ANSYS Workbench Episode 13: Mesh Refinement Strategies and Convergence Studies for Accurate FEA (Updated May 2026)

India's manufacturing sector at AURIC attracted ₹71,343 crore in investment and 62,405 jobs — and every structural component in those factories must pass a simulation sign-off before manufacture. The sign-off is only as good as the mesh. A coarse mesh gives fast results that are wrong. A fine mesh gives accurate results but takes forever. Convergence studies — the process of systematically refining the mesh until results stop changing — are how professional simulation engineers find the mesh sweet spot. Episode 13 goes deep on this: mesh refinement controls, convergence criteria, and the quality metrics that tell you whether to trust your FEA result before you show it to your manager.

▶ Watch on YouTube

Why Mesh Quality Is the Most Important Variable in FEA Accuracy

Here is a simulation truth that every beginner learns the hard way: you can have perfect boundary conditions, accurate material properties, and correct load directions — and still get a wrong answer because your mesh is too coarse. FEA solves equations at element nodes; the accuracy of the solution between nodes depends entirely on how closely spaced those nodes are. A coarse mesh with large elements misses stress peaks at geometric features like holes, fillets, and notches — the exact locations where parts fail in real life. The rule: never trust a first-run mesh. Always refine it, re-solve, and compare. When two consecutive meshes give the same peak stress within 5%, you have a converged result you can trust. This discipline is what separates a simulation analyst from a simulation button-pusher. Companies like Tata Tech, KPIT, and L&T have internal FEA guidelines that specify convergence criteria — if you can show you follow these guidelines in an interview, you stand out from 80% of other applicants.

ANSYS Workbench Mesh Controls: Sizing, Body Sizing and Face Sizing

ANSYS Workbench provides three levels of mesh control: Global Controls (Mesh branch in the tree — default sizing for the entire model), Body Sizing (right-click Mesh > Insert > Sizing, scope to a body — controls element size for one solid), and Face Sizing (scope to specific faces — concentrates elements at critical features). The Relevance Centre slider in Global Mesh Controls is the quickest way to globally coarsen or refine: slide from Coarse to Fine and click Generate Mesh. For targeted refinement: right-click Mesh > Insert > Sizing, select the fillet face or hole bore, set Element Size to 1 mm (versus 5 mm global size), and regenerate. The Elements and Nodes counts in the Mesh Statistics panel tell you the computational cost. A rule of thumb: doubling the element count doubles the solve time approximately. So always start coarse globally and refine only the high-stress regions identified in the first solve.

Inflation Layers: Critical for Stress Concentrations and Curved Surfaces

Inflation layers are layered, thin elements added to surfaces where accurate stress gradients are critical — fillets, notches, sharp corners, and contact faces. Without inflation, a stress concentration at a 2 mm fillet radius can be underestimated by 30–50% with a coarse mesh. To add inflation in ANSYS: right-click Mesh > Insert > Inflation. Scope it to the body, select the fillet faces in the Boundary field, set Maximum Layers to 5, and Growth Rate to 1.2. ANSYS will automatically create five thin layers at the selected faces, each 1.2x thicker than the previous. These layers capture the steep stress gradient without requiring an extremely fine global mesh. After adding inflation, regenerate the mesh and check that the inflated layers are smooth and do not create negative Jacobian elements (a red warning in Mesh Metrics — fix by reducing Growth Rate to 1.1). Inflation layers are mandatory for any fatigue life prediction because fatigue failure almost always initiates at stress concentration surfaces.

Running a Convergence Study: Proving Your Results Are Mesh-Independent

A convergence study runs the same analysis three or more times with progressively finer meshes and plots the critical result (typically peak von Mises stress) against element count. The point where the curve flattens is the converged mesh. Practical procedure: Run 1 — coarse mesh (Relevance Centre: Coarse). Note peak stress. Run 2 — medium mesh (Relevance Centre: Medium). Note peak stress. Run 3 — fine mesh (Relevance Centre: Fine) plus Face Sizing at critical feature. Note peak stress. If Run 2 and Run 3 results are within 5%, Run 2 mesh is sufficient — no need to use the expensive fine mesh for production analysis. If the gap is greater than 10%, add a Run 4 with further refinement. ANSYS Workbench supports Parametric Studies (via the Parameter Set tool) to automate this process — set Element Size as a parameter, define a range, and let ANSYS run all mesh variants automatically overnight. This automation is used routinely at KPIT and Mahindra for automated design validation workflows.

Mesh Quality Metrics in ANSYS: What to Check Before You Solve

ANSYS Workbench provides built-in Mesh Metrics to assess quality before you solve. Access via Mesh branch > Details > Mesh Metric dropdown. The five most important metrics: Orthogonal Quality (target: above 0.1, ideal: above 0.7 — measures how close elements are to perfect perpendicularity); Skewness (target: below 0.9, ideal: below 0.5 — measures element distortion); Jacobian Ratio (target: positive for all elements — negative Jacobian means an inverted element that will cause solver errors); Aspect Ratio (target: below 20 for structural analysis — measures element elongation); and Element Quality (0 to 1, higher is better — composite metric). When the Mesh Metric histogram shows outlier elements in the poor range, click on those elements in the histogram to highlight them in the geometry view and identify which geometric feature created them. Most poor-quality elements come from sharp corners, thin walls, or complex curved surfaces — fix by adding local sizing controls or simplifying the geometry at that feature.

Advanced Meshing in Industry: How Tata Tech and KPIT Use FEA Mesh Standards

Industry-grade FEA mesh standards at companies like Tata Technologies, KPIT, and L&T are documented in internal simulation guidelines: Tata Tech's CAE team in Pune (Pimpri-Chinchwad facility) uses a maximum element size of 3 mm for automotive structural parts with 1 mm local refinement at welds and fillets; convergence is required to within 3% before results are reported to the customer. KPIT in Hinjewadi Phase 1 runs ANSYS on AWS HPC clusters for large assemblies and requires automated convergence reports as part of deliverable packages. These practices are taught in ABC Trainings' advanced ANSYS program. For fresher simulation engineers at Sambhajinagar's Skoda Auto Volkswagen (Plot A-1/1, Shendra AURIC) and Bajaj Auto (Plot G-137, Waluj MIDC), entry requirements typically include Static Structural and Thermal analysis with mesh convergence validation. AmbitionBox shows CAE Simulation Engineers in Pune earning ₹4.5–7 LPA at 0–2 years. Adding mesh expertise and convergence methodology raises placement offers to ₹6–9 LPA. Call 7039169629 or WhatsApp 7774002496 for ABC Trainings' ANSYS advanced batch schedule.

FAQs

What is a convergence study in ANSYS Workbench and why is it necessary?

A convergence study runs the same FEA analysis with progressively finer meshes and checks when the result stabilises. When two consecutive mesh refinements produce the same critical result (typically peak von Mises stress) within 5%, the mesh is converged and the result is mesh-independent. Without convergence verification, your results may be significantly under-predicting stress and you could be signing off on a part that will fail in service.

How do I know if my ANSYS mesh is fine enough for accurate results?

Check the Mesh Metrics panel in ANSYS: Orthogonal Quality should be above 0.1 (target 0.7+), Skewness below 0.9, and no negative Jacobian elements. Then run a convergence study — compare peak stress at coarse, medium, and fine mesh settings. If the medium and fine mesh give results within 5%, the medium mesh is sufficient for your analysis. This process typically takes 20–30 minutes for simple parts.

What causes negative Jacobian elements in ANSYS and how do I fix them?

Negative Jacobian elements occur when inflation layers are too aggressive — the Growth Rate is too high or Maximum Layers is too many for the thickness of the feature. Fix by reducing Growth Rate from 1.2 to 1.1 and Maximum Layers from 5 to 3. Also check for very thin walls or sharp corners in the geometry — these create inverted elements. Virtualise or chamfer sharp corners in DesignModeler before meshing if they are not structurally significant.

Do ANSYS simulation engineers need to know meshing theory or just the software?

Both — and the good news is you do not need a deep theoretical background in FEM to be a productive simulation engineer. You need to understand convergence logic (results must be mesh-independent), mesh quality metrics (Orthogonal Quality, Skewness), and how to apply local refinement controls. The theory reinforces the practice. ABC Trainings' ANSYS program teaches the practical workflow first and explains the theory in context, which is how most working engineers actually learn simulation.