A Prognostics Method Based on Back Propagation Neural Network for Corroded Pipelines

Abstract

A method that employs the back propagation (BP) neural network is used to predict the growth of corrosion defect in pipelines. This method considers more diversified parameters that affect the pipeline’s corrosion rate, including pipe parameters, service life, corrosion type, corrosion location, corrosion direction, and corrosion amount in a three-dimensional direction. The initial corrosion time is also considered, and, on this basis, the uncertainties of the initial corrosion time and the corrosion size are added to the BP neural network model. In this paper, three kinds of pipeline corrosion growth models are constructed: the traditional corrosion model, the corrosion model considering the uncertainties of initial corrosion time and corrosion depth, and corrosion model also considering the uncertainties of corrosion size (length, width, depth). The rationality and effectiveness of the proposed prediction models are verified by three case studies: the uniform model, the exponential model, and the gamma process model. The proposed models can be widely used in the prediction and management of pipeline corrosion.

Interpretable machine learning for maximum corrosion depth and influence factor analysis

“…The candidate for the number of estimator is set as: [10,20,50,100,150,200,250,300]. The candidates for the loss function, the max_depth, and the learning rate are set as ['linear', 'square', 'exponential'], [3,5,7,9,12,15,18,21,25], and [0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1], accordingly. The total search space size is 8×3×9×7.…”

“…The predicted values and the real pipeline corrosion rate are highly consistent with an error of less than 0.1%. Xie et al 15 and Liao et al 16 employed the BPNN to predict the growth of corrosion in pipelines with different inputs. Meanwhile, other neural network (DNN, SSCN, et al) models were widely used to predict corrosion of pipelines as well [17][18][19][20][21][22] .…”

Surface Roughness Prediction of Titanium Alloy during Abrasive Belt Grinding Based on an Improved Radial Basis Function (RBF) Neural Network

“…The nonlinear mapping issue in feedforward networks can be resolved by an intermediary layer, according to research and analysis [28][29][30]. As a result, there was just one intermediary layer in the RBF neural network used in this study.…”

Editorial for the Special Issue on Advanced Manufacturing Technology and Systems

“…Yu et al [ 16 ] proposed an efficient algorithm using Field Programmable Gate Array (FPGA)-accelerated You Only Look Once (YOLO) v3 based on an attention mechanism. Xie et al [ 17 ] proposed a prognostic method based on a back-propagation neural network for corroded pipeline systems. The rationality and effectiveness of the proposed prediction models were verified.…”