CFD and thermofluid modeling with experimental optimization of shear-thickening abrasive flow finishing for additively manufactured cooling channels

Modeling and experimental optimization of shear-thickening media-based abrasive flow finishing (STMAFF) were performed to achieve sub-micron surface refinement (Ra < 0.05 μm) on internal cooling channels produced by selective laser melting of 316 L stainless steel. A biodegradable shear-thickening polishing fluid (cornstarch, SiC abrasives, xanthan gum, boric acid, vegetable oil, and water) exhibited pronounced non-Newtonian behavior, with viscosity increasing beyond 10 3 Pa·s at shear rates of 0.1–100 s -1 . The rheological response was accurately modeled using the Carreau–Yasuda equation, and thermal stability up to 150 °C was confirmed by TGA and FTIR analyses. CFD-DPM simulations in ANSYS Fluent, employing the Navier–Stokes equations, SST k–ω turbulence model, and a modified Preston material removal equation, predicted wall shear stress (>8 kPa) and abrasive particle trajectories with 0.90) for the percentage improvements in roughness parameters (%ΔRa and %ΔRz). A multilayer perceptron artificial neural network (ANN) surrogate model was subsequently trained on the experimental dataset and coupled with the Harris Hawks Optimization (HHO) algorithm to globally refine the process parameters beyond the RSM local optimum. The optimal conditions identified by the hybrid RSM–ANN–HHO framework were 35 wt. % abrasive concentration, 1.8 rps pump rotation speed, and 50 min finishing duration, yielding a final Ra < 0.05 μm and >99 % roughness reduction. The novel hybrid magnetic abrasive flow finishing (HMAFF) setup enabled both magnetic-assisted and conventional modes, ensuring uniform material removal across complex internal geometries. This integrated experimental–computational–optimization framework establishes STMAFF as a sustainable and highly effective post-processing route for ultra-smooth internal surfaces in additively manufactured cooling channels.