Modern Modeling Paradigms Using Generalized Disjunctive Programming^†

Round 1

Reviewer 1 Report

The paper illustrates that GDP can be a useful model abstraction that separates algebraic and logical relationships within process design problems. The paper is overall well-written, and I suggest acceptance as is.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a very interesting manuscript covering a very important issue related to model-based optimization techniques, especially for process synthesis where use of GDP is very useful. The manuscript is very well written and gives an excellent overview of the important issues and topics. However, I miss discussion on a fet additional issues that in my opinion would make the published manuscript more useful to the reader. They require minor additions of text.
1. The authors only consider one form of superstructure representation (Turkay and Grossmann, 1996). However, the state-task representation (Kondili et al, CACE 1993) or processing step-interval approach (Bertran et al, CACE 2017) could also have been mentioned. The Turkay-Grossmann representation can be rather complex if the size of the problem is large, which is not the case for Bertran et al.
2. It is not exactly clear how the authors propose to determine the superstructure (or network) model? Is it a simple mass balance model (as in Kondili et al or Bertran et al) or a rigorous process simulator based model? In the later case, how is the connection between the simulator and the optimizer handled? In the case of the former, how are the model parameters obtained (apriori before the optimization step)?
3. Could the authors briefly mention the limitations or difficulties in the current state of the art for the approach proposed by the authors? For example, can a totally new processing route to convert from a collection of raw materials, the optimal product with the optimal processing route be determined?
4. At what stage or how issues like environmental impacts, safety, operability, etc., are considered? Can they be added as additional constraints (by supplying additional model equations) or should they part of a multi-parameter objective function?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf