Advanced reliability enhancement and cost reduction techniques for wind energy generation

Wind turbines are complex mechanical-electrical systems. Their design, operation, and maintenance directly affect the cost of energy generated. Possible failure modes of wind turbines include wear, crack, misalignment, lubrication, fatigue, etc. The challenges in these areas include better understanding of the dynamic behavior of wind turbines under dynamic wind loading conditions, early detection and diagnosis of fault initiation within the moving mechanical components, prediction of fault propagation, and decision making in terms of maintenance and operation. In this project, we will develop (1) multi-body dynamic models to generate vibration responses of wind turbines under time-varying loading conditions and various health conditions; (2) advanced signal processing algorithms to analyze vibration signals generated by wind turbines for early fault detection, accurate fault diagnosis, and assessment of fault degree; (3) machine learning, data mining, and artificial intelligence techniques for pattern recognition and prediction of the remaining useful life of running wind turbines; and (4) statistical and intelligent reasoning approaches for decision making under uncertainty for operation and maintenance of on-shore and off-shore wind farms utilizing the knowledge developed in the first 3 research areas. A specific research topics are listed next. Advanced multi-body simulation tools can be developed for studying the complex dynamic behavior of gearboxes that exist in wind turbines. End-to-end deep neural networks may be utilized to map directly from the collected vibration signals to the hidden source fault signals. Stepwise signal component separation algorithms may be tested for impulsive feature identification in multisource coupled signals. Stochastic optimization models may be developed for off-shore wind farm maintenance decision making under wave-height and sea state conditions.