MDM Project: Analysis of the Micromechanics Damage Model Suite




			MDM PROJECT: ANALYSIS OF THE MICROMECHANICS DAMAGE MODEL SUITE David O'Neal, NCSA/UIUC R. Luczak, UTK N. Pagano, AFRL G. Schoeppner, AFRL G. Tandon, UDRI H. Brown III, AFRL This file is available for download in *.pdf* format. Click here This report extends a previous study of core algorithms and portability considerations associated with the Micromechanics Damage Model computer software developed by the AFRL Materials Directorate at Wright-Patterson Air Force Base.1 The MDM analysis suite is comprised of four fracture mechanics codes used to establish design criteria for newly developed composite materials, particularly for brittle, heat-resistant composites associated with the fabrication of turbine blades. Introduction: The MDM codes are based on a semi-analytical approach in which single coated fibers and the surrounding composite material are represented as a nested set of concentric shells. Stresses and displacements at prescribed points along the length of the cylinder are determined by evaluation of a polynomial function. Numerically challenging conditions arise during the construction of this function. The calculation of matrix exponentials in terms of the full spectrum of eigenvalues associated with the primitive shell model (prior to the imposition of boundary conditions) can exceed the range limitations of common floating point systems. The method is relatively stable for coarse discretizations, but such analyses can result in unacceptable levels of error, so the ability to handle very thin shells is essential. This need to produce extremely precise estimates has heretofore restricted the use of the MDM codes to Hewlett-Packard platforms specifically because the H-P compilers support a 128-bit floating point system that features an unusually wide range of exponent values. The implication is that the portability of the MDM applications was sacrificed in order to achieve the highest possible level of accuracy, but the fundamental numerical problem is inherent in the algorithm itself. Performance also suffers because quad-precision operations rely on a software layer. Following a review of the project's origins and some of the more interesting aspects of the science that the MDM suite addresses, we will attempt to summarize the sources of numerical difficulty and performance problems in very basic terms and then illustrate how our alternative approach will dispatch with these problems. Objectives: An alternative approach was recently proposed by NCSA and UTK in the form of a collaborative project involving the developers of the MDM codes and the PET CSM and PT/ES groups affiliated with the Aeronautical Systems Center. The primary goal of this project is to create a portable version of the MDM solver kernel capable of extracting equivalent levels of accuracy from a standard 64-bit floating point system. Because the proposed numerical scheme is quite different from the current semi-analytical approach, we have concentrated our efforts on the advancement of our understanding of the current model and its primary data structures. The pre- and post-processing elements of the existing code are to be salvaged. Methods/procedures/apparatus: To date, a common HPUX-IRIX installation and a comprehensive set of validation cases that conform to, push, and ultimately exceed the limitations of the 128-bit floating point system supported by the SGI compilers have been developed and deployed. Identification of key data structures and program entry points has also been completed. Recent focus on boundary conditions as they are formulated and applied by the MDM codes represents the final step towards specification of an appropriate replacement for the solver kernel, the first of three deliverables associated with the academic (NCSA, UTK) side of the project. Results and Discussion: Results are not yet available. This project is currently in-progress. However, detailed problem analyses and the collaborative solution techniques we apply in resolving them will ultimately be used to promote the quality and depth of technical support provided by the NCSA and UTK teams. We expect that our approach will simultaneously overcome both the portability and performance problems, thus producing a significant effect on the way in which fracture analysis is performed by the AFRL Materials group. Sequences of analyses might then be turned towards optimization of new material designs. The portability of the new code will also permit the MDM codes to proliferate within the composite materials research community. Conclusions and Recommendations: Conclusions and recommendations are not yet available. This project is currently in-progress. References: D. O'Neal, N. Pagano, G. Schoeppner and G. Tandon, A Case Study in Porting Strategy: The Axisymmetric Damage Model, Proceedings of the DoD High Performance Computing Modernization Program Users Group Conference, Rice University, Houston, TX, 1998. Pagano, N. J., Axisymmetric Micromechanical Stress Fields in Composites, Proceedings of the 1991 IUTAM Symposium on Local Mechanics Concepts for Composite Material Systems, Springer Verlag, Berlin, pages 1-26 (1991) Pagano, N. J., and Tandon, G. P., 2D Damage Modes in Unidirectional Composites under Transverse Tension and/or Shear, v. 1, Mech. of Comp. Mater. and Struc., pages 119-155 (1994) Reissner, E., On a Variational Theorem in Elasticity, J. Math. Phys., v. 29, pages 90-95 (1950) Schoeppner, G. A., and N. J. Pagano, Stress Fields and Energy Release Rates in Cross-Ply Laminates, Int. J. Solids, Structures, v. 35, p. 1025-1055 (1998) Tandon, G. P., and Pagano, N. J., Matrix Crack Impinging on a Frictional Interface in Unidirectional Brittle Matrix Composites, Int. J. Solids, Structures, v. 33, pages 4309-4326 (1996) Related documents and images: http://www.ncsa.uiuc.edu/EP/CSM/publications/2000/MAPINT00_MDM_abs.pdf http://www.ncsa.uiuc.edu/EP/CSM/presentations/MAPINT00_MDM_PPT.pdf http://www.ncsa.uiuc.edu/EP/CSM/projects/mdm.html Acknowledgements: ASC PET, AFRL/MLBM (Materials Directorate), University of Dayton Research Institute, University of Tennessee, NCSA