Recrystallization-Controlled Strength Evolution and Prediction of Compacted Rock-Salt Fills.

Locally available rock salt is a potential fill material for infrastructure construction in arid salt-lake regions, where conventional soil and aggregate resources are often scarce. However, its mechanical performance is strongly affected by water migration, brine concentration, salt dissolution, and possible recrystallization-induced bonding. This study investigates the unconfined compressive strength (UCS) of compacted rock-salt fills under different water contents, initial brine contents, relative compaction levels, and brine concentrations. The results show that UCS increases nonlinearly as water content decreases during air drying, with slow strength development at the early drying stage followed by rapid strength gain at low water contents. This behavior is consistent with a transition from capillary liquid bridges to stronger salt-related interparticle bonding during brine evaporation and possible recrystallization. The effect of initial brine content is state-dependent: higher initial brine content reduces UCS in the as-compacted state but markedly increases UCS after drying to constant mass, suggesting enhanced interparticle bonding after drying. Under constant-mass air-dried conditions, UCS first increases and then decreases with increasing relative compaction, reaching a peak of approximately 8.7 MPa at a relative compaction of 0.99. Lower brine concentration reduced the UCS of dried specimens, indicating that saturated brine may be more favorable for specimen preparation and field compaction under dry salt-lake conditions. An empirical UCS prediction equation incorporating water content, dry density, initial brine content, and brine concentration was developed and validated, giving good agreement with the test results with R2 = 0.91. The findings provide a material-level basis for evaluating and controlling compacted rock-salt fills under controlled dry conditions, while further durability and field validation are required before extending the results to long-term railway subgrade performance.