With the demand for clean energy is increasing, wind power generators are becoming more common. Although the wind power generation industry has broad prospects for development, there are still problems in wind generator maintenance, especially faulty detection and repair. In order to increase the efficiency of maintenance, the digital twin technology can be used. The mathematical model and the fault tree can help to spend less time alarming. In this article, the use of digital twin in the maintenance and faulty detection of wind generators is analyzed. The methods of faulty detection and the model of fault tree in the maintenance of the wind generator are summarized. In addition, a faulty detection method is proposed through YOLOv5. This method can detect the fault without contacting the gears. The database of faulty detection can be optimized as well. This article may offer a reference for digital twin and fault detection in wind generators.