Chapter 2 — LRFD Design Philosophy & ASCE 7-22 Load Combinations

Chapter focus. Every structural engineer's job — whether the project is a warehouse, a hospital, or a highway bridge — is to make sure the structure carries every load it will ever see with a defensible margin of safety, without over-spending on steel. Designing means balancing five competing considerations at once: safety (nobody gets hurt when the crowd, wind, or earthquake arrives), serviceability (no cracked drywall, no bouncing floors, no doors that jam), economy (the lightest section that still works), constructability (fabricators and erectors can build it), and durability (it survives 50–100 years of weather and use). Safety governs — collapses like Hyatt Regency (Ch 1) are the reason design codes exist and the reason licensure is required to sign drawings.

Two code-legal methods produce those safe designs from the same AISC steel spec:

• LRFD (Load and Resistance Factor Design) — factored strength method. Increase the loads by load factors γ (e.g. 1.2 D + 1.6 L), reduce the calculated capacity by a resistance factor φ ≤ 1, and check Ru ≤ φRn. Each factor is calibrated separately to real statistical data (mill reports, occupancy surveys, weather records).
• ASD (Allowable Stress Design) — service-load method. Keep the loads unfactored (D + L), divide the capacity by a single overall safety factor Ω ≥ 1 (typically 1.5–2.0), and check Ra ≤ Rn/Ω. Historically the default; still permitted for hand checks and retrofits.

This chapter builds those definitions into a full design method: probabilistic basis, the LRFD inequality, ASCE 7-22 load combinations, and the φ table you will reference in Chapters 4–19.

1. Why LRFD, Not ASD?

AISC 360-22 publishes both methods side by side, and both are code-legal. This course teaches LRFD because every modern U.S. building, bridge, and industrial project is designed that way. Here is the direct comparison.

ASD vs. LRFD design inequalities

(1)

$$R_a\le R_n/\Omega \text{(ASD — service loads, one global safety factor }\Omega \text{)}$$

$$R_u\le \varphi R_n\text{(LRFD — factored loads, split load and resistance factors)}$$

Issue	ASD (Allowable Stress)	LRFD (Load & Resistance Factor)
How is safety applied?	One number Ω hides all uncertainty in the capacity side.	γ scales up each type of load (D vs. L vs. W); φ scales down each type of resistance (yielding vs. rupture vs. bolt).
Does it recognize that dead load is more predictable than live load?	No — D and L are added at their service values, treated equally.	Yes — γD = 1.2 (small, D is well-known) vs. γL = 1.6 (larger, L is more variable).
Does it recognize a brittle failure needs more safety than a ductile one?	No — same Ω regardless of failure mode.	Yes — φyielding = 0.90 (ductile) vs. φrupture = 0.75 (brittle).
Is it calibrated to a target reliability?	Empirical — inherited from 1920s working-stress practice.	Yes — β ≈ 3.0 (members), β ≈ 4.5 (connections). Every φ and γ is back-solved from data.
Economy on live-load-dominated members?	Conservative — treats highly variable L the same as reliable D.	Sharper — larger γL is offset by removing hidden conservatism, so members are typically 5–15% lighter.
Consistency with ACI 318 and ASCE 7?	Poor — ACI abandoned working-stress design in 1971. Load combinations differ from ASCE 7 §2.3.	Full — LRFD combinations in ASCE 7 §2.3 match ACI 318 strength design. One load path for the whole building.
Seismic and wind design?	Requires special ASD combinations (§2.4) that reintroduce factors on E and W anyway.	Native — 1.0W and 1.0E already at strength-level in the LRFD combos.
Which one is the standard in industry today?	Used for quick hand checks, retrofits of pre-2005 buildings, and some petrochemical clients.	The default for essentially all new commercial, institutional, and infrastructure work in the U.S.

Bridge between methods. On the same section, the two safety factors are related by $\Omega \approx 1.5/\varphi$. Plug in the two most common resistance factors:

Ω from φ — bridge equation

(2)

$$\Omega =\frac{1.5}{\varphi }$$

$${\Omega }_{yielding}=\frac{1.5}{0.90}=1.67$$

$${\Omega }_{rupture}=\frac{1.5}{0.75}=2.00$$

When the live-to-dead ratio is around 3 (a typical office), LRFD and ASD pick essentially the same W-shape — but LRFD tells you why that shape works.

Bottom line. LRFD is a rational, probabilistically calibrated, code-aligned method. ASD is a legacy method preserved for continuity. If you can only master one — master LRFD.

2. Probabilistic Basis

LRFD calibrates φ and γ so the failure-region overlap area gives a target reliability index β ≈ 3.0 (members) and β ≈ 4.5 (connections).

Plain-language version. Nothing in a real building is a single fixed number. Think of a typical office floor beam:

• The load Q is uncertain. The dead load depends on how thick the concrete slab actually gets poured, what ceiling and MEP the tenant installs, and whether partitions are moved. The live load depends on the occupants — a light office one year could become a filing archive the next. We don't know these numbers exactly; we know a range. Statisticians represent that range as the blue "Load Q" bell curve.
• The resistance R is also uncertain. Mill-rolled A992 steel is specified as F_y = 50 ksi, but a coupon test typically measures 55–58 ksi. The web thickness of a W18×35 is nominally 0.300 in but ships anywhere from 0.290 to 0.315 in. Fabrication tolerances shift the true capacity a few percent. That range is the green "Resistance R" bell curve.
• Failure is the overlap. A beam fails when a specific load draw is bigger than that beam's specific capacity — i.e., a value from the right tail of Q lands past a value in the left tail of R. That is the shaded overlap region in the figure.

What LRFD actually does. The code picks the load factors γ (which push the design load up to the right tail of Q) and the resistance factors φ (which pull the design capacity down to the left tail of R) so the two tails almost never overlap. The remaining probability of failure is expressed as a reliability index β. Larger β = safer.

(3)

$$CalibrationTarget$$

$\beta \approx 3.0$ for members  ⇒  probability of failure $P_f=\Phi (-\beta )=\Phi (-3.0)\approx 1.35\times 10^{-3}$ — about 1 in 1,000 over the 50-year design life.

$\beta \approx 4.5$ for connections  ⇒  $P_f=\Phi (-4.5)\approx 3.4\times 10^{-6}$ — about 3 in 1,000,000.

How $P_f$ is obtained. Define the safety margin $Z=R-Q$. Under the LRFD reliability model both $R$ and $Q$ are treated as normal random variables, so $Z$ is also normal with mean ${\mu }_Z={\mu }_R-{\mu }_Q$ and standard deviation ${\sigma }_Z=\sqrt{{\sigma }_R^2+{\sigma }_Q^2}$. Failure means $Z\le 0$, i.e. resistance is exhausted by the load effect: $$P_f=P(Z\le 0)=\Phi \left(\frac{0-{\mu }_Z}{{\sigma }_Z}\right)=\Phi (-\beta ),\beta =\frac{{\mu }_R-{\mu }_Q}{\sqrt{{\sigma }_R^2+{\sigma }_Q^2}}$$ where $\Phi (\cdot )$ is the standard normal CDF. AISC back-solves the $\varphi$ and $\gamma$ factors so that the resulting $\beta$ hits the target above for every limit state in the Specification.

Why connections get a higher β (and therefore a smaller φ). A single beam has redundancy — the slab, adjacent beams, and neighboring bays can help redistribute load if it starts to yield. A connection has no redundancy: if the bolts shear or the weld cracks, the beam falls. So AISC calibrates φ_bolt = 0.75 and φ_rupture = 0.75, versus φ_yielding = 0.90 and φ_bending = 0.90 for members. The Hyatt Regency 1981 and FIU 2018 collapses (Chapter 1) are exactly why this asymmetry exists.

3. The Fundamental LRFD Inequality

ASCE 7-22 — the standard that gives us γ (load factors) and every load combination in this chapter

Load and Resistance Factor Design (LRFD) requires that the factored demand R_u not exceed the factored capacity φR_n for every relevant limit state:

• R_u — required strength from factored load combinations (γ_DD + γ_LL + …).
• R_n — nominal strength computed from AISC provisions using nominal material and section properties.
• φ — resistance factor (≤ 1.0) that penalizes the calculated capacity for the uncertainty of the limit state.

3A. Loads and Actions

The three families of actions on a building: permanent (dead), variable (live) and environmental (wind, snow, seismic).

An action is anything that causes stress, strain or displacement in a structure. Before a single member can be sized you must know every action, its magnitude, and how likely it is to occur at the same time as the others. ASCE 7-22 is the standard that answers all three questions.

Permanent actions — dead load, D

• Self-weight of the structural frame, slab and deck; superimposed finishes, ceilings, partitions, cladding and fixed MEP equipment.
• Computed from geometry and unit weights: steel 490 pcf, normal-weight concrete 150 pcf, lightweight concrete 110–120 pcf.
• Well known, so its load factor is small (γ_D = 1.2) — and deliberately reduced to 0.9 in the uplift combinations where dead load is the stabilising action.

Variable actions — live load, L and L_r

• Occupancy loads from ASCE 7-22 Table 4.3-1: offices 50 psf (plus 15 psf partition allowance), corridors above the first floor 80 psf, classrooms 40 psf, assembly with fixed seats 60 psf, light storage 125 psf.
• Roof live load L_r covers maintenance access, typically 20 psf reduced by the R₁R₂ factors of §4.8.2.
• Live-load reduction (§4.7): L = L_o(0.25 + 15/√(K_LLA_T)). The larger the tributary area, the less likely it is to be fully loaded everywhere at once. Never reduce below 0.50L_o for members supporting one floor, or 0.40L_o for multiple floors.
• Highly variable, so γ_L = 1.6 — a third larger than the dead-load factor.

Environmental actions — S, W, E

• Snow, S (Ch. 7): p_f = 0.7 C_e C_t I_s p_g, then drift and sliding surcharges at parapets and roof steps — drifts, not the flat load, cause most snow failures.
• Wind, W (Ch. 26–30): q_z = 0.00256 K_zK_ztK_dK_eV², applied as pressure on the main wind-force resisting system and, at higher local values, on components and cladding. Wind governs uplift on light roofs and drift on tall frames.
• Seismic, E (Ch. 11–12): V = C_sW with C_s = S_DS/(R/I_e). Note the inertial nature: seismic demand is proportional to mass, so making a building heavier makes the earthquake worse.
• Both W and E enter the LRFD combinations at a factor of 1.0 because their mapped values are already strength-level events.

Other actions you must not forget

• Rain, R: ponding on flat roofs is a progressive instability — deflection admits more water, which increases deflection.
• Lateral earth and fluid pressure, H and F: basement and retaining walls.
• Restrained thermal, shrinkage and settlement effects, T: only structurally significant when movement is restrained.
• Construction and erection loads: often the worst case a member ever sees, before the diaphragm and bracing exist.

3B. Design Philosophies and Limit States

Strength limit states protect life; serviceability limit states protect function, finishes and comfort.

A limit state is a condition beyond which the structure no longer fulfils an intended function. Design consists of identifying every applicable limit state and demonstrating that the demand stays on the safe side of each one.

• Strength (ultimate) limit states — yielding, rupture, buckling, sliding, overturning, connection failure. Checked with factored loads: R_u ≤ φR_n. Exceeding one risks lives, so the margin is large.
• Serviceability limit states — deflection (L/360 live, L/240 total), lateral drift (H/400 to H/500 under wind), floor vibration (AISC Design Guide 11), and permanent cracking or damage to finishes. Checked with unfactored service loads, because these are everyday conditions.

Do not mix the two. Applying factored loads to a deflection check produces an absurdly heavy beam; applying service loads to a strength check is unsafe.

3C. Structural Idealization

From a real bay to an analysable model: geometry becomes a line, the floor becomes w, and the connections become supports.

No one analyses a building as it is actually built. Engineers replace it with an idealized model that is simple enough to solve and conservative enough to trust. Four decisions define that model:

1. Geometry. Members become lines along their centroidal axes; spans are measured centre-to-centre of supports.
2. Supports and connections. A shear tab is idealized as a pin, a fully welded moment connection as fixed, and a bearing seat as a roller. Real joints are partially restrained — the idealization must err toward the conservative demand.
3. Loads. Surface pressures become a uniform line load w over the tributary width, and beam reactions become point loads on the girder.
4. Material behaviour. Linear-elastic analysis for demands, with plastic section capacity for resistance — the standard combination in AISC 360-22.

For the simply supported case in the figure the resulting demands are the two expressions used repeatedly throughout this course:

4. ASCE 7-22 LRFD Combinations (§2.3.1)

5. Common Resistance Factors φ

6. Safety Factor vs Resistance Factor — Practical Difference

Why safety factors exist at all — the practical story. Nothing you build is ever exactly what the drawings say. Steel comes in slightly heavier or lighter than nominal. Concrete cures a bit weaker on a rainy day. Occupancy loads change when a tenant swaps a call center for a filing archive. Wind gusts arrive from directions the code did not model. Safety factors are the code's way of forcing a margin between what you compute and what could actually happen — without requiring the engineer to become a statistician for every project.

A safety factor is a single scalar > 1 that inflates the demand (or deflates the capacity). It rolls together every source of uncertainty — load variability, material variability, workmanship, analysis simplifications — into one number. Historically Allowable Stress Design (ASD) used one big factor of safety Ω (typically 1.5–2.0) applied to R_n: R_a ≤ R_n/Ω. Every uncertainty was averaged into that single Ω, even though live load is genuinely more variable than dead load and steel is more predictable than concrete.

Resistance factors are what LRFD replaced the single safety factor with. Instead of one number for the whole design, LRFD splits the safety into two families:

Why splitting matters — a practical example. A member has D = 100 kip, L = 40 kip, R_n = 220 kip. Under the old single-factor ASD approach (Ω = 1.67): capacity allowed = 220/1.67 = 132 kip > 140 kip demand? Fails. Under LRFD: R_u = 1.2·100 + 1.6·40 = 184 kip; φR_n = 0.90·220 = 198 kip > 184 kip Passes, with a defensible reliability index β. LRFD "found" capacity because the dead load piece is charged a lighter 1.2× (concrete-weight variability is small) while the live-load piece is charged a heavier 1.6× (occupancy is genuinely uncertain). ASD's one-size-fits-all Ω over-penalized the well-known dead load.

7. Design Workflow

1. Assemble unfactored loads: D, L, L_r, S, R, W, E (ASCE 7 Chapters 3–13).
2. Apply live-load reduction where KLL·A_T ≥ 400 ft² (§4.7).
3. Evaluate every combo; take the envelope R_u = max over combos.
4. Compute nominal strength R_n from AISC 360-22.
5. Verify φR_n ≥ R_u; report utilization ratio R_u/(φR_n).