ABSTRACT
This chapter illustrates the application of Gaussian process modeling, informed by chemical aqueous speciation to develop predictive capabilities. However, to the authors' knowledge, it has not been applied to geochemical systems, particularly for hybrid data-driven and physics-informed settings. In addition, the concentration is expected to vary smoothly as a function of various aqueous chemistry conditions, making the Gaussian process method appropriate to use. Each time the loop is executed, the hyperparameters associated with the virtual instrument are updated, allowing for each iteration of the autonomous workflow to enhance the performance of the virtual instrument. The chemistry constraints are applied via a scoring function that optimizes for out-of-chemistry bounds defined by predictions that exceed the total U(VI) concentration possible in the system (i.e., more adsorbed to the mineral than added to the system).
