Machine-leaning assisted optimization of pressure swing adsorption processes
Project Description
Adsorption technologies are highly important in solving many grand challenges in modern society, including energy storage, carbon capture, gas separation and purification. With the great tunability of properties in many advanced adsorption materials, in particular metal organic frameworks, they have the potential of significantly lowering the energy requirement for gas separation applications compared to traditional distillation based processes. The development of adsorption technologies involves both the search for high-performance adsorption materials and the design of adsorption processes. The optimal operating condition needs to be chosen for the ideal performance of adsorption materials. This project will use artificial intelligence (AI) to aid the design of adsorption materials and optimization of adsorption processes.
Pressure swing adsorption (or vacuum swing adsorption, depending on the pressure range) is a dynamic process commonly used for gas separation. It will separate different components at high pressure based on different adsorption properties of different molecules, and recover the adsorbent at low pressure. In order to optimize this process, we will develop a mathematical model to describe this dynamic mass transport phenomenon, that can calculate the recovery and purity of the desired separation product based on the process conditions. Then an optimization framework will be established for determining the best process conditions.
Due to the large design space for process operating conditions, it is desirable to minimize the number of simulations that we need to run. We will use supervised learning models to approximate the process simulation model and active learning strategies to sample the process design space in order to efficiently perform the optimization.
Finally, we would like to incorporate materials design into process optimization, by adding materials properties as additional process optimization decision variables. This can further inform materials design for optimal separation processes.
Pressure swing adsorption (or vacuum swing adsorption, depending on the pressure range) is a dynamic process commonly used for gas separation. It will separate different components at high pressure based on different adsorption properties of different molecules, and recover the adsorbent at low pressure. In order to optimize this process, we will develop a mathematical model to describe this dynamic mass transport phenomenon, that can calculate the recovery and purity of the desired separation product based on the process conditions. Then an optimization framework will be established for determining the best process conditions.
Due to the large design space for process operating conditions, it is desirable to minimize the number of simulations that we need to run. We will use supervised learning models to approximate the process simulation model and active learning strategies to sample the process design space in order to efficiently perform the optimization.
Finally, we would like to incorporate materials design into process optimization, by adding materials properties as additional process optimization decision variables. This can further inform materials design for optimal separation processes.
Supervisor
GAO Hanyu
Quota
1
Course type
UROP1100
Applicant's Roles
To develop the framework mentioned above, you will use propane/propylene as a test system. For this UROP project, the expectation is that you will develop a process simulation model for pressure swing adsorption and use it to model the adsorption process under different operating conditions. Depending on the progress, you might be able to start establishing the machine-learning based optimization framework.
Applicant's Learning Objectives
After working on the project, you will be expected to:
Gain experience in programming and process modeling;
Participate in cutting edge research related to energy and materials;
Use chemical engineering knowledge in a challenging real-world problem;
Explore optimization and machine learning techniques.
Gain experience in programming and process modeling;
Participate in cutting edge research related to energy and materials;
Use chemical engineering knowledge in a challenging real-world problem;
Explore optimization and machine learning techniques.
Complexity of the project
Challenging