From microstructure to multivariate prediction models: Decoding the biomechanical properties of tea stems via PLSR-Ridge regression and multifactorial orthogonal design

Abstract

Efficient harvesting and field management of tea is closely related to the mechanical properties of tea stems; however, there have no research on variables and models that can be used to predict them. In this paper, the relationship between factors (number of segments, diameter, stem length, density, moisture content, moment of inertia, and fracture deflection) affecting the mechanical properties (tensile and bending strength) of tea stems was analysed using a combination of partial least squares regression and ridge regression analyses, using three tea varieties, namely, Jinxuan, Yinghong, and Liannan, as the research subjects. Ultra depth of field electron microscopy was used to aid in the interpretation of the mechanical properties. The regression fitting results showed that the R² values of the prediction models for the tensile and flexural strengths of Jinxuan tea stems were 0.9361 and 0.9054, respectively, and the root mean square errors of prediction (RMSEP) were 0.7709 MPa and 1.9083 MPa, respectively. The R² values of the tensile and bending strengths of Liannan tea stems were 0.9161 and 0.9240, respectively, and the RMSEP were 0.3948 MPa and 1.2973 MPa, respectively. The R² values of Yinghong tea stems were 0.9292 and 0.9196, and the RMSEP values were 1.3207 MPa and 1.7489 MPa, respectively. The RPD of all the prediction models was greater than 3, indicating high prediction accuracy. The results of the comprehensive orthogonal test using four factors and three levels showed that the factors affecting the shear strength of tea stems were in the following order: stem segment, variety, moisture content, and shear speed. The results of the study will help to understand the biomechanical properties of tea stems, improve their resource utilisation and provide a reference for the optimal design of industrial cutting devices.

Plant phenotyping relevance

茶茎の tensile／bending strength という植物器官特性を、形態・物性変数からPLSR-Ridge回帰で予測するモデルを開発・評価しており、特性推定法が研究の中心である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.