When the wheel-legged robot is in the state of obstacle crossing, it is inevitable to change the attitude and position (centroid position and pitch and roll attitude). In order to control its position and attitude, the multi-link suspension system of automobile is adopted to design a new wheel-legged robot to ensure the stability in complex environment. Firstly, the single leg kinematics model of the wheel-legged robot is established, and a single leg test bench is built to verify the correctness of the model. Then, the pitching and rolling model is established for the position and attitude of the wheel-legged robot, and the position and attitude of the wheel-legged robot are decoupled. the closed-loop control of the robot is realized in the condition that the position components of the center (x, y, z) of the wheel-legged robot are constants. Then, the proportional control (P-control) is used to build the control strategy of the wheel-legged robot in Simulink. Finally, the virtual prototype of the wheel-legged robot is created in Adams, and the PAC tire model is built for the large outer angle, and the joint simulation model of Simulink and Adams is used to confirm the effectiveness of the control strategy on the three-dimensional slope. Simulation results show that the centroid position and attitude of the wheel-legged robot have good tracking effect, and the position error and attitude error are controlled within 4.3% and 5% respectively, so the effectiveness of the control strategy of the wheel-legged robot is verified.

In the application of large eddy simulation in CFD numerical simulation, higher-order discrete schemes can improve the numerical accuracy, but higher mesh resolution need will greatly increase the computational complexity. By improving the existing discrete schemes reasonably to establish more efficient and accurate discrete schemes, the numerical accuracy can be guaranteed with lower requirements of mesh resolution. In this paper, based on two commonly used discrete schemes of Gauss linear and upwind, an improved discrete scheme is established. Based on the experimental results of Dshap, the simulation results of large eddy simulation under the three discrete schemes are compared and analyzed. The comparison results show that the improved discrete scheme is more efficient and accurate. Further, the improved discrete scheme is applied to analyze the outflow field of Dshap under different ground clearance. The analysis shows that the ground can improve aerodynamic performance of the model, and appropriate ground clearance can provide the model a lower aerodynamic drag while reducing the aerodynamic lift. The research of the improved discrete scheme in this paper may provide an reference for the application of large eddy simulation in vehicle external flow field simulation.

Aiming at driver panic or other factors causing the lack of braking force and decreasing the driver's braking operating burden when driving on a long ramp, HBA and HDC control strategy is carried out by Stateflow state machine theory and Matlab/Simulink. The brake pedal displacement and speed are selected as the identification parameters to better identify the driver intention. Meanwhile, the ramp recognition is designed by Newton's second law. Considering the braking problem of high temperature failure, the temperature model is built. Based on the above, MIL and HIL are carried out, the results verify the correctness of the control strategy. In addition, road tests' results show the HBA function can meet the ECE standard and the HDC function can keep vehicle speed near the target speed, proving that the control strategy achieved good control effect.

An electromagnetic brake with new structure was studied in this paper. The electromagnetic braking torque can be improved as the air gap width can self-regulate with the change of temperature. The heat conduction model, air gap width model and electromagnetic braking torque model were established. The structure parameters of the electromagnetic brake were optimized, including the air gap width, the size of the iron-core and the thermal-deformation component. The finite element model of the electromagnetic brake was established in ANSYS, which was used to research the temperature change of the brake. The long downhill condition and emergency braking condition were studied by simulation. The change of the air gap width was analyzed in Matlab. The current requirement of the new electromagnetic brake was compared with that of an old brake in the same brake task. The results show that increasing the air gap width of the self-regulation structure, the new electromagnetic brake can smoothly work in the long downhill braking condition and emergency braking condition, when the air gap width is 0.4 mm to 1 mm. The brake torque of the new electromagnetic brake is increased significantly.

This research aims to solve the problems of vehicle soiling during wading and a modified square-back MIRA model is selected as the research object. The discrete phase model and Eulerian wall film are used to obtain the droplet trajectory and vehicle surface pollution distribution and to analyze the mechanism of surface contamination. Based on the original model, three improvements were proposed, the bottom spoiler, top tail deflector and wheel house shield, to change the structure of the flow field so as to change the droplet trajectory and reduce the surface contamination of the vehicle body. Simulation results show that the top tail deflector can better control the flow direction, thereby reducing the degree of contamination in a specific area. The bottom spoiler changes the shape of the wake and reduces the contamination of the back surface. The wheel house shield reduces the number of droplets flying out of the wheel cavity, effectively reducing body side wall contamination.

To solve the problem of struggling to obtain the inertial parameters of complex mechanical systems, the static equations for solving the coordinates of the center of mass and the dynamic equations for solving the inertia parameters are established respectively on the basis of studying the computational dynamics of inertial parameters. In order to eliminate the influence of friction on identification accuracy, a series of experimental methods are proposed, including pre?adjusting the moving direction of the platform, eliminating the positive and negative friction forces at the same angle, processing the experimental data by data filtering and data fitting to eliminate the high frequency noise and harmonic noise of the data. Finally, the recursive least square method is used to identify the coordinates of the center of mass and inertia tensors, and the validity of the method is verified by experiments. The test platform can test the inertia parameters of automobile parts such as engine and powertrain.

The theory of vehicle dynamics is applied to solve the problems that the lateral slope estimation relies on the precision of the tire model and is influenced by the road surface adhesion coefficient, the dynamic response of the vehicle and the information of the vehicle body are not fully considered. Considering the dynamic characteristics of a two-degree-of-freedom (2-DOF) vehicle model on a pavement with a lateral slope, a lateral slope estimation method is proposed, which can decouple vehicle state and road information. Considering the complexity of the vehicle model and the identification error, the superposition terms of the inclination angle and the lateral slope of the vehicle are derived by using the 2-DOF vehicle model, and the inclination angle of the vehicle is solved by the suspension height information. Finally, the accurate estimation of the lateral slope of the road is realized. The vehicle model and lateral slope estimation algorithm are built by Simulink-Carsim, and the off-line simulation is carried out to verify the effectiveness of the method.

To solve the problem of road roughness identification, a NARX neural network identification method and its applicability are studied based on vehicle responses. A four degree of freedom plane model of vehicle vibration system is established, thus, the vehicle responses and road roughness of wheel can be obtained by simulation. The application selection, input scheme optimization and evaluation index of NARX neural network are studied, and the solutions of vehicle response selection and its combination optimization are put forward. The NARX neural network is used to identify the road roughness at front wheel of a car under the common road grade B and 60 km/h driving speed, for which the correlation coefficient and root mean square error are 96.75% and 0.003 3, respectively. The influences of training sampling points, vehicle response random noise, vehicle speed, and road grade on the NARX neural network are considered, and the adaptability of NARX neural network method for road roughness identification based on vehicle responses is illustrated. The results show that the use of orthogonal test design to determine the optimal input scheme of the NARX neural network and the identification of road roughness based on vehicle responses can achieve satisfactory performance and good applicability.

To estimate the parameters of the mixed distribution models which are widely used in reliability modeling, a new parameter estimation method was proposed. The kernel density estimation was introduced to non-parametric fitting of the data. With the minimized mean integration square error, the optimal bandwidth of kernel density estimation (KDE) was obtained. The original data were grouped by the optimal bandwidth of KDE as the group distance, and the statistical histogram was made. The mixed multiplicity of the distribution density function was determined by the histogram. The K-means clustering method was employed to cluster the histograms. The weights of the mixed distribution model were calculated from the clustering results. Then the gravitational search algorithm (GSA) was applied to identify the parameters of the sub-distributions' parameters for the mixed model. Taking the measured displacement signal of vehicle axle as an example, mixed reliability model was developed, in which the model parameters were estimated by the proposed method. Based on the identification results of model parameters, the determination coefficient, Kolmogorov-Smirnov coefficient and average relative error were calculated respectively, with which the effectiveness of the proposed parameter estimation method was verified. This research would lay the foundation for the fatigue load spectrum preparation and laboratory reliability bench test of commercial vehicle. Moreover, it can provide reference for related reliability modeling and model parameter identification.

In order to study the peeling failure behavior of steel-aluminum self-piercing riveted (SPR) joints, peeling tests and simulation studies were carried out on two kinds of dissimilar SPR joints in 5754 aluminum alloy and Q235 steel sheets. The finite element model of T-type SPR joint was established based on the modified Mohr-Coulomb (MMC) failure criterion. The tensile failure process of the two joints under peeling condition was simulated. The reliability of the finite element model was verified by comparing test and simulation results, and the peeling failure behavior of the joint was analyzed. The results show that the joint has better mechanical properties when the steel sheet is used as the lower sheet, and the finite element model can better predict the failure mode and mechanical properties of the joint under peeling condition. The failure mode of the aluminum-steel joint is the separation of the rivet from the upper sheet, the stress concentration is located at the rivet tail and rivet head area of the loaded side during the peeling process, and the riveted area of the upper sheet at unloaded side first failed. The failure mode of the steel-aluminum joint is the separation of the rivet from the lower sheet, the stress concentration is located at the edge of rivet head and inside of rivet tail of the loaded side during the peeling process. Delamination damage occurred in the rivet area of the bottom sheet, and the delamination damage region extended to the rivet area of the unloaded side.

The trafficability of CE-3 Lunar rover can be evaluated by the sinkage of the wheels. In this research, the static and dynamic sinkage tests were conducted. The results show that the apparent sinkage is smaller than actual sinkage, and the difference of sinkage is affected by the slip ratio, wheel load and the regolith conditions. The maximal difference of sinkage was 19.8mm and the minimal difference was 3.7mm. A prediction model of the sinkage difference was built according to the test data, and verified by the field test of CE-3 rover. Results show that the accuracy of model 3 is 85%, and the accuracy of model 4 is 84.5%. The results are helpful to evaluate the trafficability and to plan the path for the CE-3 and CE-4 rovers.

To improve the precision of time interval partition at an isolated control intersection, first, two adjacent statistical intervals are taken as examples to determine the numbers of transition cycles under different timing plan transition conditions. Second, the negative impacts caused by plan transition are considered and the calculation models of total vehicle delays when adjacent intervals are combined and not combined are developed, respectively. The difference of total vehicle delays under the two conditions is selected as the index to judge whether two adjacent intervals can be combined or not. Finally, the time interval partition procedure is developed for all statistical intervals in the whole day. A case study is conducted by taking a field intersection in Fuzhou city as an example and the comparisons are conducted between the proposed method and traditional partition method. The results show that the proposed method can produce reasonable number of time intervals and reduce total vehicle delay by 8.9%.

In order to intelligently solve the problems existing in urban road networks and improve the accuracy of short-term traffic flow prediction, a short-term traffic prediction model is established by using the least square support vector machine (LSSVM). Specifically, the immune algorithm is adopted to optimize the penalty factor and kernel parameters of the LSSVM, thus obtaining the optimal prediction model. The prediction simulation experiment takes the average speed and occupancy rate of vehicles as the input of the model to predict the traffic flow. The experimental results show that the prediction error of the optimized LSSVM model used in the simulation experiment is reduced, and the output result is closer to the real value.

In order to provide reference for setting on-street parking space that occupies vehicle lane, traffic flow parameters survey was conducted under the interruption of on-street paring. The changing rules of cars’ saturated headway with the number of on-street parking cars were analyzed. The models between the saturated headway with the number of on-street parking cars on adjacent lane and interval lane were established respectively using SPSS software. Based on established models, suggested thresholds of traffic volume for adjacent lane and interval lane were given to set on-street parking lot, and case study was conducted. It shows that cars' saturated headway on adjacent lane and interval lane increases with the number of on-street parking cars and there are quadratic function relations between them. With the increase in the number of on-street paring cars the road design speed increases, and the adjustment factor decreases for possible capacity, so the setting of on-street parking lot is more improper. When the number of on-street parking car is the same, adjustment factor for possible capacity of interval lane is larger than that of adjacent lane, so the influence of on-street parking on traffic operation of interval lane is less than that of adjacent lane.

According to the similarity theorem, two kinds of reduced-scale models of integral box culvert and four-member combined box culvertare are made, and the displacement and strain responses of key sections under various load conditions are tested using four-point bending loading method until the model is destroyed. The results of numerical simulation and test are compared to verify the validity of the finite element analysis model. Then the theoretical analysis model of the actual engineering structure is established, and the static responses of the two kinds of box culvert structures are numerically simulated based on the theoretical analysis model. The results show that the rigid-knot box culvert is better than the hinged model in resisting deformation, and the four-member composite box culvert can reduce the hinged section stress of the side wall at the cost of increasing the roof stress.

From the aspect of reducing seismic response of fixed pier, the vibration absorption measures of multi-span and long-unit continuous girder bridge were exploded. Taking the main bridge of Hanjiang Bridge with the span of 55+4×90+55 m as an engineering background, the seismic performance of the bridge influenced by viscous damper were studied. The finite element model for dynamic analysis of the bridge was established based on ANSYS. The mechanical behavior of viscous damper was simulated by Maxwell Model. Three artificial seismic waves with the exceedance probability of 2.5% in 50 years were chosen as the earthquake excitation. By using nonlinear dynamic time history analysis method, the parameter sensitivity analysis on velocity exponent α and damping coefficient C of the viscous damper was investigated. At last, the results were compared with the seismic response without considering the viscous damper model. All the results shown that saturated and closed lag circle of viscous damper under the action of earthquake was formed; The lag circle has regular shape, large envelope area, so that the energy dissipation effect had been proved successful. The internal force of bottom of fixed pier and displacement of top of fixed pier decreased with the increase of damping coefficient. And they also increased with the increase of velocity index. The velocity exponent α and damping coefficient C of the viscous damper were recommended as 0.4 and 4000 kN/(m?s^-1)^0.4 by considering the damping effect of viscous damper. Compared with the model without viscous damper, the bending moment and shearing average damping rate of the fixed pier were 44.35% and 42.52%, respectively. The overall horizontal seismic load on the fixed pier was also reduced by about 25%, and all piers bear more uniform horizontal seismic load, the load of all piers were more reasonable. In addition, it also prevents or reduces the collision between the superstructure along the bridge direction. And the fall-beam destruction in the earthquake is avoided.

In order to explore the internal mechanism of the natural vibration frequency of externally prestressed simply supported beams, the calculation formula of the natural vibration frequency of externally prestressed simply supported beams with eccentric linear reinforcement is derived based on the basic principle of structural vibration and the Hamilton principle. The finite element model is established in ANSYS and its modal analysis is carried out. The analysis results show that the calculation of the natural vibration frequency of the externally prestressed simply supported beam for the eccentric linear reinforcement is reliable according to the calculation theory given in this paper. It provides a new calculation method for the calculation of the natural vibration frequency of externally prestressed simply supported beams, and it is of great significance to reveal the variation law of the natural vibration frequency of externally prestressed simply supported beams.

Based on the finite element method of SCF of CHS-CFSHS T-joints with reliable accuracy, 160 finite element models were established to calculate SCF of CHS-CFSHS T-joints when brace is under axial tension which are made up of circular hollow section (CHS) braces and concrete-filled square hollow section (CFSHS) chords. Parameters affecting SCF of the joint were analyzed. In addition, while grasping the influence law of each parameter on SCF, it proposed a calculation formula for SCF when brace is under axial tension, and made a comparison between the calculated value and finite element data. The results show that the six positions of 0°, 60° and 90° on the brace and the chord basically cover the position at which maximum SCF may occur when brace is under axial tension; and quadratic functions of β and 2γ as well as power function of τ should be included in the calculation formula of SCF. Moreover, the influences of the three parameters are coupled with each other, and the regressive calculation formulas of SCF have reliable accuracy.

Based on the particle discrete element theory, five loading rates corresponding to different levels of strain rate were selected in this paper, and the corresponding loading rates were 0.001 m/s, 0.01 m/s, 0.1 m/s, 1.0 m/s and 3.0 m/s. The uniaxial compressive strength test of normal concrete (NC) and crumb rubber concrete (CRC) was simulated, and the effect of loading rate on the constitutive relation, crack propagation, meso-damage and energy evolution was systematically analyzed. The results show that the uniaxial compressive strength and the corresponding peak strain of the material present a nonlinear growth relationship with the increase of the loading rate. The loading rate sensitivity of CRC is more obvious than that of NC. There is a negative correlation between loading rate and NC crack generation rate, but a positive correlation between loading rate and CRC crack generation rate. The fitting equations between the mesoscopic damage variables and the loading rates of the two materials are proposed, and the fitting result is good, and the effects of different loading rates on the evolution law of elastic strain energy and kinetic energy of materials are analyzed.

The purpose of this paper is to explore the distribution of in-cylinder exergy destruction and its influence factors in combustion process. A heavy-duty turbocharged diesel engine is taken as the research object to study exergy change rule, exergy destruction distribution in cylinder and its influence factors based on test platform, STAR-CD computational fluid dynamics software and subsequent theoretical calculation. The results show that first, exergy destruction in cylinder mainly happens in combustion process, the exergy destruction caused by chemical reaction and its proportion to in-cylinder exergy destruction are positively related with engine load at the same speed, the proportion is up to 94.5% at B75 working condition. Secondly, exergy destruction in cylinder of diesel engine is mainly caused by chemical reaction irreversibility and it mostly occurs at the edge of oil beam. Third, the largest chemical reaction exergy destruction rate happens in dense mixing zone whose equivalence ratio range is 1~1.5; however, the value is relatively low in dilute mixing region or denser mixing zone whose equivalence ratio is lower than 1 or greater than 1.5. With the development of combustion process, the reaction exergy destruction rate decreases gradually in different equivalence ratios zone. Fourth, in a certain temperature range, the chemical reaction exergy destruction rate is almost proportional to the heat release rate of fuel, and in a certain heat release rate region, the chemical reaction exergy destruction and its proportion of the fuel exergy gradually reduces with the temperature increase. Finally, as the combustion process proceeding, the relative intensity of equivalence ratio, temperature and chemical reaction exergy destruction decrease gradually.

The effects of butanol/diesel blends on engine combustion and particulate emission characteristics under different operating conditions and blended ratios were investigated on a four-cylinder, common rail, turbocharged diesel engine. The results indicates that the increase in butanol blended ratio leads to prolonged ignition delay, increased premixed combustion and delayed combustion phase, the effect of which is more significant under low load conditions. Butanol/diesel fuel can dramatically reduce exhaust smoke and particulate emissions without increasing NO _x emissions, however, the thermal efficiency is slightly lower than that of pure diesel. Most of the particulate emissions from butanol/diesel blended fuel are ultrafine particles, with a ratio of more than 99%. With the increase in butanol ratio, the number concentration of particles in each particle diameter range decreases, which is more distinctive under high load condition. The concentration of accumulated particles and total particles increase significantly with the introduction of Exhaust Gas Re-circulation (EGR). However, the addition of butanol enhanced the tolerance of the mixed fuel to EGR. Butanol/diesel blend fuel combined with EGR strategy can reduce NO _x and particulate emissions simultaneously.

Aiming at the problems of slow combustion speed, high exhaust temperature, high heat load and prone to knock for natural gas engine with large cylinder bore, a visual constant volume combustion system suitable for dual-spark plug ignition is designed in this paper. Combining the rapid combustion technology and the natural flame luminescence method, the visualization experiment of natural gas combustion was carried out on a constant volume combustion bomb by changing the fuel-air equivalence ratio. The images of dynamic combustion process in cylinder were collected by high-speed camera，the effects of single and dual-spark plug ignition on flame propagation, combustion pressure in cylinder and combustion duration were studied by using the digital image processing technology of MATLAB. The results show that under different equivalence ratios, compared with single ignition, the dual ignition can shorten the flame propagation distance, accelerate the combustion speed in cylinder, shorten the flame development period and the main combustion period, and thus improve the thermal efficiency of combustion. Especially for the lean combustion of natural gas, the effect is very obvious, which can restrain the occurrence of misfire and knock, broaden the lean combustion limit, and is conducive to combustion stability and emission reduction.

In order to study the flow structure of different width of radial strut, a single-cavity trapped vortex combustor with diffuser and casings was designed. Unburnt flow field experiments were carried out by changing the width of radial strut on the combustor. Experiments were accomplished at atmospheric pressure and temperature. The main parameters of the experiments were shown as follows: the width of radial strut was 40,30,20 mm while the number of radial strut was constant. The results showed that on the central section of mainstream (PM), there were two vortices in the cavity. The main vortex was located in the middle of cavity, and the secondary vortex was located between the main vortex and mainstream. On the central section of radial strut (PA), different width of radial strut would form two different flow structures. When the width of radial strut was wider, the cavity had a single-vortex flow structure, which meant that only the main vortex existed in cavity. The reflux airflow from main vortex flowed along the radial strut to the lower wall of the liner. When the width of radial strut was narrower, the cavity had a double-vortex flow structure, which meant that the meanstream flowed to the behind of radial strut.

Nucleation boiling term of Chen model was modified by considering the influence of roughness and material of the heating surface. The modified model can effectively fit the experiment data of copper heating surface with roughness less than or equal to 4.62 μm and aluminum heating surface with roughness less than or equal to 8.12 μm, and the overall prediction error is within 30%.

The structure and operating principle of the electro-mechanical continuously variable transmission (EMCVT) is analyzed. The coupling relation between the speed ratio and the driven pulley clamping force was illustrated and the model of the EMCVT actuator was developed. The speed ratio control strategy of the EMCVT was developed based on the coupling characteristic between the speed ratio and driven pulley clamping force. The control effect of this ratio control strategy was compared with that of conventional strategy using a simulation model of the EMCVT in MATLAB/Simulink. The results indicates that the response time of speed ratio is reduced by 14% when the speed ratio increases; and although the energy consumed by the EMCVT is increased, the vehicle dynamic performance is improved.

In order to study the influence of Proper Orthogonal Decomposition (POD) basis functions on the accuracy of flow field reconstruction, the velocity and pressure proper orthogonal bases were obtained respectively for the flow field on the stator interaction surface of hydraulic retarder under full-filled condition. The influences of the instantaneous image factors and rotational speed factors on the proper orthogonal basis were studied, and the flow field is predicted by means of proper orthogonal basis and surrogate model technology. Results show that both the spiting of snapshots database and the frequency of sampling snapshots have significant effects on flow field reconstruction accuracy by POD basis functions while the features of POD basis functions are immune to the speed of the rotor, and the flow field reconstruction based on POD basis is of certain accuracy.

This paper presents a 2-DOF dynamic vibration absorber to reduce the vibration of 2-DOF main system which possesses translation and rotation around the centroiding. The main system obtains a good damping effect by optimizing the stiffness and damping of the dynamic vibration absorber. The evaluation and optimization indexes are displacement, acceleration, angular displacement and angular acceleration. The nondominated sorting genetic algorithm (NSGA-Ⅱ) is used to solve the multi-objective optimization problem including four optimization objectives and three design variables. The 60 Pareto optimal solutions are ordered by entropy weight TOPSIS. The optimal solution is selected as the dynamic vibration absorber optimization design solution and a simulation analysis is carried out to obtain the frequency domain simulation results and time domain simulation results. The results show that the dynamic vibration absorber with optimized parameters greatly reduce the four vibration evaluation indexes of the main system and the vibration control effect of the two-DOF dynamic vibration absorber is good. The NSGA-Ⅱ and Entropy TOPSIS joint optimization method is suitable for optimal design of two-DOF dynamic vibration absorber.

Through the research of 6-PSS parallel manipulator, this paper developed the kinematics model based on the relationship between spherical joints' spatial positions in three kinds of coordinate systems, which respectively are the global coordinate system, the local coordinate system and the hinge-point coordinate systems fixed on sliding blocks. A composed numerical method based on multivariate Newton's method and the steepest descent method was proposed. The theories of numerical analysis and machine computation were used to improve the convergence speed and decrease the amount of operations in a single-step iterative process. Finally, the forward kinematics solution program was written for the numerical simulation. The results show that the improved composed numerical method can not only decrease the computational time and converge to the solution accurately, but also reduce the dependence on the initial estimate and expand the convergence range. This method can be effectively applied in robot system based on 6-PSS parallel manipulator.

The autonomous inflow control device (AICD) is a downhole flow control device for oil and gas exploitation. It can inhibit the water flow into wellbore, delay water coning and prolong the service life of oil and gas wells. In this paper, an improved AICD with two layer baffles is designed for downhole water flow restriction. The mechanism of AICD was analyzed for the water flow restriction and the gas flow preservation. The gas flow and water flow in AICD are simulated numerically. The pressure field is calculated, and the inlet-outlet pressure difference of water and gas are calculated respectively. A curve fitting method is adopted to obtain the relationship between the pressure difference and the flow rate. Numerical results show that the pressure difference of water flow is much greater than the pressure difference of natural gas, indicating that our AICD has a significant effect for blocking water.

Compared with the first-generation of fracture-splitting material C70S6, the tensile and yield strengths of the new material 36MnVS4 are significantly improved. In order to reduce the weight of the connecting rod to meet the requirements of the engine's low energy consumption and low vibration, the structure of the 36MnVS4 connecting rod is improved. Based on the design of the original C70S6 connecting rod structure, the high-strength steel 36MnVS4 is used instead. Three-dimensional inverse reconstruction of fracture surface of C70S6 and 36MnVS4 connecting rods are conducted, and an effective analysis model of connecting rods is established. The fatigue strength analysis by finite element simulation shows that the fatigue life and safety factor of the 36MnVS4 connecting rod are significantly higher than that of C70S6 connecting rod, and there is sufficient safety reserve for lightweight improvement. With the constraints of connecting rod strength and stiffness, the structural design of the 36MnVS4 connecting rod is improved that the mass of the connecting rod is reduced by 21%.

To improve the surface properties of the TC4 titanium alloy, Cr alloying layers are formed on the TC4 samples' surfaces using high-current pulsed electron beam (HCPEB). The microstructure and phase analysis of the surface alloying layer were studied by X-ray diffraction (XRD), optical microscopy (OM), scanning electron microscopy (SEM) and transmission electron microscopy (TEM). The experimental results demonstrate that after HCPEB alloying, Cr alloyed layer about 1.5μm in thickness is formed on the surface. The diffusion of Cr element occurred on the surface of the substrate, the Ti atoms combined with the chromium to form tiny Laves phase Cr₂Ti. Martensitic transformation occurred in the irradiated layer and a large number of lath martensite structures were observed in the irradiated surface. Moreover, both the surface microhardness and corrosion resistance of the HCPEB treated samples were obviously improved.

TiC powder was modified by high-energy ball milling and added to the melt of nodular cast iron. The tissues were analyzed by optical microscopy (OM), scanning electron microscopy (SEM) and energy dispersive spectrometer (EDS). The results show that both graphite and matrix of the ductile iron are improved. As the heterogeneous nucleation core of graphite, some TiC particles promote the nucleation of graphite, increase the number of graphite spheres by 40.5%, refine the size of graphite spheres, indirectly increase the ferrite content by 4.5%, and increase the spheroidization rate by 9.1%. On the one hand, the morphology of the surrounding cementite is changed, and even the unique “pearlite colony” structure can be formed. On the other hand, the ferrite grain is refined due to the hindrance of TiC particles to grain boundary movement. The mechanical properties of the as-cast ductile iron were tested by tensile test machine and impact test machine. The total elongation, impact toughness at room temperature and impact toughness at low temperature (－20 °C) were increased by 19.1%, 31.1% and 21.8% respectively by adding a small amount of TiC particles.

To solve the problem of quality detection of large-scale fruit crops, this paper proposes an Internet architecture for monochrome fruit growth status detection based on comprehensive index quality evaluation algorithm. Through the RGB color model, convolution exploration method and Gabor filtering, the color depth, size and appearance damage degree of the fruit were analyzed. The three results were comprehensively analyzed, and the fruit quality was divided into 6 grades for the growing fruit quality evaluation. The algorithm was integrated into the Internet architecture. Based on the traditional industrial system structure, the network interconnection, data intelligence and security between the system and the outside world were improved. The structural optimization of the fruit growth detection Internet architecture was realized, and the structure flat development was mad and structure network, data interconnection, security level integration were promoted, which was used in real-time monitoring of crops and recording the fruit growth process. Experiments conducted by common monochrome fruits in Changchun show that the algorithms in this architecture can formulate different grading evaluation standards according to specific types, improve the interaction ability and quality evaluation efficiency between the system and the outside world, and reduce manual input.

To solve the issue of the single ability for robot in the nuclear power plant, an autonomous exchange mechanism of the end-effector for robot is developed. The interface of multi-torque output with the concentric and parallel mechanism with 3-Hooke joints is used to provide power input for the end-effector with related operations. Based on analysis of the passive compliance and tolerance condition for the docking process of the double Hooke joints between the end-effector exchange and the tool rack, an automatic docking control strategy of the end-effectors exchange is proposed based on the contact force threshold. The automatic replacement of different end-effectors is realized by experiments, and the feasibility of the control strategy is verified. The mechanism prototype for end-effector exchange is designed. The exchange experiments of different end-effectors are carried out by the SCHUNK 6 DOF manipulator mobile experiment platform, which demonstrate the effectiveness of the control strategy, the tolerance performance of the mechanism. The experimental results show that the mechanism has completed docking process with larger tolerance compliant and realized the power transmission of the end-effector.

Fabric defect dataset usually has high resolution but small defect area, which is different from common image classification datasets (such as ImageNet, etc). When applying exist classification algorithms to fabric defect dataset directly, it cannot achieve expected accuracy. To address this problem, this paper proposes a new classification algorithm based on improved Fast RCNN. For an image with small defect area, we follow the pipeline of object detection, generating lots of ROI (Region of Interest), extracting feature map with deep convolutional neural network and predicting each ROI's class probability. On the last stage, we combine all ROIs' class probabilities to get the full image's class probability. Experiments performed in a fabric defect dataset which has 3331 high resolution images show that our algorithm outperforms OurNet and exist classification algorithm.

Aiming at the network training of multi-input samples, a similarity retention instance retrieval method is proposed. Firstly, the input image features are extracted by the convolution structure in the depth network and pooled. Then, according to the benchmark order, the similarity relationship between the low correlation image and the query image is corrected, and the low correlation image contrast loss coefficient is obtained, and the loss value within the loss reference value is retained. The loss value is performed to maintain contrast loss training based on similarity. Finally, the post-training network is used to extract image features for instance-level image retrieval. The experimental results show that the loss comparison function based on similarity is feasible, and the method significantly improves the accuracy of instance-level image retrieval.

A heterogeneous computing technique based on Field－Programmable Gate Array (FPGA) is proposed in this paper. The Viola-Jones face detection algorithm is accelerated based on concurrent and pipelining methods to improve data throughput and increase the parallelism of cascaded classifiers and convolution neural network is accelerated by concurrent convolution and pipelined feature maps. The experimental results show that the hardware platform achieves a speedup of 2.9 times compared with the software platform.

To effectively enhance network service capability, Extensible Network Service Model (ENSM) extends the connotation of network services from bearer network to the entire network, and defines a unified service extension model based on dynamic deployment and combination of protocol meta-functional units, which can effectively improve network services capacity and service expansion capacity. In this paper, a new network called Service Dynamic Extended Network (SDEN) is proposed based on ENSM. Combined with the network architecture idea of SDN and NFV, SDEN node system model is researched and designed. In addition the experiment based on customization and dynamic expansion of routing services is implemented on a simple SDEN node system which turned out that the SDEN node system is feasible and has good support for ENSM services extensibility.

In order to improve the efficiency and productivity of the assembly line, the assembly line area was optimized and the space occupied by the tools and changeover for each task were taken into consideration when balancing the robotic assembly line. The two conflicting objective functions considered in this article were to minimize the cycle time of the number of workstations and to minimize the area of each workstation. A mathematical model was first established. In order to effectively solve this problem, a coding method based on separator genes was designed and then an improved multi-objective immune clonal algorithm was proposed for the constrained multi-objective optimization problem. Pareto frontier sorting, elite strategy, and global search were introduced to enhance the performance of the algorithm. Finally, the problems of different scales were optimized and the results were compared with other algorithms to evaluate the efficiency and superiority of the proposed algorithm. The results show that the algorithm is effective on both quantity and quality of solutions, and feasible for different scales of problems.

The fingerprint indoor localization method based on the Received Signal Strength (RSS) is vulnerable to multipath effect and noise interference, resulting in low positioning accuracy. To solve this problem, an indoor positioning method based on location fingerprinting of imitating the mechanism of scorpion vibration source location is proposed. Firstly, the method imitates the n/1 neuron configuration of scorpion to construct the neuron structure, in order to encode the vibration signal and transform the vibration signal into pulses. Secondly, pulses are extracted as the location fingerprint feature, and then the location fingerprint feature database is established by the number of pulses. Finally, the Weighted K-Nearest Neighbours algorithm is used to estimate the position of vibration source. To verify the performance of the proposed algorithm, a vibration signal acquisition system is set up to imitate the vibration perception of scorpions. It is used to collect the user's step signals in the indoor environment. The experimental results indicate that the proposed method can improve the average positioning accuracy by 0.148 4 meters compared with the location fingerprinting based on RSS.

A state-space model method was proposed for the joint estimation of attenuation factor and frequency of multi-component exponentially damped sinusoids. Firstly, a state-space model which includes the system matrix was established, and the system matrix was comprised of the information of estimated attenuation factor and frequency. Secondly, by solving the state-space model equation, the estimated value of generalized observable matrix was derived. Finally, based on the generalized observable matrix, the estimated values of the system matrix were obtained, which were further used to estimate the attenuation factor and frequency. The algorithm has the advantages of low computation complexity and automatic matching in parameters. The simulations show that the state-space model algorithm is superior to ESPRIT algorithm in low SNR.

Recent researches have shown that deep convolutional neural networks can significantly boost the performance of Single-image super-resolution (SISR). In particular, residual network and densely convolutional network can improve performance remarkably. Both path topologies are proposed to alleviate the vanishing-gradient problem of deep convolution networks. Since the residual network enables feature re-usage and the dense skip connections enables new features exploration. A dual path network is proposed for single-image super-resolution by combining the residual network and the dense skip connections in a very deep network. In the proposed network, the feature maps are split into two paths, one path is propagated in the form of residual, and another path is propagated by dense skip connections. In addition, the deconvolution layers are integrated into the network to upscale the feature map which can significantly speedup the network, and the mapping is learn from the low-resolution image to the high-resolution image directly. The network is evaluated with four benchmark datasets. The simulation results demonstrate that the proposed network has much higher peak signal-to-noise ratio(PSNR), in contrast to most conventional state-of-art methods.

An spectral clustering segmentation method for high spatial resolution (HSR) remote sensing image based on combination of pixels and multi-scale objects is proposed. The algorithm first focus on building a graph that integrates multi-scale information as well as improve similarity computation method between objects. Then, the similarity matrix is computed on the graph, and normalized cut criterion is used to similarity matrix eigen-decomposition so that the original data is mapped into the low-dimensional subspace. Finally, the clustering algorithm is used to complete the image segmentation after the selected subset of the eigenvectors. In order to prove the effectiveness of the algorithm, we choose high spatial resolution remote sensing images to conduct experiment and compare to the state-of-the-art techniques. The experimental result show that three of four experimental indexes are superior to other algorithms, which proves the effectiveness of this proposed method segmentation method.