Soil detachment is a key subprocess of soil erosion and is significantly influenced by freeze-thaw cycles. However, studies on soil detachment capacity under alternating freeze-thaw conditions are limited, and most existing research adopts time-consuming full-factorial designs. This study employed three experimental designs (Taguchi method, orthogonal design, and full-factorial design) to investigate soil detachment capacity under freeze-thaw conditions. Freeze-thaw simulations and flume scouring tests were conducted to quantify the effects of freeze-thaw cycles, initial moisture content, flow discharge, and soil type on soil detachment capacity. The modeling and optimization performance of the three methods were also compared. The results showed that discharge, freeze-thaw cycles, and initial moisture content were positively correlated with soil detachment capacity, while soil type showed a negative correlation. All three methods identified consistent optimal conditions: 16% initial moisture content, 14 L min-1 discharge, 15 freeze-thaw cycles, and sandy loam soil. Discharge contributed 42%-63% to the variability in soil detachment capacity, followed by initial moisture content (19%-28%), freeze-thaw cycles (11%-22%), and soil type (5.5%-7.6%). Compared to the orthogonal design (coefficient of determination [R2] = 0.92, relative error [RE] = 22%), the Taguchi method (R2 = 0.94, RE = 20%) demonstrated superior predictive performance. These findings provide a theoretical basis for promoting the application of the Taguchi method in freeze-thaw erosion research.