machine learning models for predicting characteristics of pvam membranes for post-combustion co2 capture application

Facilitated transport membranes fabricated by polyvinylamine show great potential to compete with convenient carbon capture technologies in a post-combustion co2/n2 separation, with lower energy consumption and zero toxicity. precise mathematical models are needed to predict membrane characteristics for designing suitable membrane equipment and optimizing process configuration. two main features of a membrane are the permeance of co2 gas and co2/n2 selectivity that shows the overall performance of each membrane by considering the solution-diffusion model. two known machine learning algorithms were employed to predict the permeance and selectivity of a recently developed membrane based on its four major parameters. both mlp-ann and svm functions had great potential to fit experimental data, while the mlp-ann method works better for permeance (r2 equal to 0.975) and the svm method fits selectivity better (r2 equal to 0.948).