Abstract:Pre-welding rail inspection is a significant step to ensure the safe operation of railway vehicles. Rail straightness is an important index to measure rail quality. Aiming at the problems of cumbersome traditional rail flatness detection steps, limited single measurement length and low measurement efficiency, according to the triangulation measurement principle, the overlapping of laser profiler is used to obtain the rail contour data, and the improved ICP algorithm is used to quickly register the point cloud data to complete the three-dimensional reconstruction of rail. Then the adaptive median filter is used to optimize the flatness parameter curve, and the analog ruler method is used to solve the rail flatness. The experimental results show that the flatness extraction algorithm based on rail shape reconstruction has the advantages of high speed, high precision and good stability. The maximum measurement error with manual detection is 0. 021 mm, the maximum error with high-precision electronic leveling ruler is 0. 011 mm and the maximum standard deviation is 0. 006 mm, which meet the requirements of flatness detection before rail welding.

Abstract:Aiming at the problem of carriage speed asynchronization caused by external disturbance during the operation of high-speed trains (HST), a distributed speed coordinated tracking control (DSCTC) method for HST based on an improved disturbance observer (IDO) is proposed. Firstly, a distributed HST model is established, secondly, combining HST with multi-agent system, and a DSCTC algorithm is designed; and an IDO is used to estimate the unknown disturbance of the system, and the observed value is fed back to the controller as compensation, and the Lyapunov stability analysis proves the stability of the controller and observer. Finally, the proposed method is verified by simulation, the speed tracking accuracy of the proposed method reaches 99. 9%.

Abstract:Highly accurate positioning is an important prerequisite for automatic train driving. In terms of the problems that the existing machine learning is used for train positioning, such as the insufficient theoretical basis for feature selection and difficulty in determining the proper structure of model, which lead to the unstable and inaccurate data about train positioning. A new positioning method about urban rail train is proposed based on an ensemble deep belief network (DBN). This method firstly preprocesses the original dataset, then uses the Pearson coefficient to filter the features, finally utilizes the quantum particle swarm algorithm (QPSO) to optimize the structure of the DBN-based learner. Comparing the proposed QPSO-DBN model with the ensemble model about the classical machine learning methods and the traditional optimized algorithms, respectively, the positioning accuracy of the train is further improved. Finally, the superiority of the proposed model is verified by simulation experiments.

Abstract:At present, it is necessary to measure the rail verticality and end slope by contact tools such as feeler gauge and angle gauge before welding rails in China. Due to the complicated manual measurement steps and low measurement efficiency, a non-contact rail measurement system based on laser profilometer is developed. Four laser profilers collect the information of the whole end face and side of the rail, then use the principle of dihedral angle to complete the calculation of the angle between the end face and side, end face and bottom surface of the rail. Finally, the end face slope is measured by the right-angle ruler model. After the field tests, the repeated measurement error of rail verticality is less than 0. 2°, the measurement accuracy can reach up to 0. 3°, and the limit error of end face slope reaches 0. 04 mm. The experimental results prove that the designed system can efficiently and stably complete the inspection of rail verticality and meet the actual test requirements.

Abstract:Aiming at the problems of complex fault types and low diagnosis accuracy of section jointless track circuit, a fault diagnosis method of least squares support vector machine(LSSVM)optimized by deep belief network(DBN)and marine predators algorithm (MPA) is proposed from the two aspects of fault feature extraction and feature classification. Firstly, the centralized monitoring data and status labels are input into DBN, and the dimensionality reduction feature extraction is carried out in a semi supervised way, so as to mine the different fault feature information of track circuit. Then, the intelligent algorithm MPA is used to optimize the penalty factor and kernel function parameters of LSSVM, and the optimal MPA-LSSVM diagnosis model is established. Finally, the feature samples extracted by DBN are introduced into the diagnosis model for fault classification and identification of track circuit. DBN-MPA-LSSVM diagnostic model makes full use of the advantages of layer by layer extraction of DBN in the process of feature extraction and the advantages of LSSVM in solving high-dimensional pattern recognition in the case of small samples. Experimental validation and comparative analysis show that the DBN-MPA-LSSVM model test set accuracy is 98. 33%, and the MPA optimization algorithm improves the diagnosis accuracy by 6. 11%, 3. 89%, and 3. 33% compared with PSO, GWO, and GA algorithm models, respectively, with an average accuracy of 97. 98%, which provides a new data-driven rail circuit fault diagnosis technology based on method.

Abstract:In view of the shortcomings of the traditional asymmetric half bridge (AHB) power converter of switched reluctance motor ( SRM) in electric vehicle drives, a novel DC/ DC converter cascaded at the front-end of the AHB power converter is proposed to realize the regulation of DC-link voltage and the controllable feedback of braking energy. Firstly, the operating modes of the power converter are analyzed theoretically. Then, the look-up tables of the excitation voltages, the demagnetization voltages and the turn-off angles about speeds and loads are established, respectively, so as to match the proper DC-link reference voltage according to the motor working conditions. And the overall closed-loop control scheme is given. Finally, the experiments are carried out on a three-phase 6 / 4 switched reluctance motor and the experimental results verify the effectiveness of the proposed power converter.

Abstract:In view of the problem of edge segmentation caused by workpiece adhesion and noise interference when the point cloud data obtained by the traditional 3D industrial camera is used for workpiece detection, considering the factors that the large amount of point cloud data affects the real-time detection and the inaccurate selection of 3D feature points leads to large measurement error, a preprocessing method based on 2D edge detection is proposed to realize the rapid segmentation and measurement of point cloud. In the first place, the improved Canny algorithm is applied to detect the edge of the texture image of the ordered point cloud, and the detected image is separated by mathematical morphology operation and contours detection, which avoids the segmentation process in 3D space and effectively reduces the number of point clouds. In the second place, combined with the shape characteristics and placement mode of the workpiece, the ordered point cloud data was extracted by mask operation, and the adaptive threshold filtering was performed on the segmented point cloud based on the RANSAC and conditional filtering method to effectively remove the noise point cloud. Finally, the workpiece size and normal vector are calculated based on the bounding box of PCA for the preprocessed target point cloud. We could know from results that compared with the traditional 3D algorithm, it can extract the target point cloud more accurately, efficaciously decrease the amount of point cloud data, and improve the segmentation efficiency by about 20%. The average relative error of workpiece size is 1. 24%, which can meet the needs of measurement.

Abstract:Obtaining accurate carrier attitude information is the key to improve carrier navigation and positioning. In view of the nongravity acceleration interference caused by accelerated motion, which cannot be solved by the traditional filtering algorithm, a posture solution method based on Mahony and improved Kalman fusion is proposed. Firstly, the data obtained by inertial measurement unit is used for Mahony solution, and the result is used as the measurement information of improved Kalman. Secondly, through the gyroscope solution, the result is used as the state information of the improved Kalman to realize the attitude solution. The experimental results show that compared with the traditional method, the solution accuracy of the method proposed in this paper is improved by more than an order of magnitude, which can effectively suppress the high-frequency noise of drift error, greatly improve the solution accuracy of carrier attitude angle, and has good convergence.

Abstract:In order to realize the optimal coordination of trajectory tracking and lateral stability of emergency rescue vehicles, a vehicle nonlinear parallel control method based on Hamiltonian function is proposed. The vehicle dynamics model and trajectory tracking model are established respectively, the vehicle dynamics model and trajectory tracking model are expressed as state equations with the same control input through model transformation, thus, the trajectory tracking and lateral stability coordinated control problem is transformed into a class of nonlinear parallel control problem. The Hamiltonian functions of trajectory tracking control and lateral stability control are designed respectively, and the existence conditions of controller design based on vehicle characteristics are discussed and proved, then, a nonlinear parallel control method considering the trajectory tracking and lateral stability control performance of emergency rescue vehicles is proposed, and the stability of control system is proved. The results show that, the trajectory tracking accuracy and stability control accuracy under the parallel control are increased by 10. 13% and 13. 79% respectively, which verifies that the designed method can better realize the coordinated control of trajectory tracking and lateral stability for emergency rescue vehicle.

Abstract:Aiming at the difficulty of accurate monitoring of the discharge pressure of rotary vane compressor, a method of sensing the discharge pressure of rotary vane compressor based on fiber Bragg grating (FBG) improved relief valve was proposed. Firstly, according to the structural characteristics of the pressure relief valve of the rotary vane compressor and the FBG sensing theory, the model of the diaphragm pressure sensor was established, and the mapping relationship between the pressure and the central wavelength was revealed. Secondly, a three-stage FBG pressure temperature composite sensor based on pressure relief valve was designed. Theoretical analysis shows that the sensitivity of the sensor is 304. 26 pm/ MPa. Finally, through simulation experiment analysis, the influence of diaphragm radius and thickness on sensor sensitivity was studied, and the optimal combination of radius and thickness (R = 5 mm, h = 0. 6 mm) was obtained. The experimental sensitivity is 330. 78 pm/ MPa. The research has important theoretical and practical engineering value for rotary vane compressor condition monitoring and power consumption control.

Abstract:School of Electrical Engineering, Shenyang University of Technology

Abstract:According to the working principle and structural characteristics of solenoid valve, this paper analyzes performance response characteristics of solenoid valve, builds simulation model in AMESim, simulates and analyzes pressurization and pressure relief characteristics of the solenoid valve. Combined with theoretical simulation data and test requirements, formulates performance test methods, develops a set of solenoid valve test system with virtual instrument technology, which realizes the function of automatically testing the response characteristics of solenoid valve. Measured the pressurization and pressure relief characteristics of on-off valve and linear valve, the pressurization rate of on-off valve is 835 bar/ s and the pressure relief rate is 914 bar/ s. Through relevant tests, it is verified that changing the control current can adjust the opening of linear valve. The simulation and test results are in line with expectations, which verifies the feasibility and reliability of the solenoid valve performance test system.

Abstract:Aiming at the problem that the existing difference matching detection algorithm of magnetic anomaly targets has poor effect under low SNR, a matching detection algorithm based on similarity measure is proposed. The similarity function is used to match the real-time signal and the background field signal, then the wavelet packet denoising is used to further improve the SNR. Finally, the processed signal is input into the OBF detector to complete the real-time target detection. The research indicates that when the false alarm rate is 0. 42%, and the SNR of the input signal is -9 dB, the detection rate of the algorithm is still around 90%, and its detection effect under low SNR is obviously better than that of difference matching detection.

Abstract:Point clouds have unique advantages due to its rich geometric information in computer vision field. Most of the existing point cloud classification and segmentation methods based on deep learning can identify the objects with canonical orientations. In real applications, there are problems of rotation transformation. In this paper, we propose a lightweight framework EFRI-N, namely, rotation invariant network of point cloud based on ellipsoid fitting, focusing on pointset rotation problems. We design a pre-network module to extract the rotation-invariant features. The ellipsoid fitting algorithm is used to identify the direction of the point clouds and obtain the rotation-invariant coordinate. Then the original features are mapped to the coordinate, and the rotation-invariant features were obtained by encoding the spatial and angular information. In order to obtain richer geometric information, multi-level feature connection is added to the network to enhance feature propagation and reuse. The classification and segmentation experiments are carried out by using the famous public datasets ModelNet40 and ShapeNet Parts. The results show that this method demonstrates better performance than state-ofthe-art methods in the task of processing rotating point cloud, and the network is improved by 1% ~ 62. 63%. Moreover, the computation amount and the number of parameters of the network have an order of magnitude advantage. It can meet the requirements of rotation invariance of point cloud in single object scenario and has good application value.

Abstract:In order to further improve the heat transfer efficiency of flat plate micro hearts, the effects of airflow channels of different beam structures on gas flow characteristics and heat transfer performance of flat plate micro heat pipes are compared and studied by simulation and experimental research. In the simulation software, the gas flow through the four structures at a time of 50 Pa at the inlet pressure of the airflow channel is 7. 451 6, 21. 915 3, 19. 239 2 and 23. 192 8 m 3 , respectively. The experimental working medium is Methanol, the liquid filling rate is 80% ~ 100%, and the heat input is 0. 5~ 2 W. Experimental results show that the airflow channel of the arc structure is more conducive to reducing the thermal resistance of the plate micro heat and improving the heat transfer capacity. At a filling rate of 100%, the thermal resistance of RSC-FMHP, RMC-FMHP, CSC-FMHP, and CMC-FMHP is 10. 137 5, 9. 125, 9. 575, and 8. 887 5 ℃ / W, respectively. Compared with the thermal resistance of RSC-FMHP, RMC-FMHP, and CSC-FMHP, the thermal resistance of CMC-FMHP decreased by 1. 25, 0. 237 5 and 0. 687 5 ℃ / W, respectively.

Abstract:The sensing distance and spatial resolution of phase-sensitive optical time domain reflectometer ( Ф-OTDR) are mutually restricted. In order to further improve the system performance, a random number coding method of optical pulse is proposed in this paper. The linearization theory analysis of the coherent detection Ф-OTDR based on coding and the principle of random number encoding and decoding are described, and the gain mathematical model of random number encoding is deduced. Experimental results show that compared with the single-pulse equivalent, the sensing distance can be increased from 25 km to 50. 26 km by using 128 bit random number code with 40 ns pulse width and 1 kHz pulse repetition frequency, and the signal to noise ratio can reach 17. 51 dB at 50. 22 km, where the vibration signal can be well restored. The research shows that the random number coding method provides an effective solution for improving the performance of Ф-OTDR.

Abstract:The near-surface detection capability is improved by analyzing the crosstalk mechanism and crosstalk isolation method between ultrasonic phased array probe elements. The influences of ultrasonic crosstalk between the elements before and after installing the wedge on the testing signal and testing sound field are analyzed based on the numerical simulation model. Then the capability of the probe wedge on isolation of the crosstalk of array element and near-surface testing signal is analyzed. The wedge block optimization scheme is proposed, and verified by the experimental method. The results indicate that the numerical model can visually display the ultrasonic crosstalk between the array elements, and the proposed wedge block optimization scheme can isolate the crosstalk signal between the array elements in the time domain and the near-surface detection signal from each other. The phased array probe after optimizing the wedge can perform imaging testing on flat-bottomed holes with a depth of 2 mm and a diameter of 0. 3 mm, the near-surface detection capability is significantly improved.

Abstract:Aiming at the problems of high risk and poor accuracy in common blowout fluid velocity measurement methods, in order to measure the blowout fluid flow velocity safely and accurately, a calculation method of blowout fluid velocity based on sequence images and pinhole imaging was proposed and carried out corresponding experiments. The method uses a camera and a self-developed blowout simulation device to obtain a large number of sequence images of simulated blowouts, selects two frames of images before and after, and uses the SIFT algorithm to extract the feature points of the two images respectively, and then uses the FLANN algorithm to extract the feature points for feature matching, and finally the RANSAC algorithm is used to eliminate the mismatched feature points. After obtaining the accurate matching feature points between the two images, the coordinate positions of these pixel points are extracted and the displacement distance between the matching feature points is calculated. Finally, the flow rate of blowout fluid is calculated by the method based on pinhole imaging. The experimental results show that a more accurate flow rate of blowout fluid can be obtained, so as to more accurately estimate the harm of blowout and ensure the personal safety of field operators.

Abstract:In the ultrasonic detection of corrosion defects of natural gas steel pipelines, the conventional pattern recognition method adopts the method of manually extracting echo signals, which has the problems of strong subjectivity and low universality. Based on this, this paper proposes a method to extract the features of echo signals by using one-dimensional convolutional neural network and classify the features by combining with random forest. Firstly, according to the noise of the echo signal, the wavelet packet transform is used to denoise the signal. The denoised signal is decomposed and reconstructed by variational modal decomposition to obtain a smooth signal. Finally, the processed echo signals are extracted by 1D-CNN network features and classified by random forest. The experimental results show that the identification accuracy of the method based on VMD-1D-CNN-RF is 85. 71% for artificial defects and 71. 05% for pipeline defects in natural gas stations, indicating that the pipeline condition can be preliminarily identified without expert identification.

Abstract:To achieve continuous interrogation of fiber Bragg grating (FBG) based on arrayed waveguide grating (AWG), a method of joint interrogation using two AWGs is proposed. The insertion of the spectrum of the corresponding channel of another AWG in the middle of the spectra of two adjacent channels of one AWG, forming the minimum spectral period. Each measurement selected the two channels with the strongest light intensity from the three channels, and used the relative light intensity interrogation algorithm to accurately measure the FBG center wavelength based on its wavelength-power relationship. The experimental platform was built by using two 100 GHz AWGs, and the interrogation of the temperature sensors was investigated. The experimental results show that the continuous and accurate interrogation of FBG is achieved within the 0. 8 nm minimum dynamic range period of the system, and the interrogation linearity of the system reaches 0. 999 1 and the wavelength accuracy reaches ±4 pm. The C-band range can be divided into multiple wavelength cycles by mathematically analyzing the data and experimental results, and the system can achieve continuous interrogation of a single FBG in the C-band 40 nm full-cycle range. The method not only realizes continuous interrogation of FBG based on AWG in the C-band range, which makes it possible to continuously sense the external physical quantity changes using FBG and improves the practicality of the system. Moreover, the method can accurately interrogate the wavelength information, which provides reference information for the realization of continuous and accurate interrogation of FBG and is conducive to the further expansion of the application field of FBG.

Abstract:In order to meet the special requirements of the front-end readout system of cadmium zine telluride (CZT) detector for analog to digital converter (ADC), a foreground digital self-calibration successive approximation register (SAR) ADC with 12 bit 1. 6 MS / s is presented. The working mode of calibration followed by output is adopted to improve the real-time performance of data conversion. In the calibration mode, the core part of SAR-ADC injects low amplitude disturbance to the differential ramp input voltage, then uses the least mean square (LMS) adaptive algorithm to calibrate and fix the Sub-radix-2 capacitance weight in the digital domain. In the normal working mode, the digital code is output normally according to the calibrated weight value. A digital-analog hybrid simulation platform is built for simulation and verification. The results show that when the clock signal frequency is 20 MHz and the input signal frequency is 239. 1 kHz, the signal-to-noise and distortion ratio (SNDR) of SAR-ADC after calibration is increased from 45. 59 dB to 72. 35 dB, and the effective number of bit (ENOB) is increased from 7. 28 bit to 11. 73 bit. The performance of SAR-ADC is obviously improved, and the linearity δ between front-end readout circuit and SAR-ADC is-0. 29% ~ 0. 34%, which can meet the requirements of CZT detector front-end readout system.

Abstract:Aiming at the problem that automatic parking system needs to correctly identify the parking space under multiple working conditions, this paper proposes a parking space recognition method based on distance and visual information fusion of cameras and ultrasonic sensors. This method first identifies the parking space through cameras and ultrasonic sensors, and the parking space type is output if the identification is successful. Otherwise, the information of cameras, ultrasonic sensors and wheel speed sensors is fused to obtain the vehicle parking position characteristic parameters, and the corresponding parking space type is output through fuzzy reasoning. Then, the simulation model of parking space identification under multiple working conditions is built on MATLAB, and the simulation analysis is carried out for different parking space scenarios. The simulation results show that the method is reasonable and reliable. Finally, the real vehicle experiment is carried out under multiple working conditions, and the parking space recognition accuracy is more than 90%, indicating that the method is feasible for practical application.

Abstract:The Kalman filter method based on residuals is proposed to improve the accuracy of electromagnetic flowmeter for the problem of poor measurement accuracy under strong interference conditions. In this paper, a sliding average filter is used to pre-process the experimental data and reduce the impact of strong interference noise on the electromagnetic flowmeter measurement during measurement. Analysis of the electromagnetic flowmeter in the process of solidification by the interference, the proposed residual-based Kalman filtering method to achieve process noise covariance Q with the flow rate changes quickly switch, to improve the response speed of Kalman filtering. The experimental results show that the uncertainty at constant flow rate is reduced by 14. 6% to 22. 6% and the response time is reduced by 5 to 18 seconds when the flow rate changes after the above algorithm processing. At the same time, the accuracy of measuring cumulative displacement reaches 0. 12%, meeting the construction requirements of solidification projects. The filtering algorithm designed in this paper effectively reduces the influence of noise and makes the measurement results more stable and reliable.

Abstract:In the optical marker-assisted drogue positioning system, the rapid and high-precision extraction of the optical mark is an important foundation for the positioning of the drogue. In practical application scenarios, the longer the working distance, the weaker the imaging of the point features. And the complex background interference greatly reduces the accuracy of centroid extraction. By analyzing the characteristics of optical marking points and circular features, combined with the structural characteristics of the drogue, the circular feature of the optical marking is extracted so that the effective feature area, which is used to replace the global search, can be segmented to eliminate the background interference accurately. Furthermore, this paper mainly proposes a two-level high-efficiency point feature centroid extraction algorithm that combines the FAST-like threshold coarse localization method and the two-dimensional Gaussian residual centroid fitting method. The method identifies the spot features and obtains the rough position quickly by setting the brightness threshold. On this basis, the sub-pixel centroid coordinates are extracted by fitting calculation in the pixel distribution neighborhood of the spot. Experiments results show that the proposed algorithm meets the real-time extraction requirements. Compared with other classical algorithms, the algorithm can maintain a high accuracy and achieve a high level of stability.

Abstract:Aiming at the problem that the existing parameter measurement methods are difficult to effectively measure the parameters of cross-linked polyethylene cable joint, a cable joint parameter measurement method based on three-dimensional point cloud segmentation is proposed. Firstly, radius filtering and random sample consensus (RANSAC) algorithm are used to remove the noise points and preprocess the coordinate alignment of the cable joint point cloud obtained by the composite 3D scanner. Then, the RANSAC algorithm is used to fit the cable joint point cloud to a circle, and the rough segmentation is realized according to the mutation characteristic of the ratio of radius variance of adjacent fitting circles at the regional junction, so as to obtain multiple local point clouds containing regional junction points. Next, the normals of the local point clouds were estimated using principal component analysis, and the regional junction values were derived from the jump characteristics of the axial angles of the point clouds at the regional junction points and the adaptive threshold algorithm. Finally, a statistical analysis of the junction values of the same area obtained from multiple strip point clouds was carried out to achieve a fine segmentation of the cable joint point cloud and complete the parameter measurement. The results of the measurement experiments on several cable joints show that the absolute error of the proposed method is less than 1. 0 mm and the relative error is less than 4%, which demonstrates the validity and accuracy of the method for measuring the parameters of cross-linked polyethylene cable joint.

Abstract:Electronic analytical balances generally use the method of pulse width modulation current to precisely adjust the electromagnetic force, and it is difficult to achieve a million-degree accuracy in engineering. After analyzing the influence factors of traditional methods on measurement accuracy, an improved pulse width modulation current circuit with double closed-loop control is designed, which decouples the previous PID control circuit and current measurement hybrid circuit, reduces the difficulty of PID parameter tuning and solving the difficulty of the system transmission equation of the mechanical sensor and balance detection channel, and improves the response speed and measurement accuracy. Experimental data show that, without nonlinear correction and only with moving average filtering, the original measurement data of an electronic analytical balance with a 31 g-range can achieve an accuracy of one million divisions with standard deviation of 2×10 -5 g, which is ten times better than the original 220 g-range electronic analytical balance with the same circuit and sensor

Abstract:Aeronautical auto-transformer rectifier unit (ATRU) is the key power conversion device of aircraft high-voltage DC power grid. It is continuously affected by high temperature, mechanical stress, load fluctuation and other factors during operation, then its internal components may appear corresponding failure, which can lead to threaten the reliable operation and continued airworthiness of the aircraft. The spectrum of the fault signal in the rectifier part of ATRU is difficult to distinguish and the diagnostic accuracy is low, a fault diagnosis method based on genetic algorithm ( GA) combined with Bayesian regularization back propagation neural network (BRBPNN) is proposed. Firstly, an ATRU fault simulation model is implemented and then the collected signals are processed by means of time-frequency analysis so as to mine the feature information of different fault states. Subsequently, genetic algorithm is used to optimize the initial weights and thresholds of BRBPNN and the optimal GA-BRBPNN diagnosis model is established. The feature samples are introduced into the diagnosis model for fault identification and model performance testing. Finally, the experiment platform of fault simulation is built and the measured fault data is used to validate the method. The experimental results show that the diagnostic accuracy of the proposed method can reach 99. 46% for the simulated faults and the method can diagnose and identify all the samples to be tested for the actual faults. Therefore, the method based on GA-BRBPNN has good diagnostic effect and high practical value.

Abstract:There exist some problems in the mechanical equipment condition monitoring system based on cloud computing framework. The problems of the extension of data transmission, poor real-time performance of early warning and diagnosis etc. Usually occur in practical application. This paper presents a mechanical equipment condition monitoring system for edge computing, which has three-tier architecture: Equipment layer, edge layer and cloud layer. High real-time computing tasks are deployed in multiple edge computing nodes, and data feature extraction, dimensionality reduction, intelligent diagnosis, data saving and uploading are carried out in the edge layer. The proposed method is verified on the spindle test-bed of high-speed machine tool. The experimental results show that the condition monitoring system based on edge computing reduces the output delay by 29. 5% compared with the condition monitoring system based on cloud computing, saves 81. 3% cloud storage space, and significantly improves the real-time performance of the system under the condition of ensuring a high diagnosis rate.