WORST-CASE IDENTIFICATION IN STRUCTURED PROCESS SYSTEMS

An improved method for identifying the worst-case set of parameter combinations in controllable systems is presented that is based on branch and bound search in the space of the uncertain parameters. A characteristic of controllable structured systems frequently appears wherein the limiting condition is localized in a part of the system that has only limited coupling to the other parts. While there may be many uncertain parameters varying in parts upstream of the limiting constraints, often there will remain sufficient control degrees of freedom in those other parts to compensate. The branch and bound method in its simplest form cannot recognize this situation, and will usually engage in searching among combinations of such parameters, when in the end they have no effect on the limiting condition. Modifications to the basic method are presented that allow these parameters to be recognized a priori. The complexity of the branch and bound search is reduced accordingly, while the rigor of the original method is maintained. As shown in three heat exchanger network examples, a substantial reduction in effort is often achieved.