Abstract:Inductive angular displacement sensors have good environmental adaptability, high measurement accuracy and stability. They are widely used in high-end equipment, electric vehicles, robots, aircraft, weapons and other fields. This article reviews three types of inductive angular displacement sensors, including resolver, inductosyn and the magnetic field type time-grating displacement sensor which is independently developed by Chinese scholars. Their measurement principles and key technologies are reviewed in detail, and their advantages and limitations are also analyzed. According to the development of inductive angular displacement sensors in recent years, their applications in the fields of machinery, automobile, industrial robot, aviation, aerospace and national defense are discussed in detail. Finally, it is concluded that the inductive angular displacement sensor technology should be developed in the direction of high precision, high reliability, embedded measurement, composite function measurement and intelligence.

Abstract:Aerodynamic testing plays an important role in evaluating the design method and optimizing the significant system characteristic parameters of flapping-wing vehicles. This article presents a newly developed three-dimensional force sensor with simple structure, low cost and high performance, which is used in the aerodynamic testing system of flapping-wing vehicles. It focuses on the design and calibration method of the three-dimensional force sensor. Firstly, the feasibility of the structure design is verified by the model of the micro-variation kinematics of the elastic mechanism. Then, the decoupling calibration experiment of the sensor shows that the nonlinear error of the sensor is smaller than 1. 19% , the coupling error is smaller than 3. 21% , and the sensitivity is larger than 1. 1 V/ N. Finally, by testing the aerodynamic force of a small flapping-wing vehicle, the quantitative evaluation data of the main wing aerodynamic force of the flapping-wing vehicle with the influence of the parameters are obtained, which can provide an effective help for the vehicle to achieve indoor and outdoor maneuverable flight.

Abstract:The multi-frequency detection is an important way of interference fading suppression in the optical fiber sensing system of Φ-OTDR. But, there is the problem of complex structure caused by multi-frequency generation. A multi-frequency pulse generation method based on the optical fiber frequency-shifted delay loop is proposed in this article. Optical fiber loop structure is designed with frequency shift, amplification, delay, filtering, isolation and other functions. The generation and multiplexing of multi-frequency detection pulses are realized by using this simple structure. The data alignment and data aggregation of Rayleigh backward scattering beat signal are realized by using the pulse delay compensation algorithm and rotated-vector-sum algorithm, respectively. The experimental results show that the sensing system can optimize the fading probability from 26. 43% to 0. 93% with the 10. 33 km sensing fiber by the generation of multi-frequency detection pulses containing six frequency components, and achieve the accurate location of the vibration signal with a signal-to-noise ratio of 5. 29 dB and the linear demodulation of the vibration signal with a correlation coefficient of 0. 997. It provides a new solution of anti-fading demodulation of vibration signal for Φ-OTDR in complex environment.

Abstract:Multi-target tracking is the basis to ensure the safety and efficiency of autonomous driving, and the obtained data are widely used in the upper applications of autonomous driving, such as motion planning and driving decision making. The traditional multi-target tracking method often has the phenomenon of target loss and dislocation in the occlusion environment. To solve the problem, a robust tracking method based on heterogeneous radar fusion and occlusion prediction model is proposed. Firstly, based on the local observation consistency equation of lidar and millimeter-wave radar, a multi-sensor dynamic self-calibration algorithm based on multi-target motion constraint and global maximum matching is proposed. Secondly, a hybrid supervised target position prediction method based on heterogeneous radar fusion unscented Kalman filter and long and short time series neural network is proposed to solve the problem of tracking interruption caused by missing observation data in complete occlusion environment. Experiments show that the proposed method can effectively complete at least 81% of the broken multi-vehicle target tracks in the complete occlusion environment, which can achieve more reliable multi-vehicle target tracking compared with the most advanced methods.

Abstract:A surface temperature measurement method based on FBG-FP sensor is proposed to meet the demand of structural surface temperature measurement. This method can compensate the cross-sensitivity of strain to temperature by acquiring the temperature and strain information of the measured structure through FBG and FP sensing simultaneously. This article analyzes the surface temperature measurement principle of FBG-FP and designs the main parameters of the sensor through simulation. The two-parameter minimum mean square estimation algorithm for signal demodulation of FBG-FP sensor is proposed. Then, a comparison experiment between FBG sensor and FBG-FP sensor is conducted. Experimental results show that the linear fitting correlation coefficient of temperature measurement of FBG-FP temperature sensor is 0. 998 4, and the maximum error percentage is 1. 46% in the range of room temperature to 400℃ , both of which are better than the single fiber grating temperature sensor.

Abstract:High energy efficient independent method for optimal sensor placement

Abstract:To meet the test requirements of large out-of-band suppression filter transmission and reflection characteristics, this article proposes an in-suit enhancement method for the dynamic range of a VNA. The method has add a low-noise amplifier to the test chain without changing the hardware architecture of the instrument, extend dynamic range by amplifying tiny signals, and the cross-correlation signal processing algorithm is adopted. The noise floor of the instrument is further reduced by performing multiple operations on the twochannel test data. To improve the test accuracy and solve the problem that the error term of the receiving port cannot be directly obtained due to the introduction of the low noise amplifier, this article designs a two-step calibration method based on a 12-term error model. The extraction and de-embedding of the connection adapter parameters is finished. Finally, the characteristics of the large out-of-band rejection filters in two different frequency bands are measured by using the proposed method. Experimental results show that the dynamic range of the IF bandwidth of 200 Hz can reach 130 dB, which can be improved by more than 20 dB compared to the dynamic range of conventional instruments. The effectiveness of the effectiveness of the proposed method is verified.

Abstract:As one of the key components of CNC machine tools and precision measuring instruments, the rotary table can move parts to appropriate position to achieve efficient processing and measuring task. Its own accuracy is particularly important. Based on the principle of “three-rosette method”, an optimization measurement method for multi-DOF geometric errors of rotary table based on a ball plate artifact is proposed in this article. The measurement principle, separation and settlement method of geometric error of rotary table are described in detail. Experimental results show that the measurement results using 12 standard balls are consistent as those of 6 standard balls. The absolute difference of translation error of the table is not larger than 0. 13 μm, and the absolute difference of rotation error is not greater than 0. 25″. Finally, an autocollimator is used to measure the indexing error of the rotary table, and the results are compared with those based on ball plate artifact. Results show that the absolute difference of the measurement results obtained by the above different method is not greater than 3. 9″, which evaluates the accuracy of the multi-DOF geometric errors measurement method based on ball plate artifact.

Abstract:There is the problem of machining quality caused by the thermal displacement of the motorized spindle at variable speed. To address this issue, an experimental method of natural speed reduction for the motorized spindle under different preload forces is proposed by building an experimental platform of the variable pressure preload motorized spindle. The frictional heat generation model of bearings based on the energy conservation theory is formulated, and the function relationship between preload forces and bearing heat generation are constructed. On this basis, the influence law of bearings temperature leading to the thermal displacement of the spindle is further investigated. The temperature data of bearings and the time of the motorized spindle with preload forces of 1 450, 1 550 and 1 700 N are used as input to construct a BP neural network thermal displacement prediction model of the motorized spindle. Results show that the fomulated thermal displacement prediction model can effectively predict the thermal displacement of the motorized spindle, and the residuals of the prediction model are within 0. 5 μm. The research results provide a new method for the intelligent compensation of thermal error in high-precision machine tool spindles.

Abstract:The nano-step height reference materials can deliver accurate and traceable nanometer height characteristic parameters. To investigate the fabrication technique of high-quality sub-50 nm step height reference materials in China, a method based on atomic layer deposition combined with wet etching is proposed. The sub-50 nm step height reference materials with a minimum height of only 5 nm are fabricated through process optimization, which achieves the precise control of the step height in the order of the sub-nanometer. The calibration results are traceable to the national meter-defined wavelength reference, and the extended uncertainty is less than 2. 0 nm. Meanwhile, the reference materials has excellent uniformity and stability, and the consistency level of different measurement instruments is quite high. The results show that the nano-step height reference materials can be used for transmission of sub-50 nm height values and comparative measurement between multiple instruments. The prospect of their industrialized production will also provide a perfect metrology guarantee for the semiconductor industry.

Abstract:The pose accuracy is an important index to evaluate the performance of the robot. The establishing of an effective compensation algorithm guarantees the improvement of the robot pose accuracy. This article takes a 2TPR&2TPS parallel robot as the research object, and formulates an error model based on the positive solution. According to the error model, the influence relationship between the position parameter error of the dynamic and static platforms and the length error of the drive rod zero point on the robot end pose error is obtained. A compensation algorithm based on the inverse solution is established. The minimum value of the error function is optimized by the particle swarm algorithm, and the compensation amount of the robot′s drive rod and the pose compensation amount are achieved. The simulation shows that the average position accuracy of the robot is improved by 98. 148% , and the compensation amount of the drive rod corresponds to the ideal pose and the superposition of the driving rod length of the robot are adopted as the driving rod input of the robot for experimental verification. The experiment shows that the average position accuracy of the robot is improved by 87. 457% , and the compensation effect is remarkable.

Abstract:Based on the McClements model for the analysis of acoustic propagation in two-phase flow of suspended masses, the effects of solid phase concentration, particle size and ultrasonic frequency on the acoustic attenuation coefficient in suspended mass solutions are investigated by numerical simulation. Among them, the 0. 5th power of ultrasonic frequency and the solid phase concentration have the positive correlation with the attenuation coefficient, respectively. The solid phase particle size and the attenuation coefficient have a negative correlation with the first order. The relative error between the inverse model and the McClements model is less than 2% . An experimental system is established to measure the attenuation coefficient and concentration of the suspension solution, and experiments are implemented to measure the solid-phase concentration under the conditions of particle size r = 75 μm and ultrasonic frequency of 700 kHz and 1. 1 MHz, respectively. Experimental results show that the maximum relative error between the concentration values deduced from the inverse model and the actual set concentration is 11. 2% ( 700 kHz) and 9. 6% ( 1. 1 MHz). The inverse model constructed in this article can provide some theoretical reference for the study of suspension concentration measurement equipment.

Abstract:Almost precision laboratories and semiconductor production plants have vibration isolated design. But, there may be still some vibration out of limit around. It turns into micro-vibration interference source after passing through vibration isolated installation. To find the interference sources around the site, it is necessary to classify and identify the detected blind sources signals. Due to the characteristics of low frequency, low amplitude, and short duration of the transient micro-vibration signal, traditional vibration signal analysis methods have difficulty to handle this. Thus, an identification method of micro-vibration source based on K-medoids is proposed. The endpoint detection algorithm is used to cut off transient micro-vibration signals from long-term data after preprocessing. Then, the normalized Mel filter coefficient (EN-Fbank) feature is extracted and used to constitute feature matrix. In addition, the data are clustered by K-medoids with dynamic time wrapping (DTW) distance. Finally, Gaussian mixture models are created for clustered data to identify the inspection data of the suspected vibration source with model probability threshold to find serious interference sources. In the experiment with 24 h data, two types of vibration sources with the largest average amplitude and the highest frequency of occurrence are found, and the classification accuracy reaches 90. 57% besides the identification rate reaches 96. 8% , which proves the effectiveness and accuracy of the method.

Abstract:To solve the hysteresis caused by using torque sensors to control muscle force training equipment, a joint torque prediction model based on a group of antagonistic surface electromyography (sEMG) is designed in this article. Firstly, the rehabilitation training equipment is built to provide conditions for signal acquisition and experimental verification. sEMG is preprocessed and the variance characteristic of sEMG signal is selected as the neural network input. In addition, a dynamic recurrent neural network with the nonlinear auto-regressive model with exogenous inputs (NARX) is used in this study. A multi-step ahead prediction model (MSA) based on the actual values of joint moments and another model based on model prediction output ( MPO) are developed respectively. The torque prediction performance of MSA and MPO models is compared by isotonic and isometric test experiments. Experimental results show that there is a strong correlation between the predicted output value and the actual output value of the two models ( Pearson correlation coefficient is greater than 0. 95). As the number of advance prediction steps increases, the prediction accuracy of MSA model decreases. However, the advance prediction time increases. When n is less than 29 and 35, the prediction accuracy of MSA is significantly higher than that of MPO (p <0. 05). But the MPO model has advantages in cost and size. In summary, two models proposed in this article can accurately predict joint torques. In actual rehabilitation training equipment control, different torque prediction models can be selected according to application requirements.

Abstract:Aiming at the interference of the visual pose estimation method, such as cover and occlusion, a method of estimating human three-dimensional sitting posture based on the seat surface pressure image is proposed. The cross-domain relationship between seat surface pressure distribution and human three-dimensional posture is established. A posture training system based on pressure and vision is designed. The array pressure sensor is embedded in the seat surface to perceive the posture, and the time stamp is used to realize the synchronization of the visual image matching with the binocular camera. Bilateral filtering is used to eliminate the peak noise of pressure images. Nineteen 3D keypoints are extracted from binocular vision images by OpenPose estimation and triangulation. To improve the accuracy of attitude estimation, a stochastic gradient descent method to minimize the loss function is proposed to optimize the coordinates of 3D keypoints. The 3D confidence graph of keypoints is further generated by 3D Gaussian filter. A multi-layer convolutional neural network pressure-vision cross-domain deep learning model is formulated. Continuous multi-frame pressure images are used as input of the model, and 3D pose estimation results of 3D key point coordinates and their confidence graphs are used as supervision. Based on the pressure distribution of the array sensor on the chair surface, the algorithm can accurately estimate the 3D sitting posture including 19 human key points. The average error of 19 key points is 9. 7 cm on the verification set.

Abstract:The extracorporeal counterpulsation ( ECP ) device requires compression and release of the lower extremity with a high recognition rate and real-time ECG signature wave detection method. Based on traditional medical signal processing methods, the article proposes an algorithm for ECG feature signal detection by the improved differential thresholding and distribution calculation. The algorithm utilizes the low-pass filtering and moving average filtering methods to smooth the ECG signal. Then, the locations of R-wave, P-wave and T-wave are identified and determined by using the adaptive differential double thresholding method and distribution calculation method. Simulation analysis and experimental verification results base on the MIT-BIH database and ECG sampling module show that the algorithm has an accuracy of 99. 9% for the comprehensive recognition of R-wave of ECG signal and 99. 87% for the recognition of P-wave and T-wave. In addition, the average time consumed by the algorithm is only 0. 65 s. The algorithm can identify the characteristic waves of common ECG signals, which can satisfactorily meet the requirements of devices like ECP to quickly identification of ECG characteristic waves.

Abstract:The research on the Parkinson′s disease (PD) speech recognition algorithm is important for timely diagnosis and treatment. However, the existing public PD speech datasets are characterized by small sample sizes, which is one of the main challenges faced by existing PD speech recognition methods. To address this issue, a novel dual-side two-stage means clustering envelope and convolution sparse transfer learning model is proposed. First, for the dataset side, multiple groups ofconvolution kernels are trained, which is based on the source domain dataset. Then, the optimal convolution kernels are filtered by the encoded intermediate dataset. Finally, the target domain dataset is encoded by the optimalkernels. In regard to deep instances clustering envelope,an iterative mean clustering algorithm is designed to construct the deep instance space. Secondly, various classifiers are developed after sample / feature parallel selection. Finally, the classification results of different instance layers are fused. In the experiment,the representative PD speech datasets are selected for verification. Experimental results show that the main innovative parts of the proposed algorithm are effective. Compared with more than ten classical algorithms,the obvious improvements in terms of classification accuracy are achieved 97. 8% . In addition, the proposed algorithmhas potential in clinicalapplications for acceptable time complexity.

Abstract:Emotion plays an indispensable role in human behavior and cognition. It is of great practical significance to carry out research on emotion recognition. To improve the accuracy of cross-subject identification of four kinds of emotional states, the feature representation method based on the full-view and ELM-Adaboost method is proposed. Firstly, a data processing strategy based on the fused information is proposed. The sample data consisted of multiple types of physiological signals are cross-fused to help extract sample features from the perspective of full view. Secondly, the feature selection method with the maximum correlation and the minimum redundancy is used to select the fused features to achieve effective emotional state representation. Finally, the strong classifier constructed by the ELM-Adaboost method is employed to realize emotion recognition. Experimental results from multi-perspective evaluate the effectiveness of the proposed method. The identification accuracy of cross-subject with four types of emotional states reaches 83. 06% .

Abstract:The joint torque prediction plays an important role in rehabilitation medicine, clinical medicine, sports training and other fields. The continuous and real-time torque prediction can make the human-computer interaction equipment better feedback and reproduce the intention of human motion. To provide a safe, active and comfortable rehabilitation training environment for patients and enhance the compliance of the human-computer interaction equipment, a novel method of joint torque prediction is proposed, which is based on an improved recursive cerebellar model neural network. In this method, muscle synergy analysis is used to reduce the dimensionality of surface electromyographic (sEMG) signals. Then, the reduced-dimension sEMG feature vector, joint angular velocity and joint angle are used as the input data of the prediction model. In addition, recursive unit and fuzzy logic rules are introduced into the cerebellar model neural network, while the wavelet function is used as membership function. Hence, the generalization ability of the network is optimized. The RWFCMNN model realizes the time series prediction of the dynamic torque of ankle dorsiflexion and plantarflexion in three states, non-fatigue, transitional fatigue and fatigue. The average Pearson correlation coefficient and the average normalized root mean square error between the predicted torque and the actual torque are 0. 933 5 and 0. 159 8, respectively. These numerical values verify the accuracy and effectiveness of this method for continuous prediction of lower limb joint torque.

Abstract:The active guided wave structural health monitoring enables real-time online monitoring of structural condition. However, the time-varying influence can make crack evaluation more difficult and reduce the accuracy. The time-varying factors make the guided-wave monitoring signal characteristics show obvious heteroscedasticity. The variance of the distribution of the characteristics changes with time. To address this issue, this article proposes a quantile regression-assisted online crack evaluation method, which uses quantile regression to estimate the variance of the guided wave monitoring signal characteristics with service time under the time-varying influence, and realizes the treatment of the heteroscedasticity uncertainty in the monitoring data. The proposed method is evaluated by using experimental data of the notched beam structure. The maximum absolute error of the evaluation is 1. 1 mm, and the root mean square error is 0. 4 mm. The proposed method can effectively deal with the effect of time-varying heteroscedasticity, quantify its uncertainty and provide a reference value for the evaluation results.

Abstract:The intermittent fault of the electrical connector is a common fault type in equipment, and its signal performance has a strong correlation with the degradation state of electrical connector and the environmental stress level. It is difficult to evaluate the degradation state of electrical connector. To address this issue, a new method is proposed in this article. The dynamic characteristics of intermittent fault signals of electrical connectors under sinusoidal vibration conditions are analyzed, and the behaviors of relative displacement between the contact interfaces obtained from dynamic model analysis reflect that the intermittent fault’ s bimodal amplitudes and their time delay are effective parameters to evaluate the degradation state of electrical connectors. The characteristic parameter dataset is constructed. Furthermore, a state assessment model based on deep belief network and multitask learning is formulated. The loss function of the model is improved as the iteratively weighted summation of the partial losses. Based on the parameter dataset of intermittent fault, the degradation states of electrical connectors are evaluated and analyzed, and the accuracy reaches 95. 94% . The proposed method provides a new research route for the degradation state assessment of electrical connectors.

Abstract:Relay is one kind of key components for systems such as spacecraft. Accurate reliability evaluation is essential for ensuring the safety and stability of the entire system. In the existing reliability studies of relays, the recoverable shocks resulting from factors such as materials and environment are not considered, which may cause inaccurate reliability evaluation results. To solve this problem, this article proposes degradation and shock models based on the modeling theories of Wiener process and discrete time Markov chain by incorporating natural degradation process, recoverable shock effect, and degradation correlation between performance parameters. Then, a multi-stage method to estimate model parameters is proposed to handle the problem that it is difficult to estimate all the model parameters simultaneously. Furthermore, for the reliability evaluation, an approximate calculation method of reliability is proposed by using the Monte Carlo technique. Results show that the proposed reliability evaluation method can ensure a favorable accuracy, and the corresponding goodness of fitting is 0. 103 7, which is about 70% higher than the existing methods.

Abstract:To address the problems of lack of labeled data and low cross-domain diagnosis accuracy in the fault diagnosis method of rotating machinery based on deep learning under new working conditions, a domain adaptive fault diagnosis method based on Transformer is proposed. A variant of Transformer, VOLO, is used to construct the feature extractor to obtain fine-grained and better fault feature representation. The supervised learning with source domain data pretrains feature extractors on source and target domain data, and freezes source domain extractor parameters to obtain fixed source domain features. Using domain adversarial adaptive strategy and local maximum mean difference combined with target domain unlabeled data to train target domain feature extractor, the edge distribution and conditional distribution of source domain features and target domain features are aligned. The proposed fault diagnosis algorithm is evaluated by two multi-condition experiments. Results show that the proposed domain adaptive fault diagnosis method based on Transformer feature extraction is more efficient than the five traditional domain adaptive methods on gear and bearing datasets. The average diagnostic accuracy is improved by 22. 15% and 11. 67% , respectively, which proves that the proposed method can improve the cross-domain diagnostic accuracy.

Abstract:The soft-sensing model for solidification structure of continuous casting slab is complicated in algorithm, large amount of calculation and time-consuming in solution. The method based on the central processing unit (CPU) is difficult to meet the prediction needs of large-size casting. To improve the calculation efficiency, a cellular automaton ( CA) soft-sensing model based on graphic processing unit (GPU) heterogeneous parallelism is proposed. Firstly, the heterogeneous parallel algorithm of GPU-CA is designed to eliminate the data dependence and data competition among cells, which optimizes the parallelism degree among data. Secondly, a multi-stream task scheduling scheme is proposed to solve the problem of independent tasks waiting each other in single-stream, and improving the degree of task parallelism. Finally, two kinds of the steel produced by a large-scale continuous caster in a certain steel plant are used to test the model. The predicted results are in good agreement with the field experiment data, where equiaxed grain rate errors are about 1% and 1. 5% , respectively. The maximum relative error between temperature and measured temperature is 1. 37% . In the case of the same calculation accuracy as CPU, the speedup of GPU is hundreds of times, which greatly improves the computing speed of the model.

Abstract:Fatigue cracks on the rail tread seriously affect the safety of trains. To address the problem that detects rail tread cracks quickly and effectively, a method for rapid detection of rail tread cracks is proposed in this article. Firstly, the mathematical models with Gaussian white noise and sinusoidal signal plus Gaussian white noise interference are formulated respectively. The noise suppression effectiveness of three signal processing methods are analyzed, including the coded pulse compression, the synchrosqueezed wavelet transform and the pulse compression after synchrosqueezed wavelet transform method. Secondly, to evaluate the noise suppression ability of the above methods, the surface wave electromagnetic ultrasonic transducer with excitation frequency of 1 MHz is used to detect the rail tread with cracks. Finally, the ultrasonic echo of the detected crack is taken as the research object. The noise reduction ability and ultrasonic imaging effect of the Hilbert Huang method for processing the ultrasonic echo signal corresponding to a single frequency pulse and the method of synchrosqueezed wavelet transform followed by pulse compression are compared. Results show that the proposed method can obtain the location and number of cracks on the rail tread. When Hilbert Huang transform is used to process the original ultrasonic echo with no synchronous average, the signal-to-noise ratio of the echo is low, which causes the failure of the empirical mode decomposition ( EMD). Under the condition that Barker code is used as the excitation signal and there is no synchronous average acquisition, the signal-to-noise ratio of ultrasonic echo obtained by using the pulse compression after synchrosqueezed wavelet transform method is 6. 82 dB higher than that using only phase coding pulse compression. It is 11. 02 dB higher than that using only synchrosqueezed wavelet transform, which can significantly improve the detection speed and B-scan image definition.

Abstract:To address the problems of the expensive computing cost for the wave field data processing and the difficulty for damage feature extraction in laser ultrasonic detection, a guided wave field analysis method based on deep learning is proposed. First, under the framework of VGG-Net, a residual network based on VGG11 is developed for extracting guided wave features from time-space wave field data. Then, taking the local wavenumber characteristic as the physical mechanism of the model, the problem of obtaining big data for training deep learning model can be solved by using the analytic formula of guided wave propagation. Therefore, the neural network can be obtained for extracting guided wave feature. Finally, using the experimental data in the plate structure with damage through laser ultrasonic system as test samples, the capability of guided wave feature extraction and damage identification using the proposed method is validated. The damage identification accuracy is above 67% and the shape of structural damage can be visualized.

Abstract:The detection of weld defects based on the X-ray flaw detection is a key part of maintaining pipeline safety. The realization of high-precision and high-efficiency intelligent defect detection is an important aspect to promote the intelligence and modernization of nondestructive testing. At present, it is difficult to achieve high accuracy and efficiency with deep learning-based defect detection methods because they require a large number of labeled samples and are difficult to obtain. To address this problem, this article proposes an active small sample learning-based defect detection method for pipe welds. First, the defect detector is trained in a data-driven manner by extracting small sample features based on a lightweight neural network. Then, the inference of the unlabeled samples is used to calculate the detection and classification uncertainty, which could fully exploit the value samples. Finally, the network parameters are fine-tuned according to the high-value samples to obtain a high performance improvement with minimal cost. Experimental results show that the proposed method can improve the accuracy by about 8% with fewer samples and the guaranteed operational efficiency.

Abstract:Bone external fixation is an important treatment for limb orthopedics. There are some problems such as motion coupling and rotation dislocation in the existing fixator during limb rotation and stretching orthopedics. Drawing on the configuration advantages of RCM mechanism, a 2R1T centering orthopedic external fixation robot based on traditional Ilizarov external fixator is proposed in this paper. The proposed robot has centering motion characteristics and is easy to realize single freedom independent motion. Based on the Spinor theory, the mathematical analysis of the traditional fixation orthopedics is firstly carried out, then establish the Spinor system and anti-spinor system of the 2R1T centering fixation robot, and complete the branch chain selection and assembly. Finally, this paper analyzes the motion degree and workspace of the robot, and carries out the orthopedic dynamic simulation analysis and prototype experiment based on tibia motion trajectory. The experiment result shows that the rotation correction range of the robot is -15~ 15°, the error range is ±0. 5 °, the range of drafting correction was 0~ 30 mm, and the error range is ±0. 2 mm, which realizes the quantification and accurate regulation of orthosis and ensures the safety of limb orthopedics.

Abstract:The amphibious hexapod robot is characterized by its flexible legs, which aims to address the terrestrial and underwater motion control in complex environment. In this article, a deep reinforcement learning based terrestrial motion control is firstly proposed for movements on rugged lands. By building agent interaction scenarios using the MuJoCo physical engine, the proximal policy optimization algorithm is employed to obtain the optimal motion policy applied in rugged lands of different conditions. Simulation results show that the robot is capable of climbing fast and stable on rugged terrains when controlled by the generated policies. Concerning the underwater motion control issue, the hydrodynamic model is derived. Based on its analysis, the three-dimensional underwater motions can be decoupled into the planar trajectory tracking and the depth control. Especially, the LOS and PID control are used. Experimental results prove that the robot can track the Sigmoid curve with path error less than 0. 11 m. In addition, the robot is qualified to realize the PID based depth control with the accuracy of 0. 02 m.

Abstract:The current clinical orthodontic treatment process mainly relies on the doctors′ experience. It is difficult to quantify the magnitude of the applied orthodontic force. The Omega-loop bent at the end of the orthodontic archwire can realize the effect of continuous and multiple reloading on the vertical-closure-loop bent at the same orthodontic archwire. Thus, this can reduce the clinical treatment time, and improve the efficiency of orthodontic treatment. The aim of this study is to quantify the orthodontic force generated by combining vertical closure and Omega loop orthodontic archwire. The force characteristics of the one-wire multi-loop combination orthodontic archwire is analyzed, and the orthodontic force is established by using the principle of beam micro-deformation and the principle of interaction force. The influence of bending parameters of the one-wire multi-loop combination orthodontic archwire on the orthodontic force is investigated. By formulating a 3D model of the one-wire multi-loop combination orthodontic archwire and an experimental platform for orthodontic force measurement, the finite element simulation analysis and experimental measurements are performed. The correlation analysis is conducted between the calculated data obtained from the mechanical model and the simulation data obtained from the simulation analysis and the experimental data obtained from the orthodontic force measurement experiment based on the one-dimensional force sensor. The correlation coefficient is ξT≥ 98. 192% between the calculated and the experimental data, and the correlation coefficient is ξA ≥ 97. 34% between the simulated and the experimental data. The accuracy and the reliability of the mechanical model, the simulation model and the simulation process are evaluated. The mechanical model and simulation model can assist physicians to design the personalized orthodontic archwire safely and efficiently. It provides theoretical basis for effective orthodontic treatment and further lays the foundation for clinical personalized and digital orthodontic treatment.