Health monitoring, diagnostics and prognostics of mechanical systems

Health monitoring is a very important topic in terms of safety, durability and cost of mechanical components and structures in systems such as aircrafts, power plants, rotational machinery. An extensive research and application work has been done in this area. Approaches to address health monitoring and damage detection of mechanical components generally consist of two aspects: (i) sensing and signal processing; and (ii) mathematical modeling. Both approaches are addressed in this dissertation.; The need for taking full advantage of knowledge of diagnostics experts requires an effective and reliable communication mechanism to implement remote health monitoring. A remote diagnostic system based on CORBA (Common Object Request Broker Architecture) is developed and applied to rotational machinery in a large iron and steel corporation. Diagnosis results show that this is an effective method to implement remote health monitoring of machinery. A fatigue crack growth model, which can capture the effects of variable-amplitude load, is needed to predict the severity of damage of mechanical components. A state-space model is provided to address this problem. Improvements of the state-space model are presented and the model performance is compared with an extensively used model FASTRAN II. The state-space model is further validated by the experimental data on crack length and crack opening stress. A third order difference equation is then presented to give more accurate predictions of fatigue crack growth under various loading conditions. This third order state-space model has better predictions than the second order state-space model for over/under and under/over load extrusion effects. Simulation comparisons are presented between the state-space models. FASTRAN II model and experimental data. In order to investigate deeply the internal mechanism of fatigue and to understand the physics of fatigue, a stochastic model of collective motion of dislocations is presented to address this problem. Four basic dislocation mechanisms: generation, gliding, annihilation and pinning of dislocations are accounted for through a series of constitutive rules and incorporated into this model. The randomness of fatigue crack growth is addressed in this model by investigating the evolution of dislocation distribution in the grain at the crack tip with the cycling of a specimen. Fatigue crack growth at overloads and underloads are explained with dislocation perspective.