This article summarizes the connotation and development process of advanced combustion modes， introduces the current status of various advanced combustion modes of internal combustion engines， and focuses on the control strategies and system hardware platforms for advanced combustion closed-loop control. The role of machine learning in the control process is also explained. This article also looks forward to the future development direction of advanced combustion mode engine control. It can provide a reference for related research on the implementation of advanced combustion modes.

Based on the relevant research on asphalt foaming characteristics and moisture sensitivity of foam warm mix asphalt mixture， this paper expounded the asphalt foaming mechanism， the influencing factors of asphalt foaming effect， evaluation indicators of asphalt foaming effect and the influence of foaming process on asphalt properties were analyzed. In view of the fact that the resistance to water damage of foamed warm mix asphalt mixture is low， it is pointed out that after asphalt foaming， there will be some moisture and foam left in the asphalt. The impact of moisture on the asphalt-aggregate adhesion of foam warm mix asphalt mixture was analyzed from different dimensions， and the solutions to the existing problems of foam warm mix asphalt mixture were analyzed， summarized and prospected. At this stage， the indexes such as ER_max （maximum expansion rate） and HL（half-life） to evaluate the asphalt foaming effect have certain limitations， and the correlation between the asphalt foaming state and the mixture performance is not clear. The whole process of asphalt foaming can be studied by laser ranging， dynamic X-ray photography， image processing and other means， and the impact of relevant foamed parameters on the properties of foam asphalt binder and mixture can be analyzed. The residual water in foamed asphalt binder would weakens the bonding degree of foam asphalt and aggregate， which has a negative impact on the water stability of the mixture. Quantify the impact of residual water in foamed asphalt on the properties of foam asphalt binder and foamed asphalt-aggregate adhesion， and improving the water damage resistance of foamed warm mix asphalt mixture is the focus of future research.

In order to study the matching application characteristics of the shock absorber integrated in the AT （Automatic transmission） on the large-tonnage intelligent transportation equipment，the parameters of no-load，half-load and full-load heavy-duty vehicles and the adaptive engine were taken as the input， a complete vehicle transmission system model was established based on AMESim simulation platform. The influence of different damper torsional characteristic parameters at different transmission gears on the matching characteristics of the vehicle transmission system was analyzed in the fixed throttle opening and the accuracy of the simulation model was verified through a torsional vibration experimental platform.The results show that the vehicle weight has a significant impact on the matching characteristics of the shock absorber and the whole vehicle，and the equivalent stiffness safety region value of the shock absorber corresponding to different vehicle weights is different； The effect of the shock absorber at different gears is significantly different. When the shock absorber is at high gear，it is easier to amplify the resonance amplitude of the input speed； The second level torsional stiffness shock absorber can improve the speed fluctuation at the input end of the transmission in harsh working conditions，the first level torsional stiffness has a greater beneficial impact on torsional vibration compared to the second level torsional stiffness，the minimum damping moment was determined by analyzing the torsional vibration results under different damping moments.

In order to solve the problem of precise control of large robotic arms in the remote operation system of fusion reactors， a study on robust adaptive precision control of large robotic arms in fusion reactors is proposed. This study designs a robust adaptive sliding mode control strategy based on the Hamilton Jacobi equation to suppress the influence of uncertainty and time-varying parameters on the system， and proves the stability of the controller through Lyapunov theory. To solve the position error caused by non geometric parameters such as rigid flexible coupling on large robotic arms， a variable parameter accuracy control algorithm based on dynamic controllers is proposed， which integrates the principle of workspace grid based variable parameters for parameter identification and can achieve non geometric parameter error compensation of robotic arms. The experimental results show that this method effectively improves the dynamic control accuracy of the robotic arm and suppresses the influence of non geometric parameters.

Facing the prescribed finite-time trajectory tracking control problem with input buffer for a linkage manipulator system，the backstepping method is used to design the prescribed finite-time trajectory tracking controller. A new prescribed finite-time performance function is proposed， which can more easily regulate control performance. With the help of this performance function， a constraint control law is designed for the tracking error， and a better transient performance than the existing prescribed finite-time performance function can be obtained under the same parameter conditions for the trajectory tracking process of the manipulator system. At the same time， a control input buffer function is designed， which solves both the overvoltage problem of control input caused by the overlarge position error of the manipulator and the problem that the manipulator system is difficult to start under heavy load. The simulation experiments verify the effectiveness and superiority of the proposed controller.

In this paper， a multi-objective optimization method of structural static displacement based on projection priority selection method is proposed. Projection priority selection method is a parameter selection method that sorts the parameters according to the projection angle and projection length of the parameter sensitivity vector in a specific multidimensional subspace. The structural static sensitivity is calculated by the analytical method based on Epsilon algorithm and improved Newman series. In each iteration of the optimization method， the reasonable combination of parameters can be selected according to the given number of parameters and the modification of the corresponding parameters can be obtained. The optimal solution can be obtained through multiple iterations. This method is especially suitable for solving multi-objective structural optimization problems with limited number of parameters. In the example， the proposed method is used for multi-objective optimization of a truss structure. The optimization results were compared between the randomly selected parameter group and the parameter group selected by the method in this paper. The results show that the optimization results of the Projection priority selection method are more ideal. Examples further prove that the proposed optimization method has high engineering application value.

To address the problem of deterioration of the performance of the hydraulic vibration platform under parameter ingestion and the difficulty of combining the response speed and the anti-disturbance ability of the active disturbance rejection controller， an improved active disturbance rejection control method based on switching between the self-coupling PID control law and the sliding mode control law was proposed. The self-coupling PID control law based on the adaptive speed factor was designed to reduce the error quickly when the error is large. In the later stages of the response with small errors， the control law was switched according to the principle that the outputs should be equal at the switching point and thus the control output should avoid sudden changes， improving the robustness of the system resisting parameter ingestion by using the special property that the sliding mode function did not depend on the system model and avoiding the problem of overshooting caused by the large value of the adaptive velocity factor at small errors. The algorithm was proved to be stable by the direct Lyapunov method and was verified by experiment. The results show that the proposed improved active disturbance rejection control method effectively improves the response speed as well as the anti-disturbance ability of the direct-drive valve controlled hydraulic vibration platform.

This research offers a recognition approach based on a dual-branch neural network for recognizing driver behavior in the vehicle cockpit. The main branch of the network model employs ResNet 50 as the backbone network for feature extraction， and employs deformable convolution to adapt the model to changes in the shape and position of the driver in the image. The auxiliary branch aids in updating the parameters of the backbone network during the gradient backpropagation process， so that the backbone network can better extract features that are beneficial to driver behavior recognition， thereby improving the recognition performance. The results of ablation experiments and comparing experiments of the network model on the State Farm public dataset reveal that the proposed network model has a recognition accuracy of 96.23% and a better recognition effect on easily confused behavior categories. The study's findings are critical for understanding driver behavior in the vehicle cockpit and guaranteeing driving safety.

To improve the efficiency and quality of vehicle driveability evaluation， a driving evaluation model based on principal component analysis， limit gradient lifting tree and Sparrow optimization algorithm was proposed.The dynamic upshift of Dual clutch transmission（DCT） vehicle is taken as a typical working condition， and 18 objective evaluation indexes are studied and defined. Principal component analysis was used to reduce the objective evaluation index， reduce its redundancy and coupling. The Extreme Gradient Boosting algorithm model was trained to predict the subjective driving score， and the Sparrow algorithm was used to improve the accuracy and stability of the model. The road test shows that the accuracy of the model is 97% after the objective evaluation index is reduced by principal component analysis. It is better than BPNN（90%）， SVM（91%）， ELM（92%） and SSA-XGBoost （95%）. It is proved that the accuracy and stability of the proposed PCA-SSA-XGBoost model are better than other models， and can complete the driving evaluation more effectively.The evaluation model can be applied to other driving conditions and has application value to solve the problem of subjective and objective mapping in driving evaluation.

In order to study the effect of spindle vibration on the reliability of machine tools during service life， this paper takes a certain type of spindle as the research object， uses the sum of nonlinear bearing restoring force and rotational unbalance force as the external excitation of the spindle-bearing system， establishes a 14-degree-of-freedom dynamic model of the machine tool spindle-bearing system using the lumped mass method， and solves the dynamic equation using numerical methods. An experimental platform is built to compare the results of the experiment with the results of the dynamic model to illustrate the effectiveness of the system dynamics modeling. Then， considering the randomness of bearing parameters， a spindle vibration reliability model is established with the maximum vibration displacement of the spindle shaft end as the evaluation index. Finally， the adaptive Kriging combined with Monte Carlo simulation method（AK-MCS） is used for reliability analysis of machine tool spindle vibration. The results show that the dynamic model and reliability evaluation method proposed in this paper have high accuracy.

Aluminum alloy LF21 has low yield strength and is prone to plastic deformation， and is easy to produce burrs in micro-milling process. The size of the top burr is the largest， which has a great impact on the quality of the parts， and even causes the parts to be scrapped in severe cases. At present， the research on LF21 micro-milling burr has just started. Burr size is difficult to predict accurately， and the influence of processing parameters on burr size is still unknown. A prediction model of top burr size based on response surface method is established， and experiments are conducted to verify the validity of the prediction model. Based on response surface method， the interaction effect of cutting process parameters on top burr size is investigated. Finally， the optimization of process parameters is realized with the aim of minimizing the top burr size of micro-milling aluminum alloy LF21.

To alleviate the pressure on transfer station and enhance the service level of urban rail transport， the planning model for the cross-line operation scheme between the loop line and radial line was constructed by taking the combination of loop line and radial line as the research object. The objective was to maximize the number of direct passengers， minimize total passenger travel time and the distance traveled by trains. Taking into account constraints such as departure intervals， passenger load factors， and the number of trains， and considering passenger flow distribution based on the characteristics of the loop line， a genetic algorithm was designed that includes OD passenger flow distribution to solve the model. The case study shows that compared to independent operation， the cross-line operation between the loop line and radial line can optimize the allocation of capacity resources and improve passenger directness and service quality. When the coefficient weights of the objective function were set to a ratio of 3：1：1， 33.76% （11，000 passengers） of the transfer passengers can avoid transfers and achieve direct travel， resulting in a 1.11% reduction in total passenger travel time and a 1.05% reduction in distance traveled by trains.

This study provides a quantitative analysis of the relationship between the built environment and road traffic performance. We define traffic performance by measuring average travel speeds on road segments and develop a random forest model using multi-source panel data to predict speeds based on built environment variables. The results demonstrate high predictive accuracy， revealing a significant correlation between road traffic performance and the built environment， characterized by complex non-linear relationships. Our work addresses three limitations in prior research： overlooking dynamic traffic performance characteristics； potential chance bias in small-area studies； and insufficient exploration of non-linear relationships. These findings provide actionable insights for urban planners and policymakers to optimize transport infrastructure and management strategies.

This study conducted an automated driving takeover test using a driving simulator to collect vehicle trajectory data. Then， participants driving ability recovery time was derived by employing the Gaussian mixture model to classify the driving state as either stable or unstable. Then， the parameters of the intelligent driver car-following model for the driving ability recovery and the norm manual driving state were calibrated using the genetic algorithm. Finally， numerical simulation techniques were employed to explore the impacts of driving ability recovery time on the propagation characteristics of traffic oscillations under various intensities of traffic disturbance. Results indicated that drivers took an average of 27.25 s to recover their driving ability. In the driving ability recovery state， the desired acceleration and deceleration were higher， while the desired speed was lower. Furthermore， the takeover of automated vehicles resulted in traffic oscillation， and the duration was positively linked to the driving ability recovery time and intensity of traffic disturbance. However， the amplitude of traffic oscillation was not significantly affected by the recovery time. Moreover， both the duration and amplitude of oscillation were amplified during its propagation in the simulation fleet.

To quantify the impacts of high occupancy vehicles （HOV） entering the exclusive bus lane on road traffic efficiency， the car following and lane changing characteristics of bus， HOV and non-HOV were analyzed when setting the bus-HOV lane. Using the theory of cellular automata （CA）， and combining the control rules of the bus-HOV lane， the lane occupied by a vehicle and the type of a vehicle regarded as constraints were introduced into the asymmetric lane changing rules， and then a multi-lane CA model was established for the car following and lane changing of multi-type vehicles considering the bus-HOV lane. MATLAB software was used to accomplish numerical simulation to analyze the specific effects of the proportion of HOV， bus departure frequency and bus stop spacing on the traffic flow parameters and their relationships. The results show that： when the proportion of HOV is less than 0.5 on a three-lane road segment， setting the bus-HOV lane can not only improve road traffic efficiency but also ensure the speed of bus operation under the condition of high initial space occupancy； when the proportion of HOV exceeds 0.5， it is difficult to improve road traffic efficiency by restricting non-HOV from entering the bus-HOV lane； when the initial space occupancy is greater than 0.2， the higher the bus departure frequency is， the greater the decline of traffic volume of the bus-HOV lane is； when there are 5 bus routes to dispatch a bus every 3 minutes， compared with the low initial space occupancy， the traffic volume of the bus-HOV lane under the condition of the high initial space occupancy reduces by about 30%； the bus stop spacing more significantly affects the traffic efficiency of the bus-HOV lane， the longer the bus stop spacing is， the higher the frequency of HOV lane changing is， the more the increase of the congestion segments on the bus-HOV lane and the adjacent normal lane is， thus the traffic efficiency of the whole road is influenced.

Wading driving poses risks to the life and property safety of drivers and passengers. To improve the safety of wading driving， a technical solution combining binocular vision and laser projection was adopted. Based on the imaging law of underwater virtual images， a binocular vision algorithm for calculating the depth of water ahead was proposed， which only requires data collected by the vehicle's sensors. To verify the effectiveness of the algorithm， experimental verification is conducted. The results showed that under ideal conditions， the relative error of the algorithm's calculation results was within 3%. In practical scenarios， the methods of averaging and filtering can reduce the impact of water surface fluctuations， and the maximum relative error introduced by turbid water is about 7%. This proves that the algorithm proposed has good accuracy and robustness and can provide drivers with water depth information ahead to assist drivers in judging road conditions.

In order to improve the vehicle target detection accuracy and ranging stability in the driving environment， a deep learning-based vehicle target detection and ranging method is proposed. The YOLOX-S algorithm is used as the vehicle target detection framework for improvement： on the basis of the original algorithm， the CBAM attention module is introduced to enhance the network feature expression ability， and the confidence loss function is replaced with Focal Loss， which reduces the training weight of simple samples and improves the positive sample attention. The vehicle ranging model is established according to the imaging principle and geometric relationship of the vehicle camera， and the ranging results are obtained by inputting the coordinates of the ranging feature points and the internal reference of the camera. The improved YOLOX-S algorithm is trained and evaluated using the home-made Tlab dataset and BDD 100K dataset， and the static ranging experimental scenario is constructed to validate the vehicle ranging model. The experimental results show that the improved YOLOX-S algorithm has a detection speed of 70.14 frames/s on the experimental dataset， and the precision， recall， F₁ value， and mAP are improved by 0.86%， 1.32%， 1.09%， and 1.54% respectively compared with the original algorithm； the average ranging error stays within 3.20% in the longitudinal 50m and lateral 11.25 m measurement range. It can be seen that the method in this paper has good vehicle ranging accuracy and stability while meeting the real-time requirements of vehicle detection.

The bond strength at the interface between post-embedded reinforcement and caulking mortar is a crucial factor in determining the seismic strengthening efficacy of unreinforced masonry. To investigate this， 33 brick masonry specimens were constructed and reinforced with plain and deformed steel rebars， as well as cement mortar of various strengths. The bond performance of the strengthened specimens was investigated by means of a self-developed pull-out testing device. By analyzing the bond failure mode， bond slip curve characteristics， and bond strength， the bond behavior influencing by the likely impact of factors such as caulking mortar strength， reinforcement diameter， bond length， and reinforcement type on bond strength was evaluated， and a predicting formula of the ultimate bond strength considering various influencing factors was proposed while the predicted bond strength is also analyzed in an energy perspective. The test results showed that the strength of caulking mortar， bond length， and reinforcement type have significantly impact on the bond strength， while changes in reinforcement diameter have little effect. Specifically， the bond strength of deformed rebars has generally higher effect than that of the plain rebars under the same conditions. The bond strength of reinforcement increases with the strength grade of caulking mortar and decreases as the bond length increases. The rule of thumb for the variation of interfacial bond failure energy with the strength of the embedded mortar is consistent with the test data.The research findings provide important theoretical guidance for the wider use and application of embedded reinforcement.

Three-point bending static loading tests were carried out on five test beams to study the effect of basalt fiber fabric-reinforced cementitious matrix （B-FRCM ） on the shear reinforcement effect of reinforced concrete beams. The failure modes of the test beams and the shear contribution of basalt fiber reinforced polymer （BFRP） grids in the process of loading were analyzed， and the test results were compared with the collected calculation models. The results show that the shear capacity of RC beams strengthened with B-FRCM is obviously improved， and the increase amplitude is 24% ~33%. B-FRCM reinforcement can effectively inhibit the development of inclined cracks and delay the occurrence of limit state. In the four existing calculation models， the shear contribution of BFRP grids is overestimated.

In order to understand the high and low temperature viscoelastic behavior of carbon nano powder modified asphalt， this paper will analyze the creep and relaxation behavior of carbon nano powder modified asphalt based on multi-temperature and multi-stress creep recovery test and low-temperature relaxation test， combined with the principle of time-temperature equivalence. Subsequently， the low-temperature relaxation behavior of carbon nano powder modified asphalt was inverted and verified by combining the Laplace convolution integral. The results show that the carbon nano powder modified asphalt has stronger elastic stability and lower stress sensitivity at low and medium temperatures. The short-term creep flexural volume is small， the long-term creep flexural volume grows significantly， and the stress relaxation effect is significant. The Pearson correlation between the inverse relaxation modulus method proposed in this paper and the modulus curve of low-temperature relaxation test is 0.911. This study will provide experimental and theoretical references for the application of carbon nano powder modified asphalt， and at the same time， provide an effective inverse relaxation behavior of low-temperature relaxation of carbon nano powder modified asphalt.

To study the seismic behavior of fire-fired composite shear wall with double steel plates and infill concrete after reinforcement， firstly， 1 piece of composite sheer wall of room temperature and 3 pieces of composite sheer wall of high temperature （two of the combined shear walls were reinforced after fire） were tested under quasi-static loading， and the failure form， hysteresis curve， skeleton curve， energy dissipation and stiffness degradation of the specimen were evaluated. Then， based on the test results， the finite element models were established by the finite element software ABAQUS. At last， the composite shear wall was parametrically analyzed. The results show that the failure mode of the specimens are all bending failure， the limit displacement angle is between 1/63 and 1/45， which meets the requirements of the angle of the frame-core tube structure， and the displacement ductility coefficient is between 2.18 and 3.1， indicating that the composite shear wall has good seismic performance. The finite element simulation results were reasonable and reliable. Parametric analysis shows that the ultimate lateral bearing capacity of the specimen decreases with the increase of the axial compression ratio， the degradation effect of axial pressure ratio on bearing capacity will be significantly amplified under the action of fire， and increasing the cross-sectional steel matching ratio can significantly improve the lateral stiffness and shear bearing capacity of the specimen. Since the steel plate on the fire side has the highest degree of strength degradation after a fire， the fire side must be reinforced.

In order to study the influence of small dip angle（0°<θ<30°） contact surface on the strength characteristics of pile-soil composite structure under different influencing factors， the unconsolidated undrained triaxial test on clay-concrete assembled specimens of variable inclination was carried out by TSZ-2 fully automatic triaxial instrument， and the shear strength and failure mechanism of the clay-concrete combined body under different confining pressures and moisture content were discussed. Finally， according to the shear characteristics of the small inclination clay-concrete combined body， the statistical damage model of the small-angle clay-concrete composite was deduced in the whole process. The results show that when the interface inclination of the clay-concrete combined body is 0°<θ<30° and the water content of the soil is 17%， increasing the interface inclination angle of the clay-concrete combined body can improve the shear strength， and as the water content gradually increases， the larger the interface inclination angle of the clay-concrete combined body， the smaller the shear strength of the clay-concrete combined body. The deformations during triaxial loading of the small-angle combined body can be divided into contact surface slip phase， specimen elastic compression and contact surface shear coupling phase， soil sample elastoplastic shear phase. The larger the inclination of the clay-concrete combined body， the bigger the initial damage value under the same conditions， besides， its total damage value increases with the increase of axial strain， and the initial damage value is also affected by the surrounding pressure and water content. Finally， the experimental data and theoretical values of the model are compared， and the overall fit is consistent， which verifies the accuracy and reasonableness of the model.

In order to study the change characteristics of mechanical properties and frost expansion characteristics of canal subsoil under the change of moisture and temperature and the frost expansion response of lined channel under the change of moisture and temperature， the mechanical properties and frost expansion characteristics of canal subsoil were tested， and the change rules of mechanical properties and frost expansion characteristics of canal subsoil were studied under the condition of different moisture and different temperature. And based on this， using finite element numerical simulation， based on the water-heat coupling change， the lining channel frost expansion response under the frost expansion of the channel subsoil is studied. The results show that the mechanical properties and freezing and expansion characteristics of the channel subsoil show complexity under hydrothermal coupling. When the temperature difference is small， the moisture difference has little effect on the frozen deformation of lined channels. When the temperature difference is large， the hydrothermal coupling has a significant effect on the freezing and expansion deformation of lined channels. Under -2℃ and -12℃， the maximum freezing deformation of the lined channel is 54% and 70% larger than that of the bottom of the channel. The larger the temperature difference， the larger the gap between normal and tangential freezing expansion force under different temperature conditions. The maximum normal and tangential freezing expansion force at -5℃ is 175% and 173% larger than that at -2℃， and the maximum normal and tangential freezing expansion force at -15℃ is 408% and 200% larger than that at -5℃. Moisture difference has a small effect on frost heave compared to temperature difference. Under the hydrothermal coupling， the distribution of freezing and expansion of channel foundation soil has non-uniform characteristics and the development and distribution of cracks in lined channels have certain regularity. The larger the temperature difference is， the larger the gap between the freezing and expansion stresses of lined channels under different hydrothermal conditions is. The maximum freezing expansion stress at -6℃ is 57% larger than that at -4℃. The maximum freezing expansion stress at -14℃ is 183% larger than that at -6℃. The lower the temperature， the more moisture， the more significant effect on the freezing and expansion stress of lined channels. This study can provide a reference for the design and structural optimisation of anti-freezing and expansion of lined channels under the change of hydrothermal coupling.

In order to predict the 28d compressive strength and permeability coefficient of porous concrete， the water-cement ratio， cementitious material dosage， bone-cement ratio and measured porosity were determined as model input parameters through correlation analysis. Furthermore， the prediction models for the two indicators are established based on the constructed dataset using artificial neural networks and regression analysis， respectively. The results showed that the statistical regression models had significant prediction errors， lacked sensitivity to multivariate responses， and their goodness-of-fit R² were all below 0.681. The artificial neural network model is more suitable for solving complex and multivariate performance prediction problems. The neural network compressive strength prediction model optimized by the genetic algorithm has a goodness-of-fit R² of more than 0.9， which reflects the accuracy and stability of the prediction model. The research results could provide reference and guidance for the optimal design of the mix ratio and performance regulation of porous concrete.

In order to achieve rapid column-to-column connection in multi-story precast composite column structures， this study proposes a new flange connection method using flange plates and high-strength bolts for small-section steel tubular concrete core columns with a diameter not exceeding 400 mm. The method involves adjusting the transition plate and applying adhesive steel glue to compensate for the precast column length error， flange plate inclination error， and irregularities in the concrete inside the steel tube. The core concrete-filled steel tubes columns can bear vertical loads during construction， and the flange connection is simple， convenient， and has high compression stiffness and load-bearing capacity. To verify the reliability of this new flange connection， four specimens were designed in this study， representing four conditions： without connection section， normal connection， concave concrete inside the steel tube， and concave concrete inside the tube with inclined flanges. The influence of various adverse construction factors on the axial load-bearing capacity of the flange was examined. Axial tensile and compression tests and finite element analysis were conducted to obtain the axial load-bearing capacity and failure mode of the specimens. Parametric analysis of flange-connected steel tubular concrete core columns with uneven and inclined concrete inside the tube was performed using finite element numerical simulation. The research results demonstrate that the proposed precast steel tubular concrete core column with flange connection has high stiffness， load-bearing capacity， and ductility， ensuring reliable force distribution at the connection points. Increasing the steel tube wall thickness does not lead to connection failure， indicating a significant safety margin for the compression load-bearing capacity of this new flange connection. The irregularities in the concrete inside the steel tube have minimal effect on the initial stiffness and peak load-bearing capacity of the core column. However， excessive inclination of the flange plate reduces the initial stiffness and peak load-bearing capacity of the core column， which should be considered during construction.

A new method based on deep learning， UCEfold， is proposed for predicting ribonucleic acid （RNA） secondary structure using both “sequence” and “image” as input to the deep learning model to extract hidden features. It also added some prior knowledge to the model to improve the prediction accuracy. There have tested the UCEfold model on both RNAStralign and ArchiveⅡ datasets， and the results show that UCEfold outperforms the traditional method significantly， and can predict the RNA sequences with pseudoknots more accurately and has strong generalization ability， which effectively solves the bottleneck of the traditional algorithm with high complexity， low efficiency and inability to predict pseudoknots.

In order to address the entity relation extraction of overlapping triplets in complex contexts， a joint extraction method called LARE is proposed， which is based on local attention and local remote supervision. First， the pointer hierarchical sequence annotation scheme is designed to solve the problem of entity overlap of triples. Second， a local attention mechanism is proposed， which uses a sliding window to calculate attention to focus on local details. Finally， A knowledge base is established， which is annotated by remote supervision during training， and some entity relationship pairs are randomly replaced to generate new training sentences， so as to enhance the fitting ability of the model. Through comparative experiments，the F₁ value of the proposed method on the Baidu dataset and Thangka dataset reaches 81.49% and 53.07%， which improves the performance of relationship extraction in the case of entity overlap.

For the problem of allocating bandwidth resources as well as computational resources of edge servers during data collection and transmission in smart fishery， the optimal solutions of bandwidth as well as computational resources are obtained by solving the nonlinear equations with constraints for minimizing the time delay. The effect of optimization algorithms such as particle swarm algorithm， immune algorithm， simulated annealing algorithm， gray wolf algorithm， firefly algorithm， and sequential quadratic programming algorithm is compared when the computational resources are 10 Gcycles/s and the number of tasks is 40， and the sequential quadratic programming algorithm has the smallest latency of 73.16 s. By preferentially allocating resources to the task improvement algorithm with small latency to latency constraint ratio， the results show that the time delay of the improved algorithm is 56.91 s， which is 16.25 s less than the sequential quadratic programming algorithm， and the improved algorithm achieves better results in the problem of bandwidth and computational resource allocation optimization.

Due to the insufficient search accuracy of ant colony algorithm in global path planning in the real environment and its tendency to fall into local extremum， excessive iteration times， and single solving objective， this paper proposes a dual starting point ant colony algorithm. By simulating a grid map and adjusting the starting position based on the ant colony algorithm， we improved the traditional ant colony algorithm's fixed one-way moving search method by setting two starting points. On this basis， we also optimized the pheromone update strategy to guide the next iteration process， which can shorten the path search time and reduce the blindness of the ant colony algorithm in the early search， and as the number of iterations increases， the diversity of solutions that the algorithm can obtain in the early stage increases， increasing the probability of obtaining the optimal solution， and also avoiding the problem of the algorithm easily falling into local optima， accelerating the convergence speed of the algorithm in the later stage. Through simulation experiments on the dual starting point ant colony algorithm， it is shown that the algorithm has good optimization performance and iterative convergence.

In order to solve the problem of low peak signal-to-noise ratio and structural similarity of reconstructed face images， a super-resolution reconstruction method based on generative adversary-network is proposed. A two-branch generator network is created， the shallow features of the image are obtained by using the feature extraction module in the network， the high-frequency branches of the image are reconstructed， the high-frequency data of the low-resolution image is trained by the deep residual network， and the face image data is mapped. Stop updating fixed generator parameters， use relative discriminator network to identify the authenticity of face image reconstruction， introduce attention mechanism to optimize the reconstruction performance of generative adversarial network， evaluate the reconstruction reliability by perception loss function， and realize the super-resolution reconstruction of face. The experimental results show that the proposed method can effectively improve the ability of facial detail reconstruction， and the subjective and objective evaluation indexes are excellent. The subjective evaluation results show that the reconstructed facial images with the proposed method have clearer detailed data， more three-dimensional facial contours， less distortion rate， and more natural and real external features. The objective evaluation results show that the peak signal-to-noise ratio of the proposed method reaches 44.6dB and the structural similarity reaches 0.74， which fully demonstrates the robustness of the proposed method and provides an effective reference for optimizing the application performance of face recognition.

3D convolutional neural networks can achieve good performance by directly fusing spatial and temporal features but are computationally intensive. Conventional 2D convolutional neural networks perform well in image recognition， but their inability to extract temporal features leads to poor performance in video action recognition. To this end， a plug-and-play temporal and motion enhancement module is proposed to learn spatiotemporal relationships for video action recognition and can be inserted into 2D convolutional neural networks with limited extra computational cost. The extensive experiments on several action recognition datasets demonstrate that the proposed network outperforms the state-of-the-art 2D convolutional neural networks with high efficiency.

In this paper， we consider a communication system combining multi-tags ambient backscatter with non-orthogonal multiple access technology. First， on the basis of multi-tags， we propose an optimal tag selection scheme in which the best tag is selected for transmission among M tags placed between the base station and the far user， to analyze the effect of M tags on the far user outage performance. Also， the difference in the near user outage performance produced by using ideal and non-ideal successive interference cancellation is analyzed at the near user side. Then， exact and asymptotic outage probabilities and throughput expressions are derived for users， as well as it is proved that full diversity orders are available for the far and near users. Finally， the correctness of the theoretical analysis is verified by simulation， and the effects of key parameters such as the number of tags， residual interference coefficient of non-ideal SIC and target rate on user performance are obtained. When the number of tags M=\mathrm{5}， the signal-to-noise ratio \gamma =-\mathrm{28}\text{?}\mathrm{d}\mathrm{B}， under the same conditions with other parameters fixed， the outage performance of far users is improved by about 10% over single tag.

An improved hp adaptive Radau pseudospectral method is proposed for the multi-constrained attitude optimal manoeuvre planning problem of high-resolution flexural remote sensing satellites for Earth observation missions. The improved method is proposed to use a two-layer optimisation iteration strategy， with the residuals of the differential-algebraic constraint at sampling points with Gaussian distribution as the error evaluation criterion for optimisation and the quarter of the ratio of curvature maximum to standard deviation at each sampling point as the refinement criterion for quadratic optimisation， which together improve the efficiency of the solution of the method. Firstly， a rigid-flexible coupled state space equation for the optimal control method is established by combining the satellite dynamics， kinematics and the vibration equations of the flexural attachment， and secondly， an improved hp adaptive Radau pseudospectral method is used to transform the optimal control problem into a general non-linear programming. The problem is solved using a modified hp adaptive Radau pseudospectral method. The simulation results show that the method can generate optimal trajectories satisfying all constraints， and it takes less time and is more efficient than the traditional hp adaptive pseudospectral method with the same accuracy. The method provides a design basis for high-resolution remote sensing satellites to perform complex space missions， and has engineering significance.

Tractors with traditional architecture generally have a series of problems， such as low energy efficiency， difficulty in traction control， power interruption during gear shifting and so on. This paper designs a new architecture of the tractor powertrain system based on TRIZ theory to solve these problems. Firstly， the paper analyzes and establishes the "substance-field model" of key problems such as power interruption during gear shifting， and sliding of the driving wheel of tractors with traditional architecture， etc. Then， the "conflict solving principle" is adopted to solve the conflicts involved in the design of the tractor powertrain system， and the final optimization scheme of the tractor powertrain system innovative design is obtained. Finally， the feasibility of the proposed architecture is verified by simulation tests and prototype tests. This new powertrain system has the dual advantages of distributed drive and hybrid drive， and the power output of each drive shaft of the tractor is decoupled. The optimized tractor has a better energy-saving effect and more drive modes. This paper provides the theoretical basis for the architecture design of a tractor powertrain system， and the distributed hybrid electric tractor system architecture can provide a new idea for the efficient operation of tractors.

The adsorption performance of bionic sucker based on the excellent adsorption of abalone abdominal foot was studied from the perspective of bionics. The surface morphology of the abdominal foot was extracted to design the bionic sucker， and the adsorption of the sucker was simulated and analyzed. The bionic sucker with good adsorption was selected for tensile test， and the adsorption mechanism was explored according to the experimental results. The width of the sealing ring has a great effect on the Mises stress of the bionic sucker， while the distance between the groove and the center of the sucker and the number of grooves distribution have little effect. The stress on the sucker bottom with a sealing ring width of 1.5 mm is greater than that of the bionic sucker with a width of 3 mm. The bionic sucker with the width of sealing ring of 1.5 mm， the distance between groove and center of sucker of 20 mm， and the number of grooves distributed is 12 has the largest adsorption force under 40% vacuum degree， and its maximum adsorption force is 5.32% higher than that of standard sucker. The bionic sucker can improve the adsorption performance of the sucker. The bionic sucker sealing ring structure can effectively prevent the edge of the sucker from shrinking inward and the groove structure can delay the connection between the inner cavity of the sucker and the outside， which is the key to improving the adsorption performance of the sucker.

A contact bionic decontamination structure was designed for the hard marine pollutants on the surface of marine facilities. The decontamination performance of different structures was tested by theoretical model and practical experiment， and the decontamination mechanism was analyzed according to the test results. The results show that when the maximum diameter （D_m） and the longitudinal reduction ratio （Y*） are the same， the diameter reduction ratio （D*） is between 6%~8%， the structure has better decontamination performance， and when D_m and D* are the same， the decontamination performance increases with the increase of Y*. Analysis of the reasons shows that D* and Y* are the main bionic structural parameters that affect tangential scraping force in the process of decontaminating. Proper design of D* and Y* can improve the decontaminating performance of the bionic decontaminating structure. Combined with simulation and actual test， it can be seen that the structure with D*=8% has the best decontamination performance.