Muhammad Abdul Qyyum

Surrogate-assisted modeling and optimization of a natural-gas liquefaction plant

Abstract In this study, surrogate-assisted modeling and optimization of the single mixed refrigerant process of natural-gas liquefaction is presented. The mixed refrigerant liquefaction process is highly nonlinear owing to the involved thermodynamics that increase the computational burden of any optimization algorithm. To address the computational-burden issue and obtain the results in a reasonable time for the complex single mixed refrigerant process, an approximate surrogate model was developed using a radial basis function combined with a thin-plate spline approach. Even with the reduced model, all the results obtained were comparable with those by rigorous first-principle models. This confirms that all the important characteristics of the model are correctly captured, and the surrogate models of the liquefaction plant are acceptable replacements of first-principle models, especially in computationally demanding situations.

Measuring the effect on chemical processes due to uncertain input states: Uncertainty-cum-sensitivity analysis using a gPC approach

Abstract In this study, uncertainty quantification (UQ) of a surrogate model of the complex chemical process was carried out using generalized polynomial chaos (gPC)-based approach. Results show that the proposed approach can save about 90% computational time compare to conventional Monte Carlo (MC) and quasi-MC (QMC) based simulation approaches. Sensitivity analysis (SA) of parameters was also performed simultaneously, which further enhances the time efficiency. Uncertainties quantification (UQ) besides the parametric sensitivity analysis (SA) is an important part of the design of the process. As in the case of optimization, the constrained violation always occur hence, a little deviation of the optimized variables will make the system worse/unreliable.