Fast determination of total solids content (TSC) and dry rubber content (DRC) of para rubber latex using near-infrared spectroscopy

A rapid measurement method of total soluble solids content (TSC) and dry rubber content (DRC) of para rubber latex (PRL) is required for trading and production control purposes. The objective of this study was to develop the TSC and DRC prediction model of a PRL sample based on the near-infrared (NIR) spectroscopy method. The samples were scanned using the transflectance method, wherein the effect of the pathlength to model accuracy and precision were investigated. NIR spectra were collected, incorporating a wavelength region of 860–1760 nm. The models were developed using partial least squares (PLS) regression with either raw spectra or spectra pre-treated by a baseline offset, a standard normal variate (SNV), multiplicative scatter correction (MSC), and a fractional order derivative (FOD). The results of the TSC model developed using NIR spectra from a 8 mm pathlength had the highest performance, with R2p, RMSEp, Bias, and RPD of 0.97, 1.15%, 0.43% and 6.17, respectively. Meanwhile, the NIR spectra from a 4 mm pathlength developed with a FOD at 0.25 gave the best performance in DRC prediction, whose R2p, RMSEp, Bias, and RPD were 0.96, 1.20%, − 0.52% and 5.40, respectively. Therefore, the author concluded that the developed models could be used to evaluate the TSC and DRC of a PRL sample as a real-time method in factory. •The TSC and DRC NIR models for fresh para rubber latex were optimised.•Best prediction models of TSC and DRC of PRL sample provided R2p were of 0.97 and 0.96.•The TSC and DRC models were evaluated for bias which acceptable for using in production control.•Fractional order derivatives methods improved the model’s accuracy.•Models ability for future sample was validated by unknown spectra, providing RMSEP of 2.01% for TSC and 1.44% for DRC.