An iron mine in Yunnan Province primarily produces siderite. The mine is developed by horizontal adits and employs a segmented-caving mining method. The mining area is situated within a fault zone characterized by well-developed joints and fissures. The 1363-m-level haulage roadway is currently being excavated. The roadway is designed as a triple-centered arch with a width of 3 m, a sidewall height of 1.8 m, and an arch height of 1 m. Field geological survey indicated that the roadway surrounding rock is mainly fractured Grade III rock. Some sections are low-quality Grade IV rock. The surrounding rock is primarily sedimentary-metamorphic carbonate rock, and the exposed rock is predominantly limestone.

SFRC is used as support in the mined 1363-m-level roadway. The concrete is specified as grade C20. The specified support thicknesses are 30 mm and 60 mm. Field investigations identified two problems for this support strategy: short-term support performance was difficult to guarantee, and long-term durability in wet environments was inadequate. In the short term, SFRC slurry sprayed onto rock surfaces exhibits poor adhesion and inadequate consolidation. A significant amount of slurry falls downward or flows along the roadway sidewall. These deficiencies render it difficult to ensure adequate sprayed layer thickness and quality. The sprayed layer exhibits cracking and spalling upon water exposure within 3–10 days of slurry set. The severity of these issues becomes increasingly pronounced over time.

The application of SFRC support in the 1363-m-level roadway showed problems such as unreliable short-term support performance and detachment/cracking. These issues are primarily constrained by material properties and inadequate matching of support parameters.

Influence of material properties

SFRC is produced by adding steel fibers, silica fume, and other constituents to ordinary concrete. The fibers form bridges that inhibit crack development and improve the matrix’s impact resistance and durability. Accelerators are incorporated to accommodate rapid mining by shortening setting time and reducing rebound rate. However, varying accelerator dosages change concrete setting time, which in turn modifies short-term mechanical properties and reduces the immediate effectiveness of SFRC support. 2.

Support parameter mismatch

The thickness of the loosened zone was calculated by applying the loosening-zone theory together with the Mohr-Coulomb and Hoek-Brown strength criteria using Equations 1–4. P 0 = σ ci C m b − σ ci s m b + σ ci C α (1) C = m b 1 − α ln R 0 r 0 + m b P i σ ci + s 1 − α 1 1 − α (2) P i = P 0 1 − sin ⁡ φ − cos ⁡ φ (3) R 0 = R 1 1 + sin ⁡ φ 1 − sin ⁡ φ 2 ⁡ sin ⁡ φ (4) where R is the thickness of the loosened zone (m). R ₀ is the radius of the loosened zone (m). P ₀ is the initial in-situ stress (MPa) and was taken as the mean of in-situ measurements at the mining site. m _b , s, and α are the dimensionless empirical parameters of the Hoek-Brown criterion. Their values were taken as m _b = 1.2, s = 0.004, and α = 0.5for a rock-mass quality of GIS = 45. r ₀ is the roadway radius. The equivalent radius is r ₀ = (b + 2h ₀)/4 for a triple-centered arch roadway, where b is the roadway width and h ₀is the full roadway height. Substituting the roadway dimensions yields r 0 = 2.15 m .P _i is the support pressure (kN). σ ci is the uniaxial compressive strength (MPa). φ is the internal friction angle (°). cis the cohesion (MPa). Their values are listed in Table 1.

The calculated loosened zone thickness for the roadway was 1.73 m. According to the loosened-zone theory, the loosened zone in this roadway is classified as large and forms beneath unstable rock masses. A common support method for this situation is bolt-mesh-shotcrete support. However, this method performs poorly in loose and fractured rock mass with well-developed joints and fissures. Moreover, it cannot provide timely support during mining operations and involves a relatively cumbersome construction process. Therefore, after comprehensive consideration of the geological characteristics, SFRC support was chosen for this roadway.

Although the roadway in this mine employed SFRC support, the selection of support parameters was primarily based on empirical analogy and lacks theoretical justification. The New Austrian Tunnelling Method (NATM) indicates that an inadequately designed SFRC thickness can result in insufficient strength and stiffness of the support. This prevents the support structure from effectively resisting surrounding rock pressure and water erosion, leading to cracking and spalling.

The cement used was ordinary Portland cement (P.O.42.5). The mass ratio of fine to coarse aggregate was 1:1. The fine aggregate was sand with a particle size of 3–5 mm. The coarse aggregate consisted of pea gravel with a particle size of 5–15 mm. Steel fibers measured 30 mm in length and 0.5 mm in diameter. The accelerator was an HQ-240 alkali-free liquid accelerator. The effect of the alkali-free liquid accelerator on support performance is concentrated in the early support stage: it substantially increases early concrete strength, results in a slower mid-term strength gain rate, and has little effect on long-term concrete strength. The water-to-cement ratio was 0.45. Table 2 lists the quantities of each raw material required for one cubic meter of SFRC.

Figure 1 shows photographs of the raw materials used in this study, including cement, silicon powder, steel fibers, and accelerator.

The tests primarily investigated the effects of different accelerator dosages and curing ages on the mechanical properties of SFRC. These results were then used to evaluate the short-term support performance of the material. The case mine used an accelerator dosage of 7.5%. Accordingly, the accelerator dosages were set to 5%, 7.5%, and 10%, respectively, while all other mix proportions were held constant. Field practice indicated that shotcrete sets in 2–4 h and typically attains about 70% of design strength at 7 h. Therefore, curing ages of 4 h and 7 h were chosen to represent the post-spray strength-gain period and the pre-secondary-spray stability check, respectively. Six specimen groups (A–F) were prepared (Table 3). Each group comprised six SFRC specimens prepared by hand rodding and tamping according to the Standard for test methods of concrete physical and mechanical properties (GB/T 50,081-2019) (Ministry of Housing and Urban-Rural Development of the People’s Republic of China, 2019). Three specimens were used for uniaxial compression tests, while the other three were used for variable-angle shear tests.

SFRC specimens cured for 4 and 7 h were subjected to uniaxial compression and variable-angle shear tests. The uniaxial compression test directly provided UCS. Measured stress-strain curves also allowed for the calculation of the SFRC’s Poisson’s ratio and elastic modulus. The variable-angle shear tests yielded the failure load for each specimen, from which the corresponding normal and shear stresses were computed. The resulting normal-stress versus shear-stress envelope was used to determine cohesion and the internal friction angle.

The uniaxial compression specimens were non-standard in size, therefore, the raw measurements required correction. UCS and elastic modulus are subject to size effects. When specimen size is below the standard, the size effect tends to cause overestimation of both parameters. Poisson’s ratio is largely size-independent and can be approximated as equal to that of a standard specimen. The UCS and elastic modulus were corrected using the Equations 5, 6: f c u , k ′ = 0.9 f c u , k (5) E ′ = E × 150 a n (6) where f cu , k ′ is the corrected unconfined compressive strength of the standard specimen. f cu , k is the measured unconfined compressive strength of the non-standard specimen. E ′ is the corrected elastic modulus of the standard specimen. E is the measured elastic modulus of the non-standard specimen. a is the side length of the non-standard specimen. n is an exponent related to concrete properties. The n for converting the elastic modulus from nonstandard to standard SFRC specimens typically ranges from 0.4 to 0.6,this study adopts 0.5.

The uniaxial compression and variable-angle shear test results for the SFRC specimens are listed in Supplementary Tables S1, S2. The calculated mechanical test results for the standardized SFRC specimens are presented in Supplementary Tables S3. Figure 2 display the resulting mechanical parameters for the specimens under different test conditions.

Figure 2a showed that the UCS of the SFRC specimens decreases as accelerator dosage increases. Increasing the accelerator dosage from 5% to 10% reduced UCS by 1.0 MPa at 4 h and 1.4 MPa at 7 h. This effect is attributed to the use of a liquid accelerator. Higher dosages increased the mixture’s free water content, which reduced short-term solidification and early UCS. UCS increased with curing age; specimens tested at 7 h exhibited higher strength than those tested at 4 h. Overall, the maximum UCS was observed at 5% accelerator dosage and 7-h curing age.

Figure 2b indicated a positive correlation between the accelerator dosage and the elastic modulus for SFRC specimens when the accelerator dosage was 5%–7.5%. Increasing the accelerator dosage from 5% to 7.5% raised the 4-h elastic modulus from 2.5 GPa to 6.8 GPa. The elastic modulus dropped as accelerator dosage increased from 7.5% to 10%. Overall, the elastic modulus was higher at the 4-h curing age.

Figure 2c indicated a negative correlation between the accelerator dosage and the internal friction angle for SFRC specimens at 4 h. Higher accelerator dosage lowered the internal friction angle and therefore reduced the concrete’s strength. In general, the internal friction angle was more favorable at the 4-h curing age.

Figure 2d showed that the cohesion of SFRC specimens cured for 4 h was little affected by the accelerator dosage and remained around 0.7 MPa. The cohesion of SFRC specimens cured for 7 h reached a maximum of 2.0 MPa at an accelerator dosage of 5%. Cohesion was positively correlated with curing age. Overall, within the tested mix-design range, lower accelerator dosage improved the shear resistance of SFRC.

The aforementioned theoretical analysis and laboratory tests indicated that short-term SFRC support performance is primarily governed by the accelerator dosage and the support thickness. This section performed numerical simulations to assess how SFRC cured for 4 h and 7 h affects roadway support performance and to determine optimal support parameters for practical application. The accelerator dosages were set to 5%, 7.5%, and 10%, respectively. Support thickness choices were based on rock mass quality guidance. Generally, roadways with Grade III-IV surrounding rock require a shotcrete thickness of 80–100 mm, whereas those exhibiting heavily loosened zones generally call for 50–100 mm. The mechanical properties of SFRC are superior to those of ordinary concrete. Therefore, support thicknesses were set to 30 mm, 60 mm, and 100 mm, respectively. Eighteen simulation scenarios were then established using a single-factor control variable method. These scenarios are summarized in Table 4.

A 3D numerical model with dimensions of 30 m × 30 m × 30 m (length × width × height) was developed in Midas GTS NX. The model comprised 195,720 elements and 201,056 nodes. The numerical model is illustrated in Figure 3. To assess the model’s accuracy, conducted a mesh sensitivity analysis. The results indicate that displacement and stress outcomes stabilise when the number of mesh elements exceeds 150,000, with variations falling below 2%. Furthermore, the simulated displacements closely match the recorded field monitoring data, validating the numerical model. The established model was imported into FLAC3D for computational analysis. It should be noted that this study treats SFRC as an isotropic continuous medium with uniformly distributed fibres, disregarding environmental factors such as temperature and humidity. An ideal elastoplastic constitutive model with Mohr-Coulomb failure criteria was used. The magnitude of the load P₀ is determined by taking the average value from measured in-situ stress data within the mining area. The load is applied to the top of the model to simulate the self-weight stress of the overlying strata. The lateral pressure coefficient was computed as 0.43 according to the Heim-Kinnick theorem. Lateral boundary constraints normal displacement to simulate the confining effect of the infinite far-field surrounding rock. The model bottom boundary was fixed.

The SFRC support performance under different support durations was simulated by assigning the mechanical parameters listed in Supplementary Table S2 to the support elements. These parameters were measured for varying accelerator dosages at curing ages of 4 h and 7 h. SFRC sprayed onto the rock surface develops strength rapidly. Therefore, directly assigning the 4-h and 7-h material property sets to the support elements does not accurately capture the roadway’s short-term response after support. Accordingly, a FISH-language script was developed to define the time-dependent short-term strength development of the SFRC. Strength development was modeled with an exponential function using a growth parameter of 2.5. This implementation reproduced the short-term change in support performance after shotcrete support application. The core computational steps can be expressed by Equations 7, 8. S = 1 − 1 − s exp k · t 0 T , S ≤ 1 (7) t 0 ′ = t + t 0 , t 0 ′ ≤ T (8) C = S · F (9) where s is the initial strength ratio of the SFRC sprayed onto the rock surface and was set to 5%. S is the actual strength ratio of the SFRC. k is the strength growth rate and was set to 2.5. T is the support duration (4 h and 7 h in this study). t 0 ′ is the cumulative duration. The calculation terminates when t 0 ′ reaches the target support duration. t₀ is the cumulative duration from the previous calculation step, with an initial value of 0 s t is the time increment per calculation step and was set to 2 min. C is the calculated mechanical parameter value for SFRC. F is the target final mechanical parameter value for SFRC.

Time-dependent functions for strength ratio and mechanical parameters were implemented so that the sprayed SFRC’s properties evolve with curing age. Strength increased from an initial 5% to the target value. These updates were executed within the support simulation cycles and ceased once the preset support duration was reached. This approach addresses the difficulty of prescribing support duration in numerical models and captures the time-dependent evolution of material strength and support performance. Poisson’s ratio, however, was fixed at its final value and was not affected by the strength evolution function. The physical and mechanical parameters of the surrounding rock are listed in Supplementary Table S2.