Abstract:One kind of “human-instrument integration” principle for designing workstations of desktop microscopic instruments which can provide better user experience is proposed in this article. The workstation based on the principle takes the operator as the center. In this way, the efficiency and comfort level are maximized, and the operating fatigue is reduced. Aiming at the operation characteristics and work path of the biological microscope, three kinds of workstation layout schemes are designed. The software Jack is used to evaluate operating time and energy consumption. The comparison experiment for operation comfort of the working postures is completed. The RMS of surface EMG for the main muscles of the neck and shoulder is tested by paired T test. The results show that the splenius capitis on the turning side and the sternocleidomastoid on the corresponding side are the most sensitive with bigger load variation, and the average RMS change rates are 56. 4% and 8. 4% respectively. Combined with the subjective evaluation of operators, the more comfortable overlapped type is finally determined as the optimal layout. The assessment method verifies the effectiveness of the “human-instrument integration” principle, which can provide guidance for the workstation design of related microscopic instruments.

Abstract:The compensation effect of the current dynamic compensation model of the time grating displacement sensor is affected by the speed of calibration platform. To address this issue, an error compensation model based on cubic spline interpolation-Fourier harmonic compensation method is proposed. Firstly, the periodic distribution characteristics of multiple probe and whole cycle error curves of time grating displacement sensor are utilized. The short period error is affected by the speed of calibration platform, and the “ dislocation” error is introduced. The dislocation error directly affects the compensation effect of the short period error compensation model. Secondly, the cubic spline interpolation method is used to locate the error sampling position accurately and reconstruct the short period error curve accurately. Finally, according to the reconstruction of short period error curve and the method of Fourier harmonic compensation short period error compensation model is established, the error compensation effect of time grating displacement sensor is improved. Experimental results show that the peak-to-peak value of short period error decreases to 1. 7″ after using the compensation model. The compensation effect of short period error is better than the traditional compensation model based on Fourier harmonic compensation method. When the calibration platform rate is 3 r/ min, the compensation effect can be increased by 56. 0% . It cannot only meet the working efficiency of sensor dynamic calibration, but also meet the demand of sensor high-precision error calibration.

Abstract:The system errors of the rotary steering drilling tool are included in the designed points of the correction matrix. The correction matrix error occurs, which is one cause of low precision attitude angle under near-vertical position compensated by the traditional correction (TC) . Average balance correction (ABC) can compensate for the correction matrix error, but the posture error in the whole self-rotation cycle is still uneven. A kind of equiangular interval balance correction (EABC) model is formulated based on the equal angle interval, the expressions of correction matrix error are induced, those influence characteristics on the posture accuracy are investigated, and the experimental data are computed respectively with TC, ABC, and EABC. The computed results show that compensated by EABC, the average inclination angle errors are lower than 0. 012°, and the peak-to-peak inclination angle errors are lower than 0. 027°. The average tool face angle errors are lower than 0. 008°, the peak-to-peak tool faces angle errors are reduced to 54 ~ 95 percent of those by ABC, and the standard deviations of tool face angle error are decreased to 40 ~ 63 percent of those by ABC, which means that EABC further reduces the absolute error under near-vertical positions, then improves the attitude angle accuracy.

Abstract:In aspheric surface detection based on the Zygo interferometer, the precision of the aspheric surface′s location along the optical axis (Z-axis) direction has direct effect on the accuracy of the measurement results. This article proposes a Z-axis positioning method for aspheric measurements based on the minimum PV value. The theoretical model of aspheric and reference spherical waves is evaluated, and the appropriate reference spherical wave radius, vertex deviation, and theoretical measurement location are developed. The positioning procedure along the Z-axis direction is completed within 0. 001 mm precision. The proposed method can precisely identify the theoretical measurement location and efficiently decrease the measurement error introduced by the positioning error. It can also precisely determine the reference spherical wave radius, which is essential for the point cloud reconstruction and stitching method based on the Zygo interferometer. The positioning of an aspherical mirror with a diameter of 108 mm is evaluated experimentally, and the positioning results are close to the calculated results of the point cloud data of the best reference spherical wave, demonstrating the method′s correctness. The PV residuals are 0. 047 λ and the RMS residuals are 0. 019 λ compared to the null testing results, which further demonstrates the positioning method′s accuracy.

Abstract:The measuring technology of moment of inertia based on Hilbert-Huang Transform ( HHT) is conducted to improve the measurement precision of the moment of inertia of large rotary body by the torsional pendulum method and to overcome the disadvantages of the traditional linear measurement model and the periodic method for calculating the moment of inertia. A nonlinear measuring model of moment of inertia considering friction resistance moment is formulated. On this basis of the model, the principal component of angular displacement is extracted by the empirical mode decomposition and the instantaneous damping coefficient and instantaneous undamped natural frequency are identified by Hilbert transform. The nonlinear model parameters are fitted by the least square method, and the moment of inertia is calculated accurately. A system for measuring the moment of inertia of a large rotary body is developed, and the rotational inertia measurement tests verify the feasibility of calculating the moment of inertia based on HHT are implemented on different combinations of standard parts. The measurement results of a large number of standard parts show that the relative error of the system is less than 0. 2% , the measurement precision of the interia using HHT is observably better than that of the traditional period method.

Abstract:The traceability of direct current ratio based on the principle of difference measurement is usually an effective way to meet the requirement of measurement uncertainty. However, the shortboard with the ratio adjustable resolution limits the application range of practical traceability. In this article, we propose the concept of equivalent fractional turns, establish the mathematical model of deep adjustment, break through the bondage of the ratio minimum step value under the existing theory, and realize the direct current ratio conversion device with the characteristics of multi-disk linkage with the resolution 1 × 10 -9 and the error superior to ± 2 × 10 -7 . The nominal ratio value of two standard devices is adjusted after linkage to be consistent with the calibrated one and solve the problem of full step value traceability within the range. A self-calibration method based on the reference winding is proposed for this conversion device. The difference value among the primary winding, reference winding, and secondary winding is measured in proper order. We correlate the error between the disks and obtain the actual ratio value of any disk at any position, which is characterized by real-time reproduction without relying on the upper standard device.

Abstract:The magnetic compensation process of single beam SERF atomic magnetometer has the problem of three-axis magnetic field coupling. To address this issue, a method of reverse calibration of three-axis sequential compensation value of magnetometer is proposed to reduce the coupling magnetic field in the process of three-axis compensation. Firstly, a three-axis magnetic field coupling mathematical model is formulated, which is a 3×3 to describe the magnetic field coupling relationship between three axes. The triaxial coupling coefficient of the laboratory magnetometer prototype is tested. Then, a method of reverse calibration of triaxial sequence compensation value by using the triaxial coupling coefficient is proposed. Finally, the magnetic compensation effects before and after calibration are compared. Experimental results show that after the reverse calibration of triaxial sequential compensation, the average response linewidth of the magnetometer is narrowed by 2~ 10 Hz and the sensitivity is improved by 3~ 5 fT/ Hz 1/ 2 . The effectiveness of this method is verified, which lays a foundation for further optimizing the magnetic compensation technology.

Abstract:The downhole annular flow electromagnetic measurement system can obtain real-time downhole annular flow information. However, the gas invaded by the annular flow channel affects its measurement results. The measurement system cannot accurately warn the overflow and blowout problems during the drilling process. To address the problem that the annular flow electromagnetic measurement system is affected by gas intrusion, the research firstly obtains the virtual current density function of the annular flow electromagnetic measurement system by formulating the theoretical model of the influence of gas intrusion on the annular flow electromagnetic measurement system. Secondly, the finite element simulation is used. The software dynamically simulates the different existing states of the invading gas, and analyzes the virtual current distribution law of the annular flow electromagnetic measurement system. Experimental results show that as the gas content of the two-phase mixed fluid in the annular flow channel increases in the range of 0% ~ 5% , the output voltage of the annular flow electromagnetic measurement system under different two-phase mixed flow rates shows a downward trend. The relationship between the output voltage and gas content of the annular flow electromagnetic measurement system under different two-phase mixed flow rates is fitted by the quadratic function. The coefficient of determination R 2 is above 0. 98, and the fitting residual e is less than 0. 011. The relational corrected annulus measures the flow. The research results can be used to reduce the influence of gas intrusion on the measurement results and improve the measurement accuracy of the downhole annular flow electromagnetic measurement system.

Abstract:In this article, an optical fiber sensing array system based on semiconductor ring laser is proposed for impact location monitoring on aluminum plate. A fiber Fabry-Perot etalon is used to demodulate the signal of fiber Bragg grating sensor (FBG). The strain applied to the FBG is encoded as the wavelength shift of the reflected light of the FBG sensor. Then, it is converted into intensity modulation by the optic Fabry-Perot etalon with a FSR of 50 GHz. The optic F-P etalon can demodulate multiple FBG signals at the same time via wavelength division multiplexing. The optical fiber sensing array system is installed on an aluminum plate with a thickness of 1 mm. A 50 g small steel ball falls freely at a height of 20 cm and strikes the aluminum plate to generate acoustic emission signals. By using time-frequency wavelet analysis, the group velocity dispersion curve of Lamb wave is obtained from the measured transient response of the FBG sensor, and the coordinates of the impact point are determined according to the triangular positioning algorithm. Through the impact test, the average positioning accuracy error of the system and positioning algorithm is 30. 89 mm, which achieves good accuracy.

Abstract:Plantar pressure distribution is closely related to human health, and abnormal changes in plantar pressure distribution are early symptoms of some foot diseases. To help patients prevent foot diseases and monitor the distribution of plantar pressure in real-time, a dynamic plantar pressure distribution measurement system based on a tactile sensing array is developed. The system has a total of 48 sensing units, and the measurement error is less than 2. 4% , which can accurately collect the pressure of each area of the sole of the foot. The system adopts a wearable design, the data acquisition system is fixed on the ankle, then connected to the host computer through Bluetooth. In addition, the traditional pressure analysis method cannot analyze the pressure distribution change during exercise. To address this issue, a dynamic plantar pressure distribution analysis method is proposed. The plantar pressure data of the human body are classified by the support vector machine, and the classification accuracy rate is 98. 6% . On this basis, the typical pressure distribution index of the gait phase is introduced, which is combined with the traditional analysis index to realize the analysis of human health status and evaluation. Experiments have shown that the system can accurately measure the plantar pressure distribution under exercise conditions, analyze pressure distribution changes under abnormal conditions, and give early warning to abnormal areas.

Abstract:The distributed pressure recognition method based on sensing arrays is usually to characterize pressure information as an image. Then, features for classification are extracted. However, there are still two problems. The first is the limited density of sensing arrays which leads to low resolution of the formed pressure images. The second is the existence of elastic coupling in flexible sensing arrays, which results in blurred edges of the pressure images. In this article, a hybrid-order channel response symmetric bilinear convolutional neural network (HoSB-CNN) is proposed. Firstly, the channel attention response CNN is constructed for enhancing the representation of first-order features. Secondly, symmetric bilinear features are proposed to improve the sensitivity to edges. In addition, due to the structural symmetry of the symmetric bilinear features, only the triangular matrix is retained in the storage and transfer of the features, which could reduce the network complexity. Finally, a multi-order feature hybrid strategy is used to enhance the nonlinear fitting ability of the network. And a press-letter dataset is constructed by self-built data collection platform and 8 × 8 sensor array to evaluate the HoSB-CNN. Results show that the accuracy of the proposed method is 98. 11% .

Abstract:A reflective torsion sensor based on the few-mode fiber tilted Bragg grating is designed. The fundamental mode and the secondorder mode in the fiber core can be mutually coupled at the small-angle tilted grating. Through analyzing the couple reflection peak intensity, the torsion angle can be detected. The influence of the tilt angles on the mode coupling efficiency is analyzed with different small-angle tilted gratings written on the few-mode fiber. The tilted gratings are on-line written by using the phase mask method on a fiber drawing tower. The gratings with title angle of 1° is selected to perform single-point and double-point torsion experiments. Experimental results show that the reflection peak of the cross coupling between the fundamental mode (LP01 ) and the second-order mode (LP11 ) is sensitive to the twisting of the optical fiber. Thus, the twisting angle can be measured. The torsional sensitivity is 0. 52 dB/ (rad·m -1 ) during the counterclockwise twist from -50° ~ -150°, while the torsional sensitivity is 0. 34 dB/ (rad·m -1 ) in the 40° ~ 190° clockwise twist angle range. The sensor has the potential to realize multi-point torsion sensing with a single fiber, which has application prospects for multi-point torsion monitoring.

Abstract:Sensors are important components in intelligent detection and automation devices. To solve the problem of fusion under multisensor asynchronous data, an innovative sliding clustering-based multi-sensor asynchronous information fusion method is proposed. Firstly, a K-Means clustering method is introduced to tolerate the asynchronous problem, which mainly uses curve fitting to give a simple and fast rule of thumb for the calculation of k-values in the real-time clustering method. Secondly, a clustering filter kernel is designed to form a sliding pipeline for fusion in the spatial-temporal domain. In this way, the variation of data is always kept within an acceptable error, and the real-time multi-sensor information fusion method is fully implemented. Finally, the experiments validate the correctness and rationality of the designed clustering fusion method. The experiments show that the SC-MSIF method is correct and feasible and has a better performance in terms of real-time performance, and the RMSE error of the SC-MSIF method is reduced by 47. 8% and 36. 3% compared to the EKF and MEAN methods. The actual test results of multi-sensor fusion in UAVs are also better.

Abstract:The good healthy state for the components of aircraft electromechanical system ( aircraft electromechanical system components) is an important prerequisite for the safe operation of the aircraft electromechanical system, which is of great significance to ensure the normal flight of the aircraft and the safety of the crew. The health state on-line estimation of electromechanical system components can realize the on-line estimation of component health state, and effectively support the condition-based maintenance of components and then provide decision-making reference for flight mission support. In this paper, the health state estimation methods in this field are fully analyzed, in which the typical representative of them, namely Electro-mechanical Actuator and Auxiliary Power Unit, are taken as the specific research objects. The basic concept and connotation of the estimation of the health state for aircraft electromechanical system components are presented, and the data-driven health state on-line estimation method for aircraft electromechanical system components is analyzed systematically. On this basis, the overall development trend is summarized, and the future development direction is prospected to provide references for the related researchers.

Abstract:While promoting the upgrading of traditional manufacturing, the industrial internet has introduced cyber security issues which makes industrial devices such as intelligent instruments face more severe cyber security risks. To effectively clarify and analyze the failure causes and hazards of intelligent instruments under functional safety and cyber security threats, this article proposes a method framework that enables reasoning and evaluation of integrated casual failure paths. Based on the framework, this article realizes the integrated failure analysis of functional safety and cyber security, and realizes the integrated causal failure path reasoning. Meanwhile, the importance of the failure path which represents the possibility of the path is quantitatively evaluated by attributes such as structure, probability and essence. Therefore, it realizes accurate analysis of instruments failure scenarios. Finally, the effectiveness and feasibility of the proposed method are evaluated by the intelligent transmitter. Through the failure path reasoning method based on dynamic causal diagram, it is the first time to reveal the influence principle of cyber security failure penetration between the internal functional modules of intelligent instruments, and realize the integrated failure analysis process of functional safety and cyber security.

Abstract:Fusion of wide-area traveling wave (TW) information in the power network can improve the accuracy and reliability of fault location. However, there is no available multi-source fusion location method which can adaptively determine the importance of each TW measurement point at different fault locations according to network topology. Hence, a novel fault location method based on wide-area TW and graph attention network (GAT) is proposed in this paper. Firstly, all TW measuring points of the whole network and the overhead lines among them are taken as the nodes and edges to construct the graph data of wire-area TW, while the variational mode decomposition (VMD) theory is employed to obtain the nodes′ features. Then, GAT mines the wide-area TW characteristics according to the topology correlation of the power network, identifies the fault line, and outputs the adaptive weight that represents the importance of the measured points. Furthermore, the adaptive weight is used to fuse multi-source TW information to calculate the precise fault location. The results show that this approach can locate multiple types of line faults, and the location errors at many typical fault points are within 100 m. Compared with the traditional methods, this approach has more obvious advantages in case of fewer measuring points, and the location accuracy under different fault points is improved by 20~ 400 m.

Abstract:The high temperature strain electric measurement technology has been widely used in the stress state measurement of the hot end parts of aeroengines. How to rationally match the structural parameters of the high temperature strain gauge to improve the sensitivity and service life of the strain gauge is of great importance in engineering. Firstly, a simple beam-measurement error model and a cantilever beam-fatigue life model are formulated for two objectives of affecting the measurement error and fatigue life of the high temperature strain gauge. Secondly, the influence rule of each parameter variation on strain gauge measurement error and fatigue life is analyzed by finite element analysis, and the parameters to be optimized are selected. Then, the genetic algorithm combined with response surface method is used to optimize the high temperature strain gauge. Finally, the optimization results are evaluated by experiments. Results show that the single-parameter analysis method can directly reflect the influence of each parameter on the sensitivity and life of strain gauge. Based on the optimization of response surface model and the multi-objective genetic algorithm, the optimal parameter combination of five parameters of high temperature strain gauge including grid wire diameter, grid wire length, grid wire spacing, grid wire bending number and base thickness can be obtained. The optimized measurement error is 0. 255% and the fatigue life is 2. 384 6×10 7 cycles. Experimental results show that the measurement error is reduced by 89. 2% and the fatigue life is increased by 10. 14% after multi-objective optimization.

Abstract:The mechanical fault of the contact system for a conventional circuit breaker is a process from slight to severe. The accurate identification of its operating state can greatly improve the reliability of the circuit breaker. In this article, a single signal input and multi-task output MTL-SEResNet model is proposed for fault diagnosis and degree evaluation. Firstly, the raw vibration signals of the contact system are analyzed using a continuous wavelet transform. And the corresponding two-dimensional time-frequency images are obtained. Secondly, the improved ResNet18 network is combined with the SENet structure, and the multi-task learning sharing mechanism is used to formulate the MTL-SEResNet model. The model is optimized by adjusting the weight ratio of the two task loss functions for fault classification and degree evaluation. Finally, the proposed method is verified by experiments with simulated fault data of the contact system. The results show that the proposed model has better performance with 99. 78% and 99. 36% accuracy in type and degree, respectively, which can effectively evaluate the fault degree of the conventional circuit breaker.

Abstract:The failure behaviour of modern systems is complex, with both dynamics and correlation. First, in order to describe the dynamic failure behaviour intuitively and accurately, a novel continuous-time dynamic Bayesian network analysis method is proposed, which uses node sequence conditional probability table(CPT) to describe the event relationship. Then, the calculation method of child node failure probability, posteriori probability and importance measures of root node based on the rule execution degree of node sequence CPT and the sampling property of impulse function is proposed. Further, aiming at the system correlation failure behaviour caused by common cause failure(CCF), a novel continuous-time dynamic Bayesian network analysis method considering CCF is proposed to solve the overlapping problem of system failure logic dynamics and correlation. Compared with the Bayesian network, discrete-time dynamic Bayesian network analysis method, Markov chain and Monte Carlo method, the feasibility and superiority of the proposed method are verified. Finally, the reliability of dynamic failure related systems is evaluated, the results show that the proposed method can directly and effectively describe the dynamic and correlation failure behavior, obtain the accurate system reliability index, compared with ignoring CCF, considering CCF can improve the reliability analysis accuracy of the system by 29% when the task time is 5×10 6 h, which is more practical.

Abstract:Data in the process industry are highly time-varying and nonlinear. Traditional offline models can hardly cope with the changing working conditions in the actual production process, while the just-in-time learning ( JITL) is an effective online modeling method. Most of the studied similarity measurements of JITL only focus on samples’ spatial distance, which ignore the time-series characteristics of industrial data. To address this issue, a JITL method based on spatial-temporal similarity is proposed. First, the sample point is extended into a sample sequence, and the temporal-sequence distance among samples is calculated by combining dynamic time warping. Then, the spatial-temporal similarity metric (SSM) is proposed, and the SSM is constructed by nonlinearly weighting the temporal and spatial distances. Finally, the online modeling method for just-in-time learning based on spatial-temporal similarity ( SSJITL) is proposed. The algorithm is applied to a public dataset and an actual polyester fiber polymerization process. Experiment results show that the goodness of fit reaches 91. 6% and 98. 6% , which demonstrates the effectiveness and superiority of the proposed algorithm.

Abstract:Using γ photons to detect dynamic flow fields inside a cavity requires a fast image reconstruction algorithm. The traditional processing method is to collect all events first. Then, the algorithm processing is performed, such as OSEM. This study proposes an image reconstruction ( T-OSEM) algorithm that subdivides the response events according to the time stream. At the same time of continuous sampling data, the sampled data are divided into sub-sampling data sets according to the time period, and OSEM iteration is performed on the subset to achieve image reconstruction. The previous frame image is taken as the iterative input, and the correlation between frames is used to accelerate the convergence of iterative operation. The sampling of the data stream in T-OSEM is carried out simultaneously with the processing of the previous image frame. The image reconstruction process is accelerated by the multithreaded parallel operation. The relationship between the optimal number of subset events and the corresponding sampling time is studied to achieve the optimal reconstruction effect under the shortest sampling time. Experiments show that when the sampling time period reaches 1s, there is still a good particle tracking effect, and the mean structural similarity of the particle trajectory image is 0. 92. Results indicate that the T-OSEM algorithm is a good solution for dynamic image reconstruction.

Abstract:The mismatch of image features affects the basic matrix calculation and leads to poor estimation accuracy of SLAM visual odometry. To address this issue, an optimization method of visual odometry based on RANSAC is proposed. First, the initial matching is roughly filtered by the minimum distance threshold method with an appropriate threshold, and the relative transformation relationship between images is then calculated by RANSAC. The result that conforms to the transformation relationship is considered an interior point. The iteration result with most interior points is the correct matching result. Then, the homographic transformation between images is calculated, and the basic matrix is derived from the calculated results. The interior points are determined by epipolar geometric constraints and the fundamental matrix with most interior points is obtained. Finally, the TUM data set is used to validate the performance of the Visual Odometry optimization algorithm from characteristic matching and basic matrix calculation. The experiment results show that the optimized RANSAC algorithm not only effectively improves operation efficiency and removes the mismatched feature points, but also improves the accuracy of image feature point matching by 7. 7% . Meanwhile, the interior-point rate of the basic matrix estimation algorithm in this paper is increased by 3% while improving the basic matrix calculation accuracy. This algorithm provides the theoretical basis for improving the accuracy of visual odometer estimation.

Abstract:The infrared thermal imager has advantages of real-time on-line, non-contact acquisition of the two-dimensional temperature distribution of the measured object, which is widely used in steel, agriculture, power electronics and other fields. However, infrared temperature measurement results are easily affected by interference factors. To address the infrared temperature measurement error caused by the change of directional emissivity, a temperature compensation method based on directional emissivity correction is proposed in this article. Firstly, based on the principle of infrared temperature measurement, an infrared temperature measurement compensation model for the change of directional emissivity is formulated. Secondly, to determine the directional emissivity in the compensation model, a 3D thermal imaging system constructed with an infrared thermal imager and a laser scanner is used, and a reference body-based directional emissivity correction method is proposed. Then, the law of the directional emissivity changing with the viewing angle is determined by polynomial fitting. Experimental results show that the proposed infrared temperature measurement compensation method is effective in reducing the infrared temperature measurement error caused by the change of directional emissivity. After temperature compensation, the maximum error of infrared temperature measurement is reduced from 9. 64℃ to 2. 97℃ , and the standard deviation is reduced from 3. 57℃ to 0. 71℃ .

Abstract:The crowd anomaly detection is an important research direction under intelligent crowd surveillance technology. In the existing methods, the first step of anomaly detection is to obtain motion information. The traditional way of uniform chunking of video frames does not guarantee the integrity of pedestrians, and the extracted features do not accurately reflect the motion status of pedestrians. In this article, an incremental crowd grouping method is proposed, which firstly combines crowd motion field with frame difference method to segment the image to obtain crowd foreground. Then, the direction group is achieved, which is based on crowd motion direction. Finally, Spatio-temporal information is combined to re-cluster the direction group to get a more detailed pedestrian group. For each pedestrian group, the crowd energy feature is used to characterize the overall pedestrian motion information, and the ring block energy histogram feature is constructed based on the energy field to weaken the effect of pedestrian limb swing, and finally combined with the image appearance features for crowd anomaly detection. Experimental results show that the proposed method achieves 83% and 92% accuracy at frame level and 64% and 83% accuracy at the pixel level in two different scenes, which is a significant improvement compared to the traditional method.

Abstract:The object dimension measurement in small scenes has problems of low resolution, weak texture and lack of reference objects. To address these issues, a method using the 5-dof electromagnetic tracked endoscope is proposed. Firstly, the positioning principle of the endoscope based on the 5-dof sensor is analyzed. The displacement of endoscope along the main optical axis can be obtained. Then, the navigation acquisition method of the key information of the image target is analyzed. The contour information of the image target is obtained through the semantic segmentation network. Then, the coincidence degree with the main optical axis of the endoscope is determined, and the image target length information and the corresponding pose information of the key frame meeting the coincidence conditions are recorded. Finally, based on the pinhole camera imaging model, the target size measurement method is established by combining the target imaging scale relationship with the displacement of the endoscope along the main optical axis. Experimental results show that the measurement error of the method is controlled within 10% . The average measurement error for targets with a length of 1~ 9 mm is 0. 33 mm. It can meet the requirements of stable and reliable target size measurement, high precision, time-saving and laborsaving in the monocular endoscopy. Keywords:computer vision; object d

Abstract:Ice scoop surface defects include severe defects and minor defects. Serious defect detection technology is relatively mature, while mild defect detection involves the game of missed detection and false detection. Serious defects belong to the waste products that cannot be sold. Splitting and pollution in the mild defects are also waste products, while knots, mineral threads, decay, and color difference belong to qualified products, and the price is lower than that of good products without defects. This article proposes a set of solutions for the detection of mild defects: Firstly, the contour of the ice scoop is extracted and the surface area of the ice scoop is located. Then, the surface of the ice scoop is preprocess. Finally, the linear equation fitted by the least-squares method and the preprocessed gray value data subtract are used to calculate the deviation. The normal surface deviation value is obtained by iterative fitting of the least-squares method, thereby calculating its standard deviation, and finally extracting the abnormal data (defect) based on the 3σ criterion. The above algorithm can detect small-area light defects. For large-area decay, chromatic aberration, and other defects, the deviation calculation based on the gray peak level linear equation is adopted. A combination of the two methods detects surface defects on ice scoops. The ice scoop surface defects studied in this paper cover the main types of surface defects in the ice scoop production industry, including knots, mineral lines, splitting, contamination, decay, and chromatic aberration. The algorithm test on the self-built image database shows that the missed detection rate of this method is only 1. 27% , and the false detection rate is reduced to 3. 85% , which demonstrates its practical deployment value.

Abstract:Dioxin (DXN), known as “ the poison of the century”, is one of the by-products emitted by the municipal solid waste incineration (MSWI) process. Limited by the technical difficulty, time and economic cost of DXN emission concentration detection, it is difficult to obtain sufficient labeled samples for building a DXN emission soft sensor model. To effectively utilize a large number of unlabeled samples collected by the field control system and solve the problem of poor generalization performance of traditional shallow learning models, a soft-sensor method of DXN emission concentration based on Bagging semi-supervised deep forest regression (DFR) is proposed. First, multiple training subsets are obtained by resampling the original labeled dataset based on the Bagging mechanism, and multiple random forest (RF) models with diversities are formulated. Then, the RF model is iteratively updated, the nearest neighbor set is selected and the generalization performance strategies are evaluated, which are all used to obtain high-confidence pseudo-labeled samples. Finally, a DFR model is constructed based on the pseudo-labeled and original labeled sample sets. The effectiveness of the proposed method is evaluated with the actual DXN detection data of MSWI power plant in Beijing. It shows that the propose method has well prediction stability, and the root mean square errors are 0. 015 50, 0. 020 23 and 0. 019 73 for training, validation and testing datasets respectively.

Abstract:The d-q axis current of the PMSM under the vector control cannot be dynamically decoupled, which brings difficulty in adapting to the dynamic changes of the motor parameters. The stability of the motor control system is affected. To address this issue, this article proposes a deviation decoupling control method of the PMSM based on the parameter identification (PI-DDC). First, taking the coupling coefficient of the d-q axis current controller as zero, and the transfer function of the deviation decoupling could be obtained under this condition. Then, the dynamic changes of the motor parameters that affect the decoupling effect of the deviation decoupling can be considered, such as the resistance, inductance and permanent magnet flux. The recursive least square method with the forgetting factor is introduced to online identify the motor parameters, which are further used to correct the corresponding values in the deviation decoupling control model real time and the current decoupling of d-q axis can be realized. A case study has been carried out on a PMSM. The simulation and experimental results under different working conditions validate the better control accuracy and dynamic performance of the proposed PI-DDC method comparing with the traditional deviation decoupling control method. The response time, overshoot and fluctuation amplitude of the d-q axis current are decreased about 50% , 60% and 70% , and 30% and 37% respectively.

Abstract:The dual-winding bearingless flux-switching permanent magnet machines ( BFSPMM) is a non-linear and multi-variable complex system. The traditional speed PI and radial displacement PI control have disadvantage of large overshoot. The system is susceptible to external disturbances. Based on the idea of sliding mode variable structure control, this article proposes a sliding mode control (SMC) strategy for the radial displacement of Dual-winding BFSPMM to control the suspension plane. In view of the inaccurate mathematical model in sliding mode control, the gyro effect in the rotor dynamics, etc. An extended state observer (ESO) is established to observe the system disturbance, and the disturbance feedforward compensation control is added to SMC. Thus, a compound control strategy of SMC and ESO is formed. Compared with the traditional PI control, experimental results show that the proposed radial displacement SMC has advantages of faster response speed, smaller radial displacement pulsation, and strong resistance to load disturbances. Compared with the radial displacement SMC, the proposed composite control strategy of radial displacement SMC and ESO can effectively reduce the radial displacement pulsation of the rotor by about 30% , and further enhance the robustness of the system.

Abstract:The gas flow verification system with positive pressure sonic nozzle has high requirements for the stability of air source pressure. It is difficult to control the stagnation chamber pressure within the set value range, and adopt the dynamic characteristic analysis based on the mechanism model. To solve the aforementioned problems, a modeling and parameter identification method is proposed by using data and the mechanism model for the gas flow verification system with positive pressure. Aiming at the nonlinear dynamic parameters in the process of mechanism modeling, the stochastic configuration network is used for dynamic parameter identification. The RMSE of the pressure and temperature in the stagnation cabin after identification and the actual data are 1. 26×10 4 and 2. 64× 10 -4 , respectively. In view of the dead zone of the control valve or the replacement of the control valve in the actual operation, the disturbance experimental analysis of the model is carried out. The practical results show that the dynamic mathematical model reflects the change trend of pressure in the pressure regulation link, and has a certain practical application value.