The computational aspect of steady state simulation and optimization of an existing plant has been studied, using an ammonia synthesis loop as an example.
One-dimensional fundamental models for the main units in the synthesis loop - converter, waste heat boiler, air-cooled condenser and circulating compressor - have been developed.
The thesis describes various numerical techniques, computational and computer storage problems involved in simulating a multi-stage system. The computer storage problem has been overcome by replacing the fundamental model for each unit in the synthesis loop by a second order regression model. The experimental designs used were those of Box and Wilson (Ref.[2.3]).
Dynamic Programming and the Discrete Maximum Principle have been shown not to offer any computational advantages over the simpler method of considering the synthesis loop as a single unit. The main factors were: (i) the large number of state variables, (ii) the presence of a recycle stream, and (iii) the relatively low number of decision variables considered.
The optimization problem has proved to be beyond the scope of many efficient optimization techniques, such as gradient methods and Powell's conjugate directions method. This has been due to (i) the unavailability of an explicit analytical expression of the objective function in terms of the decision variables, and (ii) the constrained nature of the problem.
Simplex method is shown to be superior under these conditions. A modified Simplex method is proposed. The synthesis loop model, considered as a single unit, has successfully been optimized by the proposed method.