Polylactic acid (PLA) is widely used in fused deposition modeling (FDM) due to its excellent mechanical properties and processability. However, its inherent brittleness significantly restricts its application in load-bearing and high-toughness scenarios. To address this limitation, this study establishes a multi-objective optimization framework that integrates single-factor experiments, Box–Behnken design (BBD), response surface methodology (RSM), and entropy weight-based objective weighting to simultaneously enhance the tensile strength, elongation at break, and flexural strength of FDM-printed PLA/thermoplastic polyurethane (TPU) blends. Through single-factor experiments, the optimal PLA/TPU blend ratio was determined as 80:20, achieving a tensile strength of 38.93 MPa, an elongation at break of 14.12%, and a flexural strength of 42.33 MPa—representing improvements of 39.4%, 15.7%, and 31.6%, respectively, over the 70:30 blend. Multi-scale characterization via FTIR, XRD, and SEM reveals that this enhanced performance arises from strong interfacial hydrogen bonding, a well-retained crystalline PLA framework, and the uniform dispersion of fine TPU domains. Subsequently, a four-factor, three-level BBD was employed to investigate the effects of printing speed, nozzle temperature, raster angle, and layer height on the mechanical properties. The entropy weight method assigned objective weights of 0.4924, 0.0901, and 0.4805 to the three properties, yielding a comprehensive score as the evaluation index. Unlike conventional approaches that rely solely on software-recommended optima from RSM models, this study critically compares two decision routes: secondary RSM modeling followed by continuous optimization versus direct ranking of BBD experimental data. Direct ranking identified the optimal parameter set (90 mm/s, 200 °C, 0° raster angle, 0.2 mm layer height) with a comprehensive score of 0.9614, surpassing the RSM-recommended route (70 mm/s, 200 °C, 0° raster angle, 0.2 mm layer height; score 0.7366) by 30.5%. This comparison demonstrates that, in strongly nonlinear FDM processes, software-based continuous optimization may converge to mathematically conservative suboptimal regions, whereas direct ranking of discrete experimental data preserves full physical responses and captures higher-order synergistic effects that the RSM model fails to account for. ANOVA results further confirm that raster angle is the most influential factor affecting the comprehensive score. Overall, this work bridges material formulation and printing parameter optimization for PLA/TPU blends, offering a validated strategy for high-performance FDM manufacturing.

Fused deposition modeling (FDM) involves numerous process parameters, making it difficult to efficiently identify key influencing factors and achieve accurate prediction of straightness errors in printed parts. To address these issues, this paper proposes a two-stage experimental optimization and prediction method combining Plackett–Burman (PB) design and central composite design (CCD). First, PB experiments were conducted to screen nine process parameters and identify two significant factors affecting straightness error: layer height and line width. Subsequently, a face-centered central composite design (CCF) was employed to establish a quadratic response surface regression model for analyzing the main effects and interaction effects of the selected factors. Based on 18 sets of CCF experimental data, response surface methodology (RSM) and decision tree–genetic algorithm (DT + GA) prediction models were constructed, and their generalization capabilities were evaluated using five independent reserved experiments. The results show that the quadratic regression model exhibits a coefficient of variation of 2.21% and a signal-to-noise ratio of 45.45, indicating good fitting accuracy. On the test set, RSM achieves a mean squared error (MSE) of 7.29 × 10⁻⁶ mm² and a mean absolute error (MAE) of 0.002142 mm, both superior to the DT + GA model (MSE = 3.41 × 10⁻⁵ mm², MAE = 0.004578 mm), representing reductions of 78.6% in MSE and 53.2% in MAE. This validates the effectiveness and superiority of the CCF-based response surface method for straightness error prediction under small-sample conditions. The proposed strategy provides a quantitative reference for the control of shape accuracy and process optimization in FDM-printed parts.

Fused deposition modeling (FDM) of polylactic acid (PLA) is plagued by dimensional inaccuracies—thermal shrinkage, warpage, and geometric distortion—that restrict its application in high-precision manufacturing. Blending thermoplastic polyurethane (TPU) with PLA enhances toughness, yet the coupled effects of blend ratio and printing parameters on dimensional accuracy remain unclear. This study establishes a multi-objective optimization framework integrating single-factor experiments, orthogonal design, ANOVA, and entropy weight–TOPSIS. Single-factor experiments combined with TOPSIS first identify the optimal PLA/TPU blend ratio, and orthogonal experiments are subsequently conducted on this optimal ratio to determine the best parameter combination. The 70:30 PLA/TPU blend delivers the optimal comprehensive performance (TOPSIS closeness: 0.680), attributed to favorable phase compatibility and robust interfacial adhesion as verified by XRD and SEM. Range analysis and ANOVA on the orthogonal results identify infill rate and layer height as the dominant factors governing dimensional accuracy (p < 0.05). Under the optimized parameter set (60 mm/s, 210 °C, 90% infill, 0.1 mm layer), the total dimensional deviation reaches only 0.256 mm, substantially lower than single-objective counterparts (0.321–0.418 mm). This work offers a validated strategy for precision control in FDM through synergistic optimization of blend formulation and processing parameters.