Abstract: This study aimed to develop a longitudinal conical-frustum dual-roller progressive-cutting banana straw crushing and returning machine and identify a parameter combination that balanced crushing quality, deposition within the intended working swath, and machine power consumption. The design addressed uneven cutter engagement associated with thick banana stalks entering the crushing chamber at different heights. Two longitudinal blade rollers were combined with a progressively decreasing cutter clearance to extend stalk engagement and support successive fragmentation. The study investigated the working principle and principal performance factors and evaluated the optimized parameters through simulation and field verification. The crushing process was divided into pickup, lifting, and penetration stages, and force analysis guided parameter selection. A bonded-particle banana straw model was established using the discrete element method (DEM) and the Bonding contact model. Blade roller speed, center radius, and inclination angle were investigated using a Box–Behnken design comprising 15 simulation runs. Bond breakage proportion served as a simulation surrogate for the qualified crushing rate of banana straw. The proportion of particles deposited outside the effective working width represented the out-of-swath spreading rate for straw returning. Average operating power was calculated from roller torque and rotational speed. Quadratic regression, response surface analysis, and optimization were followed by three repeated field tests. The crushing assembly incorporated two parallel longitudinal blade rollers with small-end and large-end envelope diameters of 300 and 500 mm, respectively. Each roller carried four longitudinal cutter rows with seven blades per row. The increasing roller radius reduced cutter clearance along the feeding direction and supported successive cutting engagements. Across the simulation runs, the qualified crushing rate of banana straw ranged from 80.2% to 96.9%, the out-of-swath spreading rate for straw returning ranged from 4.1% to 8.9%, and machine power consumption ranged from 1636.86 to 3737.89 W. All three quadratic models were significant at the 0.01 level, whereas their lack-of-fit terms were not significant at the 0.05 level. Rotational speed and center radius significantly affected the qualified crushing rate of banana straw. Rotational speed and inclination angle significantly affected the out-of-swath spreading rate for straw returning and machine power consumption. The interaction between center radius and inclination angle also significantly affected machine power consumption. Response surface analysis showed that crushing performance initially increased and subsequently decreased with increasing center radius or inclination angle under the corresponding fixed-factor conditions. These trends suggested that intermediate structural settings favored stalk–cutter engagement, whereas excessive settings could restrict effective fragmentation. The out-of-swath spreading rate generally decreased and then increased as the investigated factors increased under the corresponding fixed-factor conditions. Machine power consumption increased with rotational speed and rose gradually with inclination angle. Optimization yielded a blade roller rotational speed of 1488 revolutions per minute, a center radius of 198 mm, and an inclination angle of 86.7°. The corresponding predicted qualified crushing rate of banana straw, out-of-swath spreading rate for straw returning, and machine power consumption were 91.6%, 4.2%, and 2483 W, respectively. Field verification yielded corresponding values of 92.8%, 4.5%, and 2485 W. Relative differences were 1.3% for the qualified crushing rate and 6.7% for the out-of-swath spreading rate. Fine straw fragments were observed, and no obvious blockage or severe wrapping occurred during the field tests. Compared with the reported power of 5.1 kW for a sliding-cutting anti-wrapping banana straw returning machine, the prototype consumed 2.485 kW, a 51.3% lower value. This comparison involved separate studies and did not establish a reduction under identical conditions. The optimized structure achieved satisfactory crushing and target-area deposition under the reported field conditions. The results supported coordinated evaluation of crushing quality, deposition, and power during machine optimization. Successive cutter rows and decreasing clearance were intended to distribute cutting engagement over an extended stroke. Agreement between predictions and field measurements supported parameter selection within the investigated range. The study provided a structural and operating reference for banana straw crushing and returning machinery, although broader applicability required further validation under different material and field conditions.