Flexibility Analysis Using Surrogate Models Generated via Symbolic Regression

Computing the flexibility index to quantify the extent to which disturbances and uncertainties can affect a given process can be very challenging, especially if constraints are hard to describe algebraically or if they are unavailable as closed-form expressions. Here, we tackle the challenge of solving a flexibility index problem in the presence of such constraints by using symbolic regression. In essence, we replace those constraints in the flexibility index problem by an algebraic surrogate built using symbolic regression. This facilitates the solution process of the flexibility index problem by allowing the user to apply off-the-shelf deterministic solvers. We showcase the capabilities of our approach in a case study, discussing the pros and cons of the suggested approach relative to other existing approaches.