Extending the Scope of Algebraic MINLP Solvers to Black- and Grey-box Optimization

First introduced in process systems engineering by Ignacio Grossmann in the late 70s, algebraic MINLP approaches are now a mainstream approach for many problems in process synthesis and process operations. In order to extend the scope and applicability of MINLP to simulation-based and experiment-based optimization, we have developed methodology to solve the following key problems: Converting data and simulations to surrogate models that are accurate and as simple as possible Exploitation of theoretical understanding in the building of empirical models Discovery of highly nonlinear relationships that are hidden in data This talk will discuss the above methodologies and their computational implementation in our ALAMO software.