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.

Pseudo excitation method of stationary random vibration in time domain is summarized and analyzed， the pseudo excitation method of non-stationary random vibration in space domain is established. Based on the transformation relationship of variables and their derivatives between time domain and space domain， the vehicle vibration model in space domain is derived from the vehicle vibration model in time domain， and the spatial frequency response characteristics of vehicle vibration and acceleration in time domain are built. According to the two degree of freedom single wheel model of vehicle vibration， the expressions of virtual road roughness， virtual vibration response variables and statistical characteristics of vibration response variables are determined. The parameters of a car and a typical driving condition are used to analyze the variation law of statistical characteristics of vibration response variables. The results show that the method presented is suitable for solving the statistical characteristics of the vehicle non-stationary random vibration response variables.

In this paper, the cab behind engine vehicles are taken as the research object, and the entire aerodynamic performance optimization is carried out. The optimization is composed of two steps, the initial shape optimization of the molding surface in the early modeling stage and the aerodynamic optimization in the later vehicle stage. The optimization process is based on Hyperstudy software to realize the shape parameterization, and CFD software simulation is used to evaluate the aerodynamic performance. An automatic process was built to realize the aerodynamic shape optimization of cab behind engine vehicles with these two kinds of software. Eight design variables that need to be considered (i.e. sunshield rotation, hood deformation, container deformation, fairing optimization, front wheel baffle optimization, side deflector optimization, side skirt at the container, and rear guide device of the container) were determined for aerodynamic optimization design. Finally, the drag coefficient of the vehicle was reduced from 0.432 to 0.387, compared with that of the vehicle base model reduced 45 counts by 10.4%.

In this paper, the daily operation data of a pure electric shared vehicles was taken as the research object, and 25 driving behavior characteristic parameters of five categories were determined. Vehicles′ data were divided into starting, parking and running three phases. The correlation among 25 characteristic parameters was calculated by the maximum information coefficient (MIC) algorithm, which is suitable for correlation analysis of high?dimensional big data. The effect law of driving behavior was studied in energy consumption. The characteristic parameters that have significant influence on total energy consumption and 100 km energy consumption at different vehicle service stages were identified. The results show that there are differences in driving behavior characteristic parameters that have significant effects on energy consumption in different stages. Through analyzing the influence rules of different driving behavior characteristic parameters, optimization suggestions for energy saving and consumption reduction driving behaviors in different stages are put forward.

In this paper， a method is presented to identify parameters of Bouc-Wen hysteresis model of magnetorheological damper. Based on the new meta-heuristic algorithm—Seagull optimization algorithm， the model parameters describing the nonlinear mechanical properties of the magnetorheological damper were fitted and identified. The opposition-based learning model based on the principle of refraction was introduced to improve the problem of falling into local optimal of the seagull optimization algorithm. The calculation results of the parameter identification method are accurate， simple and robust. The relationship between the model parameters and the control current is established by fitting and calculating the test data of the magnetorheological damper under different operating conditions. A general model of the magnetorheological damper including the control current is established and provided an accurate mechanical model for semi-active seat suspension control system.

The pressure fluctuation of the hydraulic variable valve actuation system may be produced in the actual operation， which will affect the performance of the variable valve actuation. In this paper， the reason and trend of pressure fluctuation are analyzed by combining experiment and simulation， and effective measures are taken to reduce pressure fluctuation. It is found that the key parameters such as valve spring stiffness， valve piston mass and the diameter size of the thin-walled hole，all have effect on the system pressure fluctuation. The research shows that when the valve-seating velocity is less than 0.5 m/s， the pressure fluctuation can be effectively reduced by choosing a larger thin-walled hole. The pressure fluctuation amplitude can be reduced by increasing the valve spring stiffness. The pressure fluctuation can be decreased by reducing the valve piston mass. When the diameter of valve piston increases， the pressure fluctuation amplitude decreases significantly， which can be reduced by 2.719 MPa at most. By reducing the total hydraulic fluid volume of the system can reduce the number of pressure fluctuations.

In order to reduce the fluctuation of the automobile engine timing chain transmission system and increase the transmission load of the engine， this paper proposes a new type of dual-phase timing bush chain transmission system. This new type of dual-phase chain transmission system effectively reduces the vibration caused by the contact between the sleeve and the sprocket by changing the position of each phase sprocket， and suppresses the transmission noise to a certain extent. Through dynamic simulation analysis， some dynamic parameters are selected for comparison with the single-phase system. The results show that the transmission performance of the dual-phase timing chain transmission system is better than that of the single-phase timing system. The wear failure study provides a scientific basis for further analysis of the service life of the timing bushing chain drive system.

According to the nature of the circle involute， a new method of variable thickness scroll composed of different circle involutes is proposed. The method of generating the profile is discussed in detail. The basic conditions for establishing the baseline are expounded. The general equations of the profile are derived. Two methods for calculating the volume are given. The range of control variables for correctly generating the inner and outer profiles is discussed. According to different requirements， a series of geometric models of circle involute variable thickness scroll can be quickly established， which broadens the research on variable thickness profile. The results show that， compared with the traditional method of establishing the variable thickness scroll geometry model， the new method not only has the advantages of traditional variable thickness scroll， but also reduces the calculation of establishing variable cross-section scroll， reduces the machining difficulty of variable cross-section scroll to a certain extent， and expands the practical application of variable thickness scroll compress or expander.

A method was proposed to simulate the micro pitting on gear tooth surface considering the lubrication contact pair based on the contour integral， and the damage characteristics of micro pitting， i.e.， the variation law of stress intensity factor and the final pitting morphology of tooth surface were studied by establishing its two-dimensional finite element numerical model. The contact loads， interactions and elasto-hydro dynamic lubrication （EHL） conditions of the gears were considered in the lubrication contact model. The stress intensity factors K_I，K_II and the propagation angle θ_C at the crack tip were calculated by ABAQUS. Then the equivalent stress intensity factor K_σ was calculated according to the maximum tangential stress （MTS） criterion and used as the criterion for crack propagation. The influence of different lubrication conditions on the formation of micro pitting was further discussed. The results indicate that K_I and K_σ are increasing rapidly with larger crack length and the size of pitting is smaller when the gears are lubricated adequately. By contrast， K_I and K_σ are changing slowly with crack length and the size of pitting is larger when the gears are sparely lubricated. The research results can provide theoretical support for contact fatigue life analysis， on-line wear monitoring and meshing stiffness calculation of micro-pitting gears.

To verify the design parameters of each component more accurately before the actual excavator production， an analysis method for predicting the digging performance of the excavator based on Discrete Element Method and Multi-body Dynamics （DEM-MBD） is proposed. A series of experiments are carried out on a domestic hydraulic excavator under flat and sloping conditions， and the load spectrum data of corresponding conditions are obtained. According to the digging process of the excavator and the shape of the material heap during the experiment， the co-simulation was carried out using the dynamic software RECURDYN and discrete element software EDEM to restore the operating state of the excavator during the excavation. By comparing the data derived from the load spectrum collected in the experiment with the data from the co-simulation， the results show that the correlation coefficients between the hydraulic cylinder thrust and the digging resistance data obtained by the co-simulation and the measured data are all greater than 0.9， which proves that the DEM-MBD co-simulation method is suitable for performance analysis of excavator.

In view of the fact that carbon fiber reinforced polymer（CFRP） partially wrapped steel tube can realize further lightweight on the premise of improving the performance of steel tube， the transverse impact simulation of CFRP reinforced steel tube was carried out， and the influence of strengthening form and size of CFRP wrapped steel tube on the transverse impact performance of steel tube was obtained. Based on the Hashin composite failure criterion and Ductile metal failure criterion， the simulation model of the drop hammer transverse impact of CFRP fully wrapped steel tube was established， and the validity of the model was verified. The relationship between the size of CFRP and the impact strength of steel tube was studied when CFRP was pasted axially on the steel tube surface， circumferentially on the impact side and circumferentially on the back of the impact side. The transverse impact simulation of drop hammer was carried out and the distribution of steel tube damage under impact was studied，the maximum displacement of the strengthened steel tube and the energy absorption of CFRP were compared under different bonding lengths. The results show that the strengthening influence of CFRP axial bonding size is more obvious；the impact resistance of the steel tube strengthened by CFRP is basically the same as that of the steel tube fully wrapped by CFRP， when the length of CFRP is pasted axially half of the span and the full length of the steel tube circumference or the length of CFRP is pasted axially 87.5% of the circumference and full length of span.

Aiming at the identification of hysteresis characteristic parameter of composite leaf springs， a linear algorithm was proposed based on the stochastic gradient decent method and the recursive formular was deduced. From the two dimensions of identification accuracy and calculation time， the linear identification method and the improved firefly algorithm in the existing work are compared and verified. On this basis， a momentum term is introduced to improve the linear identification method. The results show that the improved algorithm accelerates the convergence time and reduces the convergence error. Finally， the improved algorithm was applicated to identify the hysteresis characteristics of a certain composite leaf spring and good results were obtained. The results show that the linear algorithm， which can get results in 2.5 ms， is more suitable for hysteresis characteristic parameter identification of composite leaf springs than nonlinear algorithm and can be used in online parameter identification and digital twin simulation.

This paper analyzes the traffic characteristics of motor vehicle drivers and pedestrians in the crosswalk traffic conflict area. The drivers are divided into two types： comity and impoliteness， and the ratio of motor vehicle comity are introduced. At the same time， on the basis of the Non-game theory cellular automata （NCA） model， the dynamic game update rules are introduced to establish the Game theory cellular automata （GCA） model， which overcomes the need to describe the two parties through the spatial position relationship in NCA model. Comprehensive analysis shows that the GCA model is more realistic than the NCA model， and can better simulate the traffic conflict between drivers and pedestrians at crosswalks without signal lights.

First， using quadUP element with water-soil dynamic coupling and a multi-yield plastic constitutive model that can accurately describe the liquefaction characteristics of sand and the accumulation of shear deformation， a two-dimensional finite element numerical model of pile-soil dynamic interaction in liquefying mild inclined sloping ground is established. Then， this model is used to obtain the dynamic p-y curve under sinusoidal load. The slope angles and the depth are selected as the control indexes to establish the formula of dynamic p-y curve of saturated sand. Third， combined with the nonlinear Winkler foundation beam model， the pseudo-static analysis model of pile-soil interaction in liquefiable mild inclined ground was established. Based on the finite element analysis results， the validation of the pseudo-static analysis method was achieved before presentation of a detailed parametric study on the pile dynamic response. The results show that the displacement of pile decreases with the increase of the modulus of pile foundation and the rigidity of pile bottom connection. When the bending stiffness of pile foundation is constant， the bending moment and displacement of pile increase significantly with the increase of pile diameter.

Liquefied natural gas （LNG） storage tank belongs to urban lifeline project， adopting seismic isolation technology to ensure its seismic safety is the development trend in the future. but the seismic isolation is a problem faced by LNG storage tank built in the coastal soft site. To solve this problem， a new combined sliding isolation scheme， which can be self-centering， energy dissipation and resistance to large deformation is proposed， and a combined isolation layer with damper reaction wall is developed. In order to verify the control effect of the new combined sliding seismic isolation scheme， taking outer tank shearing force， impulse shearing force， single pile shearing force， sloshing wave height and the horizontal displacement of seismic isolation layer as the main performance indicators， the dynamic response of lead rubber bearing isolation tank， friction pendulum isolation tank and combined isolation tank under earthquake were analyzed and compared respectively. The results show that under the action of operating basis earthquake（OBE） earthquake， the above three isolation schemes all have obvious isolation effect （except sloshing wave height）， but under the safe shutdown earthquake（SSE） earthquake， only the combined sliding seismic isolation scheme can resist the large horizontal displacement of the isolation layer under large earthquake， and it has better control effect on the sloshing wave height of the LNG storage tank both under OBE and SSE earthquakes.

In order to better improve the safety level of hanging basket system of cantilever-construction and clarify the risk factors contained in the construction process of hanging basket system and reasonably avoid risks， in this paper， a new Bayesian network-based dynamic risk analysis method for hanging basket system of suspended casting multi-tower cable-stayed bridge is proposed. Also， taking the PAIRA bridge in Bangladesh as an example， the relationship between risk probability and risk intensity matrix of hanging basket system is discussed， the criterion of risk rating is clarified. The prediction results can provide a new idea and reference for the risk decision and the selection of countermeasures of hanging basket system in cantilever-construction bridge.

Estimating the frequency of each data （frequency estimation） is a critical task in stream data management. Sketch is one of the most memory-efficient structure for solving this kind of task. Existing multi-layer sketches often use counters of the same size to record the highest layer that data could be reached. However， many counters represent very small values， which is a waste of memory. On the other hand， multi-layer static sketches cause the problem of unreasonable memory allocation due to the change of data distribution. To address the issues， this paper proposes an elastic hierarchical structure， named Rhombus sketch. Rhombus sketch can assign an appropriate number of counters to each data on demand， and adjust memory allocation by estimating the data distribution. Experimental results show that， compared with other typical sketches， the accuracy of Rhombus sketch is greatly improved， especially for tight memory.

In order to solve the problems of high computational difficulty and low classification accuracy when ensemble learning algorithm is used to process multi-source data sets， a multi-source data selective integration algorithm based on frequent pattern tree is proposed. By comparing the difference between data source data and truth value， the error of multi-source data is judged by using Raida criterion， and all frequent patterns of multi-source data are extracted and converted into compressed form. The frequent pattern tree structure is created， and the concept of dynamic selection is integrated. The current data is used to test the extent to which an instance belongs to a subset of wrongly classified data by adjusting the difference and accuracy of the base classifiers by weighted harmonic average， the frequent itemsets are normalized， and finally the selective ensemble of multi-source data is completed by combining the base classifiers with high precision and big difference with other base classifiers. Simulation results show that the proposed ensemble learning algorithm has better generalization performance and efficiency， and has high classification accuracy.

With the gradual increase in the amount of books borrowed， domestic libraries have basically applied the self-help borrowing and returning system， but there are still many abnormal phenomena in the process of borrowing and returning books， such as the low accuracy of the identification of many books and the inaccurate identification of the number of books. In this paper， a multi-book recognition algorithm based on enhanced edge detection is proposed. This algorithm uses the convolution kernel of 2×2 to convolve with the edge detection image， so as to enhance the line information in the image， detect the number of book edges in the image， and assist the detection of the spine edge to carry out multi-book recognition. Among them， the accuracy of single book recognition is 94.0%， that of multi-book recognition is 92.9%， and that of comprehensive recognition is 93.8%. This method has high practicability and can be used to complete multi-book recognition task.

In order to address the problem that most people are psychologically in sub-healthy state， a bidirectional Long Short-Term Memory （LSTM） network based on spatio-temporal attention is designed to realize the feature extraction and micro-expression recognition for finding out the emotion that people try to conceal. The network consists of three main parts： the bidirectional LSTM module， the spatial attention module and the temporal attention module. Using the micro-expression video as input， the learning network can selectively focus on action units that are significantly different for each frame， and pay different levels of attention to different frames. A new regularized cross-entropy loss function is also designed to further optimize the network， taking into account the correlation between modules. Finally， comparative experiments were performed on CASME， CASMEⅡ， CAS（ME）²， and SAMM. The experimental results show that the proposed method can improve the accuracy of Micro-Expression recognition and is superior to other methods.

Because the current existing methods fail to consider the problem of text information feature extraction， the average relevance， average excellence rate and new word search accuracy rate decrease. In order to effectively solve the above problems， a text information similarity search algorithm based on segment estimation and PageRank is proposed. First， the segmentation estimation method is used to extract text features， and the PageRank value is taken as the criterion for preliminary classification of text information. Then， the similarity of different features of text information is calculated， the text information similarity is sorted. Finally， the similarity is searched based on the relevance between text information， effectively realizing the similarity search of text information. Simulation results show that the proposed algorithm can comprehensively improve the average correlation， average excellence rate and new word search accuracy， and the highest new word search accuracy reaches 98.98%， which indicates that the algorithm can obtain high quality and high stability search results.

The panoramic mosaic technology can get a larger field of observation view， based on the partial medical images obtained by ultrasonic detection device. Through stitching and fusing， it can provide more intuitive and comprehensive information for clinical diagnosis and doctor-patient communication. However， due to the influence of objective factors such as low precision image and more noise caused by medical equipment， the results of traditional panoramic mosaic method are not ideal. To solve the problems above， a template fusion method based on improved k-MEANS clustering algorithm for ultrasonic medical image mosaic is proposed. Compared with the traditional template stitching algorithm， this method can self-organize the region with dense features into flexible template， and improve the speed of adjacent image matching by using heuristic window algorithm. The experimental results show that the template detection rate and average template density are improved， the image registration accuracy is improved， and the method has high robustness and real-time performance.

Current semantic segmentation methods can obtain the outline of pedestrians， but when pedestrians block each other， the information such as the number， height and central position of the pedestrians in the figure can not be directly obtained. To solve this problem， we propose the G-UNet model algorithm that， in addition to the semantic segmentation backbone， a Gaussian ellipse density kernel detection branch of the pedestrian area is added， so that the center position， height and width of the pedestrian are detected respectively by the maximum point of the kernel， vertical axis and horizontal axis scale. The uniqueness of the maximum point of density kernel solves the detection problem of pedestrian occlusion. Besides， UNet stiffly concatenates the features of the bottom layer and the top layer in a spatial symmetric way so that 50% fixed errors are directly transmitted to the bottom layer. Then， we propose a trainable Soft Connection to obtain the optimal error distribution propagation method. Finally， to solve the problem that the value of the traditional loss function is proportional to the demarcated area of pedestrian， which makes small-scale pedestrians easy to be undetected， an Objective Enhanced Loss was proposed to improve the ability of detecting small-scale pedestrians of network. In the self-built pedestrian segmentation dataset， the experimental results show that the proposed method is effective and superior to other methods.

In order to solve the problem of large conversion error in traditional power conversion system， a design of high-speed motor power conversion system based on intelligent control algorithm is proposed. On the basis of traditional hardware system， combined with the operation principle of fuzzy neural network in intelligent control algorithm， the controller equipment is designed to realize the modification of bidirectional buck converter， phase-shift reactor， rectifier and inverter， and adjust the hardware connection circuit of the system. Taking the high-speed motor as the research object， its basic working characteristics are analyzed. A fuzzy neural network control model is established. In this model， the instantaneous power of high-speed motor is calculated based on the working data of high-speed motor， and the power conversion function of high-speed motor is realized by the way of generating reverse current or torque ripple signal gain. The experimental results show that， compared with the traditional power conversion system， the average conversion error of the designed conversion system is reduced by 1.7 kW， and the application of the power conversion system can realize the effective control of high-speed motor power， which has a high utilization value in practical application.

In response to the demand， application and development of integrated technology for reconnaissance and jamming， in order to solve the problem of simultaneous detection and jamming with large bandwidth， this article proposes an integrated design for reconnaissance and jamming based on Modulation Wideband Converter （MWC） technology. The design includes a multi-channel compression multiplexing receiving module， a compressed signal jamming generating module， a multi-channel random demixing module， and a jamming transmitting module. The design firstly uses the MWC technology to realize random mixing of the signals received by the multi-channel， and realizes the single-channel analog-to-digital converter （ADC） reception after multiplexing to get the composite compressed signal. Then a specified pattern of jamming is added to the compressed signal to generate a compressed jamming signal. Next， the random demixing of the composite compressed signal of a single digital-to-analog converter （DAC） is completed. Finally， the signal is multiplexed. The simulation results show that the design can achieve large bandwidth simultaneous reconnaissance and simultaneous jamming with fewer ADC and DAC channels， and verify the influence of the signal-to-noise ratio and the correlation of the mixing sequence on the jamming signal.

The parameters extraction of the target echo signal of linear frequency-modulated continuous-wave （LFMCW） lidar is more difficult under low signal-to-noise ratio （SNR）. To solve this problem， a signal time delay estimation method based on synchroextracting S transform based on the Hough transform （SESTH） is proposed. Firstly， integrating Hough transform， the SESTH model of a Chirp signal was constructed. Then， the parameter estimation models of initial frequency and frequency modulation rate of the Chirp signal were derived. Finally， using the time-frequency characteristics of the LFMCW lidar transmitted signal and echo signal， the time delay estimation model of a Chirp signal was constructed， and the relative distance between a target and the lidar was effectively calculated. In order to verify the effectiveness of the algorithm， simulation comparative analysis were carried out. The results show that the minimum relative error value of the proposed SESTH is 3.191×10^-6 when the time delay is 4.34 μs； the minimum relative root mean square error value is 0.014 65 in the case of \mathrm{S}\mathrm{N}\mathrm{R}=-\mathrm{6}~\mathrm{25}\text{?}\mathrm{d}\mathrm{B}， and only 0.0499 s is consumed to execute the target echo signal once.

In view of the non-linear and non-stationary characteristics of pipeline acoustic signals and the difficulty of extracting the characteristics of pipeline leakage signal， a feature extraction method for acoustic signal was proposed. Firstly， the variational mode decomposition （VMD） was used to de-noise the collected acoustic signal， during which process the mode number of VMD was determined by the method of Minimum Bhattacharyya distance. Then the Wasserstein Distance （WD） between the probability density of the （Intrinsic Mode Function） IMF components was obtained by VMD and the original signal， which was evaluated to select the effective modes so as to reconstruct the selected effective modes. Finally， the dispersion entropy value of the reconstructed signal was used as the signal characteristic parameter， and the characteristic parameter was input into the extreme learning machine （ELM） to recognize the working condition. Experimental results show that the proposed method could classify and recognize pipeline signals accurately， and the total recognition rate is up to 100%.

To solving the problems of severe vibration and poor stability of maize straw deep bury and returning machine during the field operation， the vibration characteristics and influence rules of the machine are studied. In addition， increasing its low-order natural frequency is taken as the goal to optimize the structure of the implement and improve its operating effect. First， the natural frequencies of the machines are extracted through finite element modal analysis. Secondly， through the field testing， the amplitude statistical characteristics and power spectrum of the eight test points on the machines are obtained to analyze their impact on the vibration characteristics of the whole machine. Finally， based on the ISIGHT multidisciplinary software platform， the sequence quadratic programming method is used to optimize the parameters of the main structure of the implements， avoid the main external excitation frequency and optimize the structure of the implement. The research shows that machine vibration is mainly determined by forward speed， its own structure， and the conditions of soil and ground surface. When the implements operate， resonance has occurred at the tail of the frame， the covering device and the suppression wheel have a certain adsorption effect on vibration. Through the analysis on power spectrum， the main frequency that affects the vibration of the implement is concentrated at 8~16 Hz， which is close to the first-order and the second-order natural frequencies. The optimized first-order natural frequency of the main structure of the whole machine is increased to 20.348 Hz， which effectively avoids the resonance. The field experiment shows that the optimized implements can reduces vibration and noise during operation， and the effect of the operation is good， which provides a certain theoretical reference for the design optimization of the maize straw deep bury and returning machine.