The invention discloses a method for detecting a leakage passage based on entropy weight fuzzy clustering, comprising the steps: that information parameters are non-dimensionalized and normalized wherein the information parameters provided by ''information flow'' are fully utilized; that in the information flow of supply sources, information parameters with the maximum differences are selected wherein an entropy weight method is utilized to screen and determine weight for each information parameter; that a new sampling set is established, a fuzzy clustering method is used to carry out clustering, and membership degrees of different boring sampling water to different potential supply sources are calculated wherein the maximum membership degree represents the largest correlation, thereby deducing the supply source of the boring sampling water; and that after an influential range of the leakage passage is calculated, a communicating experiment is carried out in a boring where an abnormal sampling locates so that a concrete position of the leakage passage is finally determined. The method improves judgments for the information flow to theories and determines correlations of the samplings quantitatively, thereby classifying the samplings objectively and providing scientific basis for correct determination of leakage passage position.