Design methodology decides which safety factors apply to your analysis results and how load combinations are built. The same finite element model, under the same loads, returns a different utilization ratio depending on the method behind it.
ASD puts the entire margin into one global factor on the resistance side. LRFD splits that margin between load factors and a resistance factor. Eurocode applies partial factors on both sides of the inequality.
This article works through how these three approaches build different load combinations, why their utilization ratios are not directly comparable, and what that means for code verification of an FE model.
Where the Safety Margin Sits in ASD
Allowable Strength Design (ASD) reduces the check to a single inequality: the required strength must not exceed the nominal resistance divided by a safety factor. The notation is short: Ra ≤ Rn / Ω. The whole margin is concentrated in one number, Ω, applied to resistance.
Ω is fixed by the limit state. In AISC 360, yielding (tension, flexure, shear, buckling) carries Ω = 1.67. Rupture at the net section, fillet welds, and fasteners carry Ω = 2.00. Loads stay at service level, with no amplifying factors.
Here, D is dead load, L is live, W is wind. Combinations like D + L or 0.6D + W never push the load above service level. The 0.75 reduction factors on companion actions account for the low odds of two extremes peaking at once.
That simplicity has a cost. One factor averages the uncertainty of every load and every material in the structure at once.
LRFD: The Margin Splits Between Load and Resistance
LRFD (Load and Resistance Factor Design) builds the inequality differently. On one side sit the loads, each scaled by a factor γ. On the other sits the nominal resistance, multiplied by a resistance factor φ. The notation: Σ γi Qi ≤ φ Rn.
Load factors vary by action type. Dead load is known more precisely, so its factor is 1.2. Live load swings with occupancy, so it gets 1.6. A typical combination reads 1.2D + 1.6L + 0.5(Lr or S or R). The resistance factor φ reduces the design capacity: 0.90 for yielding, 0.75 for rupture.
The logic is probabilistic. LRFD factors are calibrated to a target reliability index and tied to the scatter of each quantity: the better-characterized dead load takes a smaller factor, the variable live load a larger one. ASD works with one averaged margin across the whole structure.
Run the same model through both methods, and the ASD vs LRFD comparison stops being a matter of terminology. AISC 360 brought both into one document, listing φ and Ω side by side for every limit state. The two sets of factors are calibrated to give the same answer at a live-to-dead load ratio of L/D = 3.
From that calibration comes an exact relationship: Ω ≈ 1.5 / φ. The numbers check out: 0.90 × 1.67 ≈ 1.5, and 0.75 × 2.00 = 1.5.
Away from L/D = 3, the two diverge. On dead-load-dominated structures (parking decks, bridge spans), ASD tends to be the more conservative of the two. On light structures carrying high live loads, LRFD comes out stricter.
Eurocode: Partial Factors on Both Loads and Resistance
The European approach in EN 1990 employs a more granular system of partial factors and multiple limit-state combinations. It is a limit state method, and its partial factors apply to both sides of the inequality.
On the action side: γG = 1.35 for permanent actions, γQ = 1.5 for variable ones. On the resistance side, EN 1993-1-1 adds separate material factors: γM0 = 1.0 for cross-section yielding, γM1 = 1.0 for instability, γM2 = 1.25 for net-section fracture (several National Annexes lower it, the UK to 1.10).
Accompanying variable actions get reduced further by the combination factors ψ. Wind takes ψ0 = 0.6. Snow below 1000 meters takes ψ0 = 0.7.
The principle matches the ASD reduction factors: two independent actions rarely peak together. Splitting the margin between γ on actions and γM on resistance is what sets Eurocode apart from the single-φ scheme of LRFD.
Each limit state gets its own set of combinations. ULS uses the full partial factors and governs strength and stability. SLS takes loads at reduced or unit factors and limits deflection, vibration, and crack width. ALS covers accidental actions.
The same model is checked against several sets of limits, and each set carries its own factors and its own ψ values.
Why Results From Different Methods Are Not Comparable
The different factor mechanics carry a direct practical limit. A 0.95 utilization under ASD and a 0.95 under LRFD describe a different actual distance to the limit, because different loads and different factors sit underneath them. Comparing those two numbers directly is wrong.
Hence, a firm rule: within one check, ASD and LRFD combinations cannot be mixed. A force taken from a service-level combination does not belong in a check that uses φ. The reverse holds too: a factored load is not measured against an ASD allowable stress.
A common post-processing error traces back to the source of the stresses. An engineer pulls stresses from FEA without confirming whether they came from factored or service loads. The number is identical. The physical meaning is not.
The Number of Load Combinations Grows Nonlinearly
A plain flat roof with four action types (permanent, variable, snow, wind) generates more than forty combinations under EN 1990: roughly fourteen for ULS plus sets for three SLS regimes. That is with one load direction and one load case.
A real structure scales that figure fast. A crane runway with wind from eight directions, several trolley positions, and multiple load patterns clears a hundred combinations before seismic or fatigue cases are even added.
Every combination has to be built with the right factors, run through the model, and checked against each limit state.
The governing combination shifts from member to member. A column is driven by one combination in buckling, a beam by another in bending, and a connection by a third in fastener shear.
No single worst combination covers the whole structure, so every member is checked against the full set.
At that volume, manual assembly becomes a source of error: a governing combination left out, a factor swapped, a spot check run only on the obviously critical cases.
One Model, a Different Verification Setup
The finite element model stays the same regardless of methodology. Geometry, mesh, boundary conditions, and base loads do not depend on whether the check runs under ASD or LRFD.
What changes is the verification setup layered on top: which combinations to build, which factors to apply, which limits to treat as allowable.
This is where automation earns its place. Once the methodology is fixed by code (Eurocode 3, AISC 360, or another project standard), generating the combinations to its rules, extracting governing loads from large sets, and recomputing utilization against the right φ, Ω, or γM all run on a fixed algorithm.
The engineer is left to interpret the result. Building hundreds of combinations by hand and cross-checking factors against tables drops out of the process.
The main payoff is less routine around the analysis and the removal of an entire class of manual transfer errors. When the model changes, the combinations and checks are rebuilt on the same rules, and no load case is left on a stale revision.
