Strategies for the simultaneous optimization of the structure and the process variables of a protein production plant

Process performance models for a multiproduct batch protein plant are used to exploit alternative strategies in the optimization of both the process v ariables and the structure of the plant. Simple process performance models are used to describe the unit operations, which renders explicit expression s for the size and time factor model in the design of batch plants. In the proposed approach the process variables are optimized regardless the plant structure constraints, which are left as a posterior decision. This optimiz ation is done in a single product-free intermediate storage (SP-FIS) scenar io, unbiased with any plant structure. The approach is compared to the case of recipe values for the process variables and to the best optimal solutio n for the nonconvex mixed integer nonlinear program (MINLP), which arises w hen simultaneously optimizing the structure and the process variables. This last optimization model is hard to solve and its global solution remains a s an open problem. The proposed approach generates solutions Very close to the ones obtained from nonconvex MINLP and is quite superior than simply re sorting to recipes. We also study the role of process variables in this app roach. It is found that they behave as in continuous processes by trading o ff cost components, with a smooth dependence on the overall cost. Moreover, for feasible designs that include the size and time constraints that corre spond to the plant structure, the process variables accommodate the size an d time factors to reduce idle rimes and equipment under-occupancy. (C) 2000 Elsevier Science Ltd. All rights reserved.