2. Design change taxonomy for crash models

In crash simulation, a design change is rarely a single, isolated edit. A revised bumper beam may gain a new material grade, a different spot-weld pattern, and a relocated tow-hook bracket in the same release. Requalification therefore begins with classification: before deciding how much of the model must be revalidated, the CAE engineer must decide what kind of change has occurred. A taxonomy is not administrative bookkeeping — it is the mechanism that maps a change to the physical phenomena it perturbs, and hence to the fidelity risks it introduces.

This chapter defines five change classes — geometry, material, joining, restraint system, and load path — and for each one identifies the dominant physical effects, the model entities typically touched, and the fidelity indicators that should be watched during requalification. The classes are not mutually exclusive; most real changes are composites, and the taxonomy's value lies in forcing the engineer to enumerate each component separately.

Geometry changes

Geometry changes alter the spatial description of the structure: section dimensions, wall thickness, bend radii, crush initiators, packaging clearances, or the addition/removal of features. Their fidelity impact is governed by how strongly the change affects the deformation mode. A thickness increase of 0.2 mm on a crash box wall changes mass and bending stiffness modestly, but a relocated crush initiator can flip the folding sequence entirely, converting a stable two-lobe axial fold into global bending. Geometry changes are therefore graded by mode sensitivity, not by dimensional magnitude.

Practical classification distinguishes three sub-types: (a) parametric changes (thickness, gauge, small offsets) that can be propagated by updating existing mesh regions; (b) topological changes (new holes, added ribs, removed flanges) that require local remeshing and re-checking of element quality and contact definitions; and (c) architectural changes (new load-bearing members, relocated subsystems) that invalidate the previous mesh topology and often the previous validation evidence. A useful rule of thumb: if the change moves the predicted peak crush force by more than roughly 10% in a subsystem test, the change should be treated as architectural regardless of its nominal size.

Material changes

Material changes include grade substitution (e.g., DP590 to DP780), gauge-coupled grade changes, coating or heat-treatment changes, and the introduction of new grades such as press-hardened steels or aluminium extrusions. The fidelity risk is concentrated in the constitutive and failure description: yield and hardening curves, strain-rate sensitivity, forming-induced pre-strain, and the fracture locus. A grade substitution that looks like a one-line edit in the material card can shift the energy absorption of a rail by 15–25% if the hardening tail and the failure strain are not updated together.

Classification should record whether the change affects (i) the elastic–plastic response only, (ii) the rate-dependent response, or (iii) the failure/damage model. Changes of type (iii) are the most demanding, because they require re-identification of the fracture parameters from coupon or component tests, and because the previous calibration was tied to the old grade's microstructure. Where the new grade is supplied by a different mill or heat, the engineer should also verify that the forming simulation used to initialize thickness and pre-strain has been rerun — an inherited forming result is a silent fidelity defect.

Joining changes

Joining changes cover spot-weld count, pitch and diameter; weld type substitution (spot to laser, rivet to clinch, adhesive addition); structural adhesive bead geometry; and fastener changes. Joining is the class most often underestimated, because joints are small entities that carry large consequences. A weld pitch increase from 40 mm to 60 mm on a B-pillar-to-rocker flange reduces the local bending stiffness of the closed section and can trigger a different intrusion mode under pole impact.

For requalification, joining changes should be classified by whether they alter (a) the connectivity of the model (new or removed connections), (b) the constitutive behaviour of existing connections (weld nugget diameter, adhesive properties), or (c) the failure criterion of connections (force-based vs. displacement-based rupture, spot-weld fracture models). Connectivity changes require contact and constraint re-verification; failure-criterion changes require re-correlation against joint-level tests. A practical check is to extract the joint force histories at the affected connections from the baseline run and confirm that the new configuration does not push any joint into a regime where the previous failure model was never validated.

Restraint system changes

Restraint changes include airbag module geometry and venting, inflator output, seat-belt geometry and pretensioner/load-limiter characteristics, seat structure, and occupant-position assumptions. These changes act on the occupant rather than the structure, so their fidelity impact is measured in occupant injury metrics — HIC, chest deflection, femur load, neck Nij — and in the timing of occupant–interior interaction. A load-limiter force change from 4 kN to 3.5 kN is a small structural edit but a first-order change to chest deflection and to the phasing of the occupant against the deploying airbag.

Classification should separate restraint hardware changes (module, inflator, belt hardware) from restraint calibration changes (venting, limiter level, pretensioner timing) and from occupant environment changes (seat position, interior trim stiffness). Hardware changes usually require new component-level validation data; calibration changes can often be requalified by re-running the occupant subsystem with updated input curves, provided the structural pulse is unchanged. The critical coupling to watch is the interaction between restraint timing and structural pulse: if a geometry or material change has already altered the vehicle deceleration pulse, a restraint change cannot be assessed in isolation.

Load path changes

Load path changes are the most consequential and the hardest to bound. They occur when structural members are added, removed, or rerouted such that the route by which impact energy travels from the barrier to the occupant compartment is altered: a new lower load path, a removed cross-car beam, a relocated subframe mount, a changed suspension attachment. Unlike the other four classes, a load-path change is often not visible as a single edited entity — it is an emergent property of several edits, and it can invalidate the entire validated pulse.

The correct classification question is: does the change alter the distribution of internal forces among the primary energy-absorbing members? If yes, the change is a load-path change even if every individual edit was parametric. Requalification then requires re-establishing the global force balance — comparing the sum of member force histories at the firewall and rocker sections against the baseline — before any local correlation is attempted.

Composite changes and escalation

Real releases rarely fall into one class. A door inner panel change may simultaneously alter geometry (new embossment), material (thicker gauge), and joining (additional adhesive). The taxonomy's purpose here is to prevent the engineer from classifying the whole package by its most benign component. The recommended practice is to decompose the change into class-tagged items, assess each item's fidelity risk independently, and then apply an escalation rule: if any single item is classified as architectural geometry, failure-criterion joining, or load path, the package is treated as a full requalification regardless of the other items' severity.

A second escalation trigger is coupling. Two individually moderate changes can interact non-linearly — a material upgrade that stiffens a rail combined with a weld-pitch increase that softens its flange may produce a mode neither change would produce alone. Where the affected entities overlap spatially or share a load path, the engineer should run a combined sensitivity check rather than assessing each change in isolation.

Assessing potential impact on simulation fidelity

Impact assessment should be expressed in terms the model can actually be interrogated for. For each classified change, define one or more fidelity indicators: scalar or history outputs whose deviation from the baseline signals that the change has propagated into the prediction. Typical indicators include peak crush force and its timing, total energy absorbed at a given intrusion, the fold or hinge pattern of a rail, joint force histories at affected connections, occupant injury metrics, and the global deceleration pulse. The table below gives a compact assessment worksheet that can be completed per change item.

Worked example: classifying a front-end package

Consider a mid-cycle update to a front-end structure: the bumper beam moves 15 mm forward, the crash box material changes from DP590 to DP780, the spot-weld pitch on the rail flange increases from 40 mm to 55 mm, and a new lower load path is added between the subframe and the rocker. Decomposed by class:

• Bumper beam position — geometry, parametric. Entities: beam mesh, bumper-to-rail connections, barrier contact. Indicator: barrier contact timing and beam bending mode.
• Crash box grade — material, failure-locus type. Entities: material card, hardening curve, fracture parameters, forming pre-strain. Indicator: crash box energy absorption and onset of fracture.
• Weld pitch — joining, connectivity. Entities: weld set, flange contact. Indicator: joint force histories and rail section stiffness.
• Lower load path — load path. Entities: new member, subframe and rocker attachments, global force balance. Indicator: deceleration pulse and member force distribution at the firewall.

The escalation rule fires on the load-path item, so the package is treated as a full requalification. The other three items are still assessed individually, because their indicators tell the engineer where to look first when the requalified model deviates from the baseline. In practice, the weld-pitch and material changes are the most likely to produce a local mode shift, while the load-path addition is the most likely to shift the global pulse — and the two effects can mask each other if only the global pulse is compared.

Common classification errors

Three errors recur in requalification practice. First, classifying by edit size: a 1 mm offset is tagged parametric even though it moves a crush initiator past its trigger point. Second, classifying by file: a change is tagged "material" because the material card was edited, when the real effect is a joining failure-criterion change implemented through the same card. Third, missing emergent load-path changes: several parametric edits that individually pass their thresholds collectively reroute the force distribution, and no single item is flagged. The defence against all three is the same — classify by physical effect on the deformation and force path, verify with the fidelity indicators, and escalate on coupling rather than on nominal severity.

With the taxonomy in place, the next question is how much validation evidence each class demands. That mapping — from change class to requalification scope — is the subject of the following chapter.