Dynamic monitoring of oil particle pollutants based on machine vision is an important means of equipment operation safety evaluation. The main challenge is to distribute particles as evenly as possible in the micro-channels, thereby improving the accuracy and consistency of particle contamination detection. In this paper, COMSOL Multiphysics software was used to study 25 # Karamay transformer oil and copper particles with particle sizes of 4, 10 and 50 μm. The fluid dynamics of micro-channels of different lengths and widths and the motion trajectories and distributions of copper particles were simulated, and the dynamic detection effect of particles in the micro-channels was verified by experiments. The simulation and experimental results show that: (1) under the action of laminar flow, the flow field in the micro-channel is symmetrically distributed along the central axis, and the micro-channel width has a more significant effect on the velocity in the detection area. In addition, the inclination Angle can effectively improve the vortex phenomenon in the edge region of the micro-channel. (2) After 40 s of particle release, the particles in the micro-channel are evenly distributed, and the number of particles in the detection area is close to the total number of particles in this period after 5 times compensation; (3) The identification rate of particulate pollutants by the micro-channel model can reach more than 95%. The above analysis and experimental results provide a scientific reference for the particle detection and fault diagnosis of large equipment such as transformers.