In order to study the thermal dissipation performance of a liquid-cooled plate with parallel channels， the three-dimensional steady-state analysis was performed by using CFD method. The effects of coolant flow rate， channel width， depth and layout of enhanced heat transfer structure on the performances of a liquid-cooled plate were contrastively investigated， including the thermal dissipation， the uniform temperature and energy consumption. The results indicate that the thermal dissipation performance is improved by increasing the coolant flow rate， but excessive flow rate leads to increased energy consumption and limited improvement effect. Designs of decreasing channel width from the center to two sides， decreasing channel depth and adding enhanced heat transfer structure are all beneficial to the thermal dissipation and temperature uniformity of the liquid cooling system. In addition， the design of wholly added enhanced heat transfer structure （S1） reduces the average temperature and maximum temperature difference by 8.9 °C and 9.06 °C respectively， compared with the design of equivalent channels width （A5）. The conclusions provide a theoretical direction for structural design of battery thermal management system.

Topology identification of distribution network is an important work to ensure the safe and stable operation of distribution network. It could provide structure data for system power flow calculation， load capacity distribution， fault diagnosis， power network state estimation， which is the foundation of distribution network system analysis.The existing research of distribution network topology identification into two categories could be divided in this paper ： the first type of method is based on historical topology information， including matrix method， innovation graph method and optimal matching method. The second type of method is based on real-time measurement information， including correlation judgment method， signal injection method， linear programming method and machine learning method. Finally，the application range， main used data and characteristics of the existing methods was analyzed，the future research direction of distribution network topology identification was proposed.

Since the concept of resilience was introduced into the field of transportation， resilient transportation has received extensive attention from researchers in the transportation field. Road transportation infrastructures with good resilience can effectively deal with various natural and man-made disasters， and further meet the needs of efficient and safe transportation in the future. In order to clarify the current research status of road transportation infrastructure resilience around the world， the research results of road transportation infrastructure resilience were summarized from the definition， measurement methods， and resilience improvement technologies of road transportation infrastructure， and the future road transportation infrastructure resilience research was discussed especially the development direction and research focuses of resilience improvement technology. The analysis shows that the research on the resilience of road transportation infrastructure mainly focuses on the overall traffic network organization and planning level of the road network. The research on the resilience of the infrastructure structure is relatively scattered， and there is a lack of unified and comprehensive definitions and metrics. In addition， in the research on the catastrophic failure mechanism of structural resilience， there is a lack of a comprehensive research understanding of all elements and the coupling between structural systems， and it is difficult to reveal the chain process and catastrophe characteristics of catastrophic evolution. Therefore， the research on the resilience of future road transportation infrastructure should further reveal the theories and methods of catastrophe analysis of different types of facility structures， and establish a comprehensive and unified definition and measurement standard for the structural resilience of road transportation infrastructures. At the same time， from the perspectives of disaster monitoring， structural safety and resilience improvement， flexible operation and post-disaster recovery， etc.， more effective technologies for improving the resilience of road transportation infrastructure should be formed and furtherly promoted in the future.

In response to the problem of vibration caused by the large length/diameter ratio of the boring bar during deep hole boring， which affects the processing quality and efficiency， three vibration control methods， passive control， semi-active control， and active control， were summarized. The specific structures， vibration reduction mechanisms， characteristics， shortcomings， and development trends of the three methods have been sorted out. Comprehensive analysis shows that the structure， materials， and control methods of vibration damping boring bars are currently the focus of research. With the continuous development of structural design， material science， vibration reduction mechanism， control theory， big data， artificial intelligence and other technologies， the research on vibration damping boring bars is gradually becoming diversified， integrated， and intelligent. Meanwhile， intelligence is a new development direction for vibration damping boring bars.

Due to the complex characteristics of medical images， for example， the lesion region possesses an irregular shape and its scale can greatly vary， the intensity of surrounding tissues is inhomogeneous and the boundary is blurred， which reduce the accuracy of medical image segmentation， a medical image segmentation algorithm based on multi-scale context-aware and semantic adaptor is proposed. In order to improve the representation ability of feature learning， the multi-scale context-aware module is utilized to learn rich context information from multiple receptive fields， and dynamically assign the weight of semantic features at different scales according to the size of the target region. The multi-level semantic adaptor module is adopted to aggregate multi-level abstract semantic features and spatial details to refine the boundary of the target region and reduce the feature gaps between encoders and decoders. The algorithm proposed in this paper is compared with other algorithms quantitatively and qualitatively on three public medical image datasets of different modalities. The experimental results show that the proposed algorithm is superior to other algorithms in various complex scenarios of medical image segmentation tasks.

A heat pump air conditioning system performance simulation platform for electric vehicle is designed and built, which can be used to analyze the system performance in different working modes. The experimental results are used to verify the simulation model accuracy. The results show that the maximum error between experiment and simulation data of compressor power, heat exchange rate and system Coefficient of Performance (COP) is 4%~10.09% in different working modes. The compressor power gradually increases, the heat exchange rate gradually increases and the system COP gradually decreases with the increase of compressor speed in the cooling mode and heating mode. The influence of inlet air flow rate, inlet air temperature, inlet air mode and other factors of condenser and evaporator on the system performance in the cooling or heating mode are also researched in the paper. The results indicate that high air flow rate will improve the cooling capacity, and the effect of inlet air temperature is dependent on working condition, while the partial inlet air mode can save energy in heating mode.

CNC machine tools play an important role in the equipment manufacturing industry， and their reliability level has become the bottleneck restricting the development of the industry. Reliability modeling of NC machine tools is the basis of reliability engineering. A comprehensive review on the state of the reliability modeling technology research is given. Reliability models are mainly divided into four categories： reliability modeling method based on fault time data， reliability modeling method based on multi-source hierarchical information set， reliability modeling method based on performance degradation data and process reliability modeling method based on dynamic characteristic parameters. The research process and technical progress of various modeling methods are analyzed. On the basis of affirming the obvious progress made in the reliability modeling method and technology of CNC machine tools， this paper analyzes and points out the existing problems and shortcomings of the research work， and discusses the trends and hotspots of the reliability modeling research of CNC machine tools. Finally， the development trend of reliability modeling methods and technology of CNC machine tools is prospected from the perspective of reliability modeling development law、engineering application and industry demand.

This paper proposed a COVID-19 chest CT image segmentation method based on Federated Learning (FL) and blockchain to automatically segment the area in lung affected by COVID-19. Firstly, in the situation where the sample data of patients is limited and it is distributed in different institutions, which cannot be easily collected, we applied FL method. Then, we used blockchain network to replace the central server in FL to solve the “single point of failure” problem. Finally, we designed a lightweight separable convolution U-NET to reduce the cost of computation and time. Experimental results show that the method has good performance after training, and its dice metric can achieve 63.26%, which is helpful for diagnosis of COVID-19.

In the vehicle network environment， the traffic system will present the mixed coexistence of intelligent connected vehicles and traditional artificially driven vehicles for a long time. Focusing on the new mixed traffic flow in the intelligent connected traffic environment， the decision-making model of vehicle lane change behavior was established. The dynamic risk model during lane-changing is introduced to establish the interaction between autonomous vehicles and traditional vehicles in the mixed traffic flow. The lane-changing behavior of autonomous vehicles in mixed traffic flow is modelled based on game theory. Lane-changing behavior between autonomous vehicles is the nature of the non-cooperative game. Vehicles seek for a lane with better driving conditions by taking their own driving state as the game benefit. The simulation results show that the game lane-changing model has higher lane utilization and safety stability than the traditional gap threshold acceptance model.

The optical motion capture systems are extremely expensive for abnormal gait analysis, and the Kinect is a potential alternative equipment for its low?cost and convenience. The development of abnormal gait analysis is reviewed from four aspects: pathological characteristics of abnormal gait, abnormal gait data set, reliability of Kinect, and abnormal gait recognition method. Firstly, the abnormal gait and its pathological characteristics were summarized, and the common gait features and gait events in gait analysis were introduced. Then, the abnormal gait data sets collected by Kinect, wearable and pressure sensors are introduced. The feasibility of skeleton data collected by Kinect in gait analysis is discussed according to the existing experimental studies to verify the reliability of Kinect. Finally, the development of gait analysis is reviewed in detail from two aspects of abnormal gait feature extraction and abnormal gait classifier, and the shortcomings and development direction of current research are pointed out in practical application.

In order to prolong proton exchange membrane fuel cell （PEMFC） lifespan， it is necessary to design a state observer to monitor the internal states， and control the internal states at the expected level through feedback control. Since the anode hydrogen concentration directly determines the output performance of PEMFC， and circulation of anode hydrogen as well as nitrogen accumulation caused by diffusion across membrane leads to the difficulty of anode gas concentration estimation. Therefore， this paper focuses on the cutting-edge technology of PEMFC anode gas concentration estimation， and the existing problems as well as the future development trend of existing research are also described， hoping to make contributions to the research of gas， water and heat management for PEMFC.

In order to improve the power performance and economy of fuel cell vehicles， a fuel cell vehicle based on AVL-CRUISE was modeled， and the energy management strategy of fuzzy control based on Simulink was established. Then， the fixed gear was optimized into a two-speed AMT gearbox based on Isight/Cruise co-simulation. Simulation results show that the established fuzzy control energy management strategy is effective. Compared with the rule-based energy management strategy， the economy is improved by 16.4% and 8.5% respectively under NEDC and WLTP conditions. Compared with the fuel cell vehicle without optimized transmission ratio based on fuzzy control， the fuel cell vehicle with optimized transmission ratio based on fuzzy control has improved economy by 1.1% and 2.8% respectively under NEDC and WLTP conditions.

Aiming at the problem that ground points in point cloud data can affect the precision and speed of environment perception， an accurate segmentation method of ground point cloud based on plane fitting was proposed. Firstly， the scene point cloud was divided into several areas based on the projection distance. Secondly， according to the average height in the area， the divided ground points and the normal vector direction of the ground plane， the ground plane fitting point was determined， and the ground plane was fitted. Finally， according the distance from the point to the ground plane to achieve ground segmentation. Using the KITTI dataset and the collected point cloud data to compare the proposed algorithm with four algorithms： RANSAC、GPF、R-GPF and PatchWork， verify the effectiveness of area division， fitting point screening and ground plane normal vector direction screening for ground segmentation. The experimental results show that after the area division， the far-distance sparse ground can be divided； after the fitting point screening， the ground segmentation accuracy reaches 0.9417 under the condition of low iteration. After screening the normal vector direction of the ground plane， fitting the wall to the ground was avoided. The proposed method is better than the four compared algorithms in terms of F₁ score， recall rate and accuracy rate， and the speed can reach 42.78 Hz， which can divide the ground accurately and quickly.

In order to improve the hydrogen utilization rate of fuel cell engines， a fuel cell anode loop model was developed based on the AMESim platform,and completed model integration and calibration. Based on practical engineering experience, a new purging strategy that correlates anode loop temperature, pressure and reaction current was proposed. Using the anode loop model of the fuel cell engine, the parameter optimization boundary was calculated, and the optimization and fitting of the purge control parameters was completed. According to the simulation results， compared with the original purge strategy， the new strategy can make the hydrogen consumption rate decrease from 0.866 kg to 0.8 kg per 100 km under NEDC cyclic condition， and the theoretical economy can be improved by 7.5%. Then the new strategy was verified on a fuel cell vehicle， and the hydrogen consumption rate is 0.81 kg per 100 km， which is basically consistent with the simulation results. The driving range of the vehicle is increased from 484 km to 510 km， by 5.4%. And the effectiveness of model-based purge strategy optimization scheme is verified.

The objective of this study is to analyze the drying shrinkage behavior of cement-stabilized macadam（CSM） and improve its anti-cracking performance. In this paper， previous studies conducted on the mechanism of shrinkage and cracking in CSM and shrinkage mitigation strategies are reviewed， and summarized. The drying shrinkage performance of CSM materials is mainly studied through shrinkage mechanism and shrinkage prevention and improvement measures. As a cement-based material， the existing research on the drying shrinkage mechanism of CSM mainly refers to the available dry shrinkage mechanism of cement concrete. Four models are always used to characterize the shrinkage at different stages according to the change of relative humidity（RH） inside the mixture： capillary tension， surface free energy， disjoining pressure and interlayer water movement. The drying shrinkage mitigation strategies of CSM have achieved a series of results in recent years. According to the different mechanisms CSM shrinkage performance improvement methods can be classified into gradation， cement content， CSM molding method， and additive modification. The results show that reducing the amount of cement， adopting the skeleton of dense gradation， employing vibration stirring or vibration compaction， and adding rubber， steel slag， fiber or other modifiers can effectively reduce the drying shrinkage of CSM and improve its shrinkage performance and road serviceability. Additionally， the shrinkage mechanism， comprehensive optimization method， drying shrinkage measurement and evaluation method and shrinkage prediction model of CSM are assessed. The development trend and prospect of the shrinkage performance of CSM are generated.

In view of the current high-power fuel cell system, a high-power fuel cell test platform was designed and developed, and the performance test of the 120 kW fuel cell was completed through the design, selection, construction, and debugging of the system. The test platform integrates hydrogen flow， air flow and thermal management control systems, integrates and optimizes the layout between the various subsystems, and makes post-maintenance more convenient.At the same time, the test platform uses LABVIEW software to design the control interface of the upper computer, uses Simulink software to design the control program of the lower computer controller and writes the determined parameters of the components into the controller, and then realizes the communication through the CAN box between the upper computer and the controller. The online real-time control of the system by the host computer control interface, and can automatically optimize the operating parameters of each component according to the load change. By analyzing the test data collected by the test platform, it is possible to evaluate whether the fuel cell system meets the expected design requirements. The test platform has certain guiding significance for the research and production of high-power fuel cell systems and the development of fuel cell test platforms, and provides a guarantee for the development of fuel cell systems.

Using the deep learning methods can solve the model over-fitting caused by less observation data, and improve the prediction accuracy. This paper proposes spatio-temporal model of soil moisture prediction integrated with transfer learning. Firstly, the EAR5-land dataset is used as the source model. Then three-dimensional layer convolution is used to extract the spatial characteristics of the lag time of the soil moisture, and the long short-time memory network is integrated to extract the temporal characteristics. Third, the network model is pre-trained. Finally, the fine-tune method is applied to adjust the network parameters in the SMAP dataset for soil moisture prediction. The experimental results show that the proposed model has the better prediction results than the convolutional neural network, long short-term memory network and PredRNN. Meanwhile the method of transfer learning can improve the prediction accuracy.

For two lane traffic scenarios， a decision-making and optimization-based lane changing trajectory planning method for autonomous vehicles was proposed. Firstly， a risk assessment method based on the Bayesian probability theory was designed to obtain the conditional probability of the lane safety in the current scenario； then， a behavior decision-making method based on the safety utility was designed. According to the risk assessment Bayesian network and decision graph， the behavior decision of lane keeping or lane changing was made. An optimization-based trajectory planning method based on the nonlinear MPC was proposed at the trajectory planning layer， which imitates the excellent driver to give the weight coefficient of each optimized objective function to solve the optimal desired lane changing trajectory. At last，The effectiveness of the decision-making and trajectory planning method was verified by the simulation. The simulation results show that the risk assessment， behavior decision-making and optimization-based trajectory planning method can make the safe behavior decision and plan the optimal lane changing trajectory for autonomous vehicles in different risk scenarios， so that the autonomous vehicle can change the lane safely and quickly.

Text input is the most common interaction behavior in viture reality（VR） environment， and the mainstream text input is currently realized by laser pointing. However， the existing methods have many drawbacks， such as low efficiency， large jitter， and easy false trigger， which cannot be used to frequently input text in VR environment. Therefore， a novel text input method for VR environment is proposed. First，First, partition the keyboard, use the handle to select the area where the characters are located， and use the word disambiguation algorithm to realize text input in units of words； secondly， perform cluster analysis on the user's click coordinates to do one-key multi-word processing； Finally， three keyboard layouts that conform to user habits are designed， and the optimal layout is determined The experimental results show that the typing speed of the optimal layout is 13.44 WPM（Words Per Minute） with an accuracy of 92.26%， which is a significant improvement compared with other input methods.

According to the algorithm processing method, image dehazing technology can be divided into two categories. One is image enhancement algorithm, which involves histogram equalization, wavelet transform and Retinex algorithm based on color constancy theory. The other is the image restoration and defogging algorithm, which mainly includes the traditional multi-image restoration based on the characteristics of optical polarization and the single image restoration based on a priori assumption, and the emerging image restoration algorithm based on deep learning. In order to better study the image dehazing algorithm in the future, the main development process of image dehazing techniques was reviewed in this paper, the existing problems for the study on image dehazing algorithm was analyzed, and the development trend was tried to explore.

Based on the application scenarios of swarm intelligence in the field of UAVs， the application of swarm intelligence methods in the field of UAVs was reviewed. First， the recent application status of UAVs was reviewed， and the principles of swarm intelligence algorithms and examples of UAV applications were introduced. Second， the application scenarios of swarm intelligence in UAVs were divided into four parts： swarm intelligence-based UAV wireless communication， swarm intelligence-based UAV ad hoc network， swarm intelligence-based UAV trajectory planning， and swarm intelligence-based UAV intelligent decision-making. The progress of relevant research work for each part is introduced separately. Finally， a brief discussion is conducted on the development trend of swarm intelligence for UAVs.

The effects of fuel temperature， working load and wearing time on the lubrication performance of RP-3 jet fuel were studied on a high-frequency reciprocating rig. Further， different lubricity additives were used to improve the lubrication performance of RP-3 jet fuel. It is revealed that RP-3 jet fuel produces larger wear scar diameter than diesel fuel， and increasing working load and wearing time elevate the wear scar diameter. Increasing fuel temperature slightly reduces wear scar diameter， possibly because of the increased generation of anti-wearing products. Lubricity additives can dramatically enhance RP-3 jet fuel lubrication performance， but the improvement effects are not monotonic with the addition fraction. As the addition fraction exceeds a certain value， the lubrication performance of RP-3 jet fuel does not obviously change with further addition of lubricity additives.

In view of the deficiency of existing research on analyzing the resilience of urban rail transit network based on topological network efficiency， a resilience assessment method based on network performance response function is proposed， which is the weighted sum of OD passenger flow loss ratio and network service efficiency loss ratio. The Delphi-entropy weigh method is used to determine the comprehensive weight of two indicators， and a recovery optimization model with the maximum network resilience index is established， and the adaptive genetic algorithm is adopted to solve the developed model. Taking Xi'an rail transit network as an example， four hypothetical perturbance scenarios are proposed considering random attack and intentional attack. The differences of network resilience repair effects of target recovery strategy， random recovery strategy and preference recovery strategy under four hypothetical perturbance scenarios are compared and analyzed. The results show that the target recovery strategy has the best repair effect on rail transit network， followed by preference recovery strategy. Compared with random attack strategy， different recovery strategies have different network resilience under intentional attack strategy. When selecting the sequence of repairing damaged stations， we should not only consider the importance of damaged stations in network topology， but also consider the impact of passenger flow on network performance. Increasing the input of repair resources can shorten the recovery time and improve the repair efficiency， but the increase of repair resources is not proportional to the improvement of network resilience. The research conclusion can provide decision-making basis for the resilience assessment and emergency repair recovery of urban rail transit network.

To solve the problem of limited aperture of traditional uniform circular arrays， a co-prime circular array is designed， and on this basis， a two-dimensional direction of arrival estimation method based on phase mode excitation is studied. First， two uniform circular sub-arrays with a distance between array elements greater than half a wavelength are stacked to form a relatively prime circular array structure. Then， based on the phase mode transformation algorithm， the element space problem is transformed into the beam space. The algorithm combines the co-prime circular array into a Vandermonde structure similar to ULA and the Hermitian center symmetrical array flow pattern， reducing the dimensions of spectral peak searching， realizing array flow pattern through unitary transformation， reducing algorithm complexity. Theoretical analysis and simulation experiments show that under the same conditions， compared with the uniform circular array MUSIC algorithm， the proposed phase mode excitation MUSIC algorithm for the relatively prime circular array reduces the running time from 1.825 s to 0.622 s， increases the array aperture MN/（M+N-2） times， improves the DOA estimation accuracy and the real-time performance of the system， and it has a higher resolution for the estimation of similar sources.

As a new generation of automotive materials， smart materials provide new research directions and design ideas for the realization of automotive lightweight and intelligent design goals. In order to further promote the research process of smart materials for automobiles， this paper reviews piezoelectric materials， magnetorheological materials and shape memory alloys respectively， describes the special properties of various materials， and systematically summarizes domestic and foreign research achievements in important research fields such as automotive energy recovery， structural vibration suppression， sensors， actuators， and safety protection. Finally， the challenges of commercialization of smart materials for automobiles are analyzed， and the directions of future research are pointed out.

A long-term and short-term memory neural network （LSTM） based on particle swarm optimization （PSO） was proposed to predict the life of PEMFC. First， the degradation mechanism of PEMFC was analyzed. Then， the voltage degradation prediction model was established by using LSTM neural network and the Dropout layer was used to prevent overfitting to improve the generalization ability of the model. In addition， PSO was used to optimize the learning rate and Dropout rate in LSTM to improve the prediction effect. Finally,the actual aging data of IEEE 2014 Data Challenge Data fuel cell were used to verify. The results show that this method can accurately predict the degradation of fuel cells， and the prediction accuracy is improved by 50% compared with the traditional LSTM.

In order to accurately and reasonably predict the future trajectories of vehicles and understand the changes of surrounding traffic flow, a trajectory prediction method combined with high definition map in graph attention mode was proposed. The encoder-decoder framework based on LSTM network was designed, and the model structure with vehicle historical status and high-precision map information as input was established. A graph query mechanism combining local and global features of vehicles was proposed to output vehicle prediction trajectory. The results of experiments carried out on the nuScenes dataset show that the comprehensive prediction performance of our model is better than other state-of-the-art methods， such as Traj++， CoverNet， etc.， and it has good anti-interference.

Aiming at the problem of inconsistent feature distribution of rolling bearing vibration data collected under variable operating conditions and difficulty in obtaining the labels of the samples to be identified， a sub-domain adaptive deep transfer learning fault diagnosis method was proposed. Firstly， to make full use of the image feature extraction capabilities of the convolutional neural network （CNN）， the rolling bearing vibration signal was used to generate an image data set using continuous wavelet transform （CWT）.Secondly， the common feature extraction of the source domain and the target domain adopted the ResNet-50 model structure of improved image set pre-training， and the sub-domain adaptive metric introduced the local maximum mean discrepancy （LMMD） criterion. This metric is used for sub-domain adaptation by calculating pseudo-labels in the target domain to match the conditional distribution distance， thereby reducing the difference in the distribution of sub-categories of faults under different working conditions and improving the accuracy of model diagnosis. Finally， experiments on two public variable-condition rolling bearing fault data sets verify that the proposed method has an average recognition accuracy of about 99%. Compared with the results of different transfer learning methods， the effectiveness and superiority of the proposed method are demonstrated.

A new method for Chinese named entity recognition at word level based on transformer and hidden Markov model is proposed. The position coding calculation function of transformer model is improved， so that the modified position coding function can express the relative position information and directivity between characters. The character sequence encoded by transformer model is used to calculate the transfer matrix and emission matrix， and a hidden Markov model is established to generate a group of named entity soft labels. The soft label generated by hidden Markov model is brought into Bert-NER model， the divergence loss function is used to update the parameters of Bert-NER model， and the final named entity strong label is output to find the named entity. Through comparative experiments， the F₁ value of the proposed method in Chinese cluster-2020 data set and Weibo data set reaches 75.11% and 68%， which improves the effect of Chinese named entity recognition.

As the society emphasizes on the life safety of operators and the standard of machinery performance requirements for construction， engineering vehicles are developing in the direction of autonomy， efficiency and reliability. In order to realize the automatic transfer and operation of unmanned engineering vehicles， this paper systematically summarizes the relevant technologies at home and abroad， and analyzes the research progress of key technologies of unmanned engineering vehicles in detail in terms of environment perception， motion planning， engineering operation and condition monitoring， etc. It points out that the technologies of unstructured environment identification， path planning and trajectory tracking of vehicles with variable body structure and automated operation still need to be broken through， and proposes the adoption of mechanism/structure optimization design， advanced communication means， machine learning and digital twin， etc.， which is conducive to promoting the development of key technologies of unmanned engineering vehicles.

Most fault diagnosis algorithms for rotating machinery are labeled and need to be set manually， an unsupervised fault diagnosis algorithm with adaptive parameters for unlabeled fault data by introducing twice anomaly recognitions and clustering algorithms was proposed.The method extracts and selects signal features by improving empirical wavelet transform and Laplace score algorithm, and adopts the unsupervised method of quadratic anomaly identification combined with improved fuzzy C-means clustering for fault identification. Through the verification of the fault data of the rotor system of the electric spindle, the diagnostic accuracy of the proposed method can reach 93%. Compared with the traditional unsupervised diagnostic method, it has good accuracy and robustness.

A distributed adaptive sliding mode time-varying formation control scheme based on neural network state observer was proposed to solve the problem of unmeasurable information of multi-quadrotor unmanned aerial vehicles （UAVs） system in directed switching communication network. Firstly， by designing a distributed time-varying formation control protocol， UAVS can communicate with each other only by the status information of their neighbors， which is free from the dependence on global information. Due to the underdrive characteristic， a position-assisted controller was designed to solve the target information of two attitude angles and control thrust. At the same time， a neural network state observer was designed to observe the unmeasurable information of the system， and the observed values were feedback to the adaptive sliding mode controller in real time， which improves the robustness of the UAV system. The quadrotor UAV formation system is analyzed by Lyapunov's theorem， and it proved that the multi-UAV formation error is bounded and converges to near zero. The simulation results show that the proposed control method can realize the time-varying formation control of the multi-UAV system and verify the validity of the theoretical results.

Firstly， the definition of mobility was discussed， and then the main existing ground vehicle mobility evaluation methods， i.e.， empirical model， semi-empirical model， numerical simulation and machine learning， were analysed and summarized comprehensively， also the advantages and disadvantages of each method were compared. In order to describe the vehicle mobility completely， application of these methods are discussed，such as military vehicles， sea-floor operation， planetary exploration and agricultural vehicles. Finally， according to the problems existing in the vehicle mobility evaluation methods， this paper proposed some key technologies and exploratory research directions from real-time evaluation of vehicle mobility and real-time terrain perception， path planning on deformed terrain and autonomous mobility evaluation for unmanned systems， so as to provide a beneficial reference to the development of vehicle mobility elevation methods.

The task of object detection plays an important role in the safe driving of driverless vehicles. Currently, the object detection technology of environment percept is mostly one-class object detection or all the objects in an image are listed as the target to be detected. Numerous studies have not yet focused on object division and detection of the objects in front of the vehicle. To solve the above problems, in this paper, the objects to be detected in front of vehicles are divided into two categories. One is the dynamic targets with high risk and displacement at any time, including four-wheel vehicle, two-wheel vehicle and people. The other one is the static targets with less danger and no displacement, including traffic lights and traffic signs. For the dynamic multiple objects in front of the vehicle, an improved algorithm of object detection based on YOLOv3 is proposed, which can be transplanted to the embedded system. To overcome the shortcoming of the original YOLOv3 algorithm, that it is difficult to get real-time detection in the embedded terminal, the original backbone network Darknet-53 was replaced with MobileNetV2 to extract features, adding Group Normalization operation in the training process and using Adam as optimizer. The extracted BDD100K dataset is used for training. The model is tested with BDD100k partial dataset not involved in training and Team_test dataset produced by our research group. The results show that compared with original YOLOv3, the missing rate (MR) of the algorithm in this paper can be kept within 5%, and based on the increase of 0.020 in mAP, comparing with the basic model of YOLOv3, the parameters of YOLOv3-MobileNetV2 model are reduced by about 89%, the Inference Time is reduced by about 70% under the CPU.

Combined with the temperature change and icing in the process of fuel cell cold start， a control oriented third-order fuel cell cold start model was established. Aiming at the unmeasurable ice volume fraction of cathode and anode， an ice volume fraction estimation method based on extended state observer was proposed. On this basis， according to the characteristics of constraint and coupling nonlinearity in the cold start process of fuel cell， an optimal control method of fuel cell cold start system based on nonlinear model predictive control was proposed， which realizes the double optimization objectives of improving the rapidity of cold start system and reducing hydrogen consumption. Finally， simulation experiments verify the effectiveness of the designed optimal control system of cold start system.

This paper presents an on-line trajectory planning method based on the reinforcement learning for driving the quadrotor to its destination accurately and suppressing the swing motion of the slung-payload effectively. To deal with the unknown external disturbances, the desired trajectory of the Unmanned Aerial Vehicle (UAV) is divided into two parts: the positioning trajectory planning and the disturbance rejection trajectory planning. The positioning trajectory planning can be designed in advance to guide the UAV to reach the desired position, and the disturbance rejection trajectory planning can compensate the unknown external disturbances based on the reinforcement learning strategy and suppress the swing motion of the slung-payload simultenously. The Lyapunov based stability analysis is employed to prove the stability of the closed-loop system, the convergence of the UAV′s position and the swing motion of the slung-payload. Finally, real-time comparing experiments are performed to verify the effectiveness of the proposed trajectory generation method and its robustness to external disturbances and variation of the mass of the slung-payload.

The difficulties of failure data collection and operation data changeability hinder the application of fault diagnosis methods to turrets. Hence， an anomaly detection method using non-failure data and considering the change of working conditions was proposed for detecting turrets’ anomaly state during operation. The method studied the judgment principle of abnormal data through the multivariate Gaussian distribution（MGD） and the deviation characteristic associated with working conditions. First， the key working conditions and signal characteristics in different turret working processes were determined through statistical analysis. Second， some methods like linear regression， information gain， and generalized regression neural network were selected to model their relationships， respectively. Following that， the deviation of observation from the given signal characteristics is calculated. Finally， the operation data from turret normal state were used to train the model. Many experiments under different working conditions and abnormal simulation were conducted to verify that the proposed model can eliminate the influence of working conditions on abnormal judgment compared to the traditional MGD model.

The two-stage object detection algorithm has a slow inference speed， meanwhile， the light-weighted model has a poor performance on small datasets. In view of reasons mentioned above， an improved multi-object detection algorithm based on the algorithm YOLOv4 is proposed. Taking dishes detection as an example， multi-object detection is used to detect and classify the empty dishes. Similar characteristics of the plates' edge are key identity to each completed plate. In order to preserve the salient features， the attention mechanism and pooling methods are added. Furthermore， a light-weighted network can speed up the inference speed and a multi-scale fusion method is able to improve the precision of the model. It turns out that the algorithm proposed improved 4.25% than previous work. According to the comparison with the classic detectors such as Faster RCNN， the FPS is 8-9 times of the latter one. The algorithm in this paper improves the reduced accuracy that the light-weighted models caused， and a more convenient and quick deployment of mobile or embedded devices are able to complete tasks.

Most of the research on text detection has been conducted on natural scene datasets， few on such specific scene. And the existing models are not good enough for text detection on traditional Chinese newspaper. In order to solve this problem， a segmentation-based text detector for traditional Chinese newspaper is proposed in this paper. The model uses Resnet50 and FPN as feature extraction network， employing a segmentation instance scaling and extension algorithm to generate the binary map for predicting text boxes. And the methods of surrounding filling and loop detection plus region coverage are proposed to enhance the detection effect. In addition， a traditional Chinese newspaper dataset is built to satisfy the research needs. The experimental results of this model on traditional Chinese newspaper dataset are around 0.9 and are improved by 5% to 7% compared with DBNet， which indicates that the model is effective and accurate for text detection on traditional Chinese newspaper.

In order to ensure the surface quality and machining stability of turning， the real-time and accurate monitoring of turning tool wear state is realized. A tool wear state prediction model based on wavelet threshold denoising， Long and Short Time Memory（LSTM） network and Particle Swarm Optimization（PSO） was proposed. The improved polynomial threshold function was used to denoise the tool acceleration vibration signal， and the high quality signal input sample was constructed. The wear values of the tool rear face were predicted and the wear states were classified by training the LSTM network. The proposed PSO-LSTM model is superior to the unoptimized LSTM network in terms of prediction and classification accuracy by using PSO.