Odor Intensity Prediction Of Odorant Mixture And Its Application In Odor Evaluation

As one of the seven public nuisances, odor pollution not only makes people uncomfortable but also cause threats to both human health and environmental safety. In order to protect citizen’s health and to improve the regulation and management of odor pollution, effective detection method and evaluation technique are needed. This study mainly focused on the relationship between odor intensity (OI) of an odor mixture and its constituents’chemical concentraions on the establishment of corresponding OI prediction models. In addition, the OI prediction models were applied in electronic nose for both quantitative analysis and01evaluation of odor pollutants, which would be meaningful for rapid detection and on-line monitoring of odor pollution.At present, the main obstacles of olfactory evaluation on odor pollution were concluded as the following:(1) the accuracy and repeatability of OI rating are poor because of low precision of odor dilution and diversity of human assessors;(2) methods of precisely predicting OI on the basis of constituents’chemical concentrations, which were measured by instrumental analysis, were lacked;(3) a majority of odor interaction models are mathematical formulas, and fitting degree is the main content in related researches. However, the display of odor interaction is usually ignored;(4) the olfactory evaluation has specific requirements on assessor and testing condition, which makes fast detection and on-line monitoring methods lacked in odor pollution assessment.Accordingly, this study mainly focused on the following contents:(1) exploring new methods improving odor dilution precision and trainning assessor to improve the accuracy and precision of olfactory evaluation;(2) exploring the linear relationships between OI and chemical concentration of individual odorant in the mixture. Based on these relationships, the Additivity Model, U Model and Vector Model were modified and further employed for OI prediction on the basis of constituents’chemical concentrations;(3) through simulation of partial molar quantity, a partial differential equation (PDE) model of binary odor interaction was proposed. The relationship between01of an odor mixture and its constituents’chemical concentrations was showed in the form of PDE diagram;(4) choosing benzene, toluene, ethyl benzene as the targeted compounds, a gas sensors array and a Back-Propagation (BP) artificial neural network were combined with the01prediction models to form an electronic nose with functions of qualitative analysis, quantitative determination and01rating.The obtained conclusions and acheivements of this study were concluded as the following:(1) When using the triangle odor bag method to prepare odor samples at low concentration level, a proposed two step method could prepare odor sample with higher precision. By simulating the odor intensity referencing scale with water at difference temperatures, the accuracy of assessor’s01rating was improved after accepting the corresponding training.(2) Based on the linear relation between01and logarithm of odor activity value (InOAV) of individual odorant, the Additivity Model, U Model and Vector Model were individually modified. After measuring the concentration of each constituent through instrumental analysis, the odor intensities of both bianry and complex odorant mixtures of aromatic compounds were successfully predicted by employing these modified models. Thus, it laid a solid foundation for01rating by analytical instruments.(3) Based on the methodology of partial molar volume, the relation between OI of a binary mixture and its constituents’lnOAV values were explored. Then, a partial differential equation (PDE) model was proposed to exhibit the odor interaction in the form of PDE diagrams. The PDE model was proved effective at OI prediction for binary mixtures which was consisted of odorants with both same chemical functional group and similar molecular structure. Besides, it is proved that odor interaction was mainly influenced by constituents’mixing ratio instead of odor sample’s concentration level.(4) After testing with individual benzene, toluene and ethyl benzene, the gas sensor MC119, MQ6, WSP2106and2M008were assembled together to form a gas sensor array. By further combining the BP artificial neural network with01prediction models, an electronic nose was established. Then, the electronic odor nose was proved to be quite effective at qualitative analysis, quantitative determination and01rating for both individual benzene, toluene, ethylbenzene and their mixtures.