When a high-speed train is running, it is easy for foreign objects like rail-side plastic bags to enter bottom bogies, cables and equipment gaps, which affects the safety of driving. At present, the detection accuracy of such foreign objects is low. To solve this problem, the present study used the latest deep learning based object detection networks, such as SSD and Faster R-CNN, which combined with different feature extractors to build a detection network. Through data augmentation of a sample, the number of sizes was expanded. By the idea of transfer learning, the COCO dataset was used as the source data to transfer features to the new detection network and then retrained to build a train bottom plastic bag detection network based on deep learning. It was shown that through data augmentation, the precision of each combination significantly improved. Moreover, the results using combination of SSD and MobileNet obtained the fastest detection speed with an average time of 41 ms and a precision rate of 81.3%. The accuracy using combination of SSD+Inception V2 combination reached to 89.7%, with an average time of 53 ms.

Squeeze casting (SC) is an advanced net manufacturing process with many advantages for which the quality and properties of the manufactured parts depend strongly on the process parameters. Unfortunately, a universal efficient method for the determination of optimal process parameters is still unavailable. In view of the shortcomings and development needs of the current research methods for the setting of SC process parameters, by consulting and analyzing the recent research literature on SC process parameters and using the CiteSpace literature analysis software, manual reading and statistical analysis, the current state and characteristics of the research methods used for the determination of SC process parameters are summarized. The literature data show that the number of publications in the literature related to the design of SC process parameters generally trends upward albeit with significant fluctuations. Analysis of the research focus shows that both “mechanical properties” and “microstructure” are the two main subjects in the studies of SC process parameters. With regard to materials, aluminum alloys have been extensively studied. Five methods have been used to obtain SC process parameters: Physical experiments, numerical simulation, modeling optimization, formula calculation, and the use of empirical values. Physical experiments are the main research methods. The main methods for designing SC process parameters are divided into three categories: Fully experimental methods, optimization methods that involve modeling based on experimental data, and theoretical calculation methods that involve establishing an analytical formula. The research characteristics and shortcomings of each method were analyzed. Numerical simulations and model-based optimization have become the new required methods. Considering the development needs and data-driven trends of the SC process, suggestions for the development of SC process parameter research have been proposed.

Matching status (failure or success) between logistics vehicles and cargoes during transport operation influences the decisions of the owners of vehicles and cargoes on scheduling. It is therefore essential to predict the specific matching status probability of vehicle-cargo matching (VCM). This paper defines the VCM probability, formulates the VCM probability problem, and proposes a method to predict this probability based on Bayesian network. The business (vehicle and goods resources) distribution, VCM degree, and business priority are introduced. By mapping business distribution and matching results to network nodes and matching degree as conditional probability, static and dynamic Bayesian networks for single-time sequence and multi-time sequence VCM probability prediction are constructed. A recursive algorithm is also developed to efficiently solve the dynamic Bayesian network model. The results of a prediction example demonstrate that the proposed method and model are valid and efficient. The model shows that VCM probability increases with time, but there is no obvious rule and uncertainty exists, whereas the increase rate gradually decreases with time. Further, adjusting business distribution and priorities can change the VCM probability. The proposed method provides support for the logistics information platform for scheduling decision and controlling strategy and logistics resource selection services.