To solve the problem of conflicting between accuracy and efficiency in T-S fuzzy modeling, a method of establishing T-S fuzzy model by using the augmented input variables is proposed. After the input variables are augmented, the best clustering number and fuzzy partition are automatically obtained by using kernel fuzzy ??-means(KFCM) and a clustering evaluation function, and the consequent parameters are determined by using recursive least-square algorithm. In order to calculate the antecedent structure more efficiently, the input parameters are anti-inferenced according to consequent parameters. Finally, simulation examples illustrate the effectiveness of the method.