Abstract: The trace amount of impurities can be eliminated by feeding additional flow into centrifuge cascade. To maximize the separation ability and economic efficiency of a cascade, the optimization is necessary. Based on the practical hydraulic model of centrifuge cascade and relevant experimental results, selecting cascade length, centrifuge number and control pressure of each stage, and flow amount of additional feed as the optimization variables, the optimization of centrifuge purge cascade was accomplished by bit coded genetic algorithm to minimize the total number of centrifuges employed. Two cases, including the purification of ⁹⁹Tc from low enriched recycled uranium and scrubbing of impurities in the production of high-purity WF₆, were studied to validate the viability of this optimization method. Calculation results show that the bit coded genetic algorithm can effectively solve the optimization of centrifuge purge cascade with a relatively strong feasibility.