Abstract: The parameter optimization of vibration isolation system for refrigerator compressor is closely related to the effects of vibration isolation. A method of natural frequency discrete distribution to optimize vibration isolation was proposed to improve the shortcomings of easily falling into local optimum solution and iterative divergence for traditional optimization design with taking 6-DOF energy decoupling of the vibration isolation system as the objectives, and taking the stiffness of 4 rubber supports as the design variables. The chaos particle swarm algorithm based on penalty function constraints was developed to optimize isolation parameters for the first time. The results showed that optimized distribution of its natural frequencies is more reasonable, and the optimized decoupling ratios in main directions are improved significantly. Furthermore, it was shown that compared to the sequential quadratic programming and genetic algorithm, the particle swarm algorithm overcomes its weakness of converging to local optimal solution, and it is higher than genetic algorithm in terms of energy decoupling ratios. At last, it was verified with dynamics simulation that the proposed optimization method applied to vibration isolation design have both rationality and effectiveness.