A real-time on-line broken strand detection and diagnosis system is introduced for high-voltage transmission line automatic inspection robots. Based on the radiation field changes caused by the transmission line strand breaking and the operating environments of the robot， an appropriate sensor is selected with fine tuned sensor signals and highly reliable collecting circuits. Then， the signal analysis process is described in detail. Given the instable nature of broken strand signal and the time-frequency characteristics of wavelet base， the broken signal is decomposed into six level wavelets using db4 wavelet base. By introducing the concept of entropy energy wavelet， the problems caused by simply adopting wavelet transform， such as over-abundant information and eigenvalues， and hard to inspect online are solved. A small BP network can realize on-line precised diagnosis on the strand breaking problem. Experimental results show the feasibility and effectiveness of the proposed system.