3. Modal analysis fundamentals

Modal analysis is the computational backbone of structural vibration qualification. Before you can run a harmonic response sweep, a random vibration PSD analysis, or a sine-dwell test correlation, you need to know where the structure wants to vibrate and how it will deform when it gets there. That is exactly what modal analysis delivers: a set of natural frequencies, the corresponding mode shapes, and — critically for qualification work — participation factors that tell you which modes actually matter for a given excitation direction.

This section builds the working knowledge you need to set up, interpret, and defend a modal solution in a qualification context. We will move from the underlying eigenvalue problem through practical extraction settings, then into the interpretation layer where participation factors and effective modal mass separate the modes that drive your qualification loads from the ones you can safely ignore.

The Eigenvalue Problem Behind Every Modal Solve

For an undamped free-vibration case, the equation of motion reduces to the classic generalized eigenvalue problem:

[K] {φ} = λ [M] {φ}

where [K] is the stiffness matrix, [M] is the mass matrix, λ is the eigenvalue, and {φ} is the eigenvector (mode shape). The natural frequency follows directly:

f_n = sqrt(λ) / (2π)    [Hz]

In practice, most FEA solvers (Nastran SOL 103, ANSYS Modal, Abaqus *FREQUENCY) solve this in one of two families: transformation methods (Lanczos, Householder, Givens) that reduce the problem to a smaller tridiagonal or Hessenberg form, or tracking methods (Sturm sequence, inverse iteration) that converge on individual roots. Lanczos is the default for most qualification models because it handles large sparse matrices efficiently and extracts a block of modes in one pass.

The key input you control is the frequency range of interest. For qualification work, this is almost never "all modes." It is typically bounded by the test specification: if your sine sweep runs 5–2000 Hz, you need modes up to roughly 1.5× the upper bound (3000 Hz) to capture modes that could be excited by sweep harmonics or by coupling with fixture dynamics. Setting the extraction range too narrow is one of the most common causes of a qualification analysis that passes in FEA but fails in test.

Mode Shapes: What They Tell You and What They Don't

A mode shape is a normalized eigenvector. It has no absolute amplitude — only relative displacement between degrees of freedom. This is a frequent source of confusion when comparing FEA mode shapes to test mode shapes from a modal hammer or shaker survey. The FEA shape might show a peak displacement of 1.0 at the tip and 0.3 at mid-span; the test might show 0.85 and 0.26. The ratio is what matters, not the absolute value.

For qualification, you care about three things in a mode shape:

• Node locations — where the structure does not move. These are your natural mounting and sensor placement points, and they tell you where a shaker or accelerometer will have poor authority.
• Antinode locations — where displacement is maximum. These are where you place control accelerometers and where you expect the highest stress.
• Directionality — whether the mode is primarily bending, torsion, axial, or a coupled combination. A mode that is 90% torsional will respond very differently to a uniaxial shaker input than a bending mode.

Participation Factors and Effective Modal Mass

Natural frequencies and mode shapes alone do not tell you which modes will actually be excited by your qualification input. That is the job of the modal participation factor:

Γ_i = {φ}_i^T [M] {r}
M_eff,i = Γ_i^2

where {r} is the rigid-body influence vector for the excitation direction (e.g., a unit translation in X). The participation factor quantifies how much a given mode participates in the response to a uniform base acceleration in that direction. The effective modal mass M_eff,i is the square of the participation factor, and the sum of effective modal masses across all modes should equal the total mass of the structure in that direction.

This gives you a concrete, defensible criterion for mode truncation. A common qualification rule of thumb is to extract enough modes so that the cumulative effective modal mass in each excitation direction reaches 80–90% of total mass. If you stop at 70%, you are likely missing modes that will contribute to the test response, and your harmonic or random analysis will underpredict peak accelerations.

The table above is a representative example for a cantilevered electronics enclosure with a total mass of 12 kg. Note that the X-direction cumulative effective mass only reaches 57.4% by mode 8. This tells you the extraction range is insufficient for X-direction qualification — you would need to extend the solve to at least 1500–2000 Hz to capture the remaining participating modes. The Y-direction modes, by contrast, contribute almost nothing to X excitation, which is expected for a structure with strong directional stiffness.

Practical Extraction Settings for Qualification Models

When setting up the modal solve in your solver, the following parameters matter most for qualification relevance:

• Frequency range: Set the upper bound to at least 1.5× the highest qualification test frequency. For a 2000 Hz sine sweep, extract to 3000 Hz.
• Number of modes: If your solver requires a fixed count rather than a range, start with 2× the number of modes you expect in the range, then verify cumulative effective mass.
• Mass normalization: Use mass-normalized eigenvectors ({φ}^T [M] {φ} = 1) so that participation factors and effective mass are directly comparable across modes.
• Boundary conditions: Match your test fixture as closely as possible. Free-free modes are useful for component-level correlation, but qualification analysis almost always uses constrained boundary conditions.
• Preload: If the structure is preloaded (bolted joints, pressurized components), include the preload in the stiffness matrix. This can shift frequencies by 5–15% in some assemblies.

Reading the Results: A Qualification Workflow

After the solve completes, the workflow for qualification-relevant mode selection is:

1. Sort modes by natural frequency and identify the first mode in each excitation direction (X, Y, Z).
2. Compute cumulative effective modal mass per direction. If it plateaus below 80%, extend the extraction range and re-run.
3. Flag any mode within ±10% of a qualification test frequency (sine dwell, random PSD breakpoint, or acoustic tone) for detailed harmonic response analysis.
4. Check mode shapes for boundary condition anomalies — unexpected motion at constrained DOF, or zero motion at locations where the test will apply input.
5. Document the mode list with frequency, shape description, participation factor, and effective mass. This becomes the traceability record for your qualification report.

The SVG below summarizes the relationship between the eigenvalue solve, the participation factor calculation, and the qualification decision gate. It is not a flowchart of steps — it is a data-flow diagram showing how the raw eigen-solution feeds the metrics that drive mode selection.

The diagram shows the critical feedback loop: if cumulative effective modal mass does not reach the 80% threshold in any excitation direction, the extraction range must be widened and the solve repeated. This is not a one-time check — it is the gate that determines whether your modal basis is adequate for the subsequent harmonic or random response analysis that will actually produce the qualification margins.

In the next section, we will take this modal basis and apply it to harmonic response analysis, where the participation factors and mode shapes become the building blocks for frequency-domain transfer functions and peak response estimation.