With application of advanced sensing technology, the condition based maintenance and operation has been made possible to many industrial systems. In a wind turbine, there are a few hundreds of sensing signals used to monitor the component performance and operational condition. The condition information is utilized to operational control of wind turbines and the wind farm in order to reduce the down time and CoE (Cost of Energy). In this paper, a framework of condition based maintenance and operation of wind turbines is presented. This framework starts with data collection of sensing signals through SCADA and includes data processing and modeling, failure pattern recognition, remaining useful life/health condition prediction, load prediction (prediction of wind trend), integrated decision making for maintenance and operation of wind turbines and the wind farm, and maintenance planning. The research challenges involved in each step of the framework are discussed. The framework presented in this paper serves as a guideline which is also useful to other systems.