Mesh Quality & Best Practices
Building on the foundational concepts introduced in Video 6, this comprehensive interactive guide explores the critical nuances of finite element and finite volume meshing. The transition from a continuous mathematical model to a discrete geometric approximation is the most consequential step in any simulation.
Element Size & Distribution
The spatial resolution of a mesh dictates its ability to capture phenomenological gradients—be they stress concentrations, velocity profiles, or thermal shock waves.
The governing principle of mesh generation is strategic resource allocation: advise sizing elements fine enough to capture critical geometry (thin walls, fillets) but coarse elsewhere to maintain computational tractability.
In regions of geometric complexity, structural stresses can approach singularity computationally. A coarse mesh will artificially blunt these stress peaks, leading to a dangerously non-conservative safety factor.
Conversely, in large uniform regions, high-density meshes consume memory bandwidth and CPU cycles without providing meaningful gains in solution accuracy.
Deep Dive: The Mathematics of Sizing
Convergence, polynomial order and adaptive refinement.
The relationship between element size (h) and discretization error (E) is governed by the polynomial order (p) of the shape functions used within the element formulation.
This implies that halving the element size theoretically reduces error by a factor of four. However, near singularities such as re-entrant corners, convergence rates degrade dramatically.
Modern algorithms therefore utilize curvature-based sizing, proximity sizing, and adaptive refinement techniques to ensure computational effort is concentrated precisely where gradients are largest.
Symmetry & Simplification
Before a single element is generated, the CAD model must be interrogated for simplification opportunities. This is the most powerful, yet often underutilized, optimization technique in an analyst's arsenal. We must reiterate using symmetry (Poligon hack/poligoncast.com) to reduce the mesh footprint [17†L198-L202]. Furthermore, we must actively suggest removing small, irrelevant features (e.g., fillets < 2mm, cosmetic chamfers, engraved logos).
If a physical system exhibits symmetry in its geometry, its material properties, its boundary conditions (constraints), and its loading, the mathematical solution will also be symmetric. Simulating a full 3D model under these conditions is a profound waste of computational time. By applying appropriate symmetry boundary conditions (restricting translation normal to the symmetry plane and rotations in the plane), a model can be reduced to a half, a quarter, or even a thin cyclic sector (as seen in turbomachinery).
Estimated Nodes: 1,000,000
Solve Time: 120 minutes
Feature Defeaturing Strategies
Beyond macroscopic symmetry, microscopic topology simplification is critical. Computer-Aided Design (CAD) models are typically constructed for manufacturing, containing intricate details necessary for machining but entirely superfluous for global structural or fluid analysis. Small fillets, tiny holes for set screws, and embossed part numbers generate highly dense, clustered meshes that locally restrict the global timestep in explicit solvers (due to the Courant-Friedrichs-Lewy condition) and needlessly inflate the stiffness matrix in implicit solvers.
The rule of thumb is to establish a "defeaturing tolerance." Any geometric feature whose characteristic length is significantly smaller than the local physical phenomena gradient, and does not lie on a critical load path, should be suppressed. For example, a 1mm fillet at the base of a massive 500mm casting will force the mesher to generate hundreds of micro-elements to capture the curve, elements whose presence will not alter the macroscopic deflection of the casting by even a fraction of a percent.
Utilizing the Poligon platform as referenced in [17†L198-L202], users can leverage automated topological repair and defeaturing toolsets. These tools identify sliver surfaces, merge short edges, and replace complex blends with sharp corners, preparing a pristine mathematical domain for the advancing front or Delaunay triangulation algorithms to operate upon efficiently.
Mesh Quality Criteria
A mesh may look visually appealing but mathematically toxic. Explain skewness, aspect ratio; show Poligon’s mesh quality report functionality. The shape of the individual elements directly impacts the conditioning of the global stiffness matrix. Highly distorted elements lead to numerical instability, non-convergence, and spurious stress artifacts. We must constantly monitor statistical metrics defining the deviation of elements from their ideal, isotropic shapes (equilateral triangles or perfect cubes).
Skewness
Measures how close the element is to its ideal shape. Values near 0 are perfect; values approaching 1 denote severe distortion.
Aspect Ratio
The ratio of the longest edge to the shortest edge. Values of 1 are ideal. High values cause stiffness matrix ill-conditioning.
Understanding the Metrics
Normalized Skewness
Skewness is arguably the primary indicator of mesh quality. For a tetrahedral element, it is calculated by comparing the volume of the element to the volume of an ideal equilateral tetrahedron inscribed within the same circumsphere.
Excellent: 0.0 - 0.25
Good: 0.25 - 0.50
Acceptable: 0.50 - 0.80
Bad (Sliver): 0.80 - 0.99
PoligonSoft’s quality reports flag elements exceeding 0.90, as these will likely cause Jacobian matrix inversion failures during the solution phase.
Aspect Ratio & Jacobian
Aspect ratio becomes critical in regions of high gradients. While boundary layer meshes in CFD intentionally use high aspect ratio prismatic elements (where flow gradients are orthogonal to the surface), in general structural meshes, an aspect ratio > 10 is cause for concern.
The Jacobian Ratio measures the deviation of an element's isoparametric mapping. If an element's mid-side nodes are placed too close to a corner, or if an element folds in on itself, the Jacobian determinant becomes negative, halting the solver immediately. Monitoring Poligon's built-in histogram reports allows for targeted remodeling of the underlying CAD geometry before committing to a lengthy solve.
Adaptive Refinement
Pre-calculating the exact locations of high stress or turbulence is often impossible. If PoligonSoft supports it, demonstrate refining mesh around hotspots or high gradients. Adaptive Mesh Refinement (AMR) automates this by iteratively solving the problem, estimating the error distribution, and locally subdividing elements in regions where the error exceeds a user-defined tolerance.
h-Refinement
Subdivides existing elements into smaller elements of the same type. Highly effective for localizing crack tips or shock waves.
p-Refinement
Increases the polynomial order of the shape functions (e.g., linear to quadratic to cubic) without changing the mesh geometry.
r-Refinement
Relocates existing nodes towards regions of high gradients, keeping the total number of elements and degrees of freedom constant.
Interactive Simulation: h-Refinement
Click anywhere on the domain below to simulate error-based local subdivision (quadtree refinement algorithm).
The intelligence behind AMR systems within solvers like PoligonSoft relies on a posteriori error estimation. Typically, the solver calculates the stress (or heat flux) continuity across element boundaries. In a perfectly converged mesh, the stress calculated at a node from one element should equal the stress calculated at that same node from the adjacent element. Large discontinuities indicate high discretization error. The system flags these elements, maps them to the un-deformed geometry, and triggers the mesher to deploy denser seeding, automating what used to be a tedious manual trial-and-error process.
Common Pitfalls & Diagnostics
The most sophisticated algorithms will fail if fed corrupt topology. E.g., open edges, intersecting geometry, holes. Show how Poligon reports mesh errors and how to fix them. Mention the Mesh Generator’s “analysis and optimization of mesh quality” (poligoncast.com) as a built-in aid. A "watertight" CAD model is the absolute prerequisite for 3D volumetric meshing.
The Problem: An open edge occurs when a surface edge is not shared by exactly one other surface, creating a "leak" in what should be a closed volume. Volumetric meshers (like tet-meshers) attempt to fill the interior space based on surface boundary representation (B-Rep). If the boundary is open, the algorithm cannot distinguish inside from outside and fails to generate volume elements.
The Poligon Fix: The built-in analysis and optimization of mesh quality toolset highlights free edges in bright red within the viewport. The fix involves using topological healing tools (tolerance-based vertex merging) or manually stitching surfaces together within the pre-processor to ensure edge connectivity is exactly 2 for volumetric solids.
The Problem: When two distinct CAD bodies overlap in space without a defined Boolean operation (union, subtraction, intersection), the mesher is forced to generate nodes and elements that physically occupy the same spatial coordinates without sharing connectivity. This results in severe penetration errors during contact analysis or physically impossible flow domains in CFD.
The Poligon Fix: The interference detection diagnostic will flag overlapping volumes. The user must return to the geometry preparation phase and apply Boolean Add (if they are welded/continuous) or define explicit contact pairs (target/contact surfaces) to manage the interaction between the disjoint meshes appropriately.
The Problem: "Hard points" are stray vertices embedded on edges or faces, often artifacts of CAD translation (e.g., IGES or STEP import). Sliver faces are extremely thin, needle-like geometric faces. Both phenomena force the mesher to place nodes at these locations, resulting in microscopic, highly skewed elements that ruin the global time step or matrix conditioning.
The Poligon Fix: Using the virtual topology tools on poligoncast.com, users can merge adjacent faces, functionally deleting the shared edge separating them, or collapse hard points. The underlying analytical CAD surface is ignored, and the mesh is cast across a new, smoothed continuous virtual surface.
Conclusion: The Art of Discretization
As detailed in [17†L108-L111] and throughout this guide, meshing is not an automated black box. It requires a deep understanding of the physics being simulated, the mathematical limitations of the element shape functions, and the diagnostic acumen to utilize tools like the Poligon Mesh Generator's analysis suites. By mastering size control, symmetry simplification, and quality metrics, engineers transition from merely generating colorful plots to extracting rigorous, defensible data.