AI modelling of UF and RO membrane processes
Project Status
Ongoing
Motivation
Having a dynamic simulation of a physical system offers many advantages. It enables offline optimisation, process changes, and scenario-based analysis without the need to conduct time consuming real-time experiments. It also de-bottlenecks access to the physical equipment, enabling many users to work on the system and produce data and results simultaneously. There are two approaches to generating process simulations: 1) Physics based models, which use properties databases and fundamental equations to predict process outputs and 2) data-driven models, which use real world data and artificial intelligence (AI). Both these approaches have advantages and limitations. Physics-based models are constrained and are therefore more likely to give sensible extrapolated results. Data-driven models are more likely to represent a real-world process, but are only reliable within the range of the data used to generate them.
Approach
This project will use Aspen Plus and AspenTech AI Model Builder to compare, contrast and combine the two methods to generate the most accurate model possible for ultrafiltration and reverse osmosis water treatment processes.
Outcomes
This project is ongoing. Expected outcomes include:
- An Aspen Plus physics-based UF and RO model
- An AspenTech AI Model Builder data-driven UF and RO model
- A combined model
- An analysis comparing and explaining the differences, advantages and limitations of the individual and combined models