Failure dependence including degradation interactions and cascading failures pose challenges for RUL estimation in complex systems. For degradation interactions, a multi-stage model-based RUL estimation approach is proposed. This method accounts for the interactions between components caused by uneven working pressures, where even identical components can experience different degradation processes. Interaction coefficients and multi-stage degradation models are key to estimating RUL, calculated as the time difference between healthy operation and failure. The approach is validated on an electro-hydraulic subsea Christmas tree, demonstrating its effectiveness in evaluating system performance and RUL under interaction effects. For cascading failure, a novel RUL estimation methodology is introduced, focusing on multilevel systems. Using DBNs, the approach models cascading failures based on the position and function importance of system nodes. By monitoring node states and iteratively calculating performance impacts across system levels, the number of failure nodes and overall system performance are assessed. This method is validated through a subsea transportation system with a three-level network. Both approaches offer tailored solutions for different failure scenarios, enhancing RUL prediction accuracy and supporting more effective system maintenance strategies.