Abstract A immune genetic neural networks based on vaccine inoculation (VIG-NN) is proposed. VIG-NN is used in vibration fault diagnosis of steam turbine. The key to the algorithm lies in vaccine inoculation, antibodies selection based concentration and adaptive probability of cross and aberrance. The algorithm not only keeps the leading position of excellent antibody, but also develops the potential of rapidly growing antibody in seeking optimum. The search of algorithm to global optimum is rapid and effective. The practice shows that VIG-NN can simulate the nonlinear mapping relationship between fault symptom and fault type more accurately. The accuracy of vibration fault diagnosis of steam turbine is improved.