection is less than 0. 30% , which can be applied in actual production process.

Abstract:Aiming at the unclear mechanism of the effect of high and low temperature environment on the sensitivity of high-g accelerometer, a mathematical model of the sensitive unit with a solid support at both ends is proposed. The relationship among the sensor sensitivity, longitudinal stress, and transverse stress of the sensitive unit is determined. The relationship between sensor sensitivity and temperature is obtained by thermal-force coupling simulation analysis. In the range of -40℃ ~ 50℃ , the sensor sensitivity decreases with the increase of temperature, and the sensitivity changes fast in the range of - 40℃ ~ 20℃ . It is slow in the range of 20℃ ~ 50℃ . The sensor sensitivity data under high and low temperature environment are collected by the sensor high and low temperature test. The magnitude of sensitivity at - 40℃ is 0. 523 μV/ g, the magnitude of sensitivity at 50℃ is 0. 516 μV/ g. The experiment results are compared with the simulation results to evaluate the effectiveness of temperature change on the sensor sensitivity law. For improving the measurement accuracy of high-g accelerometer under high and low temperature environment, it has theoretical reference significance.

Abstract:To further trace the error caused by the magnetic field coupling process of the time-gate displacement sensor, the coupling characteristics of the time-gate displacement sensor in the construction field are studied. A new linear time-gate displacement sensor based on the exponential planar coil structure is developed. The mathematical model of the magnetic field constructed by sensor engineering is formulated. The influence of the sensor coupling gap on the magnetic field distribution of the coil coupling plane is analyzed, and the coupling characteristics of the coil of different shapes are studied. According to the coupling characteristics of the sensor, a new linear time-gate displacement sensor measurement model is formulated. The electromagnetic field finite element simulation and simulation error analysis are carried out on the model, and the optimal sensing gap of the structure is 0. 4 mm. The structural error of the sensor is traced and analyzed, and the structure of the sensor was further optimized. An experimental platform is established by using the double-layer PCB winding process to process sensor fixed and moving ruler. Comparative experiments are implemented on sensor prototypes before and after optimization. The experimental results show that the new linear time-gate displacement sensor based on exponential plane coil structure can effectively suppress the four errors of the sensor, and the original measurement accuracy of the newly developed sensor prototype is improved by 45. 8% on the original basis.

Abstract:Traditional eddy current based multi-cycle displacement sensors are difficult to solve the problem of absolute position identification after power failure and restart due to the periodic repetitiveness of the output signal. In this article, a new type of eddy current based bipolar linear displacement sensor is proposed. After theoretical and simulation analysis, it is verified that the amplitude of the induced voltage in the receiving coil shows a sine cosine variation with the sliding of the slide. A bipolar sensitive structure is designed to ensure accurate displacement measurement by means of a multi-cycle receiving coil at the upper pole. A single-cycle receiver coil is arranged at the lower pole to identify the cycle of the upper stage. The accuracy is improved by sensing signal offset and amplitude normalisation processing algorithms, and the sensor prototype is established in the laboratory and tested with a high precision electronically controlled translation table. The new multi-cycle bipolar eddy current linear displacement sensor has been tested to achieve absolute position measurement with a measurement error of 30 μm in the 0 ~ 60 mm range and a maximum non-linearity of 0. 08% . This breaks through the limitations of traditional multi-cycle eddy current displacement transducers where the absolute position cannot be identified.

Abstract:The localization accuracy of traditional radio frequency identification (RFID) indoor localization methods is low. To address this issue, a RFID indoor localization method based on dual-frequency carrier phase is proposed. Frequency hopping technology is employed to obtain rough range estimation with virtual large bandwidth to achieve multipath suppression. The selection of the optimal dual-frequency points is completed based on multipath suppression. The particle filter localization algorithm is designed by using the optimal dual-frequency point carrier phase. The resampling method in traditional particle filter is optimized by genetic algorithm, which effectively solves the problem of particle degradation and improves the localization accuracy. Experimental results show that the median localization error of the proposed algorithm is 5. 23 cm, which achieves 39% improvement than the traditional localization method based on Chinese remainder theorem.

Abstract:In view of the change of ultrasonic incidence angle caused by the online ultrasonic dynamic detection and the low measurement accuracy, a nonlinear probability model of two-parameter ultrasonic flooding detection error correction is formulated. The error introduced by ultrasonic incident angle is compensated by the function approximation theory. Using the combination of basis function and third-order Lagrange interpolation, the ultrasonic incidence angle and the nonlinear correlation between the incident angle and the detection error are obtained. The incident angle of the nonlinear probability model and the ultrasonic distance are iteratively calculated. The incidence angle is used to solve the refractive angle in reverse, and the uncertainty of the ultrasonic incidence angle in the detection is resolved. In the range of 30 to 45 mm, the results show that after the model compensation treatment, the ultrasonic wave at 0° ~ 8° angles of the incident, the plate detection accuracy is 1% , which provides a basis for effectively improving the corrosion accuracy estimation.

Abstract:The boundary artifacts and low-spatial resolution in reconstruction due to the ‘soft-field’ and the ill-posed nature of the inverse problems imaging with electrical tomography ( ET) are considered. This article designs a novel deep learning-based ET image reconstruction framework consisted of an unrolling iteration pre-reconstructor and a modified attention-based deep convolutional neural network ( CNN) postprocessor. Specifically, the pre-reconstructor, a four-layers deconvolution network, is unrolled by the NewtonRaphson algorithm. The U-Net is the backbone of the post-processor and two carefully designed feature connections are introduced. Firstly, the residual connection is added to the feature extraction and image reconstruction block which could alleviate the reverse gradient vanishing problems. Secondly, the residual self-attention skip connections are proposed which could better fuse the global and local information. These above-mentioned strategies can better express the nonlinear characteristics of ET inverse problems. The visual results show that the reconstruction using the proposed methods has higher spatial resolution and more clear shape representation (i. e. , sharper boundary features and clear medium distributions). The quantity results (RE= 0. 10 and CC= 0. 93 in test performance) indicate that the proposed method could improve the imaging results effectively. A reliable method for nondestructive measurement and visualization is promoted.

Abstract:The grating lobe appears when the microphone array element spacing is larger than the half-wavelength of the acoustic signal for the DOA estimation of the wideband sound source. Although the utilization of multi-frequency bins data can suppress the grating lobes to some degree, the current methods perform unsatisfactorily and are computationally inefficient. To address these issues, an improved method based on the sparse Bayesian learning is proposed for wideband DOA estimation. This method introduces the hyperprior to the multi-frequency sparse Bayesian estimation model, and then takes advantage of the fact that the source signal has the same sparsity in multi-frequency bins and combines the expectation maximization algorithm to derive the iterative form of each hyperparameter in the multi-frequency sparse Bayesian model. In addition, an off-grid model for the wideband sound source is incorporated into the proposed framework to better fit the practical scenarios. To evaluate the performance of the algorithm, simulations and field experiments are implemented. Results show that the proposed method can better exploit the multi-frequency characteristics of the wideband sound source to reduce the interference of the grating lobes, while having higher estimation accuracy and faster estimation speed compared with the multi-frequency compressive sensing method with l 1 minimization and the multi-frequency sparse Bayesian learning method. In the practical tests, the improved method shows better grating lobe suppression ability than other advanced methods, and its estimation error can reach 0. 09°. Compared with MF-SBL, the number of iterative convergence steps required is reduced by about 50% .

Abstract:At present, many factors restrict the application of the simultaneous localization and mapping ( SLAM) in real environment, and the interference of dynamic objects in indoor environment is one of the urgent problems to be solved. This article proposes a visual SLAM system based on ORB-SLAM3 and assisted by the instance segmentation network. The system places the segmentation task at the back end, and combines RGB-D camera input and KCF algorithm at the front end to detect semantic information from the back end. After tracking and transmitting, the system uses this information to track the motion state of key points in the Bayesian probability framework. Compared with current methods based on detection or segmentation, this system uses a lighter scheme to segment and track moving objects in the scene. With the further assistance of Bayesian filtering model, it not only realizes accurate dynamic interference filtering, but also optimizes the real-time problem of system operation caused by CNN network preprocessing. Experiments on the TUM RGB-D dataset show that the system can achieve high positioning accuracy at a speed of about 16 fps, with an average lead of 78. 56% compared with ORB-SLAM3 and 11. 85% compared with DynaSLAM.