Near-optimal time series sampling based on the reduced Hessian

A model-based experimental design strategy is developed to select a minimal set of measured points from a time series of experiments. Based on reduced Hessian information embedded within a mixed integer nonlinear program, an optimal time series is determined that leads to well-fitted models, with ratios of parameter standard deviation to estimated value within acceptable bounds. Demonstrated on three examples, the proposed approach is effective and efficient especially for time-consuming and expensive measurements.