Stoichiometry model identification for homogeneous reaction mixture: High‐dimension and missing measurement case studies

Abstract We address the identification of the actual stoichiometric network in reacting systems using composition measurements, extending our previous work (Fromer et al. I&ECR 2023). We generalize this algorithm for scenarios where not all species are measured and back‐calculate the missing concentrations through the reaction extents of the candidate network. In addition to the prior global accuracy comparison among candidate reaction networks, we introduce a species‐by‐species F ‐test accuracy comparison between the most accurate reaction networks from the global assessment. We examine two case studies involving 7 and 11 species participate in 4 or 8 reactions, respectively. In the second case study, the 8 reactions are linearly dependent, presenting an additional challenge. The enhanced algorithm successfully identifies the actual reaction network as the most accurate, even with 4 of the 11 species not measured.