A nonlinear planning methodology for handling qualitative mixing constraints is presented. It is intended to account for operational specifications prohibiting the mixing of two or more chemical species which form undesirable or potentially dangerous mixtures. The approach is based on the concept of the planning island and several algorithms derived from graph theory. The methodology is provably correct and all stated constraints are satisfied, thus ensuring that only feasible operating procedures are automatically synthesized. The methodology has been successfully applied to two industrial case studies: (a) the catalyst regeneration of powerforming reactors in a petroleum refinery; and (b) the planning of recovery procedures from faults in an ethylenediaminetetraacetic acid (EDTA) plant. The synthesis of operating plans for complete chemical plants is an interesting problem and one that has many practical applications. The thrust of a computer-based planning methodology should be toward increased automation, improvement in the efficiency of planning (planning should be achieved in polynomial time, wherever possible) and guarantees that the planning methodology produces correct (feasible) plans. It is also desirable that the plans generated are optimal, or at least cost efficient in their implementation. In developing this methodology, we have addressed the first three of these considerations within a modelling framework that is easy-to-use, menu-driven and which utilizes models with which process engineers are generally familiar. There has been an implicit tradeoff between safety (correctness) and optimality. Future work should address the issue of the completeness of the methodology (i.e. if a feasible plan exists, it will be found) which can not be currently guaranteed without the exhaustive (and practically infeasible) generate-and-test approach.
ASJC Scopus subject areas
- Chemical Engineering(all)
- Computer Science Applications