Studies in the computer aided design of complex heat exchange networks

• Main Author: Donaldson, Robert Adams Brown
• Published: University of Edinburgh 1976
• Subjects: 660.2832
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.649639

This thesis describes research undertaken in the synthesis of process design and the effect of uncertainty on process design, aspects of computer aided design which have received comparatively little attention. Heat exchange networks form the basis of the studies. Investigation of methods previously proposed for the synthesis of heat exchange networks are reported. Due to the immense size of the synthesis problem, these methods, mainly exhaustive search procedures, are limited to the solution of small problems. Following this study, a very fast, simple technique, based on the use of a single heuristic, is described; this technique overcomes most of the limitations of previous methods. The new method is applied to an industrial sized problem, the heat recovery section of a crude oil distillation unit, and networks up to 10% cheaper than those designed by conventional techniques result. Simulation studies of these synthesised networks suggest that they are sufficiently flexible to operate well under non-design conditions. The generation of synthesis trees indicates that the costs of these networks are consistently very close to the optimal. There are inevitable uncertainties associated with the design of equipment. Using published correlations, and data available for several exchangers, the statistical distribution of the uncertainty in overall heat transfer coefficient is investigated. Use of this distribution leads to higher overdesign factors than often quoted, and a method of finding optimum overdesign factors is presented. During preliminary investigations into uncertainty in networks of exchangers, a linearised model, based on a Taylor series expansion, which permits the calculation of system variances and covariances is developed. A scheme for the incorporation of this model into a modular simulator is outlined.