BI Ao-rui,LUO Zheng-shan,QIAO Wei,SUN Yang-yang.Prediction of Pipeline Inner-corrosion Based on Principal Component Analysis and Particle Swarm Optimization-support Vector Machine[J],47(9):133-140 |
Prediction of Pipeline Inner-corrosion Based on Principal Component Analysis and Particle Swarm Optimization-support Vector Machine |
Received:May 03, 2018 Revised:September 20, 2018 |
View Full Text View/Add Comment Download reader |
DOI:10.16490/j.cnki.issn.1001-3660.2018.09.018 |
KeyWord:inner-corrosion corrosion factors corrosion rate prediction principal component analysis support vector machine catfish particle swarm |
Author | Institution |
BI Ao-rui |
1.School of Management, Xi'an University of Architecture &Technology, Xi'an , China |
LUO Zheng-shan |
1.School of Management, Xi'an University of Architecture &Technology, Xi'an , China |
QIAO Wei |
2.Xi’an Research Institute, China Coal Technology and Engineering Group Corp, Xi'an , China |
SUN Yang-yang |
3.China Huaneng Clean Energy Research Institute, Beijing , China |
|
Hits: |
Download times: |
Abstract: |
The work aims to improve the maintenance strategy and service life by studying the prediction model for inner-corrosion of metal pipelines. The causes and factors for inner-corrosion of pipeline were analyzed and summed up. The corrosion factors were filtered by the principal component analysis. Those factors related to each other but contributing to low corrosion were abandoned. The causes for corrosion were characterized at maximum and some unnecessary analysis processes were reduced. The factors contributing to large corrosion were used as the input variables to the support vector machine prediction model, and the corrosion rate was used as the target output to build the pipeline corrosion prediction model Aiming at parameter selection of support vector machines, catfish particle swarm optimization algorithm was applied to improve the prediction accuracy. 20# pipeline was taken as an example to verify the model. The model was contrasted and compared with other prediction models. The main factors for inner-corrosion of pipeline selected by the main component analysis were H2S, CO2, Cl−, pH, pressure, medium temperature and current velocity. The average relative error between the predicted value of support vector machine model improved by catfish particle swarm optimization algorithm and the actual value was 2.82% and the correlation coefficient value was 0.9955, which were better than those in the other three prediction models. The inner-corrosion of pipeline is formed by multiple corrosion factors, and the inner-corrosion rate can be predicted accurately by the support vector machine model adopting main component analysis and particle swarm optimization, which provide a reference for the maintenance and management of the metal pipelines. |
Close |
|
|
|