One of the fundamental problems in chemical engineering is the synthesis of a process, that is, the selection of unit operations, their interconnections and operational conditions to generate a flowsheet that meets given goals and constraints. Major approaches include enumeration of alternatives, evolutionary modification, and superstructure optimization. In enumeration of alternatives, alternative designs are generated and evaluated, which is only feasible when the number of alternatives is relatively small. In evolutionary modification, designers make changes to known flowsheets for similar processes to meet new objectives and constraints. Superstructure optimization is a model-based approach that compares alternative processes simultaneously. A superstructure incorporates all potentially useful units and relevant interconnections. Our work focuses on addressing some of the challenges pertinent to superstructure optimization.
We develop a framework for the efficient representation, generation, and modeling of superstructures for process synthesis (see Figure 1).
First, we develop a UPCS representation based on three basic elements: Units (u), Ports (p), and Conditioning Streams (s) (see Figure 1A). A unit is indexed by a unit type (ut) and a unit number (un), e.g. “Dst,1” refers to the unit of unit type “Dst (distillation)” and unit number “1”. In addition to general units, source units and sink units are included as auxiliary elements whose function is to provide raw materials and collect the final product stream and wastes. Every unit type has a predefined set of inlet and outlet ports, e.g. membrane units have one inlet port- for the feed stream, and two outlet ports- one for the permeate stream and the other for the concentrate stream. The inlet ports are generally treated as mixers, and the outlet ports are treated as splitters. A stream is indexed by an inlet and an outlet port. The use of conditioning streams allows us to treat the main reaction/separation tasks and the conditioning tasks separately.
Second, we present four rules based on “minimal” and “feasible” component sets (components that are “required” and “allowed”, respectively, to be present in the ports during the normal operation of their units) for the generation of simple superstructures containing all feasible embedded processes (see Figure 1B).
Third, in terms of modeling, we develop a modular approach, and formulate models for each basic element (see Figure 1C for example). We also summarize the element models into a general canonical form using input/output variables and constrained/free variables. To cope with large process synthesis problems, promoting ease of construction and modification of the model, we further develop a modular modeling implementation architecture (Figure 2).
Our new UPCS representation allows straightforward superstructure generation, while modularity allows easy model generation and modification. Note that for design problems, which are not solved online or repeatedly, most of the effort goes into the generation of the design alternatives and the formulation of the models rather than their solution. Thus, methods that facilitate this process, as opposed to solution methods, are likely to have greater impact.
In chemical process synthesis, the heat integration between hot and cold process streams reduces utility consumption and improves energy efficiency, especially when it is carried out simultaneously with the synthesis of the process. Accordingly, we develop mixed-integer nonlinear programming (MINLP) optimization models to address simultaneous process synthesis and heat integration. For example, we recently developed a model that accounts for unclassified process streams, a feature that arises in superstructure process optimization (see Figure 1). The model accounts for (1) streams that cannot be classified a priori as cold or hot, and (2) variable process stream temperatures and flow rates. We employ binary variables to represent the cold/hot stream “identities” and, to handle variable stream conditions, we construct “dynamic” temperature intervals using an implicit ordering of stream inlet temperatures. The proposed model is not only capable of handling isothermal streams and multiple utilities, but also able to deal with process streams with potential phase changes. The proposed model can be integrated with a process design model to allow simultaneous process synthesis and heat integration (see Figure 2).
Modern biotechnology enables the use of engineered microorganisms such as E.coli, yeast and algae for the production of chemicals that are currently derived mainly from fossil fuel feedstocks. Processes that employ such biological routes (“bio-based chemicals”), as opposed to complex conversion steps from fossil fuel feedstocks, could in some cases be economically promising. Additional advantages of bio-processes include mild production conditions and selectivity toward a specific product. However, the effluent of bioreactors is dilute (containing less than 20 wt% product), and thus the downstream separation tends to be expensive (it usually accounts for 60–80 % of the total production cost). Past work on the synthesis of bio-separation processes have been mainly focused on specific products. There has been limited research towards the systematic treatment of the general process synthesis problem. To this end, we develop a general framework, based on superstructure optimization, for the synthesis of bio-separation processes (see Figure 1).
Specifically, based on general separation principles and insights obtained from industrial processes for specific products, we first identify four separation stages: Stage 1 – cell treatment, where cells are harvested and then disrupted to release intracellular products (present if the product is intracellular; bypassed if the product is extracellular); Stage 2 – product phase isolation, where the phase that contains the product is isolated; Stage 3 – concentration and purification, where water and impurities are removed; Stage 4 – refinement, where the product is further refined. Based on the four stages, we first perform a stage-wise analysis of general bio-separation processes. Then, for each stage, we systematically implement a set of connectivity rules to develop stage-superstructures, all of which are then integrated to generate a general superstructure (see Figure 2) that accounts for all types of chemicals produced using microorganisms. We further develop a superstructure reduction method to solve specific instances, based on product attributes, technology availability, case-specific considerations, and final product specifications (see an example in Figure 2). A general optimization model, including short-cut models for all types of units considered in the framework, is then formulated.