10.1007/s00521-021-06738-5

NEURAL COMPUTING & APPLICATIONS

Underground pipelines are an indispensable part of urban public facilities. However, the frequent occurrence of pipeline accidents in recent years has not only brought great inconvenience to people's lives, but also affected people's lives and property safety to a certain extent. Therefore, timely treatment and treatment are very important. Preventing sudden underground pipeline accidents plays an important role in improving urban livability. This article studies pipeline risk big data intelligent decision-making systems based on machine learning and situational awareness. In this paper, by analyzing the application scope of gas leakage and diffusion models under different modes, leakage, diffusion, fire and explosion models are determined, and a combined model framework of leakage accident consequence system analysis is formed. The system uses the pipeline failure probability model and the pipeline failure consequence analysis model to determine the pipeline failure probability, the probability and the consequences of each accident; it uses the spatial analysis ability of GIS technology to determine the accident impact area and displays the impact area in graphics form. Through the effect verification of the test set, the prediction result of the SVR model based on the grid search parameter, the relative percentage error of the predicted value of each sample and the true value fluctuate is in the range of 4%-36%, and the amplitude is not very large. Most of the error values are approximately 13.56% of the MAPE value. The results show that the optimization method using grid search parameters can have better prediction performances.

Journal paper

34

18

0941-0643

no

2021-01-01