To conduct odor analysis of the main effective components, namely volatile oils, contained in two varieties, Notopterygium incisum Ting ex H.T.Chang and Notopterygium franchetii H.de Boiss, to provide a feasible method for promptly and accurately distinguishing between the differences in volatile oils of these two varieties of Notopterygii Rhizoma et Radix. This enriches the traditional evaluation content and serves as a reference for assessing the quality of extracts predominantly governed by volatile oils.

The flavors of two samples of Notopterygii Rhizoma et Radix volatile oil were analyzed using electronic nose technology and sensory evaluation. The electronic nose data obtained were subjected to analysis and identification through principal component analysis (PCA) and linear discriminant analysis (LDA). Additionally，two nondestructive testing models-Fisher discrimination and multilayer perceptron (MLP) neural network discrimination were established for sample differentiation.

Sensory evaluation results indicated that pine resin flavor，cool flavor and woody flavor were the primary odor characteristics of both Notopterygii Rhizoma et Radix volatile oils. Additionally，the key flavor attribute influencing acceptance and differentiation was identified as spoiled yuba flavor，with the Notopterygium franchetii H. de Boiss volatile oil exhibiting a stronger presence of this attribute than the Notopterygium incisum Ting ex H. T. Chang volatile oil. The electronic nose results revealed that the nitrogen oxides’ response values in Notopterygium franchetii H. de Boiss volatile oil were significantly higher than those in Notopterygium incisum Ting ex H. T. Chang volatile oil. Meanwhile，the response values of hydrides，alcohol ether aldehydes，and ketones were slightly lower in Notopterygium franchetii H. de Boiss volatile oil compared to Notopterygium incisum Ting ex H. T. Chang volatile oil. The Fisher discriminant model demonstrated overall discrimination rates of 93.8% for the training set and 87.5% for the prediction set of the two volatile oils. In contrast，the MLP model achieved discrimination rates of 89.3% for the training set and 91.7% for the prediction set. Notably，the MLP model proved effective for identifying volatile oils，while the Fisher model exhibited greater suitability for discriminating volatile oils with broad-leaved characteristics.

The combination of artificial senses and intelligent senses can be characterized from both subjective and objective perspectives，elucidating the flavor differences between the two kinds of Notopterygii Rhizoma et Radix volatile oils. The established Fisher discriminant function and MLP discriminant models can rapidly and accurately distinguish between the two kinds of Notopterygii Rhizoma et Radix volatiles. This lays a preliminary foundation for quality control in Notopterygii Rhizoma et Radix volatiles and offers new ideas and directions.