GUANGDONG PROVINCIAL ACADEMY OF BUILDING RESEARCH GROUP Co.,Ltd.
Title: Quantitative Research Based on Parametric Simulation of Waterlogging on Bridge Deck
Abstract: With the rise of the bridge construction wave, the problem of waterlogging on bridge deck (WBD) also followed. The problems of WBD will cause serious economic losses, restrict the benign development of urban traffic and reduce the life quality of residents. The computational fluid dynamics (CFD) method is adopted to quantitatively research the WBD problem due to its high efficiency, low-cost and outstanding performance of visualization. In addition, four factors including rainfall, performance of the grates, mode of alignment change and disease characteristics have been parameterized simulated. In view of the influence of different factors on the WBD problem, corresponding suggestions in daily operation and management have been put forward in the present paper.
China University of Petroleum (East China)
Title: Short-term traffic speed prediction based on deep learning
Abstract: Short-term traffic speed prediction is an important technology for building intelligent transportation systems. This paper proposes a short-term traffic speed prediction model based on deep learning. First, based on the GPS trajectory data of Qingdao taxis, the preprocessed trajectory data is analyzed for spatio-temporal characteristics, and then according to the spatial characteristics of the trajectory data, use different methods to construct the adjacency matrix and perform weighted fusion, and use the gradient update method to construct a learnable spatial weight matrix. Finally, the matrix is applied to the graph neural network, and the prediction model is constructed together with the LSTM with the added attention mechanism. Since the space and time characteristics of the trajectory data are considered at the same time, the prediction accuracy of this model is much higher than that of traditional prediction methods, and the effectiveness of the model is also verified in the comparison of ablation experiments.
South China University of Technology
Title: Two-dimensional vehicle load spectrum of highway and urban road bridges
Abstract: Two-dimensional monitor system and method were proposed with the combination of weigh-in-motion (WIM) system and license plate recognition system of vehicles in this paper. Vehicle load information was continually recorded including vehicle license, type/ axles, weight, speed, passing time, driving lane. The optimal sample size of traffic flow series was determined according to the quantity of overload vehicles. Then, the probabilistic properties of the traffic flow series in each lane were analyzed. Basing on probability distribution transformation and trigonometric series harmonic synthesis method, simulations of the traffic flow series in each lane were carried out respectively, and vehicle load-time series (one-dimensional load spectrums) were obtained. Finally, the two-dimensional vehicle load spectrum, which is consist of the vehicle load – time/ location series, was generated with the basis of vehicle load data on the same time and section, different traffic lane. Comparing with those load spectrums basing on traditional simulation method, the two-dimensional vehicle load spectrum can better reflect the actual traffic load of multi-lane bridges. Moreover, high-accuracy safety evaluation and risk management can also be carried out.