Bio-inspired structures are well-known as porous structures with remarkable strength and toughness; hence they are proven to have various applications in many fields, namely medical equipment, tissue engineering, lightweight material, etc. Recent studies on 3D printing plastic triply periodic minimal surfaces (TPMS) which were utilized as a reinforcement approach for cement beams have shown such significant enhancements. It is indicated that the structures created multiple local confinement cement volumes without applied axial force; consequently, higher maximum load, lower deflection, improved ductility, and corrosion-resistant ability could be noticed. However, the aforementioned results should be conducted from simulations or experiments, which might be an excessive computational and time-consuming process. Therefore, in this study, a surrogate model based on artificial neural networks (ANN) will be established to predict the mechanical behaviors of the plastic Primitive TPMS reinforced beams. Finite element analysis (FEA) simulation results of different numbers of reinforcement layers and volume fractions were adopted as the model data, the robust model have been owing to a hyperparameter tuning investigation.