For the Electronic Stability Controller(ESC) of commercial vehicle, the partial parameters such as longitudinal velocity, lateral velocity, and sideslip angle are difficult to obtain directly, and the sensor process noise of the vehicle system is generally time-varying and unknown. To solve the problems, an Adaptive Cubature Kalman Filter(ADCKF) algorithm was proposed to estimate the vehicle state parameters. First, the standard Cubature Kalman Filter(CKF) algorithm was combined with the suboptimal Sage-Husa estimation algorithm to estimate the parameters of vehicle. Then, according to the ESC control requirements and considering the modeling uncertainty and external disturbances, the commercial vehicle ESC control of Radial Basis Function(RBF) neural network Sliding Mode Control(SMC) algorithm was proposed. Finally, the disturbances were estimated by RBF neural network. The co-simulation results of MATLAB/Simulink and TruckSim show that the ADCKF algorithm is accurate in estimating vehicle state parameters, and the RBF neural network SMC based on vehicle state estimation of commercial vehicle ESC has good control effect.

Based on Hybrid Particle Swarm Optimization(HPSO), a parameter optimization method is proposed for a Dual Clutch Transmission(DCT) of a multi-mode axle-split Hybrid Electric Vehicle (HEV). Combining the enhanced Particle Swarm Optimization(PSO) and chaotic local search PSO, the HPSO algorithm is proposed to optimize the minimum transmission ratio and partial geometric progression gear ratio of DCT by adopting the equivalent consumption minimization strategy. The optimization is based on a co-simulation platform, in which, the HEV is modeled in AMESim and the control/optimization algorithms are programmed in MATLAB/Simulink. The simulation results show that the optimized transmission parameters can reduce fuel consumption by 2.04%, not only ensuring the dynamic demand, but also having good fuel economic performance.

The complex interaction process between the tire and the ground determines that the inverse solution of the model does not exist adequately. It is impossible to directly solve the vehicle model, in which tire characteristics are considered, to obtain the desired control amount from the trajectory. To solve this problem, a trajectory tracking control strategy based on tire force prediction and fitting is proposed. According to the physical boundary, the trajectory tracking control is decomposed into two layers. The upper layer control uses a terminal sliding mode control algorithm. To realize the pre-judgment of the desired front wheel lateral force, the body dynamics is solved according to the target curvature. The lower layer control uses the system identification method. According to the feature point data, the desired steering angle is obtained by fitting the characteristics of the tire force through the spline function. A 12-DOF vehicle dynamics model with body and tire is established. According to the test value of sliding rate required to maintain different speeds, a semi-empirical sliding mode longitudinal control strategy is established to control the vehicle speed. A contrast algorithm is constructed with neural network algorithm. In the step curvature condition for testing stability and the multi-variable curvature condition for simulating continuous turning roads, the effectiveness and global robustness of the proposed algorithm are verified.

To meet the research demand of lithium-ion batteries under high temperature, cylindrical lithium-ion battery cells with different State Of Charge (SOC) were placed at elevated temperature for rest and cycle, and then these battery cells were compressed in radial direction in this paper. The experiment results indicate that the cells with a high SOC tend to be short-circuited early under compression after resting at a temperature of 40 °C or higher. This is due to the structural softening of the cells. The higher the rest temperature, the lower the resistance of the high SOC cells to deformation. Short cycling under high temperature would increase the resistance of the cells to deformation.

In wind tunnel test, sound source drifts due to jet structure when locating in outfield. The measured sound drift error is enlarged due to inaccurate locating focus distance caused by uncertainty of sound source location. Base on geometric acoustics theory, this paper proofs that the acoustic drift error is proportional to the locating focus distance error. The position function of sound source is established using the surface geometric relationship of test piece according to the spatial distribution of the target sound source. More accurate locating focus distance is obtained by approximation method, thus, the sound drift error was reduced. The effectiveness of the method is verified by wind tunnel test. This approximation method can reduce the locating focus distance error apace and eliminate the sound drift error caused by location uncertainty.

Based on the rearview mirrors of sedans, the Lattice Boltzmann method is used to study the flow field characteristics of the rearview mirrors. Combining aerodynamics with multiphase flow theory, the Lagrangian method is applied on the basis of the lattice Boltzmann method to study the water phase distribution on side windows caused by the rearview mirror on rainy days, and to investigate the pollution mechanism. Then the pollution effects of the mirror column and the mirror cover on the side window are studied. The mirror column factors include the length and size, and the mirror cover factors include the windward angle, the inside diffuser angle, the leading edge chamfer angle and the rear departure angle. The Results show that these factors have significant different influences on pollution location and pollution degree of the side windows. Increasing the column length, changing the inside diffuser angle of the cover to 0°, reducing the leading edge chamfer angle of the front cover and changing the rear departure angle to 0° will reduce the pollution level. The influence of the size of the mirror column and the windward angle of the mirror cover on the side window is complicated, but a lower pollution degree can also be obtained. The influence of each factor on the side window pollution is not independent.

The change of the railway track during rail vehicles travel will affect the vertical random vibration response of the vehicle body. Based on the vertical acceleration of the vehicle, the angular velocity of the vehicle body and the attitude angle, the vertical vibration mechanics model of the rail vehicles is established. The vertical vibration response of the vehicle body during the operation of the rail vehicles is obtained by numerical solution of the differential coefficient differential equations. The rationality of the vertical vibration mechanics model is tested by comparing with the experimental test vibration acceleration signal. The variation of vertical random vibration of vehicle body with train motion parameters is analyzed. The results show that there is a certain degree of correlation between vertical random vibration of vehicle body and vertical acceleration. It lays the foundation for the safe operation of rail trains and further fault diagnosis based on motion and vibration signals.

Heavy-loaded Anti-Vibration Platform (AVP) is mainly used to carry high-precision instruments and equipments to isolate vibration and impact caused by uneven complex road surface, thus, providing a stable working environment. In this paper, taking the 6 DOF series AVP as the research object, the parameters of the AVP damping system are optimized by the usage of the fast elitist multi-objective genetic algorithm (NSGA_II) and multi-objective particle swarm optimization algorithm (MOPSO) to significantly improve the anti-vibration performance of AVP. The damping and stiffness parameters of the damping system are optimized by NSGA-II and MOPSO under various road surface analog signals. The optimization results show that the anti-vibration rate of the two optimization targets of vertical acceleration and vertical displacement of upper platform under different excitation can be up to 56.21%. In addition, NSGA-II is applied to optimize the damper support angle parameters of the AVP, which increases the anti-vibration rate of the two optimization targets by 7.00%. The sine excitation is verified by experiments, and the error between experiment and simulation is as small as 8.30%, thus verifying the effectiveness of optimization parameters.

Fault diagnosis technology can be used to detect the potential fault of the equipment by analyzing and detecting the signal, so as to ensure the operation safety and effectively improve the operation efficiency of the equipment. The bearings are widely used in rotating machinery and equipment. The fault of the bearings may seriously affect the normal operation of equipment, and inflict economic damage, even endanger the safety of staff. Therefore, it is of great theoretical and practical significance to monitor the health status of the bearing, find out the fault location and analyze its severity in time. In the actual engineering conditions, the operating environment and workspace of mechanical equipment are characterized by complexity and variability. The intelligent fault diagnosis method based on Artificial Neural Network (ANN) can effectively identify the health status of equipment, but the traditional ANN requires a large number of labeled samples for training, which greatly limits its application in equipment fault diagnosis. Also its adaptability to different working conditions is poor. In order to solve this problem, this article proposed a model of bearing fault diagnosis based on transfer learning theory. The model consists of stacked sparse AutoEncoder (SAE) and flexible maximum function (Softmax) regression. In this model, high order KL divergence (HKL) is used to train domain adaptive ability, which can transfer the working condition with a large number of known data to the similar condition with a small amount of data. Only a small amount of data is needed to train the model to adapt to the new working condition. The experimental data set of bearing from Case Western Reserve University was used to verify the effectiveness of the model.

On the basis of in-depth study of scroll compressor profile construction, a new method for constructing scroll profiles using Frenet moving frame was proposed, a new type of variable wall thickness scroll profile was constructed by this method. The basic geometrical theory of the profile was established and the working chamber volume was also derived. On this basis, four performance indexes including compression ratio, area utilization coefficient, axial gas force and tangential gas force were introduced to evaluate its performance. The results show that under the same geometrical parameters, the newly constructed variable wall thickness scroll profiles have better performance than traditional variable wall thickness scroll profile. Not only the compression ratio and the area utilization coefficient are increased by 7.15% and 4.84% respectively, but also the axial gas force and tangential gas force are also significantly smaller. It can be seen that the use of this variable wall thickness scroll profile helps to further improve the performance of the scroll compressor.

The wake of 1/4 scale square back Ahmed body with or without boat-tail structure was investigated at ReH=9.6×104 by pressure sensor and particle image velocimetry(PIV). The statistical analysis of the instantaneous pressure on the back shows that the wake presents a bi-stable behavior in the horizontal direction, characterized by alternating occurrence of two symmetry-breaking stable states, each can maintain a relatively long-time duration. During the two stable states, a lower pressure on the back presents a high probability, while in the switch state the pressure on the back is higher. The bi-stable behavior was also observed by instantaneous and conditional average PIV results of flow field in the horizontal plane. With boat-tail structure added at the model base, the bi-stable behavior of the wake is gradually suppressed with the increase in the boat-tail length, and the wake vortex structure tends to be symmetrical, the wake width becomes narrow, the vortex shedding strength weakens, leading to higher back pressure and lower drag coefficient.

2205 duplex stainless steels containing 0.07% and 0.01% Ag were respectively prepared by adding Cu-Ag alloy particles. Meanwhile, the base material 2205 and Cu-bearing 2205 were prepared for comparison with Ag-bearing 2205. The microstructure, mechanical properties, corrosion resistance and antibacterial properties of the four materials were investigated by Optical Microscope(OM), Scanning Electron Microscope(SEM), tensile test, electrochemical workstation and film mulching method. Results reveale that the increase of Ag content increases the amount of α phase and decreases the amount of γ phase in the microstructure. Ag-containing phases exist in both Ag-containing materials. Ag-phases with particle sizes ranging from 1 to 6 μm are distributed at the interface of α matrix and in α/γ phase, while Ag-phases with particle sizes ranging from 80 to 400 nm are distributed in γ phase. Compared with base material 2205, adding of Cu can improve the tensile strength of the materials, but decrease its elongation. The tensile strength and elongation of the materials both increase with the Ag content. The material containing 0.1% Ag exhibits the best comprehensive mechanical properties. In 3.5% NaCl medium, Cu-bearing 2205 exhibits the best corrosion resistance. With the increase in Ag content, the stability and corrosion resistance of passivation film first increases and then decreases. The antibacterial rate of Ag-bearing 2205 materials increases with the contact time. The high Ag-bearing materials exhibit a good antibacterial property, and the base material 2205 and Cu-bearing 2205 do not have antibacterial properties.

The effects of different Ti contents on the microstructure and mechanical properties of ductile cast iron were investigated by adding Ti-C-Fe pre-forms. The results show that Ti mainly exists in the form of TiC and Ti (C,N) in molten iron, and the structures of pearlite and ferrite are refined. Ti content of 0.023 wt.% can increase the spheroidization rate, graphite number and ferrite content. With the increase of Ti content, The generated TiC particles will hinder the diffusion of C atoms, resulting in distortion of the graphite spheres, a decrease in the spheroidization rate and ferrite content. The mechanical properties of the specimens were tested and found that the ductile iron specimen with Ti content of 0.023 wt.% possessed good comprehensive mechanical properties, and its yield strength and tensile strength were increased by 5.43% and 3.73% respectively, and the elongation percentage was increased by 10.64%. When the Ti content was between 0.054 wt.% and 0.072 wt.%, the tensile strength of the sample was higher. The yield strength and brinell hardness of the sample with Ti content of 0.135 wt.% reached the maximum.

Al alloy was deposited by using variable polarity cold metal transfer technique. The welding parameter of EP/EN Balance was varied to control the ratio of positive numbers to negative numbers in a weld cycle. During welding process, a high speed camera and an electrical signal acquisition system were used for observing the dynamic droplet transfer behavior and obtaining the electrical signals, respectively. The effect of EP/EN Balance on the formation of the welded joint was analyzed by the metallographic pictures. The results show that the heat input in positive period was higher than that in negative period. The alternation of “cold” and “heat” was realized during the variable polarity cold metal transfer welding process. The numbers of positive and negative periods were adjusted by altering EP/EN Balance, which could dominate the heat input to the welded joints. The weld bead with good appearance can be fabricated using variable polarity cold metal transfer technique. Weld depth and contact angle increased and reinforcement decreased with the increase in EP/EN Balance.

The high-power ultrasonic spot welding machine was used to successfully weld T2 pure Cu and Ti6Al4V dissimilar metals. The effects of different welding energies on the interface temperature, cross-sectional macroscopic morphology, interface formation and mechanical properties of the joints were analyzed. It is shown that with the increase in welding energy, the peak temperature and high temperature residence time increase, the indentation depth increases gradually, the joint interface is flat and no intermetallic compounds appear, There is a diffusion layer with a maximum thickness of 2 μm at the interface. The tensile shear force gradually increases with the bonded area, and the maximum shear force is 2322 N. The joint fractures along the interface of the Cu side of the base material, and the fracture mode changes from brittle fracture to ductile-brittle mixed fracture. The stress analysis of the interface under different welding energies was carried out by using ABAQUS finite element numerical simulation software. The interface forming mechanism of Cu/Ti ultrasonic welded joints was systematically expounded.

This paper proposes a new type of dynamic bus lane, which not only ensures the bus priority strategy, but also significantly improves the spatial-temporal utilization of road resources, and reduces the conflict of demand between buses and social vehicles on road resource. The paper analyzes the difference of bus arrival time and proposes the determination method of dynamic lane clear distance and lane control flow based on the dissipation process of queue at intersection. Combined with HCM2010 vehicle delay formula and BPR function, the model of vehicle average travel time and per capita travel time under different road conditions is established. Through the numerical simulation, the operational benefits of three lane organization schemes, including without bus lanes, conventional bus lanes and dynamic bus lanes, were compared and analyzed. The experimental results indicate that the time benefit of setting dynamic bus lanes is better when social flow is less than 2268 pcu/h. The study provides theoretical support for rational optimization of bus lanes.

Based on ACCTS, the features of ACCTS were firstly described in detail. Secondly, three arrangement schemes of struts grid-jumping were proposed, which were compared with the original scheme in terms of pre-stress distribution, load condition, buckling resistance, steel consumption and so on, and the optimal grid-jumping order was determined. Finally, based on the optimal scheme and original scheme, the static and dynamic performances of two schemes were systematically comparatively studied. The research results show that under the condition that the static and dynamic performances of the original structure are unchanged, the optimal scheme after grid-jumping not only saves the amount of steel and improves the anti-buckling ability of the structure, but also simplifies structure system and reduces the construction difficulty by means of grid-jumping optimization. This research can promote the application of ACCTS in practical engineering.

To investigate the interface failure mechanism between CFRP tendon and epoxy mortar in the bond-type anchorage system, a total 9 specimens were designed for static load test. Two parameters were considered in the test, anchorage length and the internal inclination of anchorage. The failure mode of the specimens, the wear degree of CFRP tendon and load-slip curve in the loading end of specimens were obtained. The interface failure mechanism between CFRP tendon and epoxy mortar in the bond-type anchorage system was described in detail. The energy dissipation in the bonding interface was investigated. The grey relational theory was applied to investigate the sensitivity of parameters on bond strength. Based on the testing data, the regression formula of bond strength was fitted and compared with other existing formulas. The results show that the chemical bonding force only plays a limited role in the initial loading stage, and the interface bonding stress is mainly borne by the friction force and the mechanical bite force. The anchorage length and the internal inclination of anchorage have no obvious effect on the relative elastic deformation energy, but the relative local failure energy is mainly affected by the internal inclination of anchorage. The elastic bond strength, ultimate bond strength and residual bond strength all decrease with the anchorage length. With the increase in the inner inclination angle of the anchorage, the ultimate bond strength of the specimens without shear failure increases to a certain extent, but the elastic bond strength and residual bond strength change irregularly. The results of grey correlation analysis show that the correlation degree of each parameter in the order from large to small are the number of CFRP tendons, anchorage length, the internal inclination of anchorage and the spacing of CFRP tendons. A practical formula for calculating bond strength is derived and the calculated results are in good agreement with the experimental results.

In order to study the viscoelasticity of the hydrophobic nano-silica modified asphalt and its mixture, the hydrophobic nano-silica obtained by surface modification of silane coupling agent was used to modify the asphalt. Penetration test, rotational viscosity test, shear rheological test and static creep test were carried out on the modified asphalt and modified asphalt mixture. The penetration index, rotational viscosity, complex shear modulus, elastic shear modulus, viscoelastic modulus, phase angle, rutting factor, strain difference and Burgers viscoelastic parameters were evaluated for viscoelasticity. The results show that the incorporation of hydrophobic nano-silica can enhance the viscoelasticity and high-temperature stability of asphalt mixture, and improve the temperature sensing performance of asphalt and the thixotropy under normal temperature. According to the main curve analysis, in the low frequency region, the incorporation of hydrophobic nano-silica can improve the phase angle of the asphalt and enhance the elastic property of the asphalt. By analyzing the Burgers model parameters of hydrophobic nano-silica modified asphalt, it is found that the incorporation of hydrophobic nano-silica can improve the viscosity of asphalt. The results of static creep test show that the incorporation of hydrophobic nano-silica can significantly reduce the strain value of the asphalt mixture at the same temperature. In addition, the incorporation of the hydrophobic nano-silica modifier can enhance the elastic portion of the asphalt mixture, increasing viscous flow deformation and viscoelastic delayed deformation.

In order to realize the controllable seismic behavior of continuous girders bridge and ensure its seismic safety effectively, the applied devices were developed based on the corresponding design criteria. A new type of isolation bearing was designed, which can be destroyed during earthquake to release the inertial force from the superstructure so that the pier and pile foundations are not damaged by the earthquake. A combination of two kinds of large-stroke energy dissipation devices was utilized to ensure that the displacement of the superstructure is within the acceptable range. A series of mechanical performance tests of the isolation bearing and energy dissipation limiting device were conducted, and the constitutive model parameters were determined. The test results show that the shear process is highly controllable, and the friction behavior after fracture is highly stable with small uncertainty. The hysteretic curves of the two energy consumption limiting devices are full. By means of simulation calculation, a continuous girder bridge was used as an example to verify the effectiveness of the applied devices proposed in this paper.

In order to provide basic data and technical support for the optimal design of prefabricated culvert, the mechanical properties of four types of culvert structures, namely, integral, flat-joint, circular hinge and tenon, are compared and analyzed by means of laboratory test and finite element analysis based on physical engineering. Firstly, the scale models of four kinds of culvert structures are constructed according to the similarity theory, and the static responses of these models in the full loading process are tested. Secondly, the mechanical properties of typical integral and circular hinged box cults are studied by finite element simulation. Finally, based on the test data and numerical simulation results, the mechanical properties of the four types of box culvert structures are analyzed and compared comprehensively, and some useful suggestions are given. The research results show that the failure modes of all four types of components are bending failure of the concrete at the bottom of the upper roof after cracking, but the setting of hinge joints significantly changes the stress state of the box culvert structure, especially the stress state of the side wall. The bearing capacity is similar to that of cracking load, that the integral box culvert structure has the maximal bearing capacity and circular hinged box culvert structure has the the minimal capacity.

Due to the lack of long-term wind speed data in some regions, the basic wind pressure can not be calculated according to the Building Structure Load Code. It is urgent to get a basic wind pressure for the design of bridges and other engineering structures according to the short-term wind speed data. In this paper, the short-term wind speed data of 54 representative cities in China in 2013~2017 are selected, respectively. The extreme value I-type distribution function is used to fit the parameter estimation with the monthly maximum wind speed data. The moment estimation method and Gumbel method are used to calculate the parameters, which are examined examined using the Kolmogorov-Smirnov test method. The results show that the monthly maximum wind speed data in 54 cities are in line with the extreme value type I distribution. When fitting the monthly maximum wind speed data with the extreme value type I distribution, the Gumbel method has a better fitting effect than the moment estimation method. Therefore, it is feasible to use the maximum monthly wind speed data to fit the extreme value I-type distribution to calculate the basic wind pressure in regions lacking long-term wind speed data. Based on this method, the basic wind pressure was calculated for a recurrence period of 6 months, 50 years and 100 years. The basic wind pressure with a return period of 6 months can provide reference for projects with short design life, such as demolition works. The basic wind pressure with a return period of 50 and 100 years can provide a reference for bridge structural design in areas without long-term wind speed data. Taking Fuyang city as an example, this paper provides an example for the basic wind pressure calculation which lacks long-term wind speed data. The basic wind pressure of lacking the long-term wind speed data and short recurrence period can be calculated.

For the problem that there are many noise elements in the medical image samples and the distinction between positive and negative samples is not obvious, this paper designed a progressive type of multi-scale information fusion system, an end-to-end image automatic recognition model is established based on the existing convolution neural network structure with the attention mechanism in the classification of natural images. The model selected the corresponding fusion strategy for the output of different layers of the convolutional neural network and comprehensively utilized the information of different layers of the image. The improved model was used to identify the X-ray images of pneumonia patients and normal human beings, with an accuracy rate of 95.11%, an precision rate of 90.75%, and a recall rate of 90.28%, then a multi-group cross-contrast experiment was designed. Experimental results show that the improved model is superior in time and space complexity and the multi-scale information fusion mechanism designed in this paper improves the accuracy of image recognition tasks.

The continuous development and widespread use of computed tomography(CT) in medical practice have caused patients and doctors’ attention to the risk of cancer due to radiation dose. However, spectral CT can generate severe Poisson noise as a result of the division of multiple energy bins while reducing the radiation dose, which has an adverse effect on clinical diagnosis. Aiming at improving the reconstructed image quality of spectral CT under low radiation dose, a denoising model based on a generative adversarial network with the hybrid perceptual loss was proposed. The method combines pixel space with the human-perceived feature space into generative adversarial network, and introduces the residual learning to reuse features from different levels to avoid the loss of important details that come from the increase in the number of network layers. Then the Wasserstein distance is chosen instead of Jensen-Shannon to improve the stability of training. The inputs are from multiple energy bins, so that the spatial correlation of the reconstructed image from a single energy bin and the energy correlation between different reconstructed images from multiple energy bins can be utilized to improve the visual sensitivity of low dose spectral CT images. The experiments were performed on the simulated phantom data sets from Duke University Medical Center. The results show that the proposed method increases the PSNR by about 5 dB, the structure similarity (SSIM) value by about 0.2, and the feature similarity (FSIM) value about by 0.06. Compared with current advanced low-dose CT denoising method, the hybrid loss function based generative and adversarial network achieves a better noise removal, while retaining the necessary details for diagnosis. Moreover, the spectral information also plays a positive role in the recovery of noise images, and the use of hybrid perceptual loss is superior to any single loss. Overall, the proposed generative adversarial network with a hybrid loss method can significantly improve the quality of low-dose spectral CT images.

In order to solve the problems of clothing landmark detection, category classification and attribute prediction, a novel deep neural network based on the combination of landmark attention mechanism and channel attention mechanism was proposed. First, the network predicts clothing landmarks by convoluting the input feature map to extract features, deconvoluting to restore the feature map size. Then, it acquires the connection between the landmarks by adding a non-local structure, thus, obtaining the landmark attention. The landmark attention module emphasizes the characteristics of the discriminative area in the clothing, and then new feature maps are generated. In addition, channel attention increases the weight of some feature maps which are more useful for category classification and attribute prediction. The experimental results on the DeepFashion dataset show that the proposed method can improve the accuracy of category classification and the recall rate of attribute prediction compared with the existing methods.

Images captured in outdoor scenes typically have lower contrast and visibility due to atmospheric particles suspended in the air, directly affecting the effectiveness of subjective vision and intelligent image processing systems. To solve this problem, we propose an end-to-end network, which jointly estimates transmission map and atmospheric light effectively. The global characteristics of transmittance and atmospheric light are acquired simultaneously by shared feature module, and the multi-scale convolution of the pyramid pooling module is utilized to extract the combination feature. Then, the transmittance and atmospheric light are separately estimated by two parallel branch networks. Finally, the fog-free image is obtained by inversion of the atmospheric scattering model. The experiment results show that proposed method removes haze more effectively compared to these state-of-the-art methods, and the restored image has higher contrast and more natural color. Moreover, the proposed network has less parameters.

Deep learning is an effective technical method to enhance the recognition accuracy of remote sensing image target. To solve the problem of the complicated steps in road recognition from Synthetic Aperture Radar(SAR) images, this paper proposes a SAR image road recognition method based on deep learning. First, based on the traditional Full Convolution Neural Network(FCNN), a New Convolution Neural Network(NCNN) is constructed by revising the activation function to effectively alleviate the loss of road information. Then, the NCNN is applied to road recognition experiments of simulated SAR images and real SAR images with self-constructed road label sets to improve the robustness of the proposed method. The experimental results show that the NCNN can be used to identify the overall characteristics of the road with higher accuracy and better reliability in comparison with the Support Vector Machine(SVM), traditional FCNN and other algorithms.

In the image compression process, the compression efficiency is largely affected by the information utilization. In order to improve the utilization of information, in this paper, we propose an end-to-end image compression method with information enhancement based on deep convolutional networks. In the encoding and decoding networks, a special network structure is adopted, which can both increase the forward and backward connections between the convolutional layers. As a result, the bidirectional information flow and the cyclic feedback of visual features are realized, which can effectively reduce the number of bits required for image compression. Experimental results show that, at the same compression ratio, the multiscale structural similarity index (MS-SSIM) of the reconstructed image of the proposed method can gain 0.08, 0.027 and 0.012 higher than JPEG, JPEG2000 and HEVC, respectively. The information enhancement structure based on deep convolutional network can effectively improve the coding efficiency when it is used for image compression.

The sub-pixel positioning accuracy of 2D image points in the large-sized single camera vision coordinate measurement system has a great influence on the accuracy of 3D coordinate measurement. To solve this problem this paper analyzes the centroid localization method of the optical feature points on the image plane. Based on the principle of pinhole imaging, the camera model is established. The extraction method of 2D image points corresponding to optical feature points is discussed, and the sub-pixel positioning of 2D image points is studied in depth. A centroid localization algorithm based on threshold iteration is proposed. This algorithm increases the number of effective pixels around the imaging center of the optical feature points by interpolation algorithm. It reduces the noise error by gray-scale square weighting. By setting the threshold value and several iterations, the error caused by asymmetry of the characteristic spot can be reduced, so as to realize the accurate positioning of the feature points. Through comparative experiments, it is proved that the positioning algorithm proposed in this paper can realize 1/100 pixel subdivision of the image point or more. The positioning standard deviation of the proposed algorithm in the direction of x and y is 0.0022 pix and 0.0019 pix respectively. Compared with gray square weighted centroid positioning method and double cubic interpolation gray square weighted centroid positioning method, the positioning standard deviation of the proposed algorithm is smaller, and the proposed algorithm has higher positioning accuracy. The positioning errors of x and y directions are within the range of 0.0039 pix and 0.0046 pix respectively, which are smaller than the previous two methods, and the proposed algorithm has better positioning stability. The localization method proposed in this paper can be applied to the sub-pixel localization of various near-circular small light source images.

Based on Constant Bandwidth Server(CBS), a real-time scheduling algorithm, named Reliability and Low Power Mixed-task Algorithm Based on CBS(RLPMABC), is proposed for mixing tasks to minimize energy consumption while ensuring system reliability. The algorithm selects protected periodic tasks according to utilization before scheduling, and creates backup tasks and voltage scaling with available idle time during operation, realizing the collaborative optimization between low power consumption and reliability. Experimental results show that the proposed algorithm has higher efficient energy saving effect and higher reliability than the existing algorithm.

In order to solve the material handling scheduling problem for mixed-model assembly lines based on line-integrated supermarkets, this paper presents an extreme learning machine and knowledge base-based dynamic scheduling method. First, the dynamic material handling scheduling problem is described and a mathematical model is established. Using this model, the weight sum of the output of the assembly line and the number of logistics workers is maximized under the condition of variable product ratio and weights of scheduling criteria. Also the random failure of the equipment and the instability of the cycle time are considered. Then, an extreme learning machine and knowledge base-based dynamic scheduling method are constructed. Considering the defects of extreme learning machine, an elite opposition learning self-adaptive differential evolution-based extreme learning machine is proposed. Finally, the simulation results prove the feasibility and effectiveness of the proposed dynamic scheduling method in the dynamic scheduling process.

A pulse wave classification algorithm of low complexity based on aliasing of time-frequency domain feature was proposed to solve the problems of low recognition rate and complex implementation in current research of pulse wave signal recognition. First, the time domain features of the pulse wave signal are extracted based on the Convolutional Neural Network(CNN), including the single-period feature characterizing the features of the signal segments in the period and the multi-period features characterizing the relationship between the cycles. Then the features in the frequency domain are expressed by the Mel cepstrum coefficients based on the wavelet transform. Finally, based on the aliasing and redundancy elimination of the time-frequency domain features using the neural network full connection layer, the pulse wave signal classification is realized by the Softmax classifier. The method can achieve feature extraction through low computational cost, due to the weight sharing and dimensionality reduction of CNN. In the simulation experiment based on Python platform, the recognition accuracy of the proposed method can reach 93%, which is much higher than the accuracy of traditional recognition algorithms based on time domain features or frequency domain features.

In order to assess data integrity more efficiently and accurately, the existing technologies and methods for data integrity assessment are investigated. Then according to the data characteristics of Linked Data, two kinds of algorithms are proposed, the one is β algorithm for data integrity assessing, the other one is Dam algorithm for implicit data mining. Third, the effectiveness and accuracy of the algorithms are proved by theoretical analysis. Finally, the educational administration data is published to Linked Data. The β algorithm is compared with two kinds of integrity assessing algorithms in literature on the Linked Data published. The results show that the β algorithm improves the integrity for about 6%, and the efficiency increases about 40 times on average. The accuracy and efficiency of the proposed algorithm are verified. The data integrity assessment algorithm based on Linked Data proposed in this paper can not only ensure the accuracy of data evaluation, but also greatly improve the computational efficiency.

In large-scale stream processing, resource reconfiguration in strict synchronization with changes in the stream data can guarantee the performance of application, however the resource reconfiguration cost of the system is relatively higher. Based on the idea of resource reservation and combining with the load forecasting model, this paper designs two proactive elastic resource allocation strategies from heuristic and intelligent perspective respectively, which aim to minimize the application reconfiguration cost and maximize the resource utilization. Both strategies calculate the latency of the application in the pipelined processing mode, and probabilistically guarantee the constraint of the application latency performance. Among them, the heuristic strategy uses the local threshold method for resource allocation, and the intelligent strategy adaptively allocates resources based on the reinforcement learning technology. Experimental results show that compared with the reactive strategies, these two proactive resource allocation strategies proposed in this paper both have low reconfiguration cost and high resource utilization, and the performance of the intelligent strategy is better, especially when the change of sudden load is relatively stable.

To solve the problem that the intersection and parallelism ratio (IOU) of two defect-aware data detection algorithms based on data mining and pattern matching is smaller, which leads to poor detection effect, a new automatic defect detection algorithm for virtual network sensing data is proposed on the basis of machine learning. First, the perceptual data is acquired and processed as the learning samples. Then, a BP neural network model is constructed and trained. Finally, the learning samples are used as the input of the trained BP neural network model to distinguish the defect data from normal data to realize automatic defect detection of perceptual data. The results show that the IOU of machine learning-based automatic defect detection algorithm for virtual network sensing data is 0.9588, which is closer to 1 than that of data mining-based and pattern matching-based defect sensing data detection algorithms, which shows that the detection effect of this algorithm is better.

The traditional adaptive models are constructed by geometric optics method, geometric similarity method and feature target measurement method. There exist some problems in such model, such as long perception delay and low accuracy of adaptive tracking. To overcome these problems, a component perception adaptive model based on attitude estimation algorithm is proposed. The model consists of three parts. Firstly, the sensor component is calibrated by using Zhang Zhengyou's camera calibration method based on planar template. Then, the three-dimensional model of the target object is established to track the moving object. Finally, the real-time attitude state of the target object is calculated by using the model-based attitude estimation method to realize the construction of the component perception adaptive model. Experimental results show that compared with the three traditional models mentioned above, the adaptive control of the proposed model can shorten the sensing delay of components, improve the adaptive tracking accuracy and improve the operation performance of the manipulator.

When the evaluation system designed by the current method is used to evaluate the quality of talents, the evaluation results that take a long time have a large error, and there are problems of low evaluation efficiency and low evaluation accuracy. The design method of fuzzy comprehensive evaluation system for talent quality based on multi-objective evolutionary algorithm is proposed. The overall structure of fuzzy comprehensive evaluation system for talent quality is constituted by input module, fuzzy comprehensive evaluation module, report processing module and system maintenance module. AHP is used to calculate the corresponding weights of talent quality evaluation indicators, and a fuzzy comprehensive evaluation model of talent quality is constructed. A multi-objective evolutionary algorithm is used to solve the fuzzy comprehensive evaluation model of talent quality, to achieve the evaluation of talent quality, and to complete the fuzzy comprehensive evaluation of talent quality System design. Experimental results show that the proposed method has high evaluation efficiency and high evaluation accuracy.

A model to estimate the inter-bank liabilities was construct based on the Bayesian method, and then use machine learning algorithms to construct a Gibbs sampler to sample on the basis of conditional distribution. The sampling is used for stress testing and give the probability of all possible test results. Finally, as a model application, this paper will derive the bank's default probability and discuss its sensitivity to a priori information contained in the network model, helping financial regulators assess financial institutions' default risk，reduce systemic financial risks and maintainfinancial market stability.

In this paper, a novel anti-fading method based on frequency hopping is proposed to solve the problem of multipath fading characteristics of troposcatter communication channels. The troposcatter communication model based on frequency hopping is established and the performance of this method is theoretically analyzed in fading channel. The algorithm flow of its received baseband signal processing is given. The frame coarse synchronization algorithm based on frequency hopping frame structure and the frame fine synchronization algorithm based on channel response search are adopted to solve frame synchronization problem in fading channel. The frequency domain equalization algorithm based on channel de-noising estimation and aliasing extraction is adopted to overcome channel multipath effectively and suppress noise amplification. Simulation results verify the validity and effectiveness of the proposed method.

Based on the normal mode theory, the spatial–temporal correlation, sound field structure and DOA characteristics of the incident sound field in the range of different sonar distances in shallow water channel are numerically analyzed. The receiver model is established according to the receiving model of the underwater target’s incoming sound field. The experiment of the spatial-temporal correlation characteristics of the incoming sound field in typical sonar frequency band is carried out. Theoretical and experimental studies show that the long-range wave sound field is mainly composed of low order simple positive waves, which approximates the horizontal incident in the vertical plane. The low frequency sound field has obvious time and space stability, and can predict the incident sound field.

The traditional multi-band signal source switching control algorithm can not achieve stable signal source switching in distributed communication network. To solve this problem, a remote switching control algorithm of multi-band signal source is proposed. The real-time signal is sampled by communication network monitoring equipment and the frequency band is calculated. Based on the transmission path of communication signal, the location of multi-frequency signal source is realized. The remote control equipment is installed and the trigger condition of switching control is set in it. Finally, the remote switching control of multi-frequency signal source is realized by switching the sequence of access points and the conversion control between voltage and current. In order to verify the feasibility of the designed remote handoff control algorithm, the anti-blocking performance and stability performance of the control were tested. The results show that the anti-blocking performance of the designed remote handoff control algorithm is 2.2% higher than that of the traditional handoff control algorithm, and the handoff control stability is higher.

To solve the problem of large-scale Distributed Denial of Service (DDoS) attack detection in Internet of Things (IoT) simulation environment, the Docker virtualized container technology is used to construct the IoT traffic simulation platform. First, four different types of attack traffic are generated by simulating Mirai botnet and executing commands, and normal traffic is generated by manual click and IoT experiment box auto execution. Then, statistical analysis is carried out on the original traffic to generate two different levels of datasets: packet-level and second-level. Third, three statistical analysis methods and indicators are proposed, including segmented HURST exponent, sliding-window based entropy and sliding-window based confidence interval. Finally, the DDoS attack traffic detection rules are generated by the training dataset. The experimental results show that the sliding-window based confidence interval abnormal traffic detection method can achieve an accuracy of 72.11%.

A model-free control method is proposed for trajectory tracking of multi-degree of freedom robotic manipulator to deal with the problems of relying too much on the precise mathematical model and low tracking accuracy. This method combines the adaptive neural network, the ultra-local model and the integral terminal sliding mode. First, an ultra-local model based on delay estimation is used to approximate the dynamic model of the manipulator. Then, a neural network is used to compensate the errors of delay estimation because of its nonlinear approximation capability. Finally, an integral sliding mode controller is designed for the ultra-local model to improve the convergence speed and control accuracy of the system and realize the high-precision trajectory tracking of the manipulator without relying on the dynamic model. The stability and finite time convergence of closed loop system are proved by Lyapunov theory. Experimental results show that the proposed control method can realize the high precision tracking control of the manipulator without depending on the model information completely.

This paper investigate the design of the controller of Automatic Train Operation (ATO) system under the consideration of multiple optimal objectives and constraints. Based on a nonlinear multi-point model, an ATO Nonlinear Model Predictive Control (NMPC) algorithm is proposed to meet the punctuality of train operation, energy saving and passenger comfort. Moreover, the theoretical analysis of algorithm feasibility and the proof of stability for closed-loop system are presented. The validity of the algorithm is verified by numerical simulation. The simulation results show that the proposed algorithm has better control effect and lower error than the Linear Model Predictive Control (LMPC) algorithm when the train meets the operational constraints.

In order to solve the problem of high damage rate of maize kernels and high un-threshing rate during the corn harvest in Huang-Huai-Hai region, a transverse axial flow corn threshing test system is designed, which uses flexible tooth and elastic short rasp bar as the threshing elements. The flexible tooth material is polyurethane, which has high abrasion performance, and the elastic short rasp uses torsion springs as auxiliary elements so that can mitigate the force between the corn ear and rasp bar. With reasonable hypothesis, based on the Hertz contact theory, the stress equation of corn ear under the flexible nail tooth collision is established. Based on the rigid flexible coupling system theory, the force equation of the elastic short stripe and corn ear collision is established. The above analysis indicates that flexible tooth and elastic rasp bar have real potential as new threshing elements to mitigate the force between the corn ear and threshing elements. Through image processing, the corn ear contour map is established. Discrete element analysis software is applied to improve parameters design of elastic short rasp. Simulation experiments were carried out in accordance with the feeding volume of 10 kg/s, and the simulation results improve the device design. Finally, single factor test and optimum parameter combination contrast test were carried out with the rotational speed of cylinder, concave clearance and feeding rate as the test factors. The test results show that the device is superior to the conventional threshing device. It lays the foundation for the next field experiment.

In order to solve the problem that the spiced beef is easy to deteriorate during storage and the shelf life is short, the low-temperature plasma is used to sterilize the spiced beef. The effects of low temperature plasma treatment power on 300 W, 350 W, 400 W, 450 W and 500 W, and treatment time of 3 min on the quality of modified atmosphere packaging (75%CO₂+25%N₂) spiced beef were studied during storage. The results show that when the power of the low temperature plasma is 400 W and treated for 3 min, the effect on the pH, color, protein and lipid oxidation of the spiced beef is small, and the moisture loss is the least, and the sterilization effect is the best. The shelf life of the spiced beef is extended by about 5 days.