In view of the traditional stochastic resonance (SR) method is difficult to achieve the optimal output, the methods of general scale transformation are proposed based on the second-order underdamped and first-order overdamped bistable system. The adaptive SR of the two methods is realized by using quantum particle swarm optimization algorithm, and the fault diagnosis efficiency is improved. The influences of time scale, damping factor，noise intensity and system parameters on the output signal-to-noise ratio (SNR) are discussed based on the numerical simulation. The proposed method is used to analyze two vibration signals of bearing faults. The results show that the second-order general scale transformation adaptive SR method is superior to the first-order general scale transformation adaptive SR in the detection of the weak fault signal. In addition, the second-order underdamped system has a strong ability to suppress and utilize noise, and the amplitude of fault frequency is obviously enhanced. Meanwhile, the output SNR is improved. The results indicate that the second-order general scale transformation adaptive SR method can improve the detection efficiency of the bearing fault signal and extract successfully the weak fault characteristics. The proposed method has obvious superiority in bearing fault detection.