Knowledge-inspired operational reliability for optimal LNG production at the offshore site

• Published in: Applied thermal engineering Vol. 150; pp. 19 - 29
• Main Authors: Ali, Wahid, Qyyum, Muhammad Abdul, Khan, Mohd Shariq, Duong, Pham Luu Trung, Lee, Moonyong
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 05-03-2019; Elsevier BV
• Subjects: Computer simulation; Computing time; Decision analysis; Failure; First principles; Liquefaction; Liquefied natural gas; LNG; Maintainability; Natural gas liquefaction; Polynomials; Reliability; Reliability analysis; Reliability enhancement; SMR process; Statistical analysis; Studies; Uncertainty; Uncertainty analysis; Uncertainty quantification; Uncertainty quantification; Natural gas liquefaction; LNG; SMR process; Reliability enhancement
• ISSN: 1359-4311; 1873-5606
• DOI: 10.1016/j.applthermaleng.2018.12.165

[Display omitted]
•Reliability of the SMR liquefaction process is effectively measured.•A gPC based surrogate modeling approach is applied for uncertainty quantification.•Sobol sensitivity indices are obtained directly from the surrogate model.•Computational time is significantly reduced compared to MC/qMC approaches.
To develop a safe and profitable process, uncertainty quantification is necessary for a reliability, availability, and maintainability (RAM) analysis. The uncertainties of 3% in each key decision variables are propagated which could bring the system into an unreliable/risk region. Hence, in this study, uncertainty quantification (UQ) with simultaneous determination of sensitivity indices (SI) is proposed using generalized polynomial chaos (gPC) modeling approach. This approach reduces about 90% of the total computational time when compared with the conventional simulation approaches required for a complex first principle based model. Subsequently, a knowledge inspired reliability analysis is carried out using the uncertainty analysis (UA). By using the statistical properties of the process, for example, mean/optimal value at 50% failure give the bound between [0.7174, 0.9496] for LNG product stream. Further, it was found that LNG with 10% end flash gas (or 90% liquefaction rate) can be obtained with a failure probability of 14.43%. This value of reliability is promising for a given specified deviation; hence, the process could be assumed to be near to its reliable optimal operational region.