Impedance control is an important method for robot manipulator to control the contact force. By expanding the traditional integer order impedance model to the fractional order model, a fractional order impedance (FO-impedance) controller is presented first in this paper. A FO-impedance controller parameters tunning strategy based on the frequency-domain is presented. A systematic FO-impedance control simulation system with robot manipulator, servo system, dynamic disturbance and dynamic feedforward is constructed. The influence of time-varying dynamic characteristics for robot FO-impedance system is analyzed. In order to achieve desired dynamic response and high robustness performance, a dynamic feedforward method is proposed and the effectiveness is verified in the simulation.
the lightweight, highly reliable assembly and testing technology of spacecraft cables is an important direction of spacecraft design, production, inspection research and development at home and abroad. Spacecraft cable network is an important hub for connecting spacecraft equipment, and is one of the key factors for the success or failure of spacecraft. With the development of materials, electronic technology and interconnection technology, there is still a lot of room for optimizing the weight of spacecraft cable network. This paper focuses on the lightweight method, typical defects and reliability evaluation technology of special cables for spacecraft.
Vision surveillance is an effective way to detect the water accumulation and corrosion in bridge cable anchorage zone. Duo to the closed dark environment and the high reflectivity of metal surfaces in the zone, the acquired images are partially too bright or too dark. It is adverse for vision surveillance. To solve this problem, we propose to capture multi-exposure images to improve the image quality of the anchorage zone by image fusion. Seven multi-exposure image fusion algorithms are tested in the experiments. Both subjective evaluation and objective evaluation show that all the fusion algorithms can effectively improve the image quality of the anchorage zone. Among the seven algorithms, MEF-OSSI is superior to the others, and it is recommended in the engineering application.
Due to the large number and variety of equipment to be tested in the chemical plant, and the limited computing power and storage resources of inspection robots in the chemical plant, the backbone network of the deep learning algorithm is huge, and the number of parameters and calculation is large. The deep learning detection algorithm runs slowly or cannot run on the robot embedded equipment, this paper makes a lightweight improvement on the deep learning detection algorithm. Firstly, the YOLOv5 network, which has higher detection accuracy and faster detection speed at present, is proposed to be used for the detection of leak-prone devices in chemical plants, and then the huge network structure of YOLOv5 is lightened and improved. The original CSP1_X backbone feature extraction network of the network is replaced by the ShuffleNetv2 network, the ordinary convolution in the original network is replaced by depth-separable convolution to reduce the size of the network model, and the neck network structure is trimmed to reduce the number of network parameters and computational effort. Using data augmentation to randomly flip, randomly crop, randomly cut, and scale the dataset data, which in turn expands the amount of data and improves the detection accuracy of the network. The improved YOLOv5-v2 network can work on embedded devices on inspection robots in chemical plants.
This paper proposes a multi-robot formation coverage method by combining the improved Artificial Potential Field (APF) with the Centroidal Voronoi Tessellation (CVT). Firstly, we added the distance factor to weaken the infinite repulsive force, so that the robots can reach the target near the obstacle. Secondly, we set the judgment condition of the local optimal, by exploring multiple directions of local optimal area, robots will get the suitable direction to leave. Finally, we propose the Centroidal Repulsion Field (CRF) to prevent the centroids generate inside the obstacle, which will improve the efficiency of obstacle avoidance during multi-robot coverage. The effectiveness of the proposed method is verified by simulation in MATLAB.
As a kind of clean and renewable energy, wind energy has been paid more and more attention by government and academics. The goal of problem one is to evaluate the utilization of wind energy. The cure between wind speed and power was obtained firstly. Then the effective utilization probability and the utilization rate of wind energy were both calculated. The evaluation index of wind energy shows that the wind energy of this wind farm is rich. For problem two, the annual average utilization rate of wind energy was calculated firstly. Then, the capacity coefficient model of the fan was established, and the corresponding capacity coefficients were compared. It is concluded that the new type of wind turbine is more suitable for the wind farm than the old one. The experimental results illustrated that the proposed model in this paper is simple and effective.
In the factory, in order to feed back some emergencies in time, it is necessary to identify the identity of workers under specific circumstances. Aiming at this problem, this paper proposes a worker identification technology based on improved Strong Baseline. Firstly, the improved feature extraction network L-ResNet101 is used as the feature extractor; Secondly, add attention mechanism modules in different feature layers to make the model pay more attention to key features; Then the extracted features are fused and further classified. The experimental results show that the classification accuracy of the improved Strong Baseline network on Market1501 data set and user-defined AVA-Market data set is more than 90%, which proves that the improved network has better classification accuracy.
Planetary gear systems are widely used in different industry machinery due to the high transmission ratio and carrying capacity. However, the unacceptable manufacturing errors and faults in the key parts of the planetary gear systems can be easily caused by the unreasonable assemble processing and poor working conditions. To overcome this problem, a vibration diagnosis method for detecting the faults in a planetary gear system is proposed in this work. An experiment rig is conducted to obtain the vibration data of the planetary gear system. The frequency-domain vibration signals are used to diagnose the faults in the planetary gear system. The characteristic frequency of the planetary gear is modulated by the double frequency of the carrier rotation frequency. It seems that the vibration characteristic frequencies caused by the axis misalignment error of the planet gears is observed. The diagnose results are also verified by the disassemble works. This works can provide some guidance for the vibration diagnosis methods for the faults in the planetary gear systems.
In the application of permanent magnet synchronous motor (PMSM), the failure of current sensor usually leads to the unstable operation of a motor system and affect the normal operation of the system. A current sensorless PMSM control method is proposed in this paper, based on improved adaptive backstepping observer. Firstly, the recursive least squares identification inductance is combined with the adaptive backstepping observer to estimate the d-q axis current and stator resistance. Then, its stability is proved by Lyapunov method and its convergence is analyzed. Finally, Comparisons are carried out between the current sensor control and the observer without inductance identification. Simulation results show that the current and speed tracking response of the current sensorless drive system achieves expected effect, and has good dynamic performance and strong robustness.
The mechanical process card has the function of operation guidance in mechanical manufacturing. The digital reproduction of the mechanical process card with stains in the actual working conditions can help the staff to operate correctly. Aiming at the problem that the characters with stains can not be accurately recognized in the mechanical process card, this paper uses the framework of “machine learning + logical reasoning” to carry out research. Firstly, the preliminary recognition of character elements is completed through CRNN. In this step, the characters with stains may be recognized as wrong results. Secondly, the semantic analysis of the recognition results is carried out according to the special term library of mechanical and electronic industry, and then the semantic combination of various parts is carried out to analyze the syntactic structure of sentences. Finally, in the sentences with grammatical problems, the incorrect characters are corrected according to fixed collocation and other methods to output the correct results. The experimental results that the accuracy of using CRNN only to recognize the characters with stains is low, but combined with the correction of logical reasoning, the accuracy can reach 90%, but it has a good effect only for the Chinese characters, English and special characters with stains, and the recognition effect of numbers with stains is poor.
In order to solve the problem of insufficient stitching point identification accuracy in the stitching process of steering wheel leather, this paper proposed a stitching point identification algorithm of steering wheel leather based on improved normalized integral-correlation template matching and machine vision. Firstly, the “perona-mailk” anisotropic filtering was used to build a gold-like tower model for the images obtained under laser light source. Then, the stitching point template was used to search the original image, and all the image features of the adaptive region were collected. Finally, the precise position of the suture point was obtained by calculating the Harris corner of the adaptive area. Experimental results show that the algorithm described in this paper can accurately identify the stitching points on the steering wheel leather, and the recognition speed is faster.
The crawler construction machinery has great destructive power to the road surface, and needs to be loaded by a flat transport vehicle on the asphalt or cement road. Usually, the driver cannot observe the relative position of the machinery crawler belt and the vehicle crawling ladder under the machinery body. It brings difficulties for driver driving crawler construction machinery up and down the flat transport vehicle. To solve this problem, a portable vision assistance system is designed based on wireless network, which assists drivers to observe the blind spot area. The system consists of image acquisition and transmission module, optional wireless routing module, and image reception and display module. An electromagnet is utilized for the adsorption of the image acquisition and transmission module on the machinery body for the convenience of installation and removal. The optional wireless routing module is utilized for multiple cameras, and the image reception and display module is set in the driving room for the driver to observe the blind spot. Experiment results verify the effectiveness of the system, and it can be used under high-intensity vibration of construction machinery.
With the rapid development of cities, the overhead cables commonly used in urban transmission and distribution projects have been gradually replaced by underground cable channels due to their vulnerability to natural disasters and unfavorable urban aesthetics. In recent years, along with the development trend of digitalization and intelligence, the digital twin system has been introduced for the operation of underground cable channels. The digital twin platform requires the monitoring of cable channels in service as well as the numerical simulation of conditions such as ground loading and proximity to construction to evaluate structural safety. This is where modeling of different structural forms under different operating conditions becomes a very important research element. In this study, a parametric modeling approach is used to automatically generate the finite element model and the mesh of the model based on Python, and then Flac3D is automatically invoked to compute and post-process the finite element model. At the end of this paper, the stress-strain state of the structure is simulated in the case of a shallow buried cable channel under ground stacking conditions. The research results can provide some reference for similar projects in the future.
According to the “Made in China 2021” white paper plan, digital twin technology will be used as a key technology and means to practice China's advanced concepts, such as intelligent manufacturing, industrial Internet, and smart cities. According to incomplete statistics, many well-known companies such as China Midea and COMAC have introduced and applied digital twin technology. In this context, in order to promote the further development of digital twin technology in manufacturing production lines, based on the development of digital twin modeling, simulation technology, and advantages of digital twin enabling technology, this paper proposes an optimized method for the digital twin-driven manufacturing production line. Focusing on the actual needs of “Made in China 2025”, the paper analyzes the characteristics and advantages of digital twin technology in the transformation of manufacturing industry, as well as the application trend of digital twin in the future.
Due to poor mobility of the asphalt mixing station, it is necessary to transport the finished asphalt mixture to the designated locations by transportation vehicles. The existing loading methods of transportation vehicles need an operator to observe whether the stack of asphalt mixture in the carriage is full manually and control the switch of the material gate. In order to realize automatic loading of asphalt mixture in the asphalt mixing station, a method of monitoring the loading status of transport vehicles based on semantic segmentation is proposed in this paper. MobileNetV2 is used to be the backbone network of DeepLabV3+ instead of Xception, and a lightweight semantic segmentation network called M-DeepLabV3+ is constructed. The experimental results show that the M-DeepLabV3+ model has an accuracy increase of 0.5% and a speed increase of 16.36%. The recognition method proposed in this paper is almost identical with the manual judgment result, and the accuracy is 98.16%, which lays a foundation for realizing automatic loading of asphalt mixture in the asphalt mixing station.
The meshing efficiency is a very important parameter for the gear system, which can greatly affect the system transmission performance. However, the unreasonable design parameters can reduce the meshing efficiency, especially for the meshing efficiency caused the gear elastic deformation. To overcome this problem, a meshing efficiency calculation method of a spur gear system with the gear elastic deformation is presented in this paper. The effects of the tooth number, transmission ratio, addendum coefficient, pressure angle, friction coefficient, center distance, tooth elastic deformation on the meshing efficiency are discussed. It seems that the meshing efficiency of the gear system can be affected by the operational conditions and gear geometric parameters. This works can provide some guidance for the improvement method of the power loss of the oil stirring of the spur gear systems.
The power loss is a very important parameter for the planetary gear train. It can significantly influence the transmission performance of the planetary gear train. Thus, the power loss analysis can be helpful for improving the transmission performance of the planetary gear train. To overcome this problem, this paper presents a numerical analysis method to analyze the churning power loss of a planetary gear train. A computational fluid dynamics (CFD) model for a planetary gear train with four planet gears are proposed. The transient oil distribution and pressure distribution in the flow field of the planetary gear train are discussed. It seems that this works can provide some guidance for the improvement method of the churning power loss of the planetary gear train.
Visible pedestrian detection technology is widely used in video surveillance, human behavior analysis, intelligent military and other fields. In order to further improve the robustness and detection accuracy of pedestrian detection algorithm, this paper improves on the basis of YOLOv5m algorithm. Firstly, K-means++ clustering is used to optimize the initial center point selection, which makes the algorithm easier to converge in the training process and improves the robustness of the algorithm; Secondly, SENet channel attention module is introduced into YOLOv5m to make the network pay more attention to highlighted targets and improve pedestrian detection accuracy. Finally, OTCBVS data set and INO data set are used for training and testing. The results show that compared with the traditional YOLOv5m algorithm, the proposed improved algorithm has faster convergence and stronger robustness, and the detection accuracy is improved to 93.0%, which can be effectively applied to pedestrian recognition in visible light.
Vehicle target recognition is an important aspect of intelligent transportation, and the accurate recognition of vehicles is crucial for traffic control and safe driving. In this paper, we propose a vehicle target recognition algorithm based on 3D laser point cloud. In order to reduce the computational cost of data processing, the random sampling consistency algorithm (RANSAC) and two-dimensional rasterization algorithm of point cloud are introduced to reduce the original point cloud data. A density-based clustering algorithm is used to cluster the disordered target point clouds into a number of predefined clusters. According to the unique structural features of the vehicle target, the multidimensional composite feature vectors of the target point cloud clusters are manually extracted as input. The KITTI dataset is used to train the multilayer perceptron (MLP) classification model to achieve accurate recognition of vehicle targets. The results of experiments on validation data sets showed that the proposed algorithm achieved an accuracy of 93.2% for vehicle target recognition.
In order to achieve high precision 3d morphology reduction, this paper adopts the projection method with digital grating as the structured light generator, builds an experimental platform for 3D measurement based on digital micromirror, and combines the coded complementary Gray code for phase unwrapping operation in the algorithm of restoring the 3D shape of the object. The experimental results show that clear 3D object shape restoration is achieved in the time domain, and the measurement uncertainty reaches $\pm \ 2.45\ \mu\mathrm{m}$ as verified by repeatability experiments.
