Four kinds of commercially ultra-high-temperature skimmed milk selected as the research object, the flavor attributes of samples evaluated by sensory descriptive analysis and electronic nose technique. Component analysis (PCA), cluster analysis (CA) and partial least squares regression (PLSR) used to analyze the correlation between electronic nose sensor performance and sensory attributes. The results obtained in the study showed that creamy and fat attributes had low sensory score. Ten sensors of electronic nose had good ability to differ four samples, the correlation of sensory attributes and electronic sensors explained by PLSR model that reflects the overall information reflected. Sensory evaluation combined with electronic nose to analyze four UHT skim milk, we can test the overall information by electronic nose in a rapid way and analyze samples by sensory evaluation specifically. Combined these two methods can make up the defects of sensory evaluation and intelligence sensory technology and available to provide a reference of improving the flavor of UHT skimmed milk.