When the automatic weather station instrument malfunctions or its operating performance declines, special equipment for maintenance and calibration is often used, but it is more difficult for routine maintenance of the wind observation instruments installed at the height of more than ten meters or dozens of meters. In order to improve maintenance capabilities, for the wind observation instruments this paper puts forward fault detection method for wind direction sensors based on the observation data. Using the observation data of hourly extreme maximum wind direction and instantaneous wind direction from 2420 automatic weather stations in China, during 2009-2011 and based on the coding theory of Gray code disk of wind direction sensors, we design the fault detection method of Gray code for wind direction sensors. Through quality control on the national wind observation data, and the simulation of wind direction changes caused by the faulty of Gray code, we assess the impact of the Gray code failure. The analysis results show that: (1) By using the observation data of hourly extreme maximum wind direction and instantaneous wind direction, Gray code failure can be detected. (2) Throughout the country, the automatic weather stations seeing faults on the bits from 2 to 7 of the Gray code, takes up the proportion of 0.4%-0.8%. The occurrence proportion of the first bit of Gray code failure is 2.6%. (3) Gray code failure impacts the quality of wind observation data, and even causes the distribution of the wind direction frequency completely distorted, especially the fault of Gray code bit from 4 to 7.