4. Prototype testing4.1. Bed expansion and power consumption A prototype was made using the optimum combination of design factors chosen from the orthogonal test and the relation-ship between bed expansion and superficial velocity was measured. Comparing before-optimization with after-optimization results ,bed expansion increased with superficial velocity, and bed expansion after optimization was enhanced (Fig. 5) by an average 18.3%.Besides, the average improvement was about 13%, while bed expansion was 40%–120% and the superficial velocity was about1.59–2.71 cm/s. And the average improvement was larger in the other bed expansion. The possibility was that the flow pattern was unsteady, with slugging at low superficial velocity and an unclear Fig. 5. Comparison of bed expansion between before and after optimization. interface of sand and water at high velocity. Generally, bed expansion is operated in the range of 40%–100% in FSBs (Summerfelt,2006) in order to avoid washing away of sand and toxic gas because of anaerobic pockets. Hence, 40–100% bed expansion was focused up on in the following analysis. The relationship between bed expansion (y) and velocity (x) was analyzed using regression of before and after optimization, separately, y1= 0.49x1− 0.29 (R2= 0.99), y2= 0.52x2− 0.21 (R2= 0.98).The power consumption per hour for the pump is critical to the cost of operating aquaculture systems. The equation wasE = HQ_gq/3600_ (Sun, 2011), where E was power consumption per hour of pump, in units of kW/h; H was hydraulic head, unit sm; Q was flow rate, units m3/h; _ was density of seawater, units1030 m3/kg; g was acceleration of gravity, units 9.8 m/s2; q was exchange rate of water, units h−1; and v was the superficial velocity, units cm/s. The equation was simplified to E = 9.09 × 10−2v2fol-lowing input of relevant data. The result was that energy savings (E1−E2E1× 100%) after optimization were 21%–28% relative to before optimization at the same bed expansion (40%–100%) (Table 4). This outcome indicated that the structure of the optimized CB FSB was more efficient.