Available at: https://digitalcommons.calpoly.edu/theses/1895
Date of Award
MS in Aerospace Engineering
A Comparison Study of Composite Laminated Plates with Holes under Tension
A study was conducted to quantify the accuracy of numerical approximations to deem sufficiency in validating structural composite design, thus minimizing, or even eliminating the need for experimental test. Error values for stress and strain were compared between Finite Element Analysis (FEA) and analytical (Classical Laminated Plate Theory), and FEA and experimental tensile test for two composite plate designs under tension: a cross-ply composite plate design of [(0/90)4]s, and a quasi-isotropic layup design of [02/+45/-45/902]s, each with a single, centered hole of 1/8” diameter, and 1/4" diameter (four sets total). The intent of adding variability to the ply sequences and hole configurations was to gauge the sensitivity and confidence of the FEA results and to study whether introducing enough variability would, indeed, produce greater discrepancies between numerical and experimental results, thus necessitating a physical test. A shell element numerical approximation method through ABAQUS was used for the FEA.
Mitsubishi Rayon Carbon Fiber and Composites (formerly Newport Composites) unidirectional pre-preg NCT301-2G150/108 was utilized for manufacturing—which was conducted and tested to conform to ASTM D3039/D3039M standards.
A global seed size of 0.020, or a node count on the order of magnitude of 30,000 nodes per substrate, was utilized for its sub-3% error with efficiency in run-time.
The average error rate for FEA strain from analytical strain at a point load of 1000lbf was 2%, while the FEA-to-experimental strains averaged an error of 4%; FEA-to-analytical and FEA-to-tensile test stress values at 1000lbf point load both averaged an error value of 6%. Suffice to say, many of these strain values were accurate up to ten-thousandths and hundred-thousandths of an in/in, and the larger stress/strain errors between FEA and test may have been attributed to the natural variables introduced from conducting a tensile test: strain gauge application methods, tolerance stacks from load cells and strain gauge readings.
Despite the variables, it was determined that numerical analysis could, indeed, replace experimental testing. It was observed through this thesis that a denser, more intricate mesh design could provide a greater level of accuracy for numerical solutions, which proves the notion that if lower error rates were necessitated, continued research with a more powerful processor should be able to provide the granularity and accuracy in output that would further minimize error rates between FEA and experimental. Additionally, design margins and factors of safety would generally cover the error rates expected from numerical analysis.
Future work may involve utilizing different types of pre-preg and further varied hole dimensions to better understand how the FEA correlates with analytical and tensile test results. Other load types, such as bending, may also provide insight into how these materials behave under loading, thus furthering the conversation of whether numerical approximations may one day replace testing all together.