Global sensitivity analysis of computer-aided molecular design problem for the development of novel working fluids for power cycles

Abstract This study compares two methods for global sensitivity analysis as a new approach for the identification and ranking of target properties in molecular design problems: A modified Morris Screening technique and Monte Carlo based standard regression. The two methodologies are highlighted in a case study involving the design of a working fluid for an Organic Ranking Cycle (ORC) design for power generation. Morris Screening is found to be favorable over Monte Carlo based standard regression. Monte Carlo based standard regression cannot be applied, because the current model cannot be sufficiently linearized. For Morris Screening techniques the critical temperature, the critical pressure and the acentric factor of the working fluid has been identified as the target properties with the highest sensitivity to the net power output of the cycle.