4. Finite element modeling strategies for industrial equipment under shock

Shock qualification of industrial equipment—motor control cabinets, skid-mounted pump assemblies, transformer enclosures, and similar structures—is usually governed by standards such as MIL-STD-810 Method 516, IEC 60068-2-27, or a customer-specific shock response spectrum (SRS). The finite element (FE) model is the bridge between the specified shock environment and the predicted stress, deflection, and acceleration response of the equipment. Three modeling decisions dominate the fidelity and cost of that bridge: mesh density, element selection, and boundary condition representation. Getting these wrong produces results that look plausible but are physically meaningless—either over-stiff, over-damped, or dominated by numerical artifacts.

Mesh Density: Resolving the Shock Wavelength, Not Just the Geometry

Under a transient shock pulse, the structure responds in a broad band of frequencies. The mesh must be fine enough to represent the modes that carry significant energy in that band. A practical rule is to resolve the highest frequency of interest with at least 6–10 linear elements per wavelength. For a bending wave, the wavelength λ relates to frequency f, plate thickness t, and material wave speed. For a steel panel (E = 210 GPa, ρ = 7850 kg/m³, ν = 0.3), the bending wavelength at 500 Hz in a 3 mm plate is roughly 0.35 m; at 2000 Hz it drops to about 0.17 m. If the SRS shows significant energy up to 2000 Hz, an element size of 20–25 mm is a reasonable starting point for that panel.

Mesh density should not be uniform. High-stress regions—bolt holes, welded brackets, corner gussets, and cable tray attachments—need local refinement to capture stress concentration. A common mistake is to refine globally, which inflates run time without improving accuracy in low-gradient regions. Use a graded mesh: coarse in flat panels, fine near discontinuities, with transition ratios no greater than 2:1 between adjacent elements to avoid spurious stiffness.

Element Selection: Shells, Solids, and Beams

Industrial equipment is rarely a single monolithic solid. It is an assembly of thin panels, stiffeners, bolted joints, and heavy components. The element type should match the dominant load path and geometry:

• Shell elements (S4R, S3R) for panels, cabinet walls, and stiffener webs where thickness is small relative to span. They capture bending and membrane behavior efficiently.
• Solid elements (C3D8R, C3D10) for thick brackets, castings, and regions where through-thickness stress matters—for example, a welded lug under out-of-plane shock.
• Beam elements (B31) for frame members, pipes, and cable trays where the cross-section is compact and bending dominates.
• Mass elements (MASS21) for heavy internal components—transformers, motors, batteries—that are stiff relative to the supporting structure. Represent them as lumped masses at their center of gravity, connected by rigid or beam elements to the mounting points.

Element selection also affects hourglassing and shear locking. Reduced-integration solids (C3D8R) need hourglass control; fully integrated shells (S4) avoid hourglassing but can shear-lock in thin bending. For shock analysis, reduced-integration elements with enhanced hourglass control are usually the best compromise. Avoid mixing element types at a shared node without a tie constraint or kinematic coupling, which can create artificial stiffness or local stress singularities.

Boundary Condition Representation: The Most Common Source of Error

How the equipment is attached to the test fixture or the vehicle floor determines the shock input path. Three representations are common:

The large mass method is the workhorse for shock simulation. A very large mass is attached to the base nodes, and a force equal to mass × acceleration is applied. This drives the base motion while allowing the structure to respond dynamically. The mass must be large enough that the structure's reaction does not significantly alter the prescribed motion—typically 10⁶ times the structure mass, but check that the resulting base acceleration matches the target within 5%.

Bolted joints deserve special attention. A fully fixed constraint at every bolt hole over-stiffens the joint and suppresses local flexibility. A more realistic approach is to model the bolt as a beam or spider constraint with a contact interface, or to use a distributed coupling that allows local deformation. For shock loads, the joint's nonlinear behavior—slip, gap opening, and bolt preload loss—can dominate the response. If the shock level is high enough to cause joint separation, a linear model will under-predict displacement and over-predict stiffness.

Putting It Together: A Worked Example

Consider a 1.2 m × 0.8 m × 0.6 m steel control cabinet, 2 mm wall thickness, with a 45 kg transformer mounted on the back panel. The shock requirement is a 40 g, 6 ms half-sine pulse applied vertically. The cabinet is bolted to a rigid fixture at four corner feet.

Mesh: Shell elements for walls and stiffeners, 20 mm nominal size, refined to 5 mm around bolt holes and the transformer mounting bracket. Solid elements for the transformer bracket. The transformer is a lumped mass at its CG, connected to the bracket by rigid elements.

Boundary conditions: The four feet are constrained in all DOF, but the bolt holes are modeled with a distributed coupling over a 15 mm radius to allow local deformation. The base acceleration is applied via the large mass method with a mass of 10⁷ kg.

Result interpretation: The first mode is a local panel mode at 85 Hz; the global cabinet bending mode is at 42 Hz. The shock pulse excites both. Peak von Mises stress occurs at the transformer bracket weld and reaches 180 MPa—below the 250 MPa yield of the steel, but the bolt preload is lost for 2 ms during the pulse. This is a pass, but the joint should be inspected for fatigue.

Verification and Common Pitfalls

Before trusting any shock simulation, perform three checks. First, a modal check: compare the first few natural frequencies against a hand calculation or a simple test. If the FE model predicts 42 Hz and a tap test gives 38 Hz, the boundary conditions are likely too stiff. Second, an energy check: the kinetic energy of the structure should not exceed the input energy from the shock pulse. If it does, the large mass is too small or the time step is too coarse. Third, a mesh convergence check at the critical stress location, as described earlier.

Common pitfalls include: using a fixed base where a resilient mount exists; neglecting bolt preload and joint nonlinearity; applying the shock as a static load instead of a transient; and using a time step larger than 1/20 of the highest period of interest. For a 2000 Hz mode, the period is 0.5 ms, so the time step should be ≤ 0.025 ms. Explicit solvers handle this naturally; implicit solvers need careful time integration settings.

The modeling strategy is not a one-time decision. It evolves with the analysis: start with a coarse model to identify dominant modes and load paths, then refine locally and re-run. The goal is not the finest mesh or the most complex boundary condition, but the simplest model that captures the physics that matters for the shock level and the failure mode being assessed.