Intelligent column chromatography prediction model based on automation and machine learning

Abstract

Efficient compound separation remains a persistent challenge in synthetic chemistry, with column chromatography serving as a critical purification tool. Traditional methods require extensive expertise and repetitive labor—areas where AI offers significant advantages. This study introduces an AI-driven platform to automate data collection and optimize separation processes. By leveraging deep learning, the system predicts key separation parameters, while transfer learning enables adaptation to diverse column specifications. A novel metric, separation probability $\left(S_{\mathit{p}}\right)$$\left(S_{\mathit{p}}\right)$, quantifies the likelihood of successful component isolation and has been experimentally validated. The approach enhances precision, reduces manual intervention, and expands the scope of chromatographic applications, offering a more efficient and scalable solution for chemical purification.