Boundary Element Analysis at Caterpillar Ling Pan Caterpillar
Boundary Element Analysis at Caterpillar Ling Pan Caterpillar Inc. 12/25/2021 NIST Green's Function Meeting 2002
EZBEA z. Easy Boundary Element Analysis z Originated from ‘ 70’s code at Kentucky State y. Owned & enhanced significantly by Caterpillar z Model external boundary only y. Meshing is much easier, especially with details & small holes y. Accurate with coarse triangles, quads 12/25/2021 NIST Green's Function Meeting 2002 2
Strength z. Excels on Geometry with Complicated - Still difficult to tet-mesh Details Looking Inside - “Easy” to surface-mesh Intersecting Holes 12/25/2021 Brake Stand for C 9 Engine NIST Green's Function Meeting 2002 3
Applications z. Engine components y. Fuel injector y. Crankshaft z. Powertrain components y. Gear y. Shaft 12/25/2021 NIST Green's Function Meeting 2002 4
Challenges z. Longer solve time compared to FEA analysis z. Nonlinear (contact) analysis z. Not suitable with thin body (fabricated) structures 12/25/2021 NIST Green's Function Meeting 2002 5
EZBEA+ z EZBEA+ - provides integration with Pro/Mesh EZBEA Menus 12/25/2021 NIST Green's Function Meeting 2002 6
EZBEA Capabilities z Material z Analysis types y steady-state heat conduction y static thermoelasticity y beam section properties z Domain types y planar y axisymmetric y Isotropic, linear, 1 per subregion z Symmetry: y planes of geometric symmetry y planes of loading symmetry/anti-symmetry y general 3 D 12/25/2021 NIST Green's Function Meeting 2002 7
EZBEA Capabilities z Loading and Boundary Conditions y Only linear, static loads and boundary conditions xpoint load constraints xedge pressure constraints xsurface displacement constraints y Loads and boundary conditions in nature have area associated with them. x. A true point load is impossible x. A knife edge constraint is impossible x. These loads would result in infinite stress 12/25/2021 NIST Green's Function Meeting 2002 8
An Example: 9881 nodes, 20 hours on HP C 240 12/25/2021 NIST Green's Function Meeting 2002 9
Current Models size vs. Time Plot on HP C 240 12/25/2021 NIST Green's Function Meeting 2002 10
Speedup 12/25/2021 NIST Green's Function Meeting 2002 11
Future Work z. Eliminate the physical limit on number of elements and nodes ydynamic memory allocation yscratch files yaccuracy issue when elements get smaller(double precision) 12/25/2021 NIST Green's Function Meeting 2002 12
Future Work z. Solver improvement (performance issue) yimprove current solver (direct solver, O(N 3)) yiterative solver (O(N 2)) x. Caterpillar has looked into this before x. Does not always converge for large problems 12/25/2021 NIST Green's Function Meeting 2002 13
Future Work z. Multipole accelerated BEM (O(Nlog. N) operations) ysignificant performance gain yhas to use iterative solver yconvergence for large problem questionable ywould involve significant work 12/25/2021 NIST Green's Function Meeting 2002 14
Future Work z. Continue gap, contact work ycapability to analyze assemblies zimplement line integral for thinner structures yimprove accuracy 12/25/2021 NIST Green's Function Meeting 2002 15
- Slides: 15