A front multi-vehicle target tracking algorithm optimized by DeepSort was proposed in order to improve the awareness of autonomous vehicles to the surrounding environment. Gaussian YOLO v3 model was adopted as the front-end target detector, and training was based on DarkNet-53 backbone network. Gaussian YOLO v3-Vehicle, a detector specially designed for vehicles was obtained, which improved the vehicle detection accuracy by 3%. The augmented VeRi data set was proposed to conduct the re-recognition pre-training in order to overcome the shortcomings that the traditional pre-training model doesn't target vehicles. A new loss function combining the central loss function and the cross entropy loss function was proposed, which can make the target features extracted by the network become better in-class aggregation and inter-class resolution. Actual road videos in different environments were collected in the test part, and CLEAR MOT evaluation index was used for performance evaluation. Results showed a 1% increase in tracking accuracy and a 4% reduction in identity switching times compared with the benchmark DeepSort YOLO v3.
A novel method based on health index similarity in multiple time scales with autoencoder (AE MTS-HI) was proposed aiming at the shortage of the traditional similarity-based method in extracting health index and similarity matching. Autoencoder was applied to construct the health index based on monitoring data, which can minimize the loss of nonlinear information. The health index in multiple time scales was developed for similarity matching by considering the fluctuation of the length of test degradation trajectories. The method can remove the accuracy limitation caused by fixed time scales and enhance the prediction robustness. Performance of the proposed method was evaluated on public turbofan engines datasets. Results demonstrate that the method can improve the remaining useful life (RUL) prediction accuracy and provide stable support for predictive maintenance.
A repetitive high-voltage nanosecond pulsed electric field (RnsPEF) generation system was independently developed based on the spark switch and transmission line transformer (TLT) technology in order to analyze the key impact parameters of the process of malignant tumors ablation by high-voltage nanosecond pulsed electric field (nsPEF). The system can stably generate nanosecond exponential pulse. The experimental results proved the effectivity and controllability of RnsPEF on tumor cells ablation. B16 melanoma cells adherently seeded in six-well plates as the research object to analyze the effects of pulse number, peak voltage, repetition frequency and electrode spacing on tumor cells ablation. Cell counting kit-8 (CCK-8) was applied to measure cell viability of B16 tumor cells suspension in the cuvette after treated by pulses. The experimental results show that the pulsed electric field intensity and injected energy density of the applied RnsPEF play the key roles in determining the ablation effect. The repetition frequency hardly affects the ablation results. The pulsed electric field intensity threshold of RnsPEF ablating B16 melanoma cells is 6.8 kV/cm, and the injected energy density threshold is 11.4 J/cm3, as well as the optimal pulse number is 500 pulses.
Chitosan aerogel has good biocompatibility, non-toxicity, easy degradation and other excellent properties, which can be used as an ideal green oil adsorbent to effectively solve the major problems of oil leakage and pollution. The researching progress of new petroleum adsorbents based on chitosan aerogel was reviewed. Firstly, the advantages and disadvantages of traditional oil treatment methods and oil absorbents were compared and the superiorities of chitosan aerogels as oil adsorbents were summarized. Then the synthesis and modification methods of chitosan aerogels and their advantage as petroleum adsorbents were analyzed and summarized. Finally, the problems existing in the current research and the future research direction were summarized and prospected.
A garbage image classification method based on improved MobileNet v2 was proposed aiming at the problems of poor real-time performance and low classification accuracy of existing garbage image classification models. A lightweight feature extraction network based on MobileNet v2 was constructed. The parameter numbers of the model were reduced by adjusting its width factor, channel and spatial attention modules were embedded in the model to enhance the network's ability to refine features, a multi-scale feature fusion structure was designed to enhance the adaptability of the network to scale, and transfer learning was used to optimize the model parameters to further improve the model accuracy. Experimental results show that the average accuracy of the algorithm on the self built dataset was 94.6%, which was 2.0%, 3.4%, 3.2%, 2.3% and 1.2% higher than that of MobileNet v2, VGG16, GoogleNet, ResNet50 and ResNet101 models, respectively. The proposed algorithm achieved good performance in two public image classification datasets, CIFAR-100 and tiny-ImageNet. The parameter numbers of the model was only 0.83 M, which was about 2/5 of the basic model. The single inference on edge device JETSON TX2 took 68 ms, which proved the improvement of inference speed and prediction accuracy.
The researches on data security management and privacy protection technologies at home and abroad were analyzed and summarized aiming at current problems in blockchain security, such as unreasonable data management mode, unreliable data sharing scheme, smart contract vulnerabilities not easily fixed and incomplete privacy protection of multiple types of data. Various security problems and reasonable solutions in current blockchain systems were outlined from four aspects: data storage security, data privacy security, data access security and data sharing security. The challenges and future research directions of data security in blockchain were discussed. Some reference for the future work of researchers was provided in the field of blockchain security.
The shared control strategy, as the main control mode of teleoperation robots based on telepresence, can make full use of the operator's perception, judgment and decision-making ability, and utilize to the robot's own unique advantages. The telerobotic telepresence technology was introduced. The development of teleoperation shared control strategy was summarized. The principles of each control strategy were introduced mainly based on tactile feedback guidance, kinematic constraint avoidance and sharing factor assignment. The bottlenecks and shortcomings in the development of telerobotic shared control strategy were analyzed, such as the singleness or rigidity of shared factors, time delay and limited autonomous judgment ability of robots. The future research trends were proposed from three aspects in view of the limitations of the current study, namely, improving the intervention level, strengthening robot intention prediction, and combining machine learning, which have a certain guiding significance.
The data is easy to be tampered and the access control of data is not flexible in the Internet of Vehicles (IoV). A secure data sharing scheme based on blockchain and ciphertext-policy weighted attribute-based encryption was proposed aming at the above problem. In this scheme, roadside units jointly maintain the generation, verification and storage blocks to achieve distributed storage of data, which ensures the data from being tampered. Attribute-based access control ensures that only authorized entities can access the content of data on the blockchian. A hierarchical access policy formulation method based on multi-attribute was proposed to reduce the complexity of access control policy aiming at the data sharing requirements among the multiple entities and roles in the IoV, by mining the association of attributes in the roles for data access. Experimental results show that the proposed scheme can realize the secure storage and flexible access control of the data in the IoV, and the hierarchical access policy formulation method can effectively reduce the calculation and transmission overhead of vehicles, and meet the access requirements of multiple entities and roles in the IoV.
The characteristics of industrial boiler design and the necessity of introducing digital twin technology were summarized. The development and research status of digital design technology for industrial boilers were comprehensively summarized, and it was proposed that the digital design technology of a new generation of industrial boilers, with the design process optimization as the core and the digital twin as the foundation, was the key to improve the design capability and comprehensive performance of industrial boilers. The application characteristics of digital twin technology in industrial boiler design were analyzed, and three key technical problems of digital twin driven industrial boiler design were summarized: digital twin modeling technology for the expression of multiple information in the design process of industrial boiler; design process optimization technology based on human-computer interaction and virtual reality intelligent verification; industrial boiler digital twin data management technology for the full life cycle. On this basis, a digital twin driven digital design technology framework for industrial boilers was proposed, which was expected to provide ideas and valuable references for the research and application of digital design technology for high-performance industrial boilers.
The research works of Chinese font style transfer were classified according to different stages of research development. The traditional methods were briefly reviewed and the deep learning-based methods were combed and analyzed. The commonly used open data sets and evaluation criteria were introduced. The future research trends were expected from four aspects, which were to improve the generation quality, enhance personalized differences, reduce the number of training samples, and learn calligraphy font style.
A mixed integer linear programming (MILP) model was proposed to optimize the spatial-temporal resources at intersections under the environment of mixed traffic flow with connected and autonomous vehicles (CAVs) and human-driven vehicles. The objective of the model is to maximize the intersection capacity, and the constraints mainly include those regarding lane channelization, flow distribution and signal timing settings. The lane channelization and signal timing scheme at intersections were optimized with different CAV driving behavior settings and different CAV penetration rates by taking a typical four-lane intersection as an example. Results show that the optimal channelization and signal timing scheme need to be adjusted with the change of CAV penetration rate and CAV car following behavior. The increase of the CAV penetration rate and the decrease of CAV headway are both beneficial to the improvement of the intersection capacity. The increase in the intersection capacity is slightly larger when the headway of CAV is not affected by the type of vehicles ahead.
The characteristics of steam heat network in actual operation were tested using enthalpy drop method and surface heat flow method. Heat loss characteristics and composition characteristics of the heating network were analyzed. Results showed that operating conditions and steam state greatly influenced on the accuracy of enthalpy drop method in evaluating heat loss of steam heating pipe. The actual heat rate of heat supply network was 135.66 W/m2, of which the heat rate of pipeline insulation was 67.67 W/m2, accounting for about 49.88%. The steam flow resistance loss accounted for 14.98%, and the local heat loss of supports, elbows and traps accounted for 35.14%. The condensation loss coefficient of the heat supply network was defined based on the relationship between actual pipe loss and condensation loss of the heat supply network. Heat loss model was constructed based on heat loss of the heat supply network combined with actual measurement data. The condensation loss coefficient of the heat supply network was 0.16.
The classical BPR impedance function model was improved in order to more accurately calculate the impedance value of road traffic. The long short-term memory (LSTM) neural network was established to predict the positive and negative of the undetermined coefficient value in the improved function. The traffic data collected from the Shangtang Elevated to Zhonghe Elevated sections of Hangzhou City were used to verify the model. The results were compared with the traditional BPR impedance function method, the classic EMME/2 cone delay function, the BP neural network prediction method and the LSTM neural network prediction method. Results show that the improved model has higher accuracy and reliability under the premise that the data accuracy meets the requirements, indicating that the road impedance calculated by using the improved model can more realistically reflect the traffic operation condition of the road.
An approach based on baseband signals was proposed aiming at the range spread problem of multi-human target tracking in ultra wide band (UWB) radar. The RF echo was down-converted and decimated, and clutters were filtered out by moving target indication. The baseband CLEAN detection was applied to extract measurements. Then the initial state of the target was determined by clotting and jumping-window method. Joint probabilistic data association and Kalman filter were adopted for tracking. The experiment was conducted in three indoor environments. Results showed that the root mean square error (RMSE) of multi-target tracking was less than 0.26 m, while the data storage was reduced by 87.5% and the processing time for target detection was reduced by 39.7% compared with directly using RF echo.
A hardware efficient Radix-22 fast Fourier transform based on a single-path delay feedback architecture was designed aiming at the problem that the rotation factor module takes up more resources in FFT hardware implementation. The method of mixing CORDIC and MCM was adopted to design the rotation factor module to realize FFT architecture without conventional multiplier and DSP48E resource. MCM method based on ternary adders was used to minimize the number of adders for the W16 rotation factor module with less rotation angles. CORDIC method was adopted for the rotation factor modules of W64, W256 and W1 024 with more rotation angles. The real-time generation module of rotation angles was designed according to the mathematical law of rotation angles. The method does not need to occupy ROM resources and avoids complex addressing logic and timing control compared with the traditional CORDIC method. The designed 16 bit 64-point FFT improves the throughput per slice by 35.20% on Xilinx Virtex-7, the 256-point FFT improves by 30.37% on Virtex-5, and the 1 024-point FFT improves by 25.38% on Virtex-7 compared with other architectures.
An underwater image enhancement algorithm was proposed based on generative adversarial networks (GAN) and improved convolutional neural networks (CNN) in order to solve the problems of haze blurring and color distortion of underwater image. Generative adversarial network was used to synthesize underwater images to effectively expand the paired underwater data set. The underwater image was decomposed by multi-scale wavelet transform without losing the feature resolution. Then, combined with CNN, the compact learning method was used to extract features from multi-scale images, and skip connection was used to prevent gradient dispersion. Finally, the fog blur effect of the underwater image was resolved. In order to improve the color correction ability of the model and overcome the problem of color distortion of underwater images, the correlation between different channels of color images was learned by using the style cost function. Experimental results show that, in subjective visual and objective indicators, the proposed algorithm is superior to the contrast algorithm in comprehensive performance and robustness.
The components of the aircraft assembly line and the business logic were analyzed, and a digital twin based modeling framework for the aircraft assembly line was proposed, in order to realize the real-time interaction and deep integration of physical space and information space in the final assembly line of aircraft manufacturers. The modeling and the implementation of the key elements of assembly line were elaborated from the six-dimensional perspective of “human, machine, material, method, environment, and measurement”. Correspondingly, the technical process of the three-dimensional visualization and the information integration platform of the aircraft assembly line was proposed. A three-dimensional digital model of the workshop was established in CATIA. Then, a virtual space was built up based on the browser-based framework and WebGL technology and the real-time mapping of physical entities to virtual space was achieved by collecting process data from the shop-floor. An aircraft final assembly line was taken as an example, the synchronous mapping between the assembly workshop and virtual visualization, webServices service and information query service were realized, which improves the assembly efficiency and can provide scientific references for manual decision.
In order to promote the transformation of industrial cyber security defense mode from static passive defense to active defense, and alleviate the contradiction between the serious shortage of security experts and the sharp increase of cyber security demands, a cyber security active defense system framework of digital twin system was built from the perspective of bionics, and then five kinds of key technologies focusing on active defense were proposed based on the digital twin security brain (DTSB), including security data interaction and systems collaborative defense based on cloud-edge collaboration, cyber security active defense model of parallel digital twin system, situation awareness of parallel digital twin systems based on digital twin security brain, active defense and control technical framework for digital twin system based on immune system, and anti-attack intelligent recognition of digital twin system based on artificial intelligence. A case study of a digital twin workshop was given to demonstrate the successful application of digital twin cyber security in smart manufacturing.
A new railway sound barrier design was proposed aiming at the problems of poor ventilation performance and heavy weight of existing railway sound barriers. The structure was formed by a unit in which a spiral channel and a hollow channel were combined with each other in parallel. The coupling between the spiral channel and the hollow channel was realized based on the Fano resonance mechanism. The structure can effectively block the sound wave energy in a specific frequency band. This railway sound barrier had the advantages of thin thickness, small weight and good ventilation performance. The isolation effect of the meta unit structure on oblique incident sound waves was analyzed. The optimization analysis of the parameters of the meta unit structure was conducted, and the influence of the number of meta units on the sound insulation performance of the designed sound barrier was analyzed. The meta unit structure was verified by experiment. Results show that the structure can achieve wide frequency isolation of sound waves incident at different angles. The meta unit structure can effectively control the main frequency of wheel-rail noise after parameter optimization. Then the actual application requirements are met. The number of meta units will not affect the sound insulation performance of the sound barrier. The simulation results accorded with the experimental results.
Aiming at the problem that the Coulomb and viscous friction model cannot really reflect the nonlinear characteristics of friction in the process of robot motion, an improved Stribeck friction model was used to describe the joint friction, and a method of model parameter identification based on the combination of hybrid genetic algorithm and cosine trajectory was proposed. First, different cosine trajectories were used to excite the robot joints, and the friction torque of the joints was determined using the known dynamic equations of the robot, so as to establish the mapping relationship between the robot joint velocity and the joint friction torque. Second, the simulated annealing hybrid genetic algorithm was used to identify the friction parameters. Finally, a multi-joint series robot was used as the research object, and the friction parameter identification experiments were carried out to verify the effectiveness of the proposed method. Experimental results indicate that compared with the traditional Coulomb and viscous friction model, the improved Stribeck friction model can reduce the calculation error of the joint torque by 17.7% to 33.6%, and can further improve the accuracy of the robot dynamic model.
