Preloaded Ball Bearings FEA Vibration Analysis

Rolling-element bearings are a masterpiece of mechanical engineering. They are also a notorious headache for Finite Element Analysis (FEA) engineers.

Because they rely on Hertzian contact, bearings are inherently non-linear systems. The rolling elements, raceways, contact angles, internal clearances, and thermal conditions all shift dynamically under load. If you try to model these behaviors directly in standard linear dynamic analyses—like normal modes, modal frequency response, or random vibration—your solver will either reject the non-linearities or grind to a halt.

So, how do industrial engineers solve this? They linearize the bearing about its operating point.  By calculating the bearing’s tangent stiffness under a specific static preload, you can substitute a highly complex contact problem with a representative linear spring matrix. Here is a comprehensive, step-by-step guide to setting up preloaded bearing models in FEA and accurately extracting the dynamic loads.

—

1. Define the Bearing Operating Condition

Before touching your FEA software, you must accurately capture the physics of your bearing assembly. The linearized stiffness you calculate is entirely dependent on the baseline state of the physical system.

Be sure to document and define the following parameters:

• Bearing Type: Deep-groove ball, angular contact, duplex pair, tapered roller, etc.
• Arrangement Back-to-back (O), face-to-face (X), tandem, or floating/fixed configurations.
• Preload Level: Essential for angular contact pairs, as preload directly dictates axial, radial, and moment stiffness.
• Static Loads: Any constant external forces (e.g., gravity, belt tension, aerodynamic loads).
• Environmental Factors: Shaft speeds, fit-up tolerances, and operational temperatures that alter internal clearances.

Caution: If your operational vibration forces are large enough to completely counteract and “unload” one of the bearings in a preloaded pair, your linearized model is no longer valid. The system will cross over into non-linear behavior.

2. Linearize the Bearing Stiffness

Once the operating condition is locked in, calculate the linearized tangent stiffness. For rough, preliminary models, engineers sometimes simplify this down to independent radial ($K_r$) and axial ($K_a$) scalar stiffnesses.

However, precision systems—especially angular contact pairs—require a coupled stiffness matrix. A robust bearing model utilizes a 5-DOF or 6-DOF matrix capturing:

• Three translational stiffnesses
• Two rotational (rocking) stiffnesses
• Cross-coupling terms (e.g., where a radial displacement induces a moment reaction)

Note: Torsional stiffness about the shaft axis is generally negligible or ignored unless explicit seal friction is being modeled.

Where to Find Stiffness Values

Do not guess these numbers. Rely on trusted sources:

•  Proprietary Vendor Tools: SKF, NSK, Schaeffler, or Timken engineering calculators.
• Specialized Software: MESYS, RomaxDESIGNER, or KISSsoft.
• Analytical Texts: Classic formulas from Harris & Kotzalas’ Rolling Bearing Analysis.

If you are forced to pull data directly from a vendor catalog, double-check that the cataloged stiffness matches your exact preload, contact angle, and mounting arrangement.

3. Representing the Bearing in FEA

In your FEA environment, the bearing acts as a spring-like bridge connecting the rotating shaft to the stationary housing.

Instead of connecting the bearing element to a single “hard point” on the shaft or housing—which creates artificial stress concentrations—use rigid or interpolating elements (like spiders/couplings) to distribute the load across the actual race and housing regions.

Element Selection Across Major Solvers

FEA Software, Recommended Element Types, Local Coordinate System Notes

Software	Element / Feature	Notes
MSC / Autodesk Nastran	CBUSH, CELAS	Ensure the CBUSH orientation matches the shaft local coordinate system.
Abaqus	CONN3D2 (Connector), Bushing elements	Define local orientation fields at both connector nodes.
ANSYS Mechanical	Bushing, Remote Spring, COMBI214	Align the local reference coordinate system carefully with the rotor axis.

Pro-Tip: Always establish a local coordinate system specifically for the bearing. Align one axis exactly along the centerline of the shaft (Axial), and the remaining two axes transversely (Radial). Mapping an axial stiffness onto a radial degree of freedom is one of the most common setup errors in rotor dynamics.

4. How to Treat Preload Correctly

A common point of confusion is how to handle the preload force itself during a dynamic run. In linear dynamics, the preload is not explicitly “solved” as a force. Its primary job is to establish the tangent stiffness matrix used by the solver.

When it comes to tracking down the absolute, worst-case bearing loads, you must combine the static preload with the incremental dynamic loads computed by your FEA run:

$$P_{\text{total}} \approx P_{\text{static}} + P_{\text{dynamic}}$$

For random vibration environments, the dynamic output is delivered as a Root-Mean-Square (RMS) value. To account for statistical peaks in a Gaussian distribution, a 3-sigma peak estimate is standard practice:

$$P_{\text{peak}} \approx P_{\text{static}} + 3 P_{\text{RMS}}$$

Always verify your specific industry framework or design standard, as some require alternative statistical scaling factors.

5. Executing the Analysis Types

Normal Modes Analysis (Check Your Sanity)

Before diving into expensive dynamic simulations, run a normal modes analysis (e.g., Nastran `SOL 103`). Use this as a diagnostic tool:

If the bearing stiffness is erroneously high, the shaft will behave as if it is rigidly clamped, skewing your natural frequencies upward.
If it is too low, the rotor will exhibit unrealistic, free-floating rigid body modes.
Look closely at shaft bending, housing rocking, and cantilevered modes to ensure they are physically plausible.

Random Vibration / PSD Analysis

To extract random vibration responses (e.g., Nastran `SOL 111`), apply your Power Spectral Density (PSD) input profiles into the mounting structure.

To calculate the total radial RMS force from your directional outputs ($F_{x,\text{RMS}}$ and $F_{y,\text{RMS}}$), use a root-sum-square calculation as a screening tool:

$$F_{r,\text{RMS}} = \sqrt{F_{x,\text{RMS}}^2 + F_{y,\text{RMS}}^2}$$

Combine this with your static preload vector to find your design peak:

$$\mathbf{F}_{\text{peak}} \approx \mathbf{F}_{\text{static}} + 3\mathbf{F}_{\text{RMS}}$$

Shock and Transient Analysis

Your approach to shock depends entirely on how the environment is defined:

If you have an explicit acceleration pulse (e.g., a half-sine or sawtooth wave), run a transient response analysis. Extract the element forces over time and envelope the results to find the peak:

$$F_{\text{peak}} = \max |F(t)|$$

Shock Response Spectrum (SRS):  If provided an SRS, you can approximate forces using equivalent static acceleration methods around dominant frequencies. For higher accuracy, synthesize an SRS-compatible time-history waveform to extract direct spring/connector force profiles.

6. What Load Are You Actually Recovering?

It is crucial to recognize the limitations of your FEA output. The forces you extract from a `CBUSH`, connector, or spring element represent the net global load transferred across the bearing support structure.

FEA does not tell you the stress on an individual internal ball or roller.

To evaluate bearing fatigue life (L10 life), contact stress, or raceway degradation, you must take the net global loads recovered from your FEA model and feed them back into your dedicated bearing calculation software (e.g., MESYS or vendor tools).

[FEA Dynamic Response]
        │
        ▼
[Net Global Bearing Loads]
        │
        ▼
[Dedicated Bearing Tool] ──► Calculates:
                              • Ball-by-ball load distribution
                              • Contact stress
                              • Life margins

7. When the Linearized Approach Fails

While the linearized bearing stiffness workflow is the industrial standard for structural qualification, it is a tool with clear boundaries. You must pivot to a fully non-linear transient analysis if you encounter:

Complete Unloading: Dynamic structural loads completely overcome the operational preload, causing components to separate.
Clearance/Deadband Play: Systems where the bearing crosses back and forth over a physical internal clearance gap.
Massive Contact Angle Shifts: High-amplitude transient events that fundamentally change the internal geometry of the bearing during operation.

For the vast majority of aerospace, automotive, and precision mechanism qualification profiles, keeping your bearing seated and utilizing the linearized tangent stiffness workflow provides a highly accurate, computationally efficient, and reliable engineering path.

Tom Irvine