Predicting the cradle-to-gate environmental impact of chemicals from molecular descriptors and thermodynamic properties via mixed-integer programming

Life Cycle Assessment (LCA) has recently gained wide acceptance in the environmental impact evaluation of chemicals. Unfortunately, LCA studies require large amounts of data that are hard to gather in practice, a critical limitation when assessing the processes and value chains present in the chemical industry. We here develop an approach that predicts the cradle-to-gate life cycle production impact of organic chemicals from attributes related to their molecular structure and thermodynamic properties. This method is based on a mixed-integer programming (MIP) optimisation framework that systematically constructs short-cut predictive models of life cycle impact. On applying our approach to a data set containing 88 chemicals, 17 molecular descriptors and 15 thermodynamic properties, we estimate with enough accuracy (for the purposes of a standard LCA) several impact categories widely applied in LCA studies, including the cumulative energy demand, global warming potential and Eco-indicator 99. Our framework ultimately leads to linear models that can be easily integrated into existing modelling and optimisation software, thereby facilitating the design of more sustainable processes.