Selection of representative molecules of asphalt aromatics based on principal component analysis and hierarchical clustering

• Published in: Molecular simulation Vol. 50; no. 13; pp. 881 - 898
• Main Authors: Zhu, Jie, Yang, Shitong, Shen, Dejian, Zhang, Fengchen, Cao, Rongji, Wu, Chunying, Jin, Baosheng, Wu, Shengxing
• Format: Journal Article
• Language: English
• Published: Abingdon Taylor & Francis 01-09-2024; Taylor & Francis Ltd
• Subjects: Alkanes; Asphalt; Asphalt aromatic; Cluster analysis; Clustering; hierarchical clustering; Indicators; Infrared analysis; Low molecular weights; Molecular chains; Molecular weight; principal component analysis; Principal components analysis; quantum chemical calculations; Quantum chemistry; representative molecules; Statistical analysis
• ISSN: 0892-7022; 1029-0435
• DOI: 10.1080/08927022.2024.2364702

Aromatics constitute the highest proportion among the four asphalt components. The existing literature contains over 20 types of small molecules used to characterise asphalt aromatics, which presents a challenge in constructing asphalt molecular models. In this investigation, quantum chemical calculations were employed to determine the molecular morphology, molecular polarity, and infrared spectrum indicators of asphalt aromatics. Based on principal component analysis and hierarchical clustering, statistical analysis of the aforementioned indicators was conducted to present a recommended scheme for selecting representative molecules of aromatics, thereby reducing the difficulty of constructing asphalt molecular models. The results indicated that N2 was recommended as the representative molecule for aromatics with high molecular weight and long alkane chains, while N3 was recommended as the representative molecule for aromatics with low molecular weight. Based on the variations in substituents of aromatics, it was advisable to select one or more of N7, N10, N14 - N16, and N19 as representative molecules. The investigation guided the selection of aromatic representative molecules in the construction of asphalt molecular models. For major projects with test data, representative molecules could be adjusted according to the test data.