Iterative model-based optimal experimental design for mixture-process variable models to predict solubility

Simple Method for Incorporating Prior Parameter Information during Model-Based D-Optimal Design of Experiments for Multi-Response Chemical Process Models

“…4,17,34−36 The A-, D-and E-optimal design criteria focus on improving the accuracy of parameter estimates, 5,17,23,32,34−38 whereas G-and V-optimal design criteria focus on obtaining accurate model predictions at user-defined operating conditions of interest. 17,29,32,36,39 The D-optimal design criterion is the most popular MBDoE criterion in the chemical engineering literature and, as such, is the focus of this article. 4,17,38,40−46 The objective function for D-optimal MBDoE is 5,28,[34][35][36]47 J FIM det( )…”

“…A variety of MBDoE techniques exist for selecting new experimental settings. ,,− The A-, D- and E-optimal design criteria focus on improving the accuracy of parameter estimates, ,,,,− whereas G- and V-optimal design criteria focus on obtaining accurate model predictions at user-defined operating conditions of interest. ,,,, The D-optimal design criterion is the most popular MBDoE criterion in the chemical engineering literature and, as such, is the focus of this article. ,,,− The objective function for D-optimal MBDoE is ,,− , J D = det ( FI boldM − 1 ) where FIM is the Fisher information matrix and det is the determinant. The FIM contains information about how sensitive model predictions are to the values of the parameters and therefore is crucial to both parameter inference and MBDoE calculations .…”

“…Performing additional experiments can be time-consuming and expensive, so researchers want to carefully select operating conditions so that as much information as possible is gleaned from the new experimental runs. Model-based design of experiments (MBDoE) is an efficient way to plan new experiments that can lead to accurate parameter estimates and improved reliability of model predictions. ,,− Unlike the traditional statistical design of experiments, MBDoE considers the model structure, parameter initial guesses, parameter uncertainties, and measurement uncertainties. ,,, Sequential MBDoE methods, which are the focus of the current article, also consider any preliminary data that might be helpful in selecting new run conditions.…”