Abstract: In context of wireless sensor networks （WSNs）,fault detection by collaborating with neighbors brings additional communication and computation load. A hybrid information increment matrix-based Fault Self-Detection Method in WSN is proposed in the paper. This method is no need for collaboration with neighbors and improves the fault diagnosis method based on local information incremental matrix forgood performance in fault detection accuracy. Simulations show that the proposed method not only can reach a higher level of the performance in fault detection accuracy,but also can be more robust to change of size of the moving window and the training sets,co MPared with the fault diagnosis method based on local information incremental matrix.