2025 AIChE Annual Meeting

(391c) Enhancing Dynamic Operability Mapping Computations

Authors

Chrysanthos Gounaris, Carnegie Mellon University
Fernando Lima, West Virginia University
Dynamic operability is an attractive method to study the inherent feasibility and controllability of a dynamic process. Dynamic operability analysis can be performed by forward input-output mapping from an initial state and analyzing snapshots of the output spaces at discrete time intervals. For an asymptotically stable process, considering all input set combinations, the output set grows in terms of area/volume/hypervolume, developing a scenario tree until it becomes invariant when the final steady state is achieved. Such a terminal state is desired to be identified to obtain a complete set of dynamic solutions. Extensive simulation methods considering state transitions for all possible inputs to obtain the output snapshots have proved intractable due to the curse of dimensionality. This has been verified in this work by considering a two-input-two-output mixed tank process, where the extensive method fails to calculate and geometrically map the output spaces beyond five discrete time intervals. To overcome this challenge, a novel approach is proposed in this work to map a subset of data points by considering only the set of states corresponding to the geometric boundaries of the output spaces, which can be identified using computational geometry techniques. Considering this state-space subset allows us to perform forward mapping in a computationally efficient manner until the terminal steady state is reached, which is identified by checking for time invariance of the hypervolume of the output sets at subsequent time intervals.