Header menu link for other important links
X
On topology optimization with embedded boundary resolution and smoothing
, A. Saxena
Published in Springer Verlag
2015
Volume: 52
   
Issue: 6
Pages: 1135 - 1159
Abstract
Honeycomb tessellation, by virtue of the edge connectivity it offers, helps eliminate various geometric singularities and thus shows promise in topology optimization. Numerous V-notches, which are sites of stress concentration and also pose difficulty in manufacturing, get retained on the continuum boundaries. Explicit boundary resolution and smoothing scheme is embedded within each gradient step when performing topology optimization with negative circular masks to assuage these serrations. This procedure, in particular, is deemed necessary when synthesizing contact aided compliant mechanisms. Continuum boundaries of each intermediate design are identified, and multiple smoothing steps are performed on the boundary nodes by shifting them systematically. Consequently, many regular hexagonal cells get morphed into concave ones. Shape functions based on Mean Value Coordinates (MVC) are implemented to cater to hexagonal finite elements of generic shapes. A continuation approach is used on the logistic material assignment model to tend to Heaviside function and thus yield close to black and white solutions in addition to avoiding early convergence and numerical instability. Number of boundary smoothing steps is increased with iterations as well. Deletion/addition of masks is performed to eliminate those not contributing in topology optimization and to introduce others in cases when singularities like local islands or layers prevail. To enhance continuity and to facilitate manufacturing, boundary nodes are interpolated using order m splines. For the examples presented, it is noted that splines of order 4 are adequate in representing the continuum boundaries accurately. © 2015, Springer-Verlag Berlin Heidelberg.
About the journal
JournalData powered by TypesetStructural and Multidisciplinary Optimization
PublisherData powered by TypesetSpringer Verlag
ISSN1615147X