The reliability problems of relays in service have been identified as the primary cause of repeated operational accidents involving missiles, rockets, and satellites. Consequently, accurately assessing the reliability of relays is a critical issue that requires immediate attention. However, addressing this issue is complicated by the multi-source variability (including unit-to-unit variability, temporal uncertainty, and measurement error) inherent in the degradation processes of relays. To fill these gaps, this paper introduces a novel reliability analysis method for relays, which is based on a nonlinear Wiener process model incorporating random effects. The analytical expression of the reliability function corresponding to the proposed model is derived. Subsequently, a statistical inference method is developed based on the maximum likelihood estimation. Furthermore, the efficacy of the proposed method is demonstrated and validated using degradation data from a specific type of relay.