The goal of the proposed research is to develop real-time non-destructive evaluation (NDE) methods for health monitoring and residual life prediction of mechanical structures and aging machinery. The specific objectives are: quantification & estimation of fatigue crack damage; and assessment of the time to onset of widespread fatigue damage. The technical approach relies on fusion of heterogeneous information derived from physics-based analytical models and real-time sensor data. The analytical part of the research will be supported by laboratory experimentation on a special-purpose fatigue test apparatus that is equipped with multiple sensing devices - optical, electromagnetic, and ultrasonic. Fusion of multiple sensor data with a stochastic state-space model of fatigue crack growth will allow non-destructive evaluation of the current state of damage and prediction of the residual life in real time. This hardware-software combination constitutes a novel NDE system that can be executed on inexpensive platforms such as a Pentium processor and is ideally suited for adaptation to operating machinery. As such the NDE system will also serve as a real-time analytical sensor of fatigue crack damage for control applications. The proposed NDE system will enhance safety and productivity in a wide range of industries and thereby reduce life cycle costs. Specifically, the NDE system will be applicable to real-time health monitoring and residual life prediction in active (e.g., turbines and pumps), semi-active (e.g., heat exchanger tubes), and passive (e.g., pressure vessels) structures that are subjected to both low-cycle and high-cycle fatigue damage. Furthermore, it will allow generation of early warnings to avert impending failures of critical plant components, which will reduce the probability of emergency shutdowns and catastrophic accidents.