Abstract: As the fault of screenplate is often detected with delay in time with the use of artificial method, a deep vision-based intelligent detection method is proposed, which can make intelligent diagnosis of fault based on the depth image of Plant using TOF camera, the distance between the station in the suspected fault area and the camera defined according to the 3-dimentional spatial relationship, and the depth image data in this area. As evidenced by test result, the use of the method can realize screenplate fault detection in a real-time and intelligent manner with a high accuracy, providing a quarantee for the normal operation of coal cleaning systems.