Abstract: An approach combining Monte Carlo-stochastic finite element method and neural network is proposed for the reliability-based sensitivity design of a turbine disk. It can solve the problem that the explicit expression of the ultimate state equation of the aero-engine turbine disk can't be directly obtained because of its complicated structure. Considering such parameters of the turbine disk as the dimension of the geometry, material characteristics and randomness of the environmental load, the relational expression between the random response and random parameters is simulated using the nonlinear mapping function of neural network. Following the trained neural network, the first order second moment method is chosen to design the reliability. The effect of the parameters' means and variances on the reliability is achieved through the reliability-based sensitivity design of the turbine disk. The proposed method is verified by the structural reliability analysis software NESSUS. It provides a theoretical reference for reliability-based sensitivity design of the turbine disk.