July 2020. Here are Victor Alves and Felipe Lima with UFCG, Campina Grande, Brazil. We created a standalone software written entirely in Python programming language capable of using Self-Optimizing Control technology with the aid of surrogate models (Kriging). It generates the best self-optimizing control structures in a comprehensive user interface. We call this software Metacontrol (From Metamodel-based Self-Optimizing Control), and it is hosted on GitHub https://github.com/feslima/metacontrol We also have created a website for Metacontrol, meta-control.net where further theoretical details and a complete user manual including case-studies can be found. Aspen is only used in Metacontrol to: 1. Generate the output data using the sampling that Metacontrol generates (At this moment, a Latin Hypercube Sampling) to generate the kriging metamodels; 2. On the optimization framework using surrogate models. For both cases, the process is fully automated. You can just push the button! However, in some very few cases (more complex cases, for instance), one may try to enhance the convergence of his/hers model. In addition, when one finds the active constraints, he must implement them in the process model in order to generate the reduced-space kriging metamodel (this is done to generate the gradients/hessians accurately). Therefore, the only manual attention that would be necessary: 1. When there are active constraints: The user implements them in the process simulator and supply a copy of the simulation file to Metacontrol, and the software handle the rest of the process; 2. If the user wants to try to enhance convergence: This is more a modelling problem, and not really from Metacontrol itself. In fact, we have a full discussion in our publication for such particular cases (section 5 - Discussion): https://www.sciencedirect.com/science/article/abs/pii/S0098135420303355 Metacontrol: A Python based application for self-optimizing control using metamodels Authors: Felipe Souza Lima, Victor Manuel Cunha Alves, Antonio Carlos Brandao Araujo Computers & Chemical Engineering, Volume 140, 2 September 2020, 106979 Victor Alves felipe.lima@eq.ufcg.edu.br Campina Grande, Paraiba, Brazil --------------- Comment from Sigurd Skogestad I can see that you have done a very impressive job! In addition to finding the self-optimizing variables, your software can also be very useful in finding the optimal operating point. This can be difficult with Aspen or Hysys. I have two more recent PhD students who have worked on related issues. I don’t know if you are familiar with their work 1. Vinicius de Oliveira Optimal operation strategies for dynamic processes under uncertainty, 2016. He worked on MIQP methods for finding optimal measurement subsets 2. Julian Straus Optimal Operation of Integrated Chemical Processes - With Application to the Ammonia Synthesis, 2018. He looked in to use of surrogate models using Hysys, but he encountered a lot of numerical problems. He used Latin hypercube for sampling and he looked into how many points you need to sample, and when you can stop samling (using Hysys or Aspen). Best regards, Sigurd