Machine Health Prediction for Condition-Based Maintenance

The failure of machine components is expected but difficult to predict. The worst-case maintenance procedures are very expensive, as all parts of a given type must be replaced after a fixed number of hours, regardless of the fatigue and wear that they have experienced. A less restrictive high-fatigue schedule reduces costs but may result in a larger number of catastrophic failures. By analyzing the vibrational characteristics and performance parameters of an operating machine, one can determine the extent of degradation due to fatigue without the necessity of dismantling the engine and performing a detailed inspection of its components. The conditionbased maintenance will result in large cost savings by extending the useful life of mechanical components, reducing unnecessary maintenance, and safeguarding against premature failure of components. ORINCON proposes to develop an automated health monitoring system to predict the development of mechanical faults in an operating turbine engine and to provide suggestions for the engine maintenance schedule, based on the predicted failure mode, The system will be implemented as a combination of advanced signal processing and neural network expert systems. A general framework for machine health monitoring and prognosis will be developed and applied in Phase 11 to a ship system.