In order to ensure the safe and stable operation of the power plant steam turbine in the high temperature and high speed environment, a sudden vibration fault identification method of high and intermediate pressure rotor of steam turbine in power plant is proposed. According to the shafting vibration characteristics of the high and medium pressure rotor vibration of the power plant turbine, the integrated eddy current displacement sensor is used to collect the corresponding fault signal, and the wavelet packet analysis is used to extract the high and medium pressure rotor vibration mutation fault characteristics, and the extracted fault characteristics are used as input. Input the fault features into the RBF neural network optimized by artificial fish swarm algorithm, and output the identification result of the vibration mutation type of the high and medium pressure rotor of the steam turbine of the power plant. During the experiment, the method in this paper was used to identify 9 common faults, such as mass unbalance, rotor thermal bending, shaft misalignment, rotating parts flying off, dynamic and static friction, steam flow excitation, structural resonance, insufficient structural rigidity, and rotor cracks. The experimental results show that the method decomposes and reconstructs the high-pressure rotor vibration mutation fault signal of the power plant steam turbine with high quality, and the obtained fault identification result is the same as the actual fault, the identification accuracy is high, and the result is reliable.