Abstract: This study aims to separate the soil and corm while harvesting water chestnuts on dry land. A spring roller device of soil-corm separation was also designed for the water chestnut harvester, in order to reduce the injury rate of water chestnut. Positive and negative rotating springs were arranged side by side. During the operation of the harvester, the soil-corm mixture was dug out and passed through the lifting device and rubber-roller soil removal device, and then fell into the spring roller soil-corm separation device, where the water chestnuts were further separated from the soil under the vibration and rubbing of the rotating positive and negative spiral spring rollers. The kinetic analysis was performed on the relative motion of the water chestnut and spring roller. The influencing factors on the separation performance were determined as the spring outer diameter, pitch, adjacent spring spacing, height difference and spring speed, line diameter and operation speed. The simulation of soil-corm separation was carried out to optimize the operating performance of the spring roller. The model was then established using EDEM software, including three particles: large soil clods, water chestnuts and fine-grained soil. A single-factor test was carried out on the structural and working parameters of the spring roller, in order to clarify the influence of each factor on the soil-corm separation. Taking the sieving rates of water chestnut and soil as the test indexes, the quadratic regression orthogonal test was also carried out. An optimal combination of parameters was obtained for the spring roller, where the outer diameter was 100 mm, the pitch was 30 mm, the spacing was 9 mm, and the rotational speed was 420 r/min. The better performance was achieved, where the sieving rates of water chestnut and soil were 80%, and 80.69%, respectively. The verification tests were then conducted to compare the prediction of the model. The average relative errors of water chestnut and soil sieving rate were 2.09%, and 2.42%, respectively. The simulated harvesting and one-way tests were carried out to take the test indexes as the rates of the open, the injured, the peeling, and the digging, with the soil moisture content and spring speed as test factors. Among them, the water chestnuts were pre-buried into the soil layer and then harvested. Three wire diameters of springs of 10, 12, and 14 mm were also selected to evaluate the mud-fruit separating of spring rollers. It was found that the 12 mm wire diameter spring shared both better vibration performance and lower damage rate of water chestnut. Actual harvesting tests were conducted to test the performance of the soil-corm separating device. The performance of the water chestnut harvester was measured as follows: the operating speed was 0.21 m/s, the operating efficiency was 0.19 m²/s, the soil breaking rate was 75.61%, the open water chestnut rate was 82.42%, the injured water chestnut rate was 14.73%, and the peeling water chestnut rate was 7.01%. The soil crushing and soil-corm separation were enhanced with the increase in rotational speed. However, the injury rate of the water chestnut also rose, and the rotational speed of the spring roller should not be more than 244.5 r/min. The broken soil rate was outstandingly improved with the reduction of soil moisture content, but the injured rate of water chestnut increased as well. The damage rate of water chestnut more outstandingly increased with the decrease of soil moisture content at the soil moisture content of 15.77% to 17.44%, compared with the soil moisture content of 17.44% to 19.40%. The finding can provide a strong reference for the development and optimization of water chestnut harvester.