Fast and precise reconstruction of natural polycyclic aromatic hydrocarbons: novel algorithm and case study

The polycyclic aromatic hydrocarbons (PAHs) are common constituents and pose significant influences on many separation and conversion processes. Herein, an efficient approach has been developed to fast and precisely reconstruct the molecular models for PAHs. The method involves stochastic reconstruction for a molecular library and artificial neural network (ANN) optimization for model refinement. Stochastic reconstruction, aided by modeling matrices, allows fast arrangement of benzene molecules, simplifying the construction of fused-ring cores. ANN analysis of multiple PAHs models reveals relationships between modeling parameters (MPs) and evaluation parameters (EPs). By optimizing hyper-parameters, a minimal loss function value of 0.085 is achieved, indicating the high precision of modeling. Case study on interfacially active asphaltene (IAA) shows successful reconstruction of its molecular structure, confirmed by experimental results (deviation error < 10%). These findings provide valuable insights into using an universal algorithm to fast and precisely reconstruct intricate natural PAHs and other natural chemical compounds