Research on rubber classification and recognition based on terahertz time-domain spectroscopy and improved honey badger algorithm

Abstract

The identification of different rubber materials is crucial to ensuring the quality of rubber products. In order to quickly and effectively identify the types of rubber, reduce the impact of counterfeit rubber on the market. This study proposes a rubber identification method based on terahertz time-domain spectroscopy (THz-TDS), Chemometry, and Improved Honey Badger Algorithm (IHBA). Initially, the absorption spectra of eight types of rubber within the 0.2–1.6 THz range are obtained and calculated using THz-TDS. This is followed by data preprocessing using Savitzky-Golay and Principal component analysis(PCA). The optimization effects of genetic algorithm (GA), grid optimization algorithm (GRID), particle swarm optimization algorithm (PSO) and honey badger algorithm (HBA) on support vector machine (SVM) model parameters were compared respectively. The HBA-SVM model achieves 96.88 % recognition accuracy on the prediction set, which is higher than other models and shows excellent parameter optimization ability.To further improve accuracy, Bernoulli chaotic mapping, cosine density factor, and Cauchy mutation are introduced for improvement. Compared with the original model, the IHBA-SVM model improves the accuracy of rubber recognition from 96.88 % to 98.96 %. Furthermore, compared with other models, the IHBA-SVM model achieved the highest classification accuracy. In summary, this study provides technical support and reference for the rapid identification of rubber, which is of great significance for ensuring the quality of rubber products.