Overview

Synergistic optimization-based framework for the conceptual design of bioprocesses

S3O facilitates & expedites conceptual process design for bioprocesses by synergistically:

• selecting product(s), feedstock and process unit operations in a bottom-up manner
• performing surrogate-assisted superstructure optimization to determine candidate process topologies
• performing simulation-based optimization for consolidating an optimal process design for the candidates under uncertainty

Workflow

Employment

The repository contains the following folders:

• easyGSA: empty folder to install the easyGSA toolbox
• modelling_ANN: folder with the functions/scripts for fitting ANN surrogate models and transferring the models to the optimization problem (main script: rs_ANNmodeling_cval.m,nk_ANNtransfer.m)
• modelling_DTR: folder with the functions/scripts for fitting DTR surrogate models (main script: main_DTR.py)
• modelling_GPR: folder with the functions/scripts for fitting GPR surrogate models and transferring the models to the optimization problem (main script: rs_GPRmodeling_cval.m,nk_GPRtransfer.m)
• models: the high-fidelity models for the case study presented in the according publications (these are interchangeable with other models for other case studies)
• MOSKopt: empty folder to install the MOSKopt solver
• optimization_MILP: folder with the functions/scripts for the MILP (main script: main_MILP.py)
• optimization_MINLP: folder with the functions/scripts for the MINLP (main script: main_MINLP.py)
• optimization_MOSKOPT: folder with the functions/scripts for the simulation-based optimization (main script: rs_MOSKopt.m)
• optimization_NLP: folder with the functions/scripts for the series of NLP (main script: rs_MultiStart_ANN.m, rs_MultiStart_GPR.m)
• sampling: folder with functions belonging to the SPDlab toolbox for flowsheet simulations
• sensitivity: folder with the functions/scripts for performing the flowsheet sensitivity analysis (main script: rs_performGSA.m)
• simulations: folder with the functions/scripts for performing single or Monte Carlo flowsheet simulations (built upon the functions of the SPDlab toolbox) (main script: nk_mcsims.m)

Prerequisites

• MATLAB 2020b or higher (https://www.mathworks.com/)
• Python 3.7.9 or higher (https://www.python.org/)
• ASPEN Plus 11 or higher (https://www.aspentech.com/)
• ALAMO Surrogate Modelling (https://minlp.com/alamo)
• BARON Solver (https://minlp.com/baron)
• Gurobi Solver 9.10 or higher (https://www.gurobi.com/)

The frameworke furthermore utilizes the easyGSA toolbox and the MOSKopt solver developed by Resul Al (resal@kt.dtu.dk). They are available on the GSI-Lab GitHub page:

• easyGSA (https://github.com/gsi-lab/easyGSA)
• MOSKopt (https://github.com/gsi-lab/MOSKopt)

Developer

Nikolaus Vollmer (nikov@kt.dtu.dk) - PROSYS Research Center, Department of Chemical and Biochemical Engineering, Technical University of Denmark

Acknowledgements

This work is part of the Fermentation-Based Biomanufacturing Initiative (http://www.fbm.dtu.dk) at the Technical University of Denmark and received funding by the Novo Nordisk Foundation (Grant no. NNF17SA0031362)