Preliminary Cross-Platform Comparison of Nep Counts in Five Industrial Batches of Mechanically Recycled Cotton: Ranking, Technical Variability, and Measurement Disagreement

Mechanically recycled cotton is heterogeneous, and nep counts obtained from different analytical platforms may not be numerically interchangeable. This study provides a preliminary cross-platform comparison of five industrial batches selected in the original industrial work to span the available production range of total nep count (TotNep). The batches differed in product category, color, supplier/feedstock designation, and production line; Batch was therefore treated as a composite material identity rather than an isolated structural-severity factor. USTER® Neptester 720 measurements comprised three bales per batch and two technical determinations per bale; USTER® LVI 920 used the same three bales and three determinations per bale; USTER® AFIS Pro 2 used one of those bales per batch and ten determinations. Descriptive variability was reported as mean, standard deviation, and coefficient of variation. The primary inferential analysis compared Neptester 720 and LVI 920 on the 15 matched bales. Although the overall mean paired difference was +45.5 neps/g (95% CI: 14.0–76.9; p = 0.0078), batch-specific mean differences ranged from −13.6 to +101.4 neps/g and differed significantly among batches (p = 0.025), showing that +45.5 neps/g is not a transferable correction factor. A secondary exploratory analysis was restricted to the five bales measured by all platforms (one bale per batch). These deliberately range-spanning bales retained the Neptester-defined order (Spearman ρ = 1.00), but the coefficient is descriptive and does not constitute independent validation of cross-platform ranking. Exploratory Bland−Altman mean differences were +32.0, +99.3, and +67.2 neps/g for LVI 920-AFIS Pro 2, Neptester 720-AFIS Pro 2, and Neptester 720-LVI 920, respectively, with wide limits of agreement. Because no reference method was available and each platform was confounded with laboratory, operator, measurement date, and procedural conditions, these values describe cross-platform disagreement rather than instrument-specific measurement bias. The results support ordinal recognition of the selected batches but do not establish validated severity classes, numerical interchangeability, or formal interlaboratory reproducibility. Larger balanced studies using matched bales, documented conditioning, harmonized procedures, and fuller fiber characterization are required.