Future prediction & estimation of faults occurrences in oil pipelines by using data clustering with time series forecasting

Future prediction & estimation of faults occurrences in oil pipelines by using data clustering with time series forecasting

Fault identification and prediction

K-means clustering

Oil pipeline

Time series forecasting and linear regression

The world of oil pipelines is subjected to serious issues due to occurrences of toxic spills, explosions and deformations like particle deposition, corrosions and cracks due to the contact of oil particles with the pipeline surface. Hence, the structural integrity of these pipelines is of great interest due to the probable environmental, infrastructural and financial losses in case of structural failure. Based on the existing technology, it is difficult to analyze the risks at the initial stage, since traditional methods are only appropriate for static accident analyses. Nevertheless, most of these models have used corrosion features alone to assess the condition of pipelines. To sort out the above problem in the oil pipelines, fault identification and prediction methods based on K-means clustering and Time-series forecasting incorporated with linear regression algorithm using multiple pressure data are proposed in this paper. The real-time validation of the proposed technique is validated using a scaled-down experimental hardware lab setup resembling characteristics exhibited by onshore unburied pipeline in India. In the proposed work, crack and blockages are identified by taking pressure rise and pressure drop inferred from two cluster assignment. The obtained numerical results from K-means clustering unveils that maximum datasets accumulated range of multiple pressures are within 16.147–10.638 kg/cm2, 14.922–12.1674 kg/cm2, 2.7645–1.2063 kg/cm2 correspondingly. Hence by this final cluster center data, inspection engineers able to estimate the normal and abnormal performance of oil transportation in a simple-robust manner. The developed forecast model successfully predicts future fault occurrences rate followed by dissimilarity rate from clustering results holds the validity of 91.9% when applied to the historical pressure datasets. The models are expected to help pipeline operators without complex computation processing to assess and predict the condition of existing oil pipelines and hence prioritize the planning of their inspection and rehabilitation.

104203

66

Bhaskaran, Priyanka E.

Chennippan, Maheswari

Subramaniam, Thangavel

Journal of Loss Prevention in the Process Industries

2020

journalArticle

0950-4230

10.1016/j.jlp.2020.104203