The design of multiproduct batch plants under uncertainty with staged expansion
Abstract
The degree of uncertainty in a process design has a direct impact on the economics of the design and the flexibility required to produce products which meet required specifications. The problem of multiproduct batch design under uncertainty is of particular interest because there are more sources of uncertainty in the design of a batch plant than in the design of a continuous plant. This is due to the high degree of uncertainty arising from the complexity and typically incomplete information about the chemical/physical steps involved, the heavy reliance on timely and appropriate operator initiated actions, the need to process a large number of products in the same facility as well as the relative volatility of the demands for the specialized products typically produced in the batch mode. Since none of the aforementioned solution approaches enable the designer to formulate the design problem in its entirety, a new formulation is developed. Four key elements are identified as being necessary to accurately represent the design problem and allow for a practical solution strategy: (1) the distinction between long and short term uncertainties, (2) the distinction between hard and soft constraints, (3) practical feasibility criteria for both hard and soft constraints, and (4) the distinction between design, operating, and structural variables. The resulting MINLP includes a penalty term which represents the expected value of lost revenues due to the inability of the facility to always meet demand requirements. The objective function therefore represents a trade-off between investment cost and lost revenues. A heuristic is developed for larger problems which cannot be practically solved with the rigorous formulation. It is based on a bounding technique on the total cost of a sequence of deterministic designs, where the total cost is the sum of the investment cost of the facility and the expected revenue loss due to unfilled orders. The expected revenue loss is calculated with the demand model developed for the rigorous formulation.
Degree
Ph.D.
Advisors
Reklaitis, Purdue University.
Subject Area
Chemical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.
