The invention discloses a vibration detection process method for a high-precision coal mill. The vibration detection process method comprises the following steps: (a) acquiring n groups of historical running data for reflecting the running conditions of the coal mill in sequence according to a time sequence and obtaining actual vibration intensity data respectively corresponding to the n groups of historical running data through actual test; (b) building a coal mill vibration detection model by adopting a phase space algorithm and an extreme learning machine algorithm based on the above data and then performing optimized solution on the detection model; (c) performing online measurement on current working data of the coal mill and then obtaining corresponding output results according to the best detection model, wherein the output results are required current vibration state results of the coal mill. Through the vibration detection process method, the real-time detection on the vibration intensity of the coal mill can be realized in ways of high convenience in operation and control, high efficiency and high precision. The vibration detection process method is particularly suitable for application occasions such as large-scale thermal power plants and the like.