Quantitative characterization of internal stress fields in fracture-dominated geological materials remains a significant challenge due to the limitations of conventional measurement techniques. This study presents the first quantitative full-field stress analysis of slightly sandy dolomite (Level I sandification) using an enhanced CT-3D printing–photoelasticity workflow. Five transparent physical models were fabricated from CT-scanned dolomite specimens to replicate the natural fracture-matrix structure and tested under diametrical compression (800 N) using ten-step phase-shifting digital photoelasticity. To overcome the severe optical noise generated by dense fracture networks, a robust phase unwrapping procedure (CPULSI) was incorporated into the data processing pipeline, enabling continuous stress parameter retrieval where conventional unwrapping methods fail. The recovered full-field principal stress-difference maps reveal that the internal stress field is dominated by meso-scale fracture geometry: Stress concentrations localize at fracture tips and narrow intact matrix bridges, reaching 3–5 times the far-field stress, while the macro-scale loading pattern becomes progressively obscured as fracture complexity increases across the five models. Quantitative validation against CT-based finite element simulations (RFPA-3D) demonstrates good agreement in intact matrix regions, with mean relative errors of 9–18%. These results provide new experimental evidence for the meso-scale stress distribution mechanisms governing the mechanical behavior of sandy dolomite—a geomaterial of significant engineering relevance in Southwest China—and establish a validated experimental pathway for investigating stress fields in other fracture-dominated geomaterials.
This study benchmarks five phase unwrapping algorithms—Goldstein’s branch-cut, Flynn’s minimum discontinuity, quality-guided, least-squares, and CPULSI—for digital photoelastic stress field reconstruction using a 3D-printed diametrically compressed disk with a known analytical solution. Quantitative evaluation under loads of 700 N and 1500 N reveals that all algorithms perform well under low-gradient conditions (correlation coefficients > 0.95), but path-following methods degrade significantly under high gradients: Goldstein’s algorithm fails with a Pearson correlation of −0.4127. The least-squares and CPULSI algorithms maintain robust performance, ranking highest in composite evaluation. The composite robustness ranking is as follows: least-squares ≈ CPULSI > Flynn’s > quality-guided > Goldstein’s branch-cut. These findings provide quantitative engineering guidance for algorithm selection in photoelastic stress concentration analysis. The main contribution of this work is not the proposal of a new unwrapping algorithm, but the establishment of a unified experimental–theoretical benchmark for evaluating representative phase unwrapping algorithms in digital photoelastic stress field reconstruction.