Punching Shear at Slab–Column Connections: A Practical Walkthrough Under CSA A23.3 and ACI 318 (and the Corner-Column Trap)

Punching shear is the failure mode that keeps flat-plate designers up at night. There is no beam to warn you, no gradual deflection — just a column pushing a cone of concrete straight through the slab. Both CSA A23.3-14 (Cl. 13.3) and ACI 318-19 (§22.6, §8.4) handle it with a deceptively compact set of clauses, but the details are where good designs are won or lost.

This post walks through how the two-way shear check actually works in both codes, then digs into a result that surprises a lot of engineers: corner columns are often penalized far more than they need to be — and CSA gives you a perfectly legal way out that ACI does not.

The check, in five steps

The skeleton of the check is the same on both sides of the border. For a column transferring a shear V_u and unbalanced moments Mux, Muy to a slab of effective depth d:

1. Critical perimeter. Draw the critical section at d/2 from the column face (identical in CSA Cl. 13.3.3 and ACI 22.6.4.1). Its length b0 and area A_c = b0·d set the baseline. Sides close to a slab edge get truncated.

2. Section properties. Find the centroid of that perimeter and its second moments (Jx, Jy, and for unsymmetrical shapes Jxy). For edge and corner columns the centroid shifts away from the free edge — which matters, because the shear V_u then acts at an eccentricity and adds a V·e moment.

3. Fraction carried by shear. Only part of the unbalanced moment is transferred by eccentric shear; the rest goes through flexure. Both codes use the same expression — CSA Eq. 13.8 and ACI 8.4.2.3.2/8.4.4.2.2:

γv = 1 − γf ,   γf = 1 / (1 + (2/3)·√(b1/b2))

4. Peak shear stress. Combine the uniform direct stress with the linearly varying moment stress and scan the perimeter for the maximum:

vf = V_u/A_c − γx·Mx·y/Jx + γy·My·x/Jy

5. Concrete resistance. This is where the two codes diverge in form (but not in spirit). Without shear reinforcement, v_c is the smallest of three expressions:

CSA A23.3-14 (Cl. 13.3.4.1), with φc = 0.65 baked in:

v_c = min{ (1 + 2/βc)·0.19·λ·φc·√f'c ,
           (αs·d/b0 + 0.19)·λ·φc·√f'c ,
           0.38·λ·φc·√f'c }

where αs = 4 / 3 / 2 for interior / edge / corner and βc is the column aspect ratio.

ACI 318-19 (Table 22.6.5.2), nominal stress (apply φ = 0.75 separately):

v_c = min{ 0.33·λs·λ·√f'c ,
           0.17·(1 + 2/β)·λs·λ·√f'c ,
           0.083·(2 + αs·d/b0)·λs·λ·√f'c }   [MPa]

where αs = 40 / 30 / 20 for interior / edge / corner, β is the column aspect ratio, and λs is the new size-effect factor (see below).

If vf > v_c (or vu > φv_c in ACI terms), you need shear reinforcement; if vf exceeds the absolute ceiling, the section itself is inadequate.

So far, so textbook. Now the interesting part.

The corner-column trap

Corner columns get the worst of every term: the smallest perimeter (αs = 2 in CSA, 20 in ACI), two truncated sides, and a critical-section centroid that sits well inside the column. When there's meaningful unbalanced moment, the γ·M term in step 4 spikes the peak stress — and the two-way check can demand a wall of studrails, or fail outright.

Here's a real example. Take a 500 × 350 mm corner column, 200 mm effective depth, f'c = 30 MPa, with V_u = 160 kN, Mux = 180, Muy = 60 kN·m. Run the standard two-way eccentric-shear check and you get a demand/capacity ratio of about 2.15 — nominally a heavy reinforcement case.

But CSA A23.3-14 Clause 13.3.6.2 lets you check a corner column with a one-way shear model instead:

V_c = β·λ·φc·√f'c·b0·d     (β per Cl. 11.3.6.2/11.3.6.3)

and states plainly that "corner columns meeting the requirements of this Clause shall be deemed to have satisfied Clauses 13.3.4 and 13.3.5." It even lets you extend the critical section into a slab cantilever by up to d.

Run the same column through the one-way model and V_c ≈ 169.3 kN. Against the 160 kN demand, the ratio is 0.94 — it passes with no shear reinforcement at all.

Method	Demand / Capacity	Outcome
Two-way eccentric shear (γx = 0.40)	2.15	Heavy studs / NG
Two-way, γx computed (Eq. 13.8)	2.13	Heavy studs / NG
One-way model (CSA Cl. 13.3.6.2)	0.94	Passes, no reinforcement

Why the enormous gap? The one-way model drops the γ·M stress amplification entirely — it's a flat V_u/(b0·d) check against a β-based resistance. When the unbalanced moment is large relative to direct shear, that amplification is exactly what's killing the two-way result.

For US designers the picture is more nuanced — and easy to get wrong. ACI 318 itself, the building code, has no corner one-way clause like CSA Cl. 13.3.6.2; within the code the two-way eccentric model governs corners. But the companion guide ACI 352.1R-11 (§5.2.1.2(b)) does give a corner shortcut: a corner connection may be taken as adequate if the factored direct shear does not exceed 0.5·φVc (0.75·φVc at an edge). It's a different mechanism — a direct-shear cap that ignores the unbalanced moment, rather than CSA's one-way capacity model — and it's committee guidance, not a mandatory code clause. For the corner above (V_u = 160 kN), that ACI 352.1R limit works out to about 165 kN, so it too would be deemed adequate. So the honest distinction isn't "Canada passes, US fails" — it's that CSA codifies the corner relief (a one-way capacity check), while ACI leaves it to a guide with a more restrictive, moment-blind trigger.

The catch: even in CSA, the one-way model is not always more favourable. For the same column at higher direct shear and low moment (V_u = 260 kN, Mux = 60), the two-way check governs at D/C ≈ 1.00, while the one-way comes out more conservative at 1.54. The takeaway is simple but easy to miss:

Under CSA, evaluate both the two-way and the one-way model (Cl. 13.3.6.2) and take the favourable one — the code permits it. Under ACI 318 the two-way model governs in the code, but check ACI 352.1R §5.2.1.2(b) before committing to studs at a lightly-sheared corner.

A code subtlety worth knowing: the γf increase is ACI-only

If you also work to ACI 318, watch this one. ACI 318 (§8.4.2.3.4) lets you increase the fraction of moment carried by flexure, γf, by up to 25% (capped at 1.0) for interior connections when V_u ≤ 0.4·φVc and the slab steel is ductile enough — which lowers the shear-transferred fraction and can drop your peak stress by ~30%.

CSA A23.3 has no equivalent provision. CSA gives a single γv from Eq. 13.8 with no adjustment. So if you're designing to CSA, that capacity simply isn't available — and you shouldn't import the ACI trick into a CSA calc. Conversely, ACI designers who never apply it are leaving capacity on the table. It's a clean example of a provision that is not portable between the two codes.

Where studrails come in

When the concrete alone isn't enough — which, for interior columns under real gravity-plus-lateral demand, is most of the time — headed shear stud reinforcement is the industry's go-to in both countries. Jordahl's DECON® Studrails® are the reference product here: rails of headed studs with a 10:1 head-to-stem area ratio that anchor without slip and develop yield over the full stud length. The technology grew out of University of Calgary research in the 1970s and has been the preferred alternative to closed stirrups in North American flat plates ever since.

The design payoff is concrete in both codes — headed studs unlock materially higher allowable stresses than stirrups:

	CSA A23.3-14	ACI 318-19
In-zone concrete v_c, stirrups	0.19·λ·φc·√f'c	0.17·λ·√f'c
In-zone concrete v_c, headed studs	0.28·λ·φc·√f'c	0.25·λ·√f'c
Max stress, stirrups	0.55·λ·φc·√f'c	0.50·λ·√f'c
Max stress, headed studs	0.75·λ·φc·√f'c	0.66·λ·√f'c

(CSA bakes φc = 0.65 into the stress; ACI applies φ = 0.75 to the nominal value separately.) Either way, the message is the same: thinner slabs, fewer failures, cleaner rebar coordination.

The "Decon method," as practitioners use the term, is just this code-based workflow as implemented in Decon/Jordahl's design tooling — the de facto industry standard for sizing studrail layouts at slab–column connections under both CSA and ACI.

Worked example: a corner that needs studrails (full CSA calculation)

The trap case above was moment-dominated, so the one-way model rescued it. Now take the same 500 × 350 mm corner (250 mm slab, d = 200 mm, f'c = 30 MPa) but shear-dominated: V_u = 320 kN, Mux = 50, Muy = 20 kN·m. Here the one-way model doesn't help (V_u/V_c = 320/169 = 1.9), so the connection genuinely needs reinforcement. Full CSA check, end to end.

Step 1 — Critical section and demand

offset_x = min(d/2, r_x_n) = min(100, 80) = 80 mm
offset_y = min(d/2, r_y_p) = min(100, 90) = 90 mm
b0 = (500 + 80 + 100) + (350 + 90 + 100) = 1220 mm
A_c = 1220 × 200 = 244 000 mm²

Section properties of the L-shaped perimeter (line integral, about its own centroid):

ex = +160.5 mm ,  ey = −155.5 mm
Ix = 7.01×10⁹ ,  Iy = 1.22×10¹⁰ ,  Ixy = 5.53×10⁹ mm⁴
θ (to principal axes) = −57.6°   →   lx = 456 mm ,  ly = 864 mm

Moment fractions for the corner: γx = 0.40, γy = 0.28. Resolving Mux, Muy onto the principal axes, adding the V·e shift, and scanning the perimeter:

v_u_max = 1.848 MPa

Step 2 — Concrete resistance (Cl. 13.3.4)

βc = 500/350 = 1.43 ,  αs = 2 ,  √f'c = √30 = 5.48 (≤ 8 OK)
Eq.13.5: (1 + 2/1.43)·0.19·0.65·5.48 = 1.623
Eq.13.6: (2·200/1220 + 0.19)·0.65·5.48 = 1.844
Eq.13.7: 0.38·0.65·5.48           = 1.353   ← governs
v_c = 1.353 MPa ,   v_max (with studs) = 0.75·0.65·5.48 = 2.670 MPa

v_u_max = 1.848 > v_c = 1.353 → reinforcement required; and 1.848 < 2.670 → studs are feasible. (The one-way model gives V_c = 169 kN < V_u = 320 kN, so it does not rescue this shear-dominated corner — two-way governs.)

Step 3 — Studrail design (Cl. 13.3.8)

Inside the reinforced zone the concrete term rises to 0.28·λ·φc·√f'c:

v_c,zone = 0.28·0.65·5.48 = 0.997 MPa
v_s      = v_u_max − v_c,zone = 1.848 − 0.997 = 0.851 MPa
A_v/s    = v_s·b0 / (φs·f_yv) = 0.851·1220 / (0.85·300) = 4.07 mm²/mm

(φs = 0.85; f_yv = 300 MPa for the studs.) Stud size follows the slab thickness — h = 250 mm → 12.7 mm (½″) studs, area ≈ 127 mm² each. The corner takes 5 rails (2 + 3 around the two reinforced faces), so one peripheral line provides A_v = 5 × 127 = 634 mm²:

s = A_v / (A_v/s) = 634 / 4.07 = 156 mm

Spacing limits (Cl. 13.3.8.6): since v_u_max = 1.848 < 0.56·φc·√f'c = 1.99 MPa, the limit is s ≤ 0.75d = 150 mm, so use s = 150 mm. First row s_0 ≤ 0.4d = 80 mm. Stud height OAH = h − covers = 250 − 25 − 25 = 200 mm. Re-check at s = 150:

v_s,prov = φs·A_v·f_yv / (b0·s) = 0.85·634·300 / (1220·150) = 0.883 MPa
v_r      = v_c,zone + v_s,prov = 0.997 + 0.883 = 1.880 MPa ≥ 1.848 ✓

Step 4 — Extent of the reinforced zone (Cl. 13.3.7.4 / 13.3.8.5)

Extend the rails outward until the factored stress at the outer critical section (d/2 beyond the last stud) drops to v_c,out = 0.19·λ·φc·√f'c = 0.676 MPa. Iterating the outer L-perimeter row by row until the check passes:

required outer perimeter  b0,out ≈ 2 370 mm   (vs 1 220 mm at the column)
rail length needed        ≈ 710 mm  →  6 studs per rail  (first at 80 mm, then 150 mm)

The moment term on the enlarged section is what pushes the extent past a naive direct-stress estimate — exactly the iteration the Decon/Jordahl software automates.

Result: 5 studrails · 12.7 mm studs · 6 per rail · s_0 = 80 mm · s = 150 mm · rail length ≈ 710 mm · OAH = 200 mm.

Takeaways

• The two-way check is five clean steps and largely shared between CSA and ACI — same critical section, same γf moment-transfer split — but the centroid shift and the γ·M term are where edge and corner columns get interesting.
• Corner columns under CSA: always run the one-way model (Cl. 13.3.6.2) alongside the two-way check. When moment dominates it can turn a "heavy reinforcement" result into a clean pass. ACI 318 has no in-code equivalent — check the ACI 352.1R guide (§5.2.1.2(b)) instead.
• When shear dominates, the one-way route doesn't help and the corner genuinely needs studs — size them per Cl. 13.3.8 as in the worked example above.
• The 25% γf increase is ACI-only — don't apply it in a CSA design.
• When reinforcement is needed, headed studrails (Jordahl DECON) unlock materially higher allowable stresses than stirrups under both codes.

*Designing flat plates and want the punching-shear check — interior, edge, and corner, to CSA or ACI, with the one-way corner alternative built in — done in seconds instead of spreadsheets? That's what we build at Strucwise.