Fitting Adsorption Isotherms with Symbolic Regression

The ability to mathematically describe adsorption of species from a bulk onto a surface is crucial to enabling adsorption process modeling for chemical separations. The isotherm is the most central piece of adsorption information and can be closely fitted by theoretical models in most cases. However, sometimes these models are unable to describe adsorption isotherms adequately. Symbolic regression (SR) is a tool with which a computer empirically creates equations to fit datasets including adsorption isotherms measured at one or more temperatures. In this work, we demonstrate the use of SR to fit a variety of challenging isotherm forms quickly and easily. With no prior physical knowledge, SR is able to accurately fit models to experimental isotherms with one or more steps, including adsorption of water vapor, cooperative adsorption of carbon dioxide, and adsorption on flexible adsorbents. SR has applications in fitting isotherm data in cases where adsorption physics are poorly understood or accuracy of model fit is of critical importance.