Bamboo phase quantification using thermogravimetric analysis: deconvolution and machine learning

The focus of this paper is to provide a fast and reliable quantification method of bamboo's main chemical components. Therefore, thermogravimetric analysis was used to determine holocellulose and lignin content in different bamboo specimens. The influence of nitrogen vs. air atmospheres was investigated on the thermal degradation behavior of Phyllostachys edulis (Moso), Bambusa vulgaris (BV) and Iranian Phyllostachys (IR) bamboos. Due to peaks overlapping, the deconvolution process was carried out to resolve hidden peaks and to allow adequate phase quantification. Also, a set of machine learning (ML) algorithms was applied to predict the composition of the studied bamboos within the 200-500 & DEG;C range in their TGA-DTG profiles. The ensembles of the ML models at R-2 > 0.99 proved a connection between the features in thermogravimetric curves with two concentrations of the main components, which were preliminarily established by means of chemical extraction from the respective samples. (AU)