In order to realize quantitative description of logical domain instead of "belong to" or "not belong to" some set in classical extenics, an extenics engineering method was combined with cluster analysis to construct classical and limited matter elements according to air quality standards and main factors affecting air quality. The logical domain was extended from (0, 1) to (-∞, +∞). Analytic Hierarchy Process (AHP) was used to determine weights of air quality factors and establish improved extenics model with varying weight for assessing air quality. In view of no-strict attributes of city atmosphere sample and betweenness of their form and class attribute, five assessment factors which have great influence on the human health i.e. PM₁₀, SO₂, NO₂, CO, and Fallen Dust are identified. The new model not only could reflect the whole level of air quality, but also identify the main air pollution facts which exert great impact. An improved extenics model is used to assess air quality in Fuxin city in the year of 2010. And then the results are compared with the results based on Shanghai air quality index method and fuzzy comprehensive evaluation method. The serviceability of new model in the applications of air quality assessment is analyzed. To a certain degree, the establishment of application-typed model dedicated to assessment and control of city atmosphere has its theoretical value and realistic guiding significance.