Steel Fibers and Smart Anchorage Design Reshape the Limits of Ultra-High-Performance Concrete

Ultra-high-performance concrete, or UHPC, has long been celebrated as one of the most remarkable materials in modern bridge engineering, offering compressive strengths that dwarf ordinary concrete along with exceptional tensile resistance, durability, and crack-control capacity. Yet a persistent engineering puzzle has limited how confidently designers can exploit it: what actually happens in the anchorage zones of prestressed structures, where enormous concentrated forces funnel from steel anchorage devices into the surrounding concrete? A new experimental and analytical study published in Case Studies in Construction Materials now provides one of the most systematic answers to date, showing that the local bearing behavior of UHPC is governed by a coupled compression–splitting–confinement mechanism rather than simple compressive strength alone.

The research team, led by Bowen Zhang, Zhi-Qi He, Jiatong Chen, and Wenjie Li, conducted eighteen local compression tests on UHPC specimens designed to represent prestressed anchorage zones. The specimens varied systematically across four key parameters: steel-fiber volume fractions of 0.5, 1.0, and 2.0 percent; bearing area ratios of 3, 5, and 7; three forms of indirect reinforcement, namely spiral reinforcement, reinforcing mesh, and none at all; and three anchorage plate configurations, including flat plates, embedded plates, and no plate. The main group of specimens was arranged in an L9 orthogonal experimental matrix, allowing the relative influence of the primary variables to be ranked without running a full factorial program, while targeted comparison groups isolated the effects of plate type and reinforcement ratio.

The material itself was a dense cementitious composite of ordinary Portland cement, silica fume, slag powder, fine quartz sand, water, superplasticizer, and straight copper-coated steel fibers 14 millimeters long and 0.20 millimeters in diameter, with an aspect ratio of 70. Measured axial compressive strengths ranged from 88.0 to 108.7 megapascals, with converted cube strengths between 122.2 and 151.0 megapascals. Direct-tension tests revealed mean tensile strengths of 3.89, 6.34, and 8.15 megapascals for the three fiber contents, a clear demonstration of how dramatically steel fibers transform the tensile character of an otherwise brittle matrix. All specimens were tested in a 500-tonne hydraulic compression machine under carefully controlled stepwise loading, with local plate indentation separated from overall specimen compression through a dual displacement measurement scheme.

Every specimen failed by splitting rather than compressive crushing, and the failure process unfolded in three recognizable phases: crack initiation, stable crack propagation, and rapid splitting collapse. Initial vertical hairline cracks appeared at loads ranging from roughly 30 to 90 percent of the ultimate capacity, most often in the upper or upper-middle regions of the specimens. Close-up examination of the fracture surfaces revealed that most steel fibers were pulled out of the matrix rather than ruptured, confirming that fiber bridging and pull-out resistance were the dominant mechanisms transferring tensile stress across cracks after the matrix had split. Specimens with lower fiber contents developed fewer but wider cracks, typically one or two dominant vertical splits, while higher fiber contents produced finer, more distributed cracking patterns.

Strain gauges on the indirect reinforcement told an equally revealing story about when different load-resisting mechanisms activate. At low load levels, reinforcement strains remained small, indicating that the UHPC matrix and its embedded fibers carried the transverse tensile demand on their own. Only as loads approached roughly 60 to 80 percent of ultimate capacity did reinforcement strains accelerate sharply, coinciding with visible cracking and lateral expansion. Specimens with 2.0 percent fiber volume kept reinforcement strains well below the monitoring reference even near failure, suggesting that abundant fibers reduce the confinement burden that transverse steel must carry. In other words, fibers and reinforcement work in sequence rather than in parallel: fibers manage microcracks and early post-cracking stress transfer, while spirals and meshes restrain macrocrack opening and lateral dilation later in the loading history.

Range analysis of the orthogonal test matrix produced a striking hierarchy of influence. The bearing area ratio, which controls the intensity of stress concentration and the geometry of load diffusion beneath the plate, was by far the dominant factor, with a range value of 65.8 megapascals in mean bearing stress. Fiber volume fraction ranked second at 49.0 megapascals, while the form of indirect reinforcement trailed at just 15.9 megapascals. The anchorage plate configuration mattered in nuanced ways: flat plates increased ultimate capacity by 15.4 percent in one matched comparison but simultaneously reduced initial stiffness and cracking load, whereas embedded plates consistently improved stiffness and cracking resistance, raising cracking loads by as much as 71.4 percent, without reliably increasing ultimate strength. Changing the reinforcement ratio between 1.5 and 3.0 percent produced no monotonic trend in either cracking or ultimate load, underscoring that its effect is conditional on the surrounding fiber and plate-transfer state.

To place these findings in context, the team compiled a database combining their eighteen tests with published UHPC and reactive powder concrete data, yielding 51 unreinforced and 34 reinforced records. When existing design provisions were evaluated against this database, their limitations became evident. AASHTO LRFD 2017 came closest on average, with a mean calculated-to-test ratio of 0.947 for unreinforced specimens, but the French UHPC standard NF P 18-710 tended to overestimate capacity with a mean ratio of 1.154, and the Chinese bridge code JTG 3362-2018 overestimated reinforced specimens at 1.092. The Chinese steel-fiber design standard JGJ/T 465-2019 performed worst, overestimating capacity by roughly 30 percent with standard deviations above 0.20, indicating that its fiber and confinement terms fail to capture the distinctive splitting resistance and fiber-bridging behavior of UHPC.

In response, the researchers developed two prediction models grounded in the observed mechanics. The first is a modified confinement-based model that expresses bearing capacity as the sum of a fiber-enhanced UHPC contribution and an additional confinement term from indirect reinforcement. A steel-fiber characteristic parameter captures the reinforcing efficiency of the fibers, while a regression-derived coefficient reveals that the confinement contribution of transverse steel actually decreases as fiber content rises, vanishing entirely beyond a fiber characteristic parameter of 1.38. The model achieved calculated-to-test ratios of 0.993 with a standard deviation of 0.081 for unreinforced specimens and 1.013 with a standard deviation of 0.105 for reinforced ones, outperforming existing literature models in both accuracy and scatter.

The second model takes a more theoretical route, treating local bearing failure as a plasticity problem. Using a modified Mohr-Coulomb criterion with a nonzero tensile cutoff to represent fiber-bridged post-cracking strength, the team idealized the failure zone as a wedge mechanism combining a shear-compression plastic flow zone with a tensile splitting zone. An upper-bound virtual work solution, calibrated with a reduction factor of 0.172 against 25 unreinforced database records, reproduced measured capacities with a mean ratio of 1.00 on the calibration set. The authors are candid that this wedge model, with its rigid-plastic idealization and simplified geometry, is best suited for mechanism interpretation and analytical assessment of unreinforced zones rather than direct design use, while the confinement-based model serves as the practical design tool within the calibrated parameter ranges.

The implications reach well beyond the laboratory. Prestressed bridge girders, stay-cable anchorages, and other heavily loaded UHPC components can now be designed with equations that explicitly recognize how steel fibers and transverse reinforcement share the job of resisting splitting, and how plate geometry reshapes the load-transfer path. The study also flags its own boundaries: the models are calibrated only within the tested ranges of fiber content, bearing area ratio, and reinforcement ratio, and further validation is needed for full-scale members, different anchorage systems, larger specimen sizes, and alternative fiber types. Still, by demonstrating that anchorage-zone failure is a progressive compression-splitting-confinement process, the work gives engineers a mechanically honest foundation for pushing one of construction’s strongest materials closer to its true limits, potentially enabling lighter, more slender, and more durable prestressed structures at a time when infrastructure demands are growing faster than conventional materials can comfortably answer.

Subject of Research: Experimental and analytical evaluation of the local bearing capacity of ultra-high-performance concrete prestressed anchorage zones

Article Title: Experimental study and analytical evaluation of local bearing capacity of UHPC anchorage zones

Article References: Zhang, B., He, Z.-Q., Chen, J., & Li, W. (2026). Experimental study and analytical evaluation of local bearing capacity of UHPC anchorage zones. Case Studies in Construction Materials, 25, Article e06508. https://doi.org/10.1016/j.cscm.2026.e06508

Image Credits: AI Generated

DOI: 10.1016/j.cscm.2026.e06508

Keywords: ultra-high-performance concrete, UHPC, anchorage zones, local bearing capacity, steel fibers, prestressed concrete, splitting failure, confinement, plasticity wedge model, bridge engineering, bearing area ratio, indirect reinforcement