Measure This, Not That: Optimizing the Cost and Model-Based Information Content of Measurements

Model-based design of experiments (MBDoE) is a powerful framework forselecting and calibrating science-based mathematical models from data. Thiswork extends popular MBDoE workflows by proposing a convex mixed integer(non)linear programming (MINLP) problem to optimize the selection ofmeasurements. The solver MindtPy is modified to support calculating theD-optimality objective and its gradient via an external package,, using the grey-box module in Pyomo. The new approach isdemonstrated in two case studies: estimating highly correlated kinetics from abatch reactor and estimating transport parameters in a large-scale rotarypacked bed for CO_2 capture. Both case studies show how examining thePareto-optimal trade-offs between information content measured by A- andD-optimality versus measurement budget offers practical guidance for selectingmeasurements for scientific experiments.