Concrete Defect Disposition

The Forms Came Off and the Column Is Honeycombed. Write the Disposition, Not the Excuse.

Published: September 6, 2026  |  By: RHCES Engineering Team  |  13 min read

Stripping day. The last panel comes off a ground floor column, somebody photographs it, and ten minutes later a group chat has four opinions in it. Cosmetic. Render it before the consultant arrives. Break it out and recast.

None of those is an answer. The file needs a disposition: what the defect is, how deep it goes, whether it matters structurally, what the repair is, who accepts it. The engineering is in the classification, not the mortar.

Why this trips people up

Honeycombing usually looks worse than it is, and occasionally far better. Exposed aggregate stopping inside the cover is a durability problem. A tidy line of grout loss at a form joint can run through to the far face. A photograph cannot separate them.

The code will not settle it either. The Philippine structural code's concrete provisions, following ACI 318 in substance, require concrete to be consolidated so it fills the forms around the reinforcement without voids, and the specification will normally require the repair to restore code cover. What neither NSCP nor ACI 318 gives you is an acceptance criterion for honeycombing or a repair method. Those sit in ACI 301 and in the project specification, in its provisions for repair of defective areas, and that is what a stripped column is: nonconforming new work, not an existing structure being assessed. Acceptance rests with the designer of record under the project specification.

Classify before you repair

One caution before the list. Class 1, 2 and 3 below are this article's own shorthand for severity. They are not terms defined in ACI, ICRI or NSCP, so do not write "Class 2" on a non-conformance report or quote it to a consultant as though it were a standard designation. Use it to order your own thinking, then describe the defect in plain words on the record.

Position decides, not size: a 100 mm patch inside a lap splice outranks a 900 mm patch of shallow honeycomb.

Find the extent instead of guessing it

Repairing without naming the cause guarantees a repeat on the next pour.

Repair families and their limits

Worked example: a 400 mm column with one bad face

Column 400 mm by 400 mm, cover 40 mm, main bars 20 mm, ties 10 mm, eight bars with three on the affected face, lift 3.0 m. Honeycombing shows 600 mm by 350 mm, sounding 700 mm by 400 mm, the full face width, and chipping reaches sound concrete at 65 mm. Rates are illustrative.

What this disposition assumes

Every number below rests on inputs a real disposition would have to state. They are listed here because several of them change the answer.

Step 1: where is 65 mm, relative to the steel?

So 65 mm is 65 − 50 = 15 mm past the outer face of the bar and sits 65 − 60 = 5 mm beyond the bar centre. Read that 5 mm correctly: it is the height of the embedded circular segment still buried, the offset of the chipped plane past the centre, not a 5 mm skin of concrete behind the bar. The plane stops 70 − 65 = 5 mm short of the back of the bar, and the core deeper than 70 mm is untouched, so nothing here is 5 mm from being punched through. The half angle of the embedded arc has a cosine of 5 ÷ 10 = 0.5, or 60 degrees, so the embedded arc = 2 × 60 = 120 degrees and the exposed arc = 360 − 120 = 240 degrees. Since 240 ÷ 360 = 0.667, two thirds of that bar, at the point that was chipped, has no sound concrete on it: a bond question, not a cosmetic one.

Step 2: screening estimate of section loss

Two fair objections. Cover outside the ties is not counted as confined core, so 16.25 per cent overstates the strength effect while ignoring the fire and corrosion protection it gave. Treat that as an engineering convention, borrowed from the way confinement provisions and ASCE 41 style assessments define the core, rather than as an established practice you can cite at somebody. And a tie is anchored by its hooks, not by bond along the leg.

Two more things about the 16.25 per cent. The 65 mm was measured at the worst point only, so applying it uniformly across the full 400 mm is an assumption presented as a measurement. And it is not the figure that governs the decision. Step 3 removes material to 90 mm behind three 60 mm strips, so the section actually open during the repair is about 400 × 65 + 3 × 60 × 25 = 30,500 mm², or 19.1 per cent of gross. That temporary condition, not the 16.25 per cent screening figure, is the number a reader should carry forward. Both remain screening figures rather than capacity checks: the real loss depends on void geometry, which only chipping confirms, and on the concrete strength, bar grade and utilisation this example never gives.

Step 3: how much repair material

Common practice is to cut back beyond the unsound boundary and leave a square shoulder rather than a feathered edge. The International Concrete Repair Institute's surface preparation guidance is the reference for the edge itself, a perimeter cut perpendicular to the surface with no feathering and a stated surface profile. It does not fix a lateral cut back distance beyond the unsound boundary, so the 50 mm used here is illustrative and has nothing behind it but convenience. Keep the perimeter saw cut shallow, of the order of 10 to 13 mm, or on this very column it saws through ties at 40 mm and main bars at 50 mm.

Step 4: the threshold

The trigger is the load path, not a percentage.

This stops being a patch repair decision the moment the defect touches the load path or the bond, not at a percentage. Two thirds of a bar is exposed at the worst point, over a debonded length of at most 700 mm, which is the sounded extent rather than the 800 mm patch, since the patch includes 50 mm of cut back into sound concrete at each end. If that length reaches the lap splice zone above the construction joint, it goes to the designer of record automatically, and this example never says where in the 3.0 m lift it sits. Either way, a defect this size needs the designer's written acceptance of classification and method.

Step 5: illustrative cost of the two options

Repair, with designer acceptance:

Demolish and recast the lift:

That is 46,648 ÷ 16,198 = 2.88 times the direct cost, and the schedule costs more again. The repair runs 3 days, the same 3 days its labour is priced at, plus 2 before the next lift is loaded, so 5. The recast runs 3 to prop, 2 to demolish, 2 to fix and form, 1 to pour and an illustrative 14 before props are struck: 3 + 2 + 2 + 1 + 14 = 22 days, so 22 − 5 = 17 days lost. At an illustrative ₱12,000 per day that is 17 × 12,000 = ₱204,000. This prices a repair the designer has accepted; it does not justify downgrading the classification.

The record that survives an audit

What the file needs, in this order.

Photograph with a scale and the element mark. Dimensioned sketch referenced to the construction joint. Sounding map, not just the visible boundary. Chipping depth and where it was taken. Classification and reasoning. Cause, with the corrective action for the next pour. Method and product. Designer acceptance, dated before work starts. Verification by re-sounding.

The same defect on an architectural fair faced column is a different decision from one that will be plastered, because the criteria then include colour, texture and edge line. And the procedure belongs in the file before stripping season, not at 5 pm with a subcontractor waiting.

Common pitfalls

Where the arithmetic gets fudged

Almost every number above is geometry or quantity: cover planes, section areas, patch volumes, bar masses, all guessed at while the pour card waits. The RHCES web tools page has calculators for them.

FAQ

Can we render it now and raise it later?

No. Rendering destroys the evidence needed to classify the defect, and any chance of proving the repair sound. A recorded defect with an accepted disposition is normal; a concealed one is a finding.

Who signs the acceptance?

Split it. The designer of record owns the classification and method for anything reaching steel. The contractor owns the workmanship, the QA/QC engineer the verification record. One signature carrying all three is how files fail an audit.