In order to reduce the tillage resistance and soil disturbance of the multipurpose machine for improving perennial alfalfa field, a new type of bionic ripper was designed based on the contour curve model of the outer edge of the anteater’s claw toe. Based on the soil characteristics of the perennial artificial alfalfa field, the discrete element model of the soil contact part-soil interaction was established by using EDEM software. The forward resistance and soil disturbance area were used as the main evaluation indicators. The bionic loosening shovel and the light standard deep loosening were performed under different operating conditions through numerical simulation of the shovel working process and field performance comparison test. The results showed that the average drag reduction rate of the bionic shovel was 7.64%, and the error between the simulated value and the measured value was less than 9%. In order to optimize the structural parameters of the wing shovel, the wing inclination angle and the wing opening angle were used as the test factors, and the sensor tension value and the groove width were used as the test indexes, and the regression mathematical models were obtained. Particle swarm optimization (PSO) was used to solve the regression mathematical model, and the Pareto optimal solution set was obtained. Finally, the sensor tension value was 8.679kN, the groove width was 144.2mm, and the wing inclination angle was 20°, and the wing opening angle was 105.6°. Field verification test result showed that the relative error between the actual value and the predicted value was less than 6%. The multi-objective parameter optimization method based on RSM and PSO was scientific and feasible.