In order to solve the problem apt to land in local minimum results for its sensitivity to the initial conditions, in view of character of test-online, this paper proposes a global fuzzy c - means (GFCM) clustering algorithm based on incremental approach to clustering. The converging speed of GFCM is improved by simplifying the algorithm and then the approach of the algorithm is given. The experimental test for unsupervised clustering and fault pattern recognition of the information channels of airborne weapon system is given by using the new GFCM algorithm. The results show that the proposed algorithm is effective in dealing with the aforementioned problem under condition of small data capacity.