Modal Analysis

FEA LEARNING CENTER

Modal Analysis

The System's Fingerprint

By Joseph P. McFadden Sr.

McFaddenCAE.com

Companion document to the FEA Learning Center

in the Abaqus INP Comprehensive Analyzer

Every physical object has a voice.

Tap a wine glass and it rings. Pluck a guitar string and it sings. Slam a car door and it booms. None of that is random. Every one of those sounds comes from the natural frequencies of that object — the frequencies it wants to vibrate at based on what it's made of, how it's shaped, and how it's supported.

That voice is the system's dynamic fingerprint. And modal analysis is how we read it.

If you've spent any time in the simulation world, you've probably run a modal analysis. It's often the first dynamic procedure an engineer learns. But here's what I've noticed over nearly four decades in this field — a lot of engineers run modal analyses without fully understanding what the results actually mean, what they can and cannot tell you, and why the procedure has restrictions that are absolutely non-negotiable.

So let's talk about modal analysis the way it deserves to be talked about. Not as a button to click or a keyword to type, but as a window into the fundamental nature of your structure.

The Why — What Modal Analysis Actually Reveals

Every structure has a set of natural frequencies and corresponding mode shapes. These are inherent properties of the system — they exist whether you excite the structure or not, just like a guitar string has natural frequencies whether you pluck it or not.

A natural frequency is the rate at which the structure will vibrate if you disturb it and then let go. A mode shape is the pattern of that vibration — which parts move a lot, which parts stay still, how the structure deforms.

Here's the key insight that too many engineers miss.

Mode shapes are relative patterns, not absolute magnitudes.

Let me say that again because it matters. A mode shape tells you the relative pattern of displacement — which parts of the structure move and in what proportion to each other. It does not tell you how much anything actually moves. The actual physical displacements, strains, and stresses depend on the excitation — what force you apply, at what frequency, for how long. The mode shape alone doesn't contain that information.

Think of it this way. A guitar string's first mode shape is always a half-sine wave — maximum displacement in the middle, zero at the ends where it's fixed. That pattern is the same whether you pluck the string gently or aggressively. What changes is the amplitude — how far the string actually moves. The pattern is the mode shape. The amplitude comes from the response analysis.

This distinction matters because I've seen engineers look at a modal analysis result, see stress contours in the mode shape visualization, and treat those as actual stresses their part will experience in service. They're not. Those are proportional indicators of where stress concentrates in that mode pattern, but the actual stress values depend entirely on how the structure is excited. To get real physical quantities, you need to go further — participation factors, modal coordinates, and ultimately a response analysis like harmonic, random vibration, or shock response spectrum.

The What — Natural Frequencies And What They Mean

So what determines the natural frequencies of a structure? Two things: stiffness and mass.

The governing equation is beautifully simple. The stiffness matrix times the mode shape equals the frequency squared times the mass matrix times the mode shape. That's the eigenvalue problem at the heart of every modal analysis. Stiffness wants to restore the structure to its original shape. Mass resists acceleration. The balance between them sets the frequency.

More stiffness means higher frequency — stiffer things vibrate faster. More mass means lower frequency — heavier things vibrate slower. Change the material from aluminum to steel and you change the stiffness-to-mass ratio. Change the geometry — make a beam thicker or shorter — and you change it again. Every design decision shifts these frequencies.

And here's why that matters in practice.

If any natural frequency of your structure coincides with an excitation frequency in your operating environment — the spin rate of a motor, the rumble of a launch vehicle, the vibration of a road surface — you get resonance. The structure amplifies the input at that frequency. How much it amplifies depends on the damping, but even well-damped structures can see amplification factors of ten or more at resonance.

Now, I want to be very deliberate about the word resonance.

Resonance is not the enemy. Resonance is a physical state — a condition where the excitation frequency matches a natural frequency. It's something to understand and design around, not something to fear blindly. Musical instruments are intentionally designed to resonate. Ultrasonic welding systems operate at resonance on purpose. Tuned mass dampers in skyscrapers exploit resonance to absorb energy. MEMS devices — the tiny accelerometers in your phone — are designed to resonate at specific frequencies.

The problem isn't resonance itself. The problem is unintended resonance — when your structure resonates at a frequency you didn't plan for, amplifying loads you didn't account for, creating stresses you didn't predict. Modal analysis is how you identify those frequencies so you can make informed design decisions. Shift the resonance away from the excitation. Add damping to limit the amplification. Redesign to change the stiffness-to-mass ratio. But you can't do any of that if you don't know where the resonances are.

Participation Factors — Which Modes Matter

A typical modal analysis might extract twenty, fifty, even a hundred modes. But they don't all matter equally. Some modes participate heavily in the response to a particular excitation direction, and some barely participate at all.

Participation factors tell you how much each mode contributes to the overall motion in a given direction. If you're shaking the structure vertically, a mode that primarily involves horizontal motion won't participate much in the vertical response, even if its frequency is close to the excitation frequency.

The effective mass associated with each mode is another way to look at this. If you add up the effective masses across all modes in a given direction, the sum should approach the total mass of the structure. If it doesn't — if you've only captured, say, 60 percent of the mass — you haven't extracted enough modes. You're missing dynamic content. Extract more modes until you've captured at least 90 percent of the total mass in each direction of interest.

This is the bridge from modal analysis to actual physical prediction. The participation factors tell you which modes will be excited. The modal coordinates — computed in a subsequent response analysis — tell you how much each mode contributes. And from there you recover actual displacements, strains, and stresses. That's the full chain: mode shapes times participation factors times modal coordinates gives you physical response.

The How — Running It In Abaqus

The mechanics of setting up a modal analysis in Abaqus are straightforward, but every choice you make has consequences.

First, the step definition. Modal analysis is a linear perturbation step — Frequency extraction. You specify how many modes you want extracted and optionally a frequency range. The Lanczos eigensolver is the standard choice. It's fast, robust, and handles most problems well. For very large models — millions of degrees of freedom — the AMS solver can be faster, but Lanczos is the safe default.

How many modes to extract? Start with the rule of thumb: extract modes up to at least twice the highest excitation frequency you care about. If your random vibration PSD goes up to 2,000 Hertz, extract modes up to at least 4,000 Hertz. More is generally better than fewer — you can always ignore high modes, but you can't recover modes you didn't extract.

Material properties must include density. This sounds obvious, but I have seen — more than once — an engineer submit a modal analysis that fails or gives absurd frequencies because density was missing from one of the material cards. Every material in your model must have density defined. Without mass, there is no modal analysis.

Boundary conditions matter enormously. A cantilever beam and a free-floating beam have completely different natural frequencies and mode shapes. If your real part is bolted to a fixture, those bolt locations are boundary conditions. If it's floating inside packaging, it's closer to free-free.

Speaking of free-free — if you run a modal analysis with no constraints at all, the first six modes will be rigid body modes at zero Hertz. Three translations and three rotations. These are not elastic modes and they're not errors. They're the natural consequence of an unconstrained body being free to translate and rotate as a rigid unit. The elastic modes — the ones that involve actual deformation — begin at mode seven. Free-free analysis is perfectly valid and is commonly used to study the inherent dynamic character of a structure independent of its mounting.

The Restriction — And It Is Non-negotiable

Modal analysis is a linear perturbation procedure. I need you to understand what that means in practice, because this is the restriction that gets violated most often by engineers who don't realize they're breaking it.

Linear perturbation means the analysis assumes small displacements around a base state, linear material behavior, and — this is the critical one — no contact.

Contact elements cannot appear in a modal analysis. This is non-negotiable. Not sometimes. Not with special settings. Not if you're careful. Contact elements fundamentally change the stiffness of the system depending on the state of contact — whether surfaces are in contact or separated, sliding or stuck. That state-dependent stiffness is incompatible with the eigenvalue problem that modal analysis solves.

If your model has bolted joints modeled with contact pairs, gaskets with pressure-dependent behavior, or press-fit interfaces with frictional contact — those interfaces must be modeled as tied constraints or merged meshes for modal analysis. A tied constraint fuses the surfaces together and provides a fixed stiffness. That's compatible with the eigenvalue problem. Contact is not.

The same restriction applies to material nonlinearity. If your model has plasticity, hyperelasticity, or damage models, those nonlinear behaviors are invisible to a perturbation analysis. The solver uses only the linear elastic stiffness at the base state.

And large deformations are not considered. The modal analysis assumes the deformations are small enough that the stiffness matrix doesn't need to be updated.

This isn't a limitation of the software. It's a property of the mathematics. The eigenvalue problem requires a constant stiffness matrix. Anything that makes stiffness state-dependent — contact, plasticity, large deformation — breaks that requirement.

What To Do With Your Results

Once you have your modes, here's how to evaluate them.

Visualize the mode shapes. Do they make physical sense? A cantilever beam should show bending modes with increasing numbers of half-waves. A plate should show drumhead-like patterns. If you see something unexpected — a mode that seems to involve only a small feature, or a rigid body-like motion in a constrained model — investigate. It might be a meshing issue, a missing constraint, or a poorly connected region.

Check for symmetry. If your structure is symmetric, your modes should be too — or they should come in symmetric/antisymmetric pairs. If a symmetric structure produces asymmetric modes, something is wrong with the mesh or the boundary conditions.

Look at the participation factors. Which modes carry the most mass in each direction? Those are the modes that will dominate the dynamic response. If you're designing to avoid a resonance problem, those are the modes whose frequencies you need to shift.

Compare to hand calculations. For simple geometries — beams, plates, cylinders — analytical solutions exist. Your finite element frequencies should converge to those solutions as you refine the mesh. If they don't, the model has a problem.

And if you have test data, compare. A correlation between FEA and experimental modal frequencies within five to ten percent for the first several modes is typical for a well-built model. If the discrepancy is larger, look at the boundary conditions first — they're the most common source of disagreement between test and analysis.

The Bigger Picture

Modal analysis is rarely the final answer. It's the foundation.

Once you know the natural frequencies and mode shapes, you can build on them. Harmonic response analysis tells you how the structure responds to a sinusoidal excitation swept across frequency. Random vibration analysis tells you the response to a broadband random input like a launch vehicle or a road surface. Shock response spectrum analysis tells you the peak response to a transient shock event. All three of these build on modal results. All three require the mode shapes and frequencies that modal analysis provides.

And all three inherit the same fundamental restriction — they're perturbation procedures. No contact. Linear material behavior only. Small deformations. If your problem genuinely involves contact, large deformations, or material nonlinearity, you need to cross into the explicit world — time-domain transient analysis where the physics are solved step by step without the eigenvalue assumption. That's a different discussion entirely.

But for a vast range of structural dynamics problems — vibration qualification, frequency avoidance, sensor placement, design optimization — modal analysis is where it starts. It tells you what the structure is, dynamically speaking. Everything else is about how it responds when you poke it.

And that brings us back to where we started. Every system has a voice. Modal analysis is how you listen.

For more on the perturbation family and how modal results feed into response analysis, see the companion discussions on harmonic response, random vibration, SRS, and the dedicated discussion on perturbation limitations — all available at McFaddenCAE.com.

This has been a Learning Center discussion on modal analysis. I'm Joe McFadden. Thanks for listening.

About the Author

Joseph P. McFadden Sr. is a CAE engineer specializing in finite element analysis, modal analysis, materials behavior, and injection mold tooling validation. With nearly four decades of experience in structural simulation, he brings a holistic perspective to engineering education — connecting how systems respond to how people think and learn.

His work at McFaddenCAE.com includes the Abaqus INP Comprehensive Analyzer — a desktop tool for analyzing, visualizing, and extracting sub-assemblies from large FEA models without requiring an Abaqus license — along with DSP tools for SRS computation, jerk extraction, velocity change analysis, and energy balance verification.

The FEA Learning Center is an integrated educational platform within the Analyzer, providing guided discussions on structural dynamics topics with working example INP files. This document series is the companion written reference for those discussions.

The four-volume FEA Best Practices audiobook series — Building the Model, The System's Natural Character, When Things Collide, and Keeping the Simulation Honest — is available at McFaddenCAE.com.