Probabilistic framework for optimal experimental campaigns in the presence of operational constraints

Methodology for designing optimal experiment campaigns under hard constraints and model uncertainty. A tractable solution technique using a synergistic combination of sampling and convex optimization. The predictive capability of any mathematical model is intertwined with the quality of experimental data collected for its calibration. Model-based design of experiments helps compute maximally informative campaigns for model calibration. But in early stages of model development it is crucial to account for model uncertainties to mitigate the risk of uninformative or infeasible experiments. This article presents a new method to design optimal experimental campaigns subject to hard constraints under uncertainty, alongside a tractable computational framework. This computational framework involves two stages, whereby the feasible experimental space is sampled using a probabilistic approach in the first stage, and a continuous-effort optimal experiment design is determined by searching over the sampled feasible space in the second stage. The tractability of this methodology is demonstrated on a case study involving the exothermic esterification of priopionic anhydride with significant risk of thermal runaway during experimentation. An implementation is made freely available based on the Python packages DEUS and Pydex.