Modular global optimisation in chemical engineering

Chemical Engineering design and analysis is dominated by the use of modular computational systems restricting the use of rigorous global optimisation techniques. Other engineering domains also exploit modularity in order to break down complex tasks to allow the use of legacy codes, to protect intellectual property, and to allow large teams to work on problems. By casting modules in a generic form such systems could be recast to incorporate interval based methods. In this paper we explore the use of five interval contraction methods to improve the performance of interval based optimization of modular process design systems: consistency methods, constraint propagation, Interval Gaussian elimination, Interval Newton and Linear Programming. It is shown that the Linear Programming contractor provides the greatest value in contracting the intervals and that constraint propagation and Interval Gaussian elimination (as implemented here) provides less of an impact. Other contractors do provide value and the LP contractor will be of less value as the problem size increases so it is necessary to include a number of contractors which can be done at small computational cost. A number of challenges are outlined which need to be addressed before there can be routine use of interval global optimization in modular systems.