Multi-Component Quantitative Analysis of Cashmere and Wool Fiber Mixtures Based on NIR Spectroscopy and IVY-DHKELM

Fiber content is one of the key indicators for evaluating fabric quality, and cashmere is much more expensive than wool due to its scarcity and excellent characteristics, leading to frequent adulteration. This study proposes a method that utilizes Near-Infrared (NIR) spectroscopy, and combining the Ivy (IVY) algorithm optimizes the Deep Hybrid Kernel Extreme Learning Machine (DHKELM) to enable the rapid and accurate prediction of multi-component fiber content in cashmere and wool blends. The study first prepared 21 different mixing ratios of cashmere and wool blend samples using the KBr tableting method and collected spectral data using an NIR spectrometer; the Iteratively Variable Subset Optimization (IVSO) algorithm was applied for band selection; Subsequently, the IVY-DHKELM quantitative model was constructed to independently predict the Cashmere Content (CC) and Wool Content (WC). Experimental results demonstrated that the IVY-DHKELM model prediction coefficients of determination (R2 p) of 0.9743 for CC and 0.9625 for WC on the test set, respectively. It has high prediction accuracy and reliability, and has important application value in quantitative analysis of fiber composition and detection of cashmere adulteration.