Development of active odor sensing system with odor blender

研制了一种用于混合香味定量的主动气味传感系统。在该系统中，多个已知蒸汽被主动和自主地混合，以便传感器阵列的输出模式可以与未知蒸汽的输出模式相等。通过系统中的混合比例可以得到香气的混合组成。使用非线性优化算法迭代地修改该比率。利用该系统，在干扰蒸气存在的情况下，成功地测量了汽油浓度。然后对梯度下降法等优化算法进行了改进，提高了算法的精度和稳定性。最后，对其在香料中的应用进行了研究。成功地进行了各种精油的鉴定和光学异构体的分离。此外，还对香精中几种基本化合物的混合组成进行了定量实验。

Active odor sensing system for quantifying mixed aromas was developed. In the system, plural known vapors are actively and autonomously blended so that the output pattern from a sensor array can be made equal to that of an unknown vapor. The mixture composition of aromas can be obtained from the blended ratio in the system. The modification of that ratio is iteratively performed using a nonlinear optimization algorithm.Using this system, gasoline concentration was successfully measured in spite of the existence of interference vapors. Then, the optimization algorithm such as gradient descent method was improved and its accuracy and the stability was enhanced.Finally, the application to fragrances was studied. The identification of various essential oils, the separation among optical isomers were successfully performed. Moreover, the experiments on quantifying the mixture composition of several fundamental compounds in fragrances were performed.