2. Human body representation in MADYMO

MADYMO offers three fundamentally different ways to represent the human body, and the choice among them determines what physical phenomena your simulation can capture, how long it will run, and how much model-building effort it demands. The three families are ellipsoid multibody models, facet models, and finite element (FE) human models. They are not competing versions of the same thing; they occupy different points on a fidelity-versus-cost spectrum, and each is the correct tool for a specific class of question.

The three representation families

Ellipsoid multibody models approximate each body segment as a rigid body whose contact surface is a set of ellipsoids. Joints between segments are kinematic constraints (revolute, spherical, translational, or bracket joints), and the occupant is assembled from a small number of bodies — typically 15 to 50. Because contact is resolved analytically between ellipsoids and the surrounding environment, these models are extremely fast and numerically robust. Their limitation is equally clear: ellipsoids cannot deform, so they cannot predict fracture, and their coarse geometry limits the accuracy of local pressure distribution.

Facet models replace the smooth ellipsoid surface with a triangulated mesh of rigid facets. The body remains a multibody system — segments are still rigid bodies connected by joints — but the outer surface is now a detailed polygonal shell, often derived from medical imaging or laser scans. Facets improve geometric realism and contact pressure distribution considerably, especially against stiff or sharp interior surfaces, while retaining most of the computational efficiency of multibody dynamics. They still do not deform, so injury prediction remains indirect.

Finite element human models discretize the body into solid, shell, and beam elements with assigned material laws. Deformation, stress, and strain are computed directly, which makes it possible to predict tissue-level injury using strain- or stress-based criteria. The cost is substantial: runtimes increase by orders of magnitude, mesh quality becomes critical, and contact definitions require careful attention. FE human models are the right choice when the research question is about injury mechanism rather than gross kinematics.

Model families and their lineage

Within the ellipsoid family, MADYMO ships several occupant variants. The 50th percentile male (often designated as the standard adult male) is the baseline for regulatory frontal impact work. The 5th percentile female and 95th percentile male cover the small-occupant and large-occupant extremes required by many test protocols. Child models are provided for the 6-year-old, 3-year-old, and 10-year-old, each with scaled segment masses and joint properties. Pedestrian variants exist as well, with a standing posture and a leg model suitable for bumper and hood contact.

Facet models are typically derived from the same multibody skeletons but with a scanned outer surface. MADYMO's facet occupant models are commonly used for belt-fit studies, seat pressure distribution, and out-of-position (OOP) airbag interaction, where the exact location of the belt relative to the pelvis and chest matters more than tissue deformation.

FE human models in MADYMO are usually imported from dedicated human FE libraries — the THUMS family (Total Human Model for Safety) and the GHBMC family (Global Human Body Models Consortium) are the two most widely used. These are delivered as LS-DYNA or PAM-CRASH decks and converted for MADYMO's FE solver. They include separate 50th percentile male, 5th percentile female, and pediatric versions, and they carry validated material properties for bone, cartilage, ligament, muscle, and internal organs.

Selection criteria

Choosing a model family is a requirements-matching exercise, not a preference. The following decision table summarizes the dominant criteria.

Three practical rules of thumb help in day-to-day work. First, if the question can be answered with kinematics and forces, use an ellipsoid model — it will let you run hundreds of design variants in the time an FE model needs for one. Second, if the question involves where the belt or airbag contacts the body and how load spreads over the surface, move to a facet model. Third, if the question is "will this bone break or this organ rupture," only an FE human model can answer it directly.

Worked example: selecting a model for a belt-fit study

Suppose the task is to evaluate whether a new seat belt geometry keeps the lap belt on the iliac crest of a 5th percentile female occupant during a 50 km/h frontal sled pulse. The input is a MADYMO sled model with a seat, a three-point belt, and a retractor. The question is geometric and contact-driven, not fracture-driven.

An ellipsoid model would give the belt force history and gross kinematics, but its smooth pelvis surface would let the belt slide in a way that does not reflect the real iliac crest geometry. A facet model of the 5th percentile female, with the scanned pelvis surface, resolves the belt path over the crest and shows whether the belt migrates onto the abdomen. The observable output is the belt segment position relative to the anterior superior iliac spine (ASIS) landmark over time, plus the contact pressure distribution on the pelvis. If the belt remains above the ASIS throughout the pulse, the geometry passes; if it slides below, the design needs revision. An FE human model would add tissue deformation and possible abdominal injury prediction, but for a belt-fit question the facet model already provides the decisive information at a fraction of the cost.

In summary, the three families form a deliberate hierarchy. Ellipsoid multibody models answer kinematic and restraint questions quickly; facet models add geometric realism for contact-dominated problems; FE human models provide direct tissue-level injury prediction when the mechanism itself is the subject of study. Matching the model family to the question — rather than to the available computing power — is the single most important modeling decision in MADYMO human body simulation.