Abstract:Silicon micro-resonant accelerometers have advantages of small size, low cost, wide dynamic range, and high-precision quasidigital frequency signal output. But, the key performance indicators, such as zero bias and scale factor, are constrained by temperature and other factors, which cannot meet the high-performance requirements of high-precision navigation guidance. Therefore, on the basis of briefly introducing the temperature characteristics and temperature error sources of silicon micro-resonant accelerometers, this article reviews researches on temperature compensation of silicon micro-resonant accelerometers in recent years, including passive temperature compensation technology and active temperature compensation technology. The common methods and the latest research of passive temperature compensation technology and active temperature compensation technology are introduced. The advantages and disadvantages of various methods are analyzed and summarized. It is proposed to explore more accurate temperature measurement methods and isolation devices with better thermal isolation effect to design a micro-oven system with low power consumption, short preheating time and strong control stability for heating and temperature control at the device layer, and to seek more possibilities of combining passive compensation technology and active compensation technology to obtain the optimal combination of temperature compensation is the key research direction of subsequent temperature compensation work.

Abstract:To describe the influence of the sensor microchannel structure on the output characteristics of the sensor, the relationship parameter of the sensor transverse and longitudinal microchannel length is introduced in this article. A mathematical model of galliumbased liquid metal flexible strain sensor is formulated by comprehensively considering the resistance change law of the liquid metal in the microchannels which are parallel and perpendicular to the strain direction in the sensor. According to the developed mathematical model, five strain sensor samples are optimized and prepared to perform the experiments within 40% strain range. The results show that the mathematical model developed in this article can well predict the output of sensors with different microchannel structures, and provide directions for the design and optimization of high sensitivity sensors. Experimental tests show that the designed strain sensor has a sensitivity of 2. 01, a hysteresis of 5. 1% at 40% strain, and a high reproducibility and output stability. The designed sensor is used for joint motion detection in the finger and elbow, exploring the application of gallium-based liquid metal flexible strain sensor in joint motion detection. This article provides theoretical support for the application of gallium-based liquid metal flexible strain sensors in the wearable field.

Abstract:Impact source location is an important research topic in the field of aircraft structural health monitoring. The impact location method based on stress wave direction perception can effectively overcome the application limitations of time difference of arrival location method which depends on wave velocity, which is suitable for impact location monitoring of aircraft′s complex structures. In this article, the response characteristics of piezoelectric fibers to impact stress waves are analyzed. Aiming at the characteristics of impact stress wave signal for plate structures, such as broad frequency, dispersion and multi-modal components, the Shannon complex wavelet transform is used to extract the narrowband signal of the peak value of the stress wave energy. Combined with simulation analysis and experiments, it shows that the main energy component of the stress wave generated in the low-speed impact plate is A0 mode Lamb wave. Based on the directional response characteristics of piezoelectric fiber to Lamb wave, the piezoelectric fiber sensor with 45° “strain rosette” structure is optimized and prepared. The propagation direction of the stress wave is represented by the propagation direction of narrowband Lamb wave with peak energy, an impact source location method based on the sensing of the stress wave direction by piezoelectric fiber sensor is proposed. Experimental results show that the proposed impact source location method based on the sensing of the stress wave direction by piezoelectric fiber sensor is accurate and effective, and the average distance location error is 17. 2 mm. The research results provide a stress wave direction sensor and a structural impact location method that is independently from the stress wave velocity, which has certain theoretical research significance and engineering application value.

Abstract:The contact force feedback at the end of the surgical probe is one of the important factors to ensure the safety of surgery. In this article, a 3D end-force measurement method based on the kernel extreme learning machine (KELM) neural network is studied to meet the need of 3D end-force measurement of a puncture surgical probe based on optical fiber sensing. First, a surgical probe structure with implantable optic fiber sensors is designed, and four fiber Bragg grating (FBG) sensors are implanted into the probe, three of which are used for force measurement and the rest is used for temperature compensation. Then, by analyzing the relationship between probe stress and strain, a 3D end-force sensing model of the probe based on FBG is formulated. To eliminate the cross-effect of temperature change on the optical fiber sensor, a temperature compensation method for optical fiber sensing is studied. Finally, to evaluate the effectiveness of the neural networks, the temperature of the probe implanted in FBG is calibrated. The force measurement effect is verified in the normal temperature and temperature changing condition. The results indicate that the KELM neural network has better results in measuring the 3D end-force of the probe, and average measurement errors of KELM network in X, Y, and Z directions at room temperature are 0. 22% , 0. 99% , and 0. 65% , respectively. Under temperature changing condition from 20℃ ~ 40℃ , the average measurement errors in the X, Y, and Z directions are 1. 32% , 1. 03% , and 2% , respectively. The 3D end-force measurement method of the KELM neural network studied in this article has small measurement error, which has broad application prospects in the field of force feedback of surgical robots.

Abstract:To reduce the control error in a quadrature closed-loop circuit of micro electro mechanical system (MEMS) disk resonant gyroscope, a full-closed loop control method of MEMS disk resonant gyroscope based on the stiffness axis declination angle prediction mechanism is proposed. The parameters of the quadrature closed-loop circuit are automatically optimized and adjusted by predicting the stiffness axis declination angle of MEMS disk resonant gyroscope. In addition, a digital full-closed loop control method based on the stiffness axis declination angle prediction mechanism is proposed to realize the digital full-closed loop control of the driving, detection, quadrature suppression, and mode matching. This method can improve the signal-to-noise ratio of quadrature suppression circuit, and enhance the gyroscope capability in the field of quadrature drift, zero output and bias instability suppression. Experimental results show that, after adopting the full-closed loop control method based on the stiffness axis declination angle prediction mechanism, the zero output of the MEMS disk resonant gyroscope is decreased from 0. 201°/ s to 0. 021 3°/ s and the bias instability is decreased from 39. 42°/ h to 1. 237°/ h, by 9. 44 times and 31. 86 times, respectively, which evaluate the effectiveness of this method for improving the performance of the MEMS disk resonant gyroscope.

Abstract:Variational mode decomposition (VMD), a very active branch in the field of adaptive signal decomposition, has become a hot research direction in the field of signal processing. VMD shows good performance in processing non-stationary and nonlinear signals. Aiming at VMD model and its parameter selection, many extended models and parameter optimization methods have been studied. This article reviews the research progress of VMD in recent decade, summarizes and analyzes the relevant literature. Firstly, it analyzes the principle advantages of VMD and its application potential in various fields. Secondly, according to the matching ability of the model to different signal types, the different characteristics and applicable scenarios of extended models of VMD are summarized by classification. Then, the research progress of parameter optimization methods for VMD and its extended models are summarized to discuss and analyze the characteristics of different model parameter optimization methods and the latest research trends. Finally, 6 prospects are put forward for the future development of VMD, pointing out the direction for subsequent research.

Abstract:Aiming to the problem of difficult construction of damage index and uncertainty of damage estimation in quantitative identification of damage localization of composite materials,this paper proposed a Bayesian update algorithm of estimating damage states based on recurrence quantization feature (RQA) observation. Without referring to the baseline signal in undamaged state, the method used the RQA features of sensed Lamb wave signal to observe the internal damage from the perspective of structural nonlinear dynamics, and performed sensitive feature screening based on the correlation and monotonicity between the features and the delamination location morphology. Combining the linear or nonlinear correlation between the RQA features of each sensing path and the delamination location and shape parameters, a quadratic interaction model between the features and the parameters was established. Considering the uncertainty of damage observation by RQA features, the observation model of delamination damage was built, based on which the Bayesian update algorithm was used to estimate the delamination damage location. The simulation results showed that the proposed method could simultaneously achieve quantitative assessment and accurate localization of internal damage in composites structure without analyzing the complex interaction mechanism of structural damage and Lamb waves, and the estimated damage location size was generally consistent with the damage parameter settings of the simulation model. And the estimated damage location size is within the 75% confidence interval of the damage parameter setting value of the simulation model, and the damage estimation imaging area is consistent with the simulated damage area. The experimental results verified that the damage area and size obtained after 24 iterations of update by this method completely cover the real damage detected by X-ray scanning. The proposed method fully takes into account the uncertainties in the damage assessment of composite panels and is more suitable for the engineering practical application of damage detection and assessment of composite structures, and has good application prospects.

Abstract:The calculation load of the traditional narrowband active noise control (ANC) system is positively correlated with the number of narrowband frequencies. Its convergence speed and noise reduction effect have the trade-off relationship. To solve the aforementioned problems, this article firstly simplifies the narrowband ANC model based on the filtered-error technology, which makes the system calculation independent of the number of narrowband frequencies. Then, an adaptive mixture parameter combines two simplified narrowband ANC systems to make up a parallel convex combination structure. One system has the fastest convergence speed, and the other system has lower steady-state error. It not only reduces the computational complexity of the traditional narrowband ANC system, but also improves its convergence speed without sacrificing noise reduction result. Experimental results show that the proposed system shortens the convergence time by 78% without sacrificing the steady-state error, which effectively improves the system convergence performance.

Abstract:To reduce the influence of noise on the high frequency ultrasonic detection of flip chip defects, a sparse denoising method of high-frequency ultrasound signals based on the multipath matching pursuit (MMP) is proposed. The MMP algorithm is used to obtain the atoms which are globally optimal. Aiming at the excessive calculation of MMP, the setting thresholds and the introducing pruning operations during iteration are introduced. To avoid the excessive calculation amount caused by the increase of the dictionary dimension, the contiguous atom dictionary is established to adjust the reconstructed signals and realize the noise suppression. Proved by simulation and experiment, the proposed method can effectively remove the noise in high-frequency ultrasonic detection signals of flip chip. Compared with MMP, the proposed method can improve the signal reconstruction accuracy and the clarity of B-scan by adding a small amount of computation.

Abstract:To solve the ill-conditioned and singular problems of complex airfoil structures in traditional deformation sensing methods, a non-singular deformation reconstruction model based on multiple airfoil features is proposed. Based on the Timoshenko beam deformation theory, the element displacement field is discretized by using the dependency interpolation technique. A least squares variational function of theoretical surface strain and measured strain is established. Then, the reconstruction model between element node deformation and measured strain is derived. The position independence of the reconstructed model effectively eliminates the singularity caused by improper selection of evaluation sections and enhances the applicability in complex structures. Meanwhile, an adaptive multiobjective particle swarm optimization model is formulated with reconstruction accuracy and robustness for overcoming environmental disturbances. Results show that the maximum absolute error is 0. 26 mm and the maximum relative root mean square error is 0. 42% when the deformation of the wing model is less than 20 mm. With the increase of deformation, the absolute error also increases. But, the relative root mean square error does not exceed 3. 5% . Therefore, the non-singular deformation reconstruction model based on multiairfoil features can meet the requirements of real-time wing reconstruction and effectively extend the application value of deformation sensing method in complex structures.

Abstract:Rolling bearing is one of the most important components of rotating machinery systems to ensure safe operation. It is important to carry out studies on rolling bearing feature recognition for theoretical and practical application. The commonly used deep learning rolling bearing feature recognition methods require supervised labeled data or unsupervised fault data to participate in the training, and labels of data and fault data are not easily accessible to meet the rolling bearing feature recognition requirements. This article proposes an edge computing method for differential evolution of generative adversarial networks rolling bearing feature recognition, namely the EC-DE method. The training process uses only healthy data to train the generative adversarial networks and learn the distribution pattern of healthy data. The edge node compares the distribution difference between the input samples and the generative samples of generative adversarial networks for identification and exits early according to the health confidence level to improve the system′ s real-time performance. The cloud node uses a differential evolution algorithm to search the generator latent space of the generative adversarial networks to obtain the latent variables corresponding to the input samples, which improves the recognition accuracy. The proposed method achieves 99. 8% accuracy on CWRU rolling bearing public data set and is insensitive to hyper-parameters, and the inference stage takes less time, which is valuable for a practical production application.

Abstract:To address the problem of early fault state identification of the nuclear rod control system and rod position system power equipment (PWE), this article proposes an identification method based on the fusion of fault features in time domain and time-frequency domain and extreme learning machine (ELM) of grey wolf optimizer (GWO). Firstly, according to the working principle of PWE and the driving current of control rod drive mechanism, the early waveform shape and early fault mode are analyzed by using the current rise time. Then, the fault time-frequency features are constructed, which are fused with current rise time, root mean square difference summation and wavelet packet singular entropy. The discriminability of the features is analyzed. Then, the GWO algorithm can optimize parameters of the ELM classifier. The GWO-ELM model is formulated to realize the identification of early fault states of PWE, which can improve the identification accuracy. Finally, through the comparison test of different feature combinations and identification models, the results show that the proposed method can effectively realize the early fault identification and diagnosis of rod control system power supply, and the average identification accuracy can reach 98. 86% .

Abstract:In the application scenario of the actual health management for complex equipment health managements, represented by electromechanical systems, health perception and maintenance decision-making depend on the mined evolution mechanism of state of health. Both of them show an obvious coupling on their base knowledge whiling operating. The corresponding binary knowledge has the value of bi-directional fusion. Inspired by the bi-directional fusion of fault detection-maintenance binary knowledge, this article proposes a bi-directional optimization method of health perception and maintenance decision-making for electromechanical systems to regularly take advantage of the limited operation records accumulated in one period to optimize the previous health perception and maintenance decisionmaking model. Finally, the proposed bi-directional optimization method is evaluated by using the simulation experiment of the antenna leveling system in the actual electromechanical system, where the health prediction error is reduced to 0. 002% . The maintenance decision-making benefit is increased to 93. 57, which verifies the effectiveness of the proposed collaborative method of health state prediction and maintenance decision-making.

Abstract:Fatigue cracks and bolt looseness are the main damage forms of steel towers, such as relay towers and transmission towers. Under time-domain loads, these damages have time-domain nonlinear characteristics such as variable stiffness. To solve the time-domain nonlinear damage detection problem, a damage detection method based on the relative entropy of the autoregressive time-domain model is proposed. First, the autoregressive model and the basic theory of model order determination and parameter estimation are described. Then, the time-domain nonlinear characteristic of structural damage is introduced, and the three autoregressive residuals formed in the undamaged basic state and the damage state are given. In addition, the relative entropy of the probability distribution of the residual series is analyzed. On this basis, the damage detection index based on the relative entropy of the autoregressive time-domain model is derived. Finally, the numerical simulation of the eight-layer shear structure and the damage detection experimental study of the relay tower model are conducted. The results show that the relative entropy index value of the autoregressive time-domain model at the damage location is more than 22. 9% higher than the traditional second-order variance index value for the rod nonlinear damage of the relay tower. For the bolt loosening nonlinear damage, the relative entropy index value of the autoregressive time-domain model at the damage location is more than 12. 7% higher than the traditional second-order variance index value.

Abstract:The leakage of gravity flow drainage pipeline will bring many adverse consequences. But, due to the particularity of gravity flow drainage pipeline, many existing pipeline leakage detection methods are difficult to be applied. To solve this problem, this article proposes a detection method based on the geo-electrical current, which uses leakage holes to connect the inside and outside of the pipeline, and determines the leakage by measuring the current between the electrode moving in the pipeline and the electrode fixed on the ground. The basic working principle of the method is described in this article, and the detection system is designed based on the method above, and the experimental platform is established to test the detection effectiveness of the proposed method. When detecting a 10 m long buried pipeline, the maximum location error is about 0. 4 m, which indicate that this detection system can accurately locate the leakage location, and with better probe positioning method, the error will be futher reduced. According to the measured spike current value, 9 circular leakage holes with the diameter gradient of 0. 5 mm can be clearly distinguished. Nine groups of measured data are used to train the neural network, the mapping relationship between the spike current value and the leakage hole size is roughly obtained, which shows that the leakage hole size can be roughly estimated according to the detection results.

Abstract:To solve the problem of the lack of public human behavior datasets, two data augmentation methods are designed, based on generative adversarial networks and 3D human infrared models, to rapidly expand existing infrared human behavior datasets in this article. An improved generative network model AC-WGAN is formulated by adding network optimization strategy to generate high-quality infrared heat maps. The Unity 3D engine is used to build a mannequin with infrared features and motion information. By simulating the imaging principle of infrared array sensors, the function is realized to automatically generate a large and diverse amount of data for a given mannequin and sensor orientation information. A convolutional neural network is established, which is based on the data-enhanced dataset. Experimental results show that the perceptual accuracy of different behaviors is up to 80% . The ability of the network to identify unfamiliar data is significantly improved and the effectiveness of the designed data augmentation method is proven for expanding the human behavior infrared dataset.

Abstract:Self-paced treadmill is the key human-robot interactive equipment for virtual reality. The current study focuses on the jumping interaction control technology for self-paced treadmill to enrich the application scenarios. For the purpose to analyze the stability of human jump landing, a novel variable stiffness spring-mass loaded inverted pendulum model is proposed, which takes into account the combined effects of lower limb bones and joint muscles. Experimental results show that the proposed model can realize the modeling of the mass center motion trajectory and the analysis of the jumping stable domain, the accuracy of stability recognition is 93. 0% . Based on the proposed model and the stability analysis, the jumping interaction control strategies for the self-paced treadmill are proposed to improve human stability during jumping landing. The simulation and experimental results show that the proposed method can improve the stability of human jump landing significantly. Meanwhile, the proposed method reduces the torque of lower limb joints effectively. The peak torque of the knee joint reduces from 230 N/ m to 210. 7 N/ m, and the peak torque of the ankle joint reduces from 143. 6 N/ m to 131 N/ m, which is expected to lower the risk of injury.

Abstract:An aging population will inevitably make a higher prevalence of Alzheimer′s disease (AD), which results in a heavy burden on family and society. The early detection is the key to delaying or reversing the course of the disease. However, the current detection methods cannot meet the needs of inexpensive, low-invasive, rapid and reliable diagnosis of AD. The detection technology based on the biofluid spectra shows great potential in medical diagnosis. However, the challenge in the detection of dementia is difficult in extracting the characteristic information related to dementia in the plasma spectra and the complicated classification problem of multiple disease courses. For this reason, the research work will be carried out from the perspective of information space construction, feature information mining, and detection system design. An adaptive screening and diagnostic model for multiple disease courses of AD driven by the model feature wavenumber is constructed. For three different stages of AD, including early, middle, and late stages, the sensitivity of detection is 90. 0% , 87. 5% , and 100% , respectively. And the specificities are 83. 3% , 93. 7% , and 100% , respectively. Experimental results show that the proposed detection model provides superior classification power for AD.

Abstract:Aiming at the design of the adjustment scheme for the hexapod external fixator and the optimization of the position and posture changing trajectory of the fixator during the deformity correction of the foot and ankle, a method of adjustment scheme for the external fixator based on the neural network is proposed. Firstly, the kinematics of the hexapod external fixator is analyzed with the neural network. Secondly, different deformity correction modes including ankle mode, miter mode, butt mode and parallel mode, are discussed. Finally, a computer-aided adjustment scheme of the hexapod external fixator is proposed and the adjustment sequence of telescopic screws is discussed in the calculation. Experimental results show that the adjustment scheme which designed based on the proposed method could effectively guide the correction of different types of foot and ankle deformities. They could achieve ideal remained deformity and the average residual displacement deformity is less than 1 mm and the average residual angular deformity is less than 1°. Meanwhile, with this adjustment scheme, the position and posture changing trajectory of the hexapod external fixator could be closer to the ideal linear trajectory, and the offset of optimized position changing trajectory and optimized posture trajectory changing are reduced by 31% and 19% respectively, which effectively alleviate the pain of patients and improved the experience of deformity correction.

Abstract:Person re-identification is a fundamental problem in the smart video surveillance system. However, the traditional RGB-based feature extraction method cannot be used in dark environment. A new method for person shape feature extraction using depth measurement is proposed in this article. The depth data are independent from lighting condition. Therefore, the proposed method can be used for person re-id in the dark. Specifically, the point cloud of person is generated from depth data after segmentation and filtering. Then, the point cloud is registered to the initial human body model. The shape and pose parameters of the body model are estimated jointly based on the registered point cloud. Finally, the re-id is achieved by calculating the Euclidean distance in the vector space of shape parameters. The author applies this method on public and self-collected datasets in the laboratory to calculate performance indicators, including Rank-n, cumulative matching curve, and mean average precision, etc. Among the indicators, the Rank-1 of BIWI datasets in Single shot evaluation mode reaches 70. 71% and the Rank-5 of BIWI datasets is up to 92. 32% , which indicate that the proposed algorithm can effectively improve the re-recognition accuracy.

Abstract:To address the problems of small targets detection with few available features and weak texture information in the context of complex sea conditions in multiple scenarios, and to improve the environmental perception capability of unmanned surface vehicles (USV), we propose a typical small targets detection method using attention mechanism and multi-scale features. Firstly, the global prior information of the target is fused in the deep layers of the network using atrous spatial pyramid pooling module. Secondly, the shallow spatial location and deep semantic information features of the target are adaptively enhanced by the attention fusion module to improve the feature representation capability of the network. Finally, the high performance target detection is achieved through multi-scale feature fusion. We construct a typical surface small target dataset, and the method is evaluated by experiments of surface small target detection under real sea conditions based on USV. Experimental results show that the proposed method in the NVIDIA platform reaches 17 FPS, which can accurately identify small target on the water surface. Compared with the original FPN algorithm, the mIoU is improved by 7. 58% , and the average detection accuracy is improved by 11. 41% to 82. 36% .

Abstract:In terms of the problem that the signals obtained by laser ultrasonic testing technology applied to metal additive parts are complex, multimodal and with low signal-to-noise ratio, the laser ultrasonic signals are obtained for time-frequency analysis to explore their frequency domain separability. The variational mode decomposition algorithm is used to separate and extract the best surface wave mode according to the frequency domain characteristics. On this basis, a technique based on B-scan of laser ultrasonic signal combined with variational mode decomposition to extract surface echo eigenvalue is proposed for the quantitative detection of the surface crack length of metal additive parts. In view of the problem that there is a large error in obtaining crack length information by directly observing the B-scan image, the change of the peak to peak value of the reflected echo of the surface wave mode extracted by the variational mode decomposition is analyzed to detect crack. The scanning position-peak to peak value diagram is drawn, and the start and end positions of the crack can be accurately obtained according to the diagram. The relative error of the test results does not exceed 8%. Compared with directly obtaining the crack length information of the original B-scan image, the detection accuracy is improved. This method is feasible in feature extraction and quantitative detection of laser ultrasonic signals of metal additive parts.

Abstract:Ensemble learning soft sensors have been widely used to estimate key quality parameters in the process industry. However, the conventional ensemble modeling methods are often limited to mining the temporal relationships between samples for building the base models while ignoring the spatial relationships between samples. This may lead to problems such as insufficient local state mining of the process and insufficient diversity among base models. In addition, traditional soft sensor methods cannot effectively deal with the timevarying characteristics of the process due to the lack of adaptive mechanisms, which leads to the degradation of the model performance. Therefore, an ensemble adaptive soft sensor method based on the spatio-temporal local learning (STLL) is proposed. Firstly, the method mines the temporal and spatial relationships of process data, and the redundant states are removed by using statistical hypothesis testing. Then, a set of diverse spatial-temporal local Gaussian mixture regression models ( GMR) is formulated. Then, the local prediction results are combined adaptively based on an online selective ensemble strategy. Besides, a dual-updating strategy is proposed for alleviating the model performance degradation. Compared to the non-adaptive global GMR, temporal local learning based ensemble GMR, spatial local learning based ensemble GMR, experimental results show that the prediction accuracy of the proposed STLL approach is improved by 70. 3% , 14. 9% , and 27. 8% in an industrial chlortetracycline fermentation process, while it is improved by 31. 9% , 21. 2% , and 19. 3% in an industrial debutanizer process.

Abstract:Electromagnetic flow measurement is critical in the industrial manufacturing process. But, it is susceptible to the influence of air bubbles in the fluid, which will impact the measurement accuracy and cause variations in the measurement findings. Enhancing measurement accuracy by using techniques is essential. This study presents a theoretical model of the influence of bubbles on electromagnetic flow measurement from the standpoint of weight function to address the measurement optimization problem of electromagnetic flow measurement accuracy affected by bubbles. Secondly, the influence of bubbles on the weight function is investigated by using finite element simulation. An optimization method based on image acquisition and processing technology is proposed to reduce the influence of bubbles on electromagnetic flow measurement based on the simulation results. Finally, to test the feasibility of the optimization method, a bubble image processing algorithm is created, and an experimental platform for measuring gas-liquid two-phase flow electromagnetic flow is established. Experimental results show that the optimization method effectively reduces the sensitivity of the electromagnetic flow measurement system under the influence of bubbles. The error reduction amplitude is larger than 82. 63% , with a maximum error reduction amplitude of 91% . After optimization, the measurement error in the presence of bubbles is within ±3. 03% . The study effectively reduces the error of electromagnetic flow measurement caused by bubbles and provides technical support for improving the measurement accuracy of electromagnetic flow caused by bubbles and realizing the electromagnetic measurement of gasliquid two-phase flow.

Abstract:The force sensor of the traditional trilateral teleoperation system has large limitation in measuring the interaction force of the master and slave. It is easy to cause system instability when the shared weight is switched. In this article, a trilateral teleoperation control architecture of the hexapod robot is designed, which is based on force estimation and weight switching control. A nonlinear interaction force estimator is designed to estimate the interaction force between ports in real time. An adaptive switching algorithm for the weight factors is designed, which is based on the shared control strategy to focus on the compliant switching problem in the dynamic adjustment of the weight of two operators. To ensure the stability and transparency of the proposed system, the controller of the teleoperation system is designed by combining the force estimator and the weight switching algorithm. The force feedback device and Vortex and ElSpider hexapod robots are respectively used to establish a semi-physical simulation platform and a physical experiment platform to evaluate the proposed control method. Compared with the traditional trilateral teleoperation, experimental results show that the speed tracking performance is improved by 45. 12% and the force tracking performance is improved by 64. 71% on flat terrain. In rugged terrain, the speed tracking is improved by 39. 02% and the force tracking is improved by 29. 41% .

Abstract:A composite dynamic movement primitives algorithm based on Dirichlet process clustering and Gaussian mixture model is proposed to address the problems of low efficiency of parameter estimation and insufficient generalization ability in demonstration learning. To achieve the real-time estimation of Gaussian mixture model parameters, the Dirichlet clustering algorithm based on the distance threshold is used to perform online clustering of demo trajectory points, and the Welford formula is introduced to update the parameters to improve the efficiency of parameter estimation. After obtaining the trajectory distribution characteristics, the Gaussian mixture regression trajectories are encoded by using the dynamic movement primitives to improve the trajectory generalization. To evaluate the effectiveness of the algorithm, trajectory reachability and similarity metrics are introduced to evaluate the learning generalization ability of the algorithm, and demonstration learning experiments based on handwritten letter trajectories and robot kinesthetic demonstrations are designed. Experimental results show that the average parameter estimation time of the proposed composite dynamic movement primitive algorithm is only 0. 052 ms, which has the ability of fast trajectory reproduction and generalization.

Abstract:In the task of pose recognition based cooperative control of dexterous hand manipulator, the occlusion of body parts and interference of non-operators are always the main factors that affect the control accuracy. To effectively eliminate the aforementioned problems, an accurate upper limb posture recognition algorithm is proposed for human-machine collaboration system. Firstly, a frame selecting scheme is applied to box upper limb region based on Finger-YOLOv4. Then, the sparse target extraction algorithm is applied to exclude body interference of the non-operators. Next, we formulate a deep learning framework DFCRF-Net which aims at accurate positioning of 48 key points′ location and solving the problem of intra-class ambiguity. Finally, the upper limb postures is predicted according to the position relationships. The proposed method can accomplish mapping the upper limb posture between humans and manipulators, which could realize the human-machine cooperation of the dexterous hand manipulators. Experiment results demonstrate excellent performance with average detection speed of 33 FPS, average key point detection accuracy of 75. 2% , and cooperative operation completion rate of 98% , could meet the practical requirement.

Abstract:Due to the characteristics of similar appearance and dense deployment of devices in industrial field, it is difficult for the inspection robot to recognize similar devices in industrial field only by machine vision, which affects the accuracy and efficiency of autonomous inspection. To solve the above problems, this article proposes a similar industrial devices recognition strategy by using machine vision and proximity estimation based on the wireless signal characteristics of industrial internet of things. Firstly, the initial pose of the inspection robot is estimated by machine vision and the efficient perspective-N-point algorithm. Then, the proximity estimation algorithm is used to realize the recognition of proximal industrial devices targets by inspection robot. On the other hand, the strategy also includes robot angle correction and position adjustment algorithm to ensure the accuracy of proximity estimation. Compared with the traditional recognition method based on machine vision, experimental results show that the designed strategy can improve the recognition accuracy of similar industrial devices by 2% ~ 49% in different devices density scenarios, which effectively solves the problem of similar devices recognition of inspection robots in industrial field.