A data-driven sensitivity analysis of maize stalk models reveals the dominant factors influencing stalk strength

Abstract

Morphological factors significantly impact maize stalk strength, but no study has fully characterized the impact of maize stalk shape on stalk strength. This study uses a data-driven and machine-learning modeling approach to characterize these relationships through a comprehensive sensitivity analysis of model inputs. Using 3D parameterized maize stalk models gave a higher level of control than previous studies by adding more parameters, but with the increase of the dimensionality. The large dimensionality of the models was greatly reduced via principal component analysis (PCA). Analysis revealed that model flexural stiffness, failure strength, and biomass were primarily determined by the first principal component. Material sensitivity analysis was also conducted on the models, and its results were consistent with past studies. The results of this study improve researchers’ understanding of the parameters that influence 3D parameterized maize stalk models. Statistical analysis indicates a strong relationship between the first principal component and section modulus, which further validates the 3D parameterized maize stalk model. Results also show that maize stalk morphology is primarily controlled by only one factor: the first principal component. This may limit researchers’ ability to increase stalk strength without increasing mass.