Constitutive modeling of split tensile behavior of 3D printing cement composite containing randomly oriented PVA short fibers and optimization using Taguchi method and RSM

• 3D printable cement composite containing PVA fibers • Constitutive modeling of split tensile strength for randomly oriented fibers • Mix design analysis and optimization using Taguchi method • Modeling and optimization using response surface methodology (RSM) While 3D concrete printing offers significant construction efficiencies, its structural application is rather limited by inherent brittleness, weaker interfacial transition zones (ITZ) and complex mechanical behavior due to multiple influencing varibles. Because these internal vulnerabilities dictate the stability of printed elements, accurately evaluating pure tensile capacity especially when short fibers are incorporated is critical. The split tensile test is uniquely required for this, as it induces uniform indirect tensile stress to rigorously isolate matrix cohesion and fiber-bridging efficiency, avoiding the masking effects of flexural bending gradients. Furthermore, traditional constitutive models fail to capture the unique probabilistic fiber orientations of 3DCP. To address this gap, the objective of this study is to predict and optimize the split tensile behavior of 3D printed cement composites incorporating polyvinyl alcohol (PVA) short fibers. The primary novelty of this work lies in the integration of experimental design and theoretical micromechanics with robust statistical optimization. First, a novel constitutive model adapting the modified shear lag theory is proposed to explicitly account for the random, probabilistic fiber alignments in the mix. For experimental validation, mixes were printed, and cylindrical cores were extracted for split tensile and ultrasonic pulse velocity tests. A sensitivity analysis validated the constitutive model, demonstrating its ability to predict the peak load and split tensile strength with a quantitative accuracy of R 2 =0.81. Second, the complex non-linear interactions of design parameters (PVA fiber volume%, water-to-binder ratio, and superplasticizer dosage) were optimized using the Taguchi method and Response Surface Methodology (RSM). Quantitative outcomes revealed that RSM delivered superior prediction performance for split tensile strength compared to the Taguchi method, achieving a correlation of R 2 =0.95 versus R 2 =0.93, respectively. Using the desirability function approach, the RSM numerical optimization identified the optimal mixture proportions as 0.71% for PVA fiber volume, 0.35 for water-to-binder ratio, and 19 for superplasticizer-to-binder ratio, which achieved a maximum predicted split tensile strength of 6.21 MPa. Ultimately, this study can provide an integrated mechanistic-statistical approach with micromechanics-based predictive framework for designing structurally robust 3DCP composites.