RFD (reverse flow divert) set is a kind of maintenance-free delivery system driving by compress-air. It intermittently discharges through its delivery pipe. RFD set has no moving part, and its optimized design is critical to efficient and stable operation. The deficiencies in optimization of traditional enumeration algorithm and the marginal utility method were indicated. Then, as a mainly point of the paper, using the genetic algorithm (GA) to solve the multi-objective multidimensional optimization problem was discussed. By comparing the results of the three optimization methods, the superiority of GA can be drawn. Based on the industrial needs for flow and head, a serialization design of RFD was done by using GA and the relationships of the structural parameters and the operational parameters were analyzed. And the conversion relationships of RFD average flow in different densities and viscosities were provided. The promoted approach will be benefit for the RFD set design in industry.