The appropriate curing methods for 3D-printed cement-based materials have not yet been clearly defined, thus restricting their durability in cold regions. This study was to investigate the effects of natural curing (NC), standard curing (SC), high-temperature curing (T₆₀), saturated Ca(OH)₂ solution curing (SCH), and internal curing withabsorbent polymer (SAP) on the performance of solid waste-based 3D printed mortar (3DFGCM). The objective was to establish quantitative relationships among curing methods, mechanical anisotropy, frost resistance, and pore structure, thereby providing a theoretical foundation for improving the service life of 3D printed structures in cold regions.

The fly ash-slag-cement based mortar (FA : GGBS : OPC=1 : 2 : 7) was prepared at a sand-to-binder ratio of 1 : 1 and a water-to-binder ratio of 0.3. After vibration casting and 3D printing of the slurry, the specimens were naturally cured for 1 day and then divided into five groups for further curing. The 28-d compressive strength and anisotropy were tested in the X, Y, and Z directions. After 28-d curing under respective conditions, the specimens were immersed in water at 15–20 ℃ for 2 d before undergoing freeze–thaw cycles. The mass loss rate, relative dynamic elastic modulus, and strength loss rate were measured. The pore structure was analyzed before and after freeze–thaw cycles to determine parameters such as air content, chord length distribution, and spacing factor. Finally, a damage evolution model based on the Weibull distribution was established to predict service life, and a neural network was used to map the relationship between pore parameters and damage.

This study systematically investigates the effect of curing method (i.e., NC, SC, T₆₀, SCH, and SAP) on the performance of 3D printed fly ash-slag-cement-based mortar (3DFGCM). The comprehensive analysis of mechanical anisotropy, frost resistance, and pore structure, combined with freeze–thaw models and service life prediction models indicates that the 28-d compressive strength and isotropy coefficient of 3D printed specimens are lower than those of cast specimens. Among the curing methods, 3D-SC demonstrates the most significant enhancement in the 28-d compressive strength, while the SAP increases the 28-d compressive strength to 36.8 MPa, which is increased by 41.17%, compared to the NC group, with the lowest anisotropy coefficient (i.e., 0.187). The frost resistance of 3D printed specimens varies significantly. The 3D-NC and 3D-T₆₀ groups both fail after 125 freeze–thaw cycles, the 3D-SAP group exhibits the optimum frost resistance, retaining 61.5% of its dynamic elastic modulus and showing a mass loss of 4.57% after 150 freeze–thaw cycles. Its closed micropores and humidity balance mechanism effectively mitigate frost heave damage. In terms of the pore structure, 3D printed specimens exhibit distinct elliptical pores. The 3D-SCH group shows a limited improvement in cavity and interlayer cold joint defects, with large pores accounting for 75% more than those in the 3D-SAP group. The incorporation of SAP improves process-induced defects and significantly inhibits pore coarsening. With the relative dynamic elastic modulus as an index, the Weibull damage and service life prediction models are applied to assess freeze–thaw damage and service life of 3DFGCM, with R² of>0.98. The 3D-SAP group performs optimally, with a predicted service life of 15.9 years. Furthermore, based on BP neural networks, the effective correlations are established between pore structure parameters and the Weibull-predicted damage and service life, with R² of>0.97.

In terms of mechanical anisotropy, the strength of 3D-printed specimens followed a decreasing order of Cast>X>Z>Y. The SAP curing resulted in the lowest anisotropy coefficient (i.e., 0.187), with a 41.2% increase in the strength, compared to the NC group. For the frost resistance, after 150 freeze–thaw cycles, the SAP group retained 61.7% of its dynamic elastic modulus and exhibited only 4.57% mass loss. In contrast, the NC and T₆₀ groups failed after 125 cycles due to pore coarsening. In terms of the pore structure, the initial air content of 3D-printed specimens was 1.5 times higher than that of cast specimens. After freeze–thaw cycles, the total chord length of the NC group was increased by 65.3%. The SAP effectively inhibited pore coarsening, with large pores accounting for only 36%, representing a 75% reduction, compared the NC group. Finally, a damage evolution equation was established based on the Weibull distribution model, predicting a service life of 15.9 years for 3D-SAP in cold northwestern regions. The BP neural networks were employed to predict damage and lifespan based on microscopic pore structure. This study could provide a reference for the design of methods and frost resistance optimization in 3D-printed construction materials.