How canitesc/igapack works
Overall System Breakdown
• Three major “application domains” or layers corresponding to the problem dimensions: 1D, 2D (elasticity2D), and 3D (elasticity3D). Each of these folders contains models (e.g., “Main1D.m”, “CantileverBeam.m”, “Hemisphere.m”) and associated utility subfolders (e.g., “utils1D”, “PHTutils” in the 2D and 3D folders).
• Recognize the central role of the NURBS toolbox. This module is essentially an external library (with its own license and source code) and should be shown as a supporting component that provides core geometry and basis computation routines.
Main Components
a. Problem/Model Layer:
– Illustrate the user-accessible MATLAB scripts that define specific benchmark problems (e.g., CantileverBeam, EdgeCrackMultiPatch, Blade, etc.).
– Indicate that each problem file uses underlying modules for mesh generation, analysis, and postprocessing.
b. Geometry, Mesh Generation, and Basis Functions:
– In each dimensional folder, highlight modules that generate geometry and define the mesh (i.e., functions like “initPHTmesh”, “initPHTmeshNoRefine” in 1D and “initPHTmesh” in 2D/3D).
– Display how basis functions are calculated (using Bernstein, Bezier extraction, deCasteljau, etc.), with a clear link to both PHT-spline generation routines and integration with NURBS routines.
c. Assembly and Solver Modules:
– Show the assembly components (for instance, “assembleGalerkin1D.m”, “assembleGalerkinSysGIFTMP.m”, etc.) where the system matrices are built.
– List solver components (such as “solution1d.m”, “stiffness1d.m” in the 1D module and similar routines in higher dimensions) that ultimately solve the discretized equations.
d. Error Estimation and Mesh Refinement:
– Identify the error estimation routines (e.g., “calcError1D.m”, “calcErrNorms1d.m”, “recoverDerivEstGalMPDorfler.m”, etc.) and adaptive refinement methods (like “refineMesh1D.m”, “refineMeshIso.m”, “refineMesh3D.m”) as central to the adaptive analysis process.
e. Visualization and Output:
– Include visualization components (plotting routines like “plotError1D.m”, “plotSolPHTElasticMP.m”, “plotStressDisp3D.m”) and the VTU file export for 3D solutions (e.g., “msh_to_vtu_3d.m”) that interface with external visualization tools such as ParaView.
Highlighted Architectural Patterns and Interactions
• Pattern Separation: Emphasize the separation between the core numerical routines (geometry, assembly, error estimation, and solver) and the problem-specific scripts. This separation facilitates reuse and easier testing.
• Performance Enhancements in 3D: In the 3D module, note the usage of mex files for the deCasteljau algorithm (e.g., “deCasteljau3DTP_mex.*”), which should be depicted as an optimized subcomponent that interacts with the MATLAB scripts.