Contact Analysis — CAE Glossary

Category: Glossary | 2026-03-28
CAE visualization for contact analysis - technical simulation diagram

Contact Analysis

🙋

When is contact analysis necessary? How is it different from standard linear static analysis?


🎓

Contact analysis is needed when parts "touch, separate, and slide." Examples include bolted flange connections, gear mesh engagement, and press forming dies. Standard static analysis has boundary conditions fixed from the start, but in contact analysis, whether parts are in contact or separated changes with applied loads. This is called boundary nonlinearity, which linear analysis cannot handle.


🙋

How is the contact handled mathematically? What equations are used?


🎓

Contact is fundamentally an inequality constraint. If the normal gap is $g_N$, the essential conditions are three:

$$g_N \geq 0 \quad \text{(no penetration)}$$ $$p_N \geq 0 \quad \text{(no tensile contact force)}$$ $$p_N \cdot g_N = 0 \quad \text{(complementarity)}$$

This is called the Karush-Kuhn-Tucker (KKT) condition, or Hertz-Signorini-Moreau condition. The third equation is key: "either the gap is open (no contact force) or the gap is closed (contact force exists) — never both simultaneously."


Definition and Fundamentals

🙋

I see "master" and "slave" surfaces in the solver settings. What is the difference? How do I decide which should be which?


🎓

The slave surface nodes are constrained not to penetrate the master surface. The general rules are:

However, modern solvers (Abaqus, Ansys) favor surface-to-surface formulations where master-slave assignment matters less. Still, following these rules is a safe bet when uncertain.


Contact Types (node-to-surface / surface-to-surface)

🙋

Which should I use: node-to-surface or surface-to-surface?


🎓

In general, use surface-to-surface (STS) unless you have a specific reason not to.

Node-to-Surface (NTS):

Surface-to-Surface (STS):

For example, a bolted flange contact analysis with NTS shows nodal chatter in contact pressure, while STS yields smooth distribution. Abaqus now defaults to STS.


Penalty and Augmented Lagrangian Methods

🙋

What is the difference between penalty and augmented Lagrangian methods for contact?


🎓

They differ in how the constraint $g_N \geq 0$ is enforced numerically.

Penalty Method generates a restoring force proportional to penetration — like a virtual contact spring:

$$p_N = \epsilon_N \cdot \langle -g_N \rangle$$

where $\epsilon_N$ is penalty stiffness and $\langle \cdot \rangle$ is the Macauley bracket (negative values clamped to zero). Advantages: fast, no extra unknowns. Disadvantages: $\epsilon_N$ must be tuned; too small allows penetration, too large degrades the condition number.


🙋

So penalty method requires finding the "just right" penalty stiffness. Does augmented Lagrangian solve this?


🎓

Exactly. Augmented Lagrangian adds a Lagrange multiplier $\lambda_N$ to iteratively drive penetration to zero:

$$p_N = \lambda_N + \epsilon_N \cdot \langle -g_N \rangle$$

Updated each iteration:

$$\lambda_N^{(k+1)} = \lambda_N^{(k)} + \epsilon_N \cdot g_N^{(k)}$$

This allows smaller $\epsilon_N$ while still achieving near-zero final penetration. Ansys Mechanical defaults to this method. It requires more iterations than penalty, but contact pressure accuracy and robustness are much better.

Decision guide:


Friction Models

🙋

Is just entering the friction coefficient $\mu$ sufficient, or is there more?


🎓

Coulomb friction is the foundation, but numerically it needs care. Coulomb friction has two distinct states: "stick" and "slip":

$$\|\boldsymbol{\tau}\| < \mu \, p_N \quad \Rightarrow \quad \text{Stick (no relative sliding)}$$ $$\|\boldsymbol{\tau}\| = \mu \, p_N \quad \Rightarrow \quad \text{Slip (tangential force at limit)}$$

The problem is the discontinuity at the transition, which damages convergence. In practice:

Typical values: steel-on-steel $\mu = 0.15$–$0.3$, aluminum $0.1$–$0.35$, rubber-on-metal $0.5$–$1.0$. Measured data is best; surface roughness, lubricant, and temperature all affect friction significantly.


🙋

A senior engineer told me: "Adding friction breaks everything." Any tricks?


🎓

Standard industry practice:

  1. Stabilize contact first with $\mu = 0$. Confirm pure contact converges
  2. Gradually ramp friction. Step 1: $\mu=0$, Step 2: $\mu=0.15$, etc.
  3. Set tangential penalty to ~10% of normal penalty. Default values are often too high

For example, a press-forming simulation that fails to converge with $\mu=0.12$ from the start will often run smoothly if you form to mid-stage with $\mu=0$, then add friction. This is a proven trick in production CAE.


Convergence Challenges and Strategies

🙋

What are the most common convergence failures in contact analysis?


🎓

"Top 5 Convergence Killers" in contact analysis:

  1. Inverted normal direction — Most common. Shell elements with flipped normals prevent contact detection. Check element orientation
  2. Initial penetration — Mesh already intersects due to CAD fillets or clearance. Use *CONTACT CLEARANCE ASSIGNMENT in Abaqus
  3. Inappropriate penalty stiffness — Too high: oscillation (chatter); too low: excessive penetration
  4. Coarse load steps — Contact state cannot toggle rapidly. Refine step size
  5. Rigid body motion — Free parts before contact lock the matrix. Use weak springs or ensure initial contact

🙋

What is "chatter" exactly? How do I fix it?


🎓

Chatter: a node alternates "in contact → out of contact → in contact" every iteration. The Newton-Raphson method switches contact status each iteration, so it never converges.

Solutions:

In Ansys Mechanical: adjust CNOF and PINB parameters.


🙋

Summarize the essential takeaways for contact analysis?


🎓

Five key insights:

  1. Contact is a boundary nonlinear problem — Do not treat it like linear analysis
  2. Use surface-to-surface formulation — Node-to-surface is outdated and less accurate
  3. Penalty is fast but penetrates; augmented Lagrangian is accurate but slower — Choose based on requirements
  4. Stabilize contact without friction first, then add friction gradually
  5. When convergence fails, suspect setup errors first (normal direction, initial penetration) before tweaking parameters

Contact analysis causes more CAE trouble than most disciplines, but these five principles solve 80% of real-world issues.


Precise understanding of CAE terminology is the foundation of team communication. — Project NovaSolver supports learning for practicing engineers.

Share your contact analysis challenges

Project NovaSolver addresses the real problems CAE engineers face—setup complexity, computational cost, result interpretation. Your practical experience drives better tool development.

Contact (Coming Soon)
Article Rating
Thank you for your feedback!
Helpful
More
Detail
Report
Error
Helpful
0
More Detail
0
Report Error
0
Written by NovaSolver Contributors
Anonymous Engineers & AI — Sitemap
View Profile