Abstract: Aromatic compounds are among the most prevalent and persistent pollutants in the environment. One of the methods for degradation of aromatic pollutant in water sources and waste waters could occur by radicals (e.g. hydroxyl radicals). The rate constants of reactions between organic aromatic pollutants and OH radical are one of the most important characteristics related to their degradation [1]. Due to the complexity of analytical methods and the high cost of experiments, the application of the theoretical predictive methods is fast and convenient for the preliminary assessment and the estimation of degradation rates and moreover cost-effective. In this study a series of compounds based on benzene structures with different substitute groups were investigated. In the present research Multivariate Image Analysis Quantitative Structure-Property Relationship (MIA_QSPR) method was applied to model the degradation rate constant of aromatic compound. In Multivariate Image Analysis Quantitative Structure-Property Relationship, the descriptors were generated from pixels of 2D structures of each compound, which can be drawn with aid of any appropriate program [2, 3]. Different chemometric models such as Partial least squares regression and Artificial Neural Networks (as two way order models) and Parallel Factor Analysis and residual bilinearization (as three way models) were applied for the quantitative analysis. These chemometric models were applied with and without variable selection procedure (Genetic Algorithm (GA)) and data compression procedure (Principal Component Analysis (PCA)). In this study we have compared the performance of different chemometric method in the estimation of degradation rate constant of aromatic compounds.

Keywords