Automated design of separation processes using implicit enumeration and interval analysis
This thesis concerns the automated synthesis of separation processes. A single multi- component stream is to processed to give one or more pure component product streams. A list of units are available for the task and the aim is to find the optimal flowsheet structure in terms of cost. Implicit enumeration (IE) has been used to tackle the synthesis problem. The main advantage of this approach is that IE does not require the development of a superstructure. A disadvantage of using IE is that it is necessary to discretise the values of unit operating conditions in order for there to be a finite search space (Fraga et al., 2000). The user may not have any idea of the effect of the discretisations on the quality of the solution. In addition, the optimal solution may be missed between the discrete values chosen. The purpose of this work is to address these issues. Interval analysis is used to bound the effects of this discretisation. This allows the cost of each particular flowsheet to be bounded based on the level of discretisation used. The technique is demonstrated by bounding the effect of discretisation on the synthesis of distillation flowsheets. The use of runs with progressively finer uniform discretisation lead to the isolation of the optimal structure. This result leads to the development of an adaptive algorithm that changes the discreti sation profile in response to bounding information from downstream in the search. The algorithm operates recursively and isolates the optimal process structure for each stream encountered. This builds up to the isolation of the overall optimal process structure for the feed process stream. The effectiveness and performance of the new algorithm are evalu ated using two very different separation problems. The first is a distillation sequence and the second a separation of a protein from a biological stream.