This study clarifies how particle-size gradation regulates part performance in binder jetting of sand molds. The large-particle fraction \(A\) , printhead grayscale setting \(B\) and layer thickness \(C\) were selected as process factors. Single-factor tests, a 3-factor × 4-level response-surface design and multi-angle build verification were carried out. Multi-response contour maps based on second-order polynomial regression, together with a Boltzmann model for strength retention and analysis of build angle versus surface roughness, were used to identify the process window. Under combined constraints of dimensional error, tensile strength and permeability, a recommended parameter range of \(A = 0{-}10\, {\text{wt}}.\%\) , \(B = 4{-}5\) and \(C = 0.50\, \text{mm}\) is obtained, with \(A = 10\, \text{wt}.\text{\%}\) , \(B = 4\) and \(C = 0.50\,\text{ mm}\) as a preferred setting. Compared with the baseline ( \(A = 0\, \text{wt}.\text{\%}, B = 5, C = 0.50\, \text{mm}\) ), this setting reduces dimensional error by 30.3 %, while tensile strength decreases by only 3.5 % and permeability increases by 5.1 %. Multi-angle tests further show that, at \(B = 5\) and \(C = 0.50\, \text{mm}\) , increasing \(A\) from \(0\; \text{to}\; 10\, \text{wt}.\text{\%}\) raises \({\theta }_{50}\) from 53.1° to 65.6° and reduces the peak-to-valley difference in roughness across build angles by 49.9 %. These results define a process-parameter window and selection strategy for high-precision, high-efficiency binder jetting under the conditions of this study.
