3. Modal reduction fundamentals

Component mode synthesis (CMS) is the family of techniques that lets an analyst replace a large finite element (FE) component with a small set of generalized coordinates while preserving the dynamic behavior that matters to the assembled structure. The three classical formulations — Hurty, Craig-Bampton, and MacNeal — differ mainly in how they partition interface and interior motion, and in which mode sets they retain. Understanding these differences is the prerequisite for choosing a reduction strategy that keeps the assembly-level modes accurate without inflating the reduced model.

Why Reduction Is Necessary

A full-vehicle body-in-white model can easily exceed 10 million degrees of freedom (DOF). Solving an eigenvalue problem of that size for every design iteration is impractical. CMS exploits the fact that a component's dynamic contribution to the assembly is dominated by a limited number of low-frequency deformation patterns plus the static flexibility of its interfaces. The reduced component is described by a small matrix pair (K_red, M_red) that is assembled exactly like a superelement.

The general transformation is:

u = Φ q
K_red = Φᵀ K Φ
M_red = Φᵀ M Φ

where Φ is the reduction basis and q the generalized coordinates. The three methods below are distinguished by the construction of Φ.

Hurty's Method: Interface Modes Plus Fixed-Interface Normal Modes

Hurty (1965) was the first to formalize component mode synthesis. He partitioned the boundary DOF into three categories: statical (constrained) boundary, redundant (connection) boundary, and interior. The reduction basis combines:

• Constraint modes — static shapes obtained by imposing a unit displacement on one boundary DOF while holding all others fixed.
• Normal modes — eigenmodes of the component with all boundary DOF fixed.
• Rigid-body modes — for unconstrained components, the six (or fewer) rigid-body vectors.

Hurty's formulation is mathematically complete but requires explicit handling of redundant boundary coordinates, which complicates assembly when more than two components share an interface. In practice, Hurty's method is now mainly of historical and theoretical importance; its ideas survive in the more convenient Craig-Bampton form.

Craig-Bampton: The Industry Standard

Craig and Bampton (1968) simplified Hurty's partition by treating all boundary DOF uniformly. The reduction basis is:

Φ_CB = [ Φ_c   Φ_n ]
       [ I     0   ]

Here Φ_c are constraint modes (one per boundary DOF) and Φ_n are fixed-interface normal modes truncated to the lowest n frequencies. The upper block I is the identity on boundary DOF, reflecting that boundary coordinates are retained exactly. The lower block 0 reflects that fixed-interface modes have zero boundary motion.

The resulting reduced matrices have a characteristic block structure:

K_red = [ K_bb   0    ]
        [ 0      Λ_n  ]
M_red = [ M_bb   M_bn ]
        [ M_bnᵀ  I    ]

where K_bb is the statically condensed boundary stiffness, Λ_n the diagonal matrix of fixed-interface eigenvalues, and M_bn the boundary–interior coupling mass. This structure is what makes Craig-Bampton so attractive: boundary DOF are physical, so components connect by simple DOF mapping, and the interior contribution is diagonal in stiffness.

MacNeal's Method: Static Condensation with Interface Flexibility

MacNeal (1971) approached the problem from the static condensation side. Rather than building constraint modes explicitly, he used Guyan reduction to condense interior DOF onto the boundary, then augmented the reduced boundary model with a few normal modes to recover dynamic effects. The key insight is that Guyan condensation alone is exact only at zero frequency; adding normal modes corrects the dynamic stiffness at higher frequencies.

MacNeal's basis can be written as:

Φ_Mac = [ I       0    ]
        [ -K_ii⁻¹ K_ib   Φ_n ]

The first column block is the Guyan static condensation operator; the second is the fixed-interface normal modes. This is algebraically equivalent to Craig-Bampton when the same mode set is retained, but the derivation and numerical implementation differ. MacNeal's formulation is often preferred when the analyst already has a Guyan-reduced model and wants to add dynamic correction terms incrementally.

Comparison of the Three Formulations

Worked Example: A Bracketed Panel

Consider a simplified example: a 1.2 m × 0.8 m steel panel (t = 2 mm, E = 210 GPa, ρ = 7850 kg/m³) with four bolted brackets at the corners. The panel is meshed with 4-node shell elements, giving approximately 18,000 DOF. The brackets are modeled as rigid spiders connecting to the panel at 24 boundary DOF.

Using Craig-Bampton reduction with 20 fixed-interface modes:

1. Apply fixed constraints at all 24 boundary DOF; solve the eigenvalue problem to obtain Φ_n and Λ_n.
2. For each boundary DOF, impose a unit displacement and solve the static problem to obtain the corresponding constraint mode column of Φ_c.
3. Form Φ_CB and compute K_red, M_red.
4. Assemble the reduced panel with the reduced brackets and solve the assembly eigenvalue problem.

Expected outcome: the first ten assembly modes below 150 Hz should match the full-model solution within 1–2% in frequency. If a mode near 180 Hz shows a 5% error, the usual cause is insufficient fixed-interface modes — increase the retention frequency and re-run. A second common symptom is a spurious low-frequency mode caused by an unconstrained rigid-body mode in a component that should be grounded; check that all boundary DOF are properly connected in the assembly.

Choosing a Method in Practice

For most production CAE workflows, Craig-Bampton is the default because of its clean assembly and predictable accuracy. MacNeal is useful when retrofitting an existing Guyan model. Hurty remains valuable for understanding the theoretical foundation and for specialized cases involving redundant interfaces. The decision should be driven by the frequency range of interest, the number of interface DOF, and whether the component is grounded or free.

Key Takeaways

Craig-Bampton is the workhorse of production CMS: it retains physical boundary DOF, produces a clean block-structured reduced system, and assembles by direct DOF mapping. Hurty provides the theoretical foundation but is rarely used directly. MacNeal offers a convenient path when starting from a Guyan-reduced model. In all three, accuracy is governed by the number of fixed-interface normal modes retained and by whether the interface flexibility is adequately represented. The verification step — comparing reduced and full-model frequencies over the target band — is not optional; it is the only reliable way to confirm that the reduction basis is sufficient for the assembly-level predictions the model is meant to support.