The Research On Reliability Of Buried Pipeline Based On The Artificial Neural Network Model

Buried pipelines, an important means of transport, are widely used in various fields. The reliability analysis on buried pipelines in liquefied soil, non- uniform settlement soil and fault sites are explored through Back Propagation Model in Artificial Neural Network Toolkit of MATLAB and Monte Carlo method.First, a database, based on the data, collected by predecessors, on buried pipelines in liquefied soil, non-uniform settlement soil and fault sites, is established through the numerical models, made by ADIN A System, about the three types of sites. The database is divided into training set and testing set. Second, the factors that damage the buried pipelines in various conditions through research of pipeline failure principle are concluded. Third, through the analysis of the sample data it finds that, with axial stress as evaluating index, in liquefied soil, the factors that affect buried pipelines are wall thickness, buried depth, pipe diameter, length of liquefied area, internal pressure, and density of liquefied soil; in non- uniform settlement soil, the factors that affect buried pipelines are thickness, buried depth, pipe diameter, soil, and precipitation level; in fault sites, the factors are pipe diameter, wall thickness, fault space, angle of fault, soil friction angle and buried depth. At last, with the Artificial Neural Network Toolkit of MATLAB, the reliability analysis of the data of the buried pipelines in the three types of conditions is examined through nonlinear limit function mapping in order to establish the Back Propagation Model and predictive model. Then let the value of predictive limit function, the forward simulation function is calculated and then the predictive stress value and actual stress value are compared by means of reliability analysis. If the error is less than 5%, then the value meets the project demand.