Session: 16-01-01: NSF-funded Research (Grad & Undergrad)

Paper Number: 77576

Start Time: Wednesday, 02:25 PM

77576 - Three-Dimensional Parameterized Model of Maize Stalk Morphology 

Prior research has shown that the maize stalk cross-section is accurately modeled as an ellipse. This approach allows the maize cross-section to be efficiently parameterized using just three parameters (major diameter, minor diameter, and rind thickness). In this study, we extend this 2D parameterization into the axial direction.

The finite element method was used to create models of maize stalk buckling failure. A modeling platform for accurately describing the anatomical features of maize stalks in a parameterized manner was created. Models were rigorously validated as a step toward global sensitivity analyses and optimization studies to identify an optimal stalk architecture. Three types of models were created: segmented specimen-specific models, specimen-specific models with elliptical cross-sections, and fully parameterized specimen-specific models with elliptical cross-sections. Segmented specimen-specific models were created by segmenting CT scan data to create an accurate, but non-parameterized representation of the original specimen. Parameterized specimen-specific models were based on statistical assessment of the maize stalk morphology. A combination of empirical and analytic relationships was used to closely approximate the original specimen geometry.

The parameterization scheme was based upon a data set of over 1000 CT scans that were collected previously. Physically distinct landmarks were identified in the axial pattern of major diameter, minor diameter, and rind thickness. Next, principal component analysis was used to identify spatial patterns in the transverse and longitudinal variation between adjacent landmarks. Principle components were chosen for this task because: (a) they are ranked according to patterns that explain the greatest amount of variation; and (b) they are mutually orthogonal. This approach allows fine-level control over individual physical features of the maize stalk. The resulting morphological models were defined by a set of 51 coefficients.  The morphological model was assessed by “fitting” the analytic functions to segmented CT scan data.

Validation was performed in several stages to provide information on the influence of various modeling assumptions. Because the primary interest in this study was the parameterization of maize stalk morphology, material properties were held constant across all models. First, segmented, specimen-specific models were compared with corresponding models in which each cross-section was simplified as an ellipse. This test quantifies the degree of discrepancy introduced by the ellipse assumption. Results indicated that the ellipse assumption introduce approximately 10-20% error. Next, elliptical cross-section models were compared with their corresponding parameterizized counterparts. This test quantifies the degree of error introduced by the parameterization process. Results indicated that parameterization introduced relatively little error (5-10%). Finally, segmented models were compared with parameterized models with elliptical cross-sections. Results indicated that, on average, both the ellipse and parameterization simplifications introduced less than 25% error. The parameterization approach enables a wide range of advanced analyses such as sensitivity and optimization studies. As such, the moderate amount of error seems to be a valuable exchange as these advanced studies can be used to provide valuable insights into maize stalk mechanics.

Presenting Author: Michael Ottesen Brigham Young University

Authors:

Michael Ottesen Brigham Young University
Nan-Wei Liu Brigham Young University
Douglas Cook Brigham Young University