Joint Modelling Strategy: Choosing the Right Level of Fidelity

The Spectrum of Modelling Fidelity

The modelling of a structural joint spans a spectrum of fidelity, from the simplest hand calculation to the most detailed explicit contact model with threaded fasteners. At the lowest fidelity, a hand calculation — a formula for the bolt shear capacity, the bearing strength, the net-section strength — gives a quick, transparent answer for the nominal capacity. At the next level, an analytical or empirical method — the Huth fastener flexibility equation, the weld-group formula, the Volkersen shear stress equation — gives a more detailed answer for the load distribution or the stress state. At the FE level, the spectrum continues: a spring or connector model (each fastener is a spring with the fastener stiffness), a beam fastener model (each fastener is a beam with the correct diameter and offset), a shell model (the plates are shells, the fasteners are beams or connectors), a solid model (the plates and the fasteners are solids, with contact at the interfaces), an explicit contact model (the contact, the friction, the clearance and the preload are modelled explicitly), a detailed thread model (the individual threads are modelled, with contact at the thread surfaces), and a local submodel (a detailed model of a single fastener, driven by the displacements from a global model). Each level has a different cost, a different output and a different accuracy. The engineer must select the level that answers the engineering question at the minimum cost.

The modelling spectrum spans from hand calculations (level 1, transparent, nominal capacity) to explicit thread models (level 7, very expensive, local stress). Each level has a different cost, output and accuracy. Select the level that answers the engineering question at the minimum cost — the most detailed model is not always the best.

Hand Calculations and Analytical Methods

Hand calculations and analytical methods are the foundation of joint analysis. A hand calculation — the bolt shear capacity (F = τ_allow × A_shear), the bearing strength (σ_br = F / (d × t)), the net-section strength (σ_net = F / ((w − d) × t)) — gives the nominal capacity of the joint in minutes, with full transparency. The hand calculation is verifiable: the engineer can check the formula, the inputs and the result, and a reviewer can repeat the calculation. The hand calculation is also the reference for the FE: the FE result should be in the same range as the hand calc, and any discrepancy should be explained. Analytical and empirical methods — the Huth fastener flexibility equation (Article 11), the weld-group formulas (Article 24), the Volkersen shear stress equation (Article 19), the classical interference fit pressure (Article 17) — extend the hand calculation to the load distribution and the stress state. These methods are based on mechanics (elasticity, plasticity, contact) with simplifying assumptions, and they give a closed-form or a spreadsheet result. They are transparent, quick and verifiable, and they are the first analysis for any joint. The limitation is that they assume simplified geometry (a single fastener, a uniform plate, a simple weld) and they do not capture the full complexity of a real joint (multiple fasteners, varying geometry, complex loading). For complex joints, the analytical methods provide the first estimate and the verification reference, and the FE provides the detailed analysis.

Hand calculations and analytical methods are the foundation — transparent, quick, verifiable. They give the nominal capacity and the first estimate of the load distribution. They are the reference for the FE — the FE result should be in the same range. For complex joints, they provide the first estimate and the verification reference; the FE provides the detailed analysis.

Springs, Connectors and Beam Fasteners

In FE analysis, the simplest representation of a fastener is a spring or a connector — a single element with the fastener stiffness. A spring model represents each fastener as a spring with the axial and shear stiffness (the fastener flexibility from the Huth equation or a simple shear stiffness). The spring model captures the load distribution among the fasteners (the stiffness determines the load sharing) and the global stiffness of the joint. It does not capture the fastener bending, the bearing deformation, or the local stress. A connector element (Abaqus connector, ANSYS joint element) is a more capable version of the spring — it can have stiffness in all six degrees of freedom (axial, two shear, two bending, torsion), and it can include nonlinear behaviour (plasticity, failure). A beam fastener model represents the fastener as a beam element with the correct diameter, the correct offset (the distance from the plate mid-surface to the fastener centre), and the correct material. The beam model captures the fastener bending (the beam bends under the eccentric load) and the local bearing deformation (the beam represents the fastener, and the contact with the plate is implicit). The beam model is the standard representation for fasteners in shell models — the plates are shells, the fasteners are beams, and the connection is at the shell mid-surface with an offset. The beam model gives the fastener forces (axial, shear, bending) that are used for the fastener strength check. The spring, connector and beam models are all "structural" representations — they model the fastener as a structural element, not as a solid. They are efficient (a few elements per fastener) and they give the forces and the load distribution. They do not give the local stress at the hole or the contact pressure — for that, a solid model is needed.

Springs, connectors and beam fasteners are the standard FE representations for fasteners in global models. They capture the load distribution, the global stiffness and (for beams) the bending and the offset. They give the fastener forces for the strength check. They do not give the local stress or the contact pressure — for that, a solid model is needed.

Shell and Solid Models

The shell model and the solid model are the two main structural modelling approaches for the plates in a joint. A shell model represents the plates with shell elements (2D elements with membrane and bending stiffness), with the fasteners modelled as beams or connectors. The shell model is efficient (a few thousand elements for a joint) and it captures the plate stress, the fastener forces and the load path. The shell model is the standard approach for large built-up structures (aircraft wings, ship hulls, bridge decks) where the plates are thin relative to the overall dimensions and the shell assumption is valid. The limitation is that the shell model does not capture the through-thickness stress — the bearing stress at the hole, the pull-through stress, the through-thickness compression from the preload. A solid model represents the plates and the fasteners with solid (3D) elements, with the contact at the interfaces modelled explicitly. The solid model captures the full three-dimensional stress state: the bearing stress at the hole, the contact pressure, the through-thickness stress, the local bending. The solid model is required for the detailed assessment of the local stress — the notch stress at the weld toe, the contact pressure at the pin-bore interface, the bearing stress distribution through the thickness. The solid model is expensive (tens of thousands to millions of elements for a single joint) and it is mesh-sensitive at the stress concentrations. The solid model is used for the critical joint detail, not for the entire structure. The submodelling approach (Article 23) combines the two: a shell global model for the load path, and a solid submodel for the critical joint detail.

A shell model (shells + beam fasteners) is efficient and captures the plate stress, fastener forces and load path — the standard for large built-up structures. It does not capture the through-thickness stress. A solid model captures the full 3D stress state — the bearing, the contact, the through-thickness — but is expensive and mesh-sensitive. Use the submodelling approach: shell global + solid local.

Explicit Contact and Detailed Threads

The highest fidelity in joint modelling is the explicit contact model with detailed thread geometry. In this model, the fastener and the plates are solids, the interfaces (the thread surfaces, the bearing surface, the clamped surfaces) are modelled with contact (with friction), the clearance between the fastener and the hole is modelled, and the preload is applied (by a bolt preload section, by a temperature change, or by a prescribed displacement). The model captures the full mechanical behaviour: the load transfer through the threads (the load distribution among the threads, the stress concentration at the first engaged thread), the bearing contact at the hole (the contact pressure, the edge loading), the clamping at the interface (the pressure distribution, the separation), and the friction at the interface (the stick-slip, the microslip). The explicit contact model is the most accurate for the local stress state, but it is also the most expensive (a single fastener can be hundreds of thousands of elements) and the most uncertain (the friction coefficient, the clearance, the preload are all uncertain, and the result can be sensitive to these inputs). The detailed thread model is used for the most critical fasteners — aerospace structural fasteners, engine mounting bolts, pressure vessel flange bolts — where the local stress at the thread is the governing quantity. For most joints, the detailed thread model is not necessary — the beam fastener or the solid model without threads is sufficient. The engineer should reserve the detailed thread model for the fastener where the thread stress is the explicit engineering question.

The explicit contact model with detailed threads is the highest fidelity — it captures the thread load distribution, the bearing contact, the clamping and the friction. It is the most accurate for local stress but the most expensive and the most uncertain (friction, clearance, preload). Use it for the most critical fasteners where the thread stress is the explicit question. For most joints, a beam or solid model without threads is sufficient.

The Trade-Offs: Cost, Transparency, Verification, Accuracy

The choice of modelling fidelity involves a set of trade-offs. Cost: the runtime, the modelling effort and the post-processing effort increase with fidelity. A hand calculation takes minutes; a spring model takes hours; a solid model takes days; a detailed thread model takes weeks. The cost must be justified by the value of the answer — a detailed thread model is justified for a critical aerospace fastener, not for a routine bracket attachment. Transparency: the ability to verify the result by hand. A hand calculation is fully transparent; a spring model is moderately transparent (the element forces can be checked); a solid model is opaque (the stress field is complex and mesh-dependent, and it cannot be checked by hand without a submodel). The transparency decreases with fidelity, and the verification must shift from hand calc comparison to mesh convergence and test correlation. Verification: the ability to check the model against a simpler model or a test. A hand calculation verifies the spring model; a spring model verifies the solid model; a solid model is verified by mesh convergence and test correlation. The verification path must be planned — a model without a verification path is not defensible. Local accuracy: the accuracy of the local stress. A hand calculation gives the nominal stress; a beam model gives the fastener force; a solid model gives the local stress at the stress concentration. The local accuracy increases with fidelity. Global stiffness: the accuracy of the load path. A spring model captures the global stiffness if the fastener stiffness is correct; a solid model captures it if the contact and the friction are correct. The global stiffness is not necessarily better with a higher fidelity — a well-calibrated spring model can be as accurate as a solid model for the global stiffness, at a fraction of the cost. Load recovery: the ability to extract the joint forces. A beam model gives the fastener forces directly; a solid model requires post-processing (section forces, integration). The load recovery is easier at the lower fidelity. Sensitivity and uncertainty: the sensitivity of the result to the uncertain inputs (friction, clearance, preload). A higher-fidelity model with explicit contact is more sensitive to the friction and the clearance than a simpler model — the uncertain inputs have a larger effect on the result. The sensitivity must be assessed by a sensitivity study, not by a single run.

The trade-offs are: cost (increases with fidelity), transparency (decreases with fidelity), verification (shifts from hand calc to mesh convergence to test correlation), local accuracy (increases with fidelity), global stiffness (not necessarily better at higher fidelity), load recovery (easier at lower fidelity), sensitivity (higher-fidelity contact models are more sensitive to uncertain inputs). Plan the verification path and the sensitivity study.

The Hierarchical Approach

The hierarchical approach is the strategy of using multiple fidelity levels in sequence, with each level verifying the previous and focusing the effort on the critical location. The approach starts with a hand calculation (level 1): the nominal capacity of the joint, the fastener forces from a simple load distribution, the transparent check. The hand calculation identifies the critical fastener (the highest-loaded fastener) and the critical mode (the lowest margin). The next level is a global FE model with beam fasteners (level 4–5): the load distribution among the fasteners, the global stiffness, the load path. The global model verifies the hand calculation (the fastener forces should be in the same range) and refines the load distribution (the stiffness effects, the bypass load, the load redistribution). The global model identifies the critical fastener location for the submodel. The final level is a local submodel (level 8): a solid model of the critical fastener, driven by the displacements from the global model. The submodel gives the local stress at the critical fastener (the bearing stress, the contact pressure, the through-thickness stress) and the local failure mode. The submodel is verified by mesh convergence and by comparison with the hand calculation (the nominal stress should be in the same range). The hierarchical approach focuses the effort: the hand calculation is quick and covers the whole joint; the global model is moderate and covers the whole structure; the submodel is expensive but covers only the critical location. The approach also provides verification at each level — the hand calc verifies the global model, the global model verifies the submodel. The hierarchical approach is the recommended strategy for most joint analyses — it is efficient, verifiable and defensible.

The hierarchical approach uses multiple fidelity levels in sequence: hand calculation (nominal capacity, critical fastener) → global FE with beam fasteners (load distribution, load path) → local solid submodel (local stress at the critical fastener). Each level verifies the previous and focuses the effort on the critical location. It is efficient, verifiable and defensible — the recommended strategy for most joint analyses.

Matching Fidelity to the Question

The ultimate criterion for the fidelity selection is the engineering question. The question determines what output is needed, and the output determines the fidelity. A question about the nominal capacity (is the bolt strong enough?) is answered by a hand calculation. A question about the load distribution (how much load does each fastener carry?) is answered by a beam fastener model. A question about the global stiffness (what is the joint deflection?) is answered by a shell model with beam fasteners. A question about the local stress (what is the bearing stress at the hole?) is answered by a solid model. A question about the fatigue life (what is the S-N life at the weld toe?) is answered by a hot-spot or notch stress model. A question about the fretting (what is the slip amplitude at the interface?) is answered by an explicit contact model. A question about the thread stress (what is the stress at the first engaged thread?) is answered by a detailed thread model. The engineer should define the question first, identify the required output, and select the fidelity that produces that output at the minimum cost. A model that is more detailed than the question requires is not more accurate — it is less efficient and less transparent. A model that is less detailed than the question requires gives the wrong answer. The fidelity selection is a engineering judgement, not a default — the engineer should justify the selection in the analysis plan and document the reasoning.

Match the fidelity to the engineering question. Define the question, identify the required output, and select the fidelity that produces that output at the minimum cost. A model more detailed than required is less efficient and less transparent, not more accurate. A model less detailed than required gives the wrong answer. The fidelity selection is an engineering judgement — justify it in the analysis plan.

Key takeaways

• The modelling fidelity must match the engineering question. A hand calculation answers a load-capacity question quickly and transparently. A beam fastener model answers a load-distribution question. A solid submodel answers a local-stress question. Using a higher fidelity than needed wastes time and obscures the result; using a lower fidelity than needed gives the wrong answer.
• The trade-offs are: cost (runtime, modelling effort), transparency (can the result be verified by hand?), verification (can the model be checked against a simpler model or a test?), local accuracy (is the local stress correct?), global stiffness (is the load path correct?), load recovery (can the joint forces be extracted?), sensitivity (does the result change with the parameters?), and uncertainty (how much does the unknown — friction, clearance, preload — affect the result?).
• A hierarchical approach is often the best: start with a hand calculation (for the nominal capacity and the transparent check), then a global model with beam fasteners (for the load distribution), then a local submodel (for the local stress at the critical fastener). Each level verifies the previous, and the effort is focused on the critical location.
• The most detailed model is not always the best model. A detailed model with uncertain inputs (friction, clearance, preload) can be less accurate than a simpler model with conservative assumptions. The uncertainty in the inputs must be considered in the fidelity selection — a model that is sensitive to an uncertain input needs a sensitivity study, not just a single "best estimate" run.