Premium
Leak detection in liquefied gas pipelines by artificial neural networks
Author(s) -
Belsito Salvatore,
Lombardi Paolo,
Andreussi Paolo,
Banerjee Sanjoy
Publication year - 1998
Publication title -
aiche journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.958
H-Index - 167
eISSN - 1547-5905
pISSN - 0001-1541
DOI - 10.1002/aic.690441209
Subject(s) - pipeline transport , spurious relationship , leak , pipeline (software) , artificial neural network , sizing , engineering , petroleum engineering , marine engineering , computer science , artificial intelligence , mechanical engineering , machine learning , environmental engineering , art , visual arts
Abstract A leak detection system for pipelines was developed by using artificial neural networks (ANN) for leak sizing and location and by processing the field data. This system can detect and locate leaks down to 1% of flow rates in pipelines carrying hazardous materials in about 100 s. A reference pipeline was considered for practical implementation of the package. The ability of the package to withstand spurious alarms in the event of operational transients was tested. The compressibility effect, due to “packing” of the liquid in the pipeline, causes many such spurious alarms. Adequate preprocessing of the data was performed by using a computer code in conjunction with the ANN to compensate for the operational variations and to prevent spurious alarms. The package detects leaks as small as 1% of the inlet flow rate and correctly predicts the leaking segment of pipeline with a probability of success that is greater than 50% for the smallest leak. In all cases, the timely response of the system was seen as a major advantage.