A predictive model based on key particle properties for segregation of free-flowing powder blends measured using Near-Infrared spectroscopy

Segregation behavior of binary blends of free-flowing powders was systematically investigated to develop a quantitative model for predicting segregation tendencies as a function of key particle properties. Binary blends, eight training and two testing, were prepared from seven materials, spanning a range of ratios of particle size, bulk density, and aspect ratios. A novel tape test sample preparation technique, developed to enhance the accuracy of blend uniformity measurements, reduce material usage, and minimize sampling errors, helps validate the use of a Near-Infrared (NIR) probe-based SPECTester to assess segregation intensity (SI) of those binary blends. Cumulative segregation area (CSA), defined as the integrated absolute deviation area between the cumulative concentration profile and the no-segregation baseline, was introduced as a novel measure of segregation that better captured the segregation tendency across a range of variables. A multi-variate power-law model, capturing the combined influence of these properties on segregation behavior, was developed. For the segregation intensity (SI)–based model, predictions closely matched experimental measurements for the training dataset (R² ≈ 0.98) and maintained good accuracy for unseen test blends (R² ≈ 0.85). When evaluated using CSA, the model exhibited strong generalization, achieving R² = 0.92 for the training dataset and R² = 0.96 for the test blends. Heatmap visualizations overlaid with experimental data confirmed that higher D₅₀ × bulk density ratios increased segregation intensity, whereas larger ratios of aspect ratio mitigated segregation. Overall, a robust quantitative framework was established for understanding and predicting segregation in free-flowing powder systems, potentially enabling improved blend uniformity.