Heat exchanger networks are used on chemical and other process plants to recover energy from hot streams at a high temperature into cold streams which need to be heated. This saves utility costs (steam and cooling water) but involves capital expenditure on the exchangers, leading to a trade-off which can be optimised.
We report a case study using a particular approach to heat exchanger modelling,in which a problem which really should be a MINLP is approximated as a NLP. The SQP code 'Filter', recently developed at Dundee, has been successfully used to solve this problem.
We shall discuss how the performance of Filter is affected by different aspects of model implementation:
In summary, Filter successfully solved most runs, though in some cases taking up to 40 SQP iterations (and in one case, over 90). There were some failures to converge the constraints (terminating during the retoration phase): in these cases Filter became trapped in infeasible structures. Multiple local minima were found for different runs of the same problem, usually corresponding to different structural alternatives for the network.
Current 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996