In order to study the vertical vibration characteristics of the elevator system, a seven-degrees-of-freedom dynamical equation was established considering its time-varying traction rope stiffness based on the Lagrange equation. The design of experimental(DOE) method was applied to obtain the quantitative relationship between random variables and system frequency responses. Then, artificial neural network (ANN) technology was used to fit the explicit functional relationship between the random variables and system responses. According to the criterion that the ratio of the excitation frequency and the natural frequency of the system structure cannot exceed a certain value, the frequency reliability mode of the elevator system in the vertical direction was defined to calculate the frequency reliability and related frequency reliability sensitivity with respect to random variables and variances based on the frequency reliability theory. Finally, the reliability-based robust design of structures was studied using the reliability based optimization model, and the numerical method of frequency reliability-based robust design was proposed. The frequency reliability sensitivity and reliability robust design was discussed for elevator systems with vibration in the longitudinal direction using the above theory. The result shows that this method has high computational efficiency and accuracy, and can serve as a theoretical method for the robust design of an elevator system and its components.