| Title | Keywords | ||
|---|---|---|---|
| Author | Authorship | ||
| Corresponding Author | Funds | ||
| DOI | Column | ||
| Summary | |||
| Timeframe | - | ||
| Title | Keywords | ||
|---|---|---|---|
| Author | Authorship | ||
| Corresponding Author | Funds | ||
| DOI | Column | ||
| Summary | |||
| Timeframe | - | ||
Aiming at the problems of low detection recall rate,high false detection rate and insufficient generalization ability caused by the subtle crack features and scarce crack samples in the appearance crack detection of high-speed rail components,a two-stage appearance crack detection algorithm based on improved YOLO and SimpleNet(YOLO-SimpleNet-FD) is proposed. Firstly,the C2f＿Faster＿EMA module is introduced into the backbone network of the YOLOv8 segmentation model to improve the accuracy and efficiency of component extraction. Secondly,the Ghost Bottleneck module is integrated into SimpleNet to enhance the feature extraction ability,and the Mixed Local Channel Attention(MLCA) mechanism is introduced to enhance the attention to subtle crack features. Finally,the YOLOv8 is used to precisely segment the component area to effectively suppress background interference,and then the improved SimpleNet is used to achieve crack recognition and localization. The experimental results show that the intersection over union(IoU),recall rate,false detection rate and inference speed of the proposed algorithm for component extraction reach 92.3%,95.6%,2.1% and 52 FPS,respectively. Compared with the original fusion algorithm,the IoU is decreased by 4.1 percentage points,the recall rate is increased by 4.4 percentage points,the false detection rate is increased by 2.2 percentage points,and the inference speed remains basically stable. Compared with the single SimpleNet algorithm,the recall rate is increased by 16.7 percentage points and the false detection rate is reduced by 10.9 percentage points. This algorithm provides an efficient and reliable technical solution for the appearance crack detection of high-speed rail components and has good engineering application value.
In response to the problems of long waiting times and insufficient coverage of small-batch,high-value goods along the cross-border train routes during regular transportation organization,which have led to high detention rates and poor timeliness,a "selective stopping" cross-border train transportation organization model is explored and proposed. This model incorporates the risk cost of goods delay into the optimization system of cross-border train stop schemes,with train capacity,goods demand,and stop time as constraints,and aims to minimize the enterprise operation cost and the risk cost of goods delay,thereby establishing a multi-objective optimization model. The model is solved using the fast non-dominated sorting genetic algorithm(NSGA-Ⅱ) with an elite strategy. A simulation analysis is conducted using a certain cross-border train channel as an example,and the Pareto optimal solution set is ultimately obtained. The results show that the top solution has the lowest total cost,reducing it by 38.43% compared to the bottom solution; while the bottom solution can reduce the risk cost of delay by 26.25% and serve 3.65% more container goods along the route,providing decision-makers with multiple reasonable stop schemes.
To achieve both excellent linear motion stability and curve passing ability for railway vehicles,research on the rubber-liquid composite swing arm node was carried out. The longitudinal stiffness of the swing arm node was changed by the flow of liquid in the compression chamber. The structural design of the rubber-liquid composite swing arm node was completed,its working principle was analyzed,and a theoretical model was established. Stiffness performance and fatigue characteristic tests were conducted using a test device. The results show that the dynamic-static stiffness ratio of the rubber-liquid composite swing arm node reaches 4.5,and the stiffness is significantly affected by the excitation frequency. The dynamic stiffness test curve is in good agreement with the theoretical calculation results. Liquid leakage can significantly reduce the dynamic stiffness characteristics of the rubber-liquid composite node. A pressure sensor can be embedded in the hydraulic chamber to monitor the internal pressure in real time to determine whether the swing arm node is leaking. After 12 million fatigue tests,the maximum decrease in the dynamic-static ratio was 10.7%,and no leakage or rubber aging was found. It still maintained good stiffness performance under high-frequency conditions,effectively verifying that the rubber-liquid composite swing arm node has excellent fatigue durability.
Traffic accident prediction is of great significance for improving the level of road traffic safety management. However,existing prediction methods generally have problems such as imbalanced data categories and low model hyperparameter optimization,which lead to insufficient prediction accuracy of the model. Therefore,an optimized random forest model integrating an adaptive hybrid sampling algorithm and an improved Bayesian algorithm is proposed for traffic accident severity prediction. Firstly,to address the issue of imbalanced distribution of traffic accident data,an adaptive weight calculation method is introduced on the basis of the traditional SMOTE-ENN algorithm. By dynamically adjusting the SMOTE oversampling rate and ENN undersampling threshold,targeted generation of minority class samples and elimination of majority class samples are achieved,effectively enhancing the classifier's sensitivity to accident features. Secondly,to enhance the model's generalization ability,the anisotropic Matern 5/2 kernel function is adopted to replace the traditional Gaussian kernel function,and a dynamic exploration factor is introduced to improve the efficiency of optimizing algorithm hyperparameters. Finally,the random forest model is used to complete the modeling of traffic accident severity prediction. The research results show that the K value of the proposed model after consistency test is 0.913 1,indicating that the model has high prediction consistency; the recall rate reaches 84.23%,achieving a significant improvement in traffic accident prediction evaluation compared to other classification models. This method can greatly improve the prediction accuracy of minority category accidents and significantly improve the model's generalization performance. Compared with the traditional SMOTE-ENN algorithm,the adaptive SMOTE-ENN algorithm improves the accuracy,recall rate,precision,and F1 value by 2.06,1.17,4.03,and 2.51 percentage points respectively,and has better comprehensive performance than other sampling algorithms. Compared with other optimization algorithms,the improved Bayesian optimization algorithm has better prediction performance and feasibility. The research conclusion confirms that this method can provide theoretical reference for traffic safety research.
With the wide application of electric buses,the location selection and capacity determination of charging stations have significantly affected the charging efficiency of electric buses and the costs of bus companies. The layout of charging stations and the capacity configuration for electric buses are of great significance in improving the operational efficiency of public transportation,reducing charging costs,and alleviating the pressure on the power grid. To address these issues,a charging station location algorithm and a capacity configuration planning model have been developed. Firstly,the dynamic adjustment progressive coverage function was introduced into the affinity propagation clustering algorithm,making the determination of centralized charging stations and their served bus stops more flexible. Secondly,the opportunity charging strategy and the battery capacity attenuation model were innovatively incorporated into the capacity configuration planning model,establishing an optimization model with the objective of minimizing the fixed costs of the bus company,the charging costs including opportunity charging,and the operational battery degradation cost including battery penalty costs. The CBC solver was used to solve the model. Finally,taking a part of the bus network in Shenyang as an example,the layout of charging stations and the number of charging piles at each station were determined. The reliability of the model was verified through sensitivity tests under different charging powers and environmental temperatures. The results show that by using the opportunity charging strategy and the battery capacity attenuation model,the comprehensive cost can be significantly reduced to 113 million yuan,and the battery life can be increased by 4.5%,providing a theoretical basis and practical reference for the construction of urban electric bus charging networks.
Passive zone reinforcement in deep foundation pits is a highly effective measure for deformation control and is widely applied in foundation pit support design,especially in soft soil areas where it is almost indispensable. Based on a deep foundation pit project in the soft soil area of the Nanjing Yangtze River floodplain,combined with geotechnical tests and on-site monitoring data,a three-dimensional refined finite element analysis was conducted throughout the entire construction process of the foundation pit. The study systematically investigated the influence of four different reinforcement forms and various reinforcement parameters on the deformation of the foundation pit. The research results show that passive zone reinforcement can effectively control the deformation of the foundation pit,and the control effect from strong to weak is full hall,skirt + strip,skirt,and strip reinforcement. As the reinforcement effect improves,the maximum lateral displacement position of the retaining structure and the maximum settlement position outside the pit move upward and inward. Increasing the reinforcement strength and depth can enhance the inhibitory effect of the reinforced soil on the deformation of the foundation pit,but the inhibitory effect will gradually weaken. When strict control of foundation pit deformation is required,full hall reinforcement is recommended. When economic considerations are also important,skirt + strip reinforcement is suggested. For smaller foundation pits,skirt or strip reinforcement can be adopted.
Under the intermittent train load,the ballast track bed is prone to problems such as ballast particle breakage and track bed compaction,which seriously affect the stability of the track structure.To study the dynamic behavior of the ballast track bed,the images of ballast particle samples of different shapes were reconstructed,and multi-shaped ballast particle samples that meet the gradation standards were designed.Then,a discrete element model of the ballast track bed was established to truly reproduce the coupling mechanism between ballast particles and between sleepers and ballast under intermittent train loads,and to obtain the dynamic response characteristics of the ballast track bed under intermittent train loads.The research results show that the displacement,acceleration and contact force of the ballast particles increase with the increase of train speed.The displacement amplitude at 150 mm below the sleeper is about 1.5 times that at 300 mm,and the acceleration is about 3 times that at 300 mm.Under the train load,the residual displacement of the ballast at the top of the ballast shoulder is relatively large,and the residual displacement of other parts is relatively small under the constraint of the outer wall.During the intermittent period of the train load,all parts of the track bed and under the sleeper rebound to varying degrees,and the internal contact force of the ballast is significantly reduced compared to the loading period,with a reduction of about 40%.
In response to the lightweight and high-strength design requirements of a certain type of diesel engine connecting rod,the structural optimization design of the connecting rod was carried out based on the response surface method combined with finite element analysis. Firstly,the force conditions under extreme working conditions were solved according to the motion characteristics of the connecting rod,providing boundary conditions for subsequent strength analysis. Secondly,static analysis and modal solution of the connecting rod were conducted,and the results showed that the existing connecting rod had high strength,large material redundancy,a high first-order natural frequency,no resonance risk,and a large structural optimization space. Finally,with the lightest mass of the connecting rod body as the optimization objective and static strength and fatigue strength indicators as constraint conditions,the response surface method was used to optimize the structure of the connecting rod. After optimization,the mass of the connecting rod body decreased from 43.814 kg to 40.623 kg,a reduction of 7.28%; the maximum equivalent stress decreased from 306.23 MPa to 293.33 MPa,a reduction of 4.21%,and the fatigue life and first-order natural frequency both increased slightly. This study achieved the expected optimization goals of lightweight and high strength for the connecting rod,and provided a reference for the structural optimization design of similar connecting rod-type components.
The arc generated during the opening process of a vacuum circuit breaker directly affects the service life of the contacts. To investigate the relationship between the arc morphology and the contact life,it is necessary to use a high-speed camera to collect the arc images at the breaking point of the circuit breaker and accurately capture the arc morphology during the opening process. Traditional image feature extraction methods are disturbed by the brightness differences of images at different stages,and have problems such as weak generalization ability,insufficient noise resistance and low accuracy. A deep learning-based arc morphology extraction network is designed,and a feature extraction network is built by combining dilated convolution and residual structure. The Dense-ASPP structure is introduced to fuse features,and an auxiliary classifier is adopted to improve the segmentation accuracy. The results show that this method improves the generalization ability of arc morphology feature extraction while ensuring accuracy. The intersection over union of the algorithm for arc images can reach 94%,providing data support for the subsequent study of the mechanism of arc morphology on the contacts.
Mountainous road conditions are poor and driving environments are complex,leading to frequent traffic accidents with severe casualties. To enhance the safety level of driving on mountainous roads,this study comprehensively assesses high-risk sections of traffic accidents based on the traffic accident data of mountainous roads in Guilin City from 2019 to 2023,using a kernel density estimation method that considers the severity of accidents. Firstly,the accident samples are divided into high-risk and low-risk groups based on the identification results of risk sections. Secondly,chi-square analysis is used to explore the key influencing factors of accident risk,and a Bayesian network is employed to construct a prediction model for high-risk points of traffic accidents on mountainous roads. The results show that road type,intersection and section type,and central isolation facilities have a significant impact on the severity of accidents,among which road type and central isolation facilities directly affect the probability of accident risk. The constructed prediction model can assess the probability of accident risk under different accident forms and severity levels. The research results can provide a decision-making basis for formulating differentiated accident prevention policies for mountainous roads.