The forms are off and the testing agency has handed over a cover survey with a lowest reading of 22 mm against a drawing that says 50. The foreman says it is one reading in thirty. The consultant writes "non-conforming". The client asks how bad it is, and nobody can answer with a number.
Cover is the most consequential dimension on a reinforced concrete drawing and the one most casually lost: a spacer collapses, a top mat is walked flat, a stirrup is tied to the wrong face. Everyone agrees this is bad; almost nobody can say by how much, which is why these arguments end in opinion.
Cover does three unrelated jobs with different governing numbers.
Corrosion protection. Cover is the barrier chlorides and carbon dioxide must cross to reach the steel and break down its passive film. Ingress is a transport problem, so cover and time are not linearly related.
Bond. A bar transfers force through its ribs and the concrete resists with ring tension. Thick cover lets the bar fail by pulling out, which is relatively ductile; thin cover splits instead, and splitting is a sudden longitudinal crack. Splitting governs whenever cover and bar spacing are small relative to bar diameter, and the changeover sits at roughly two and a half bar diameters of cover, which is exactly why the ACI 318 and NSCP confinement term is capped at 2.5: past that point the bar pulls out rather than splits, and more cover buys no further bond. At one bar diameter of cover, splitting has been governing for a long time already. Development length provisions already reflect cover and spacing, so a shortfall lengthens the anchorage required.
Fire resistance. Cover keeps the steel below the temperature at which it loses a defined fraction of its yield strength for the rated period. Documents differ: some measure to the bar centre, others to the outermost steel (the stirrup on a beam). Convert before comparing. The temperature limit, the rated period and the measuring datum all come from the fire document your project is actually designed to, and from the edition your specification names, not from a remembered rule of thumb.
A shortfall can be fine for one purpose and unacceptable for another, so check all three, not the one you remember.
Specified cover is the value the designer put on the drawing; as-detailed cover is what the bending schedule and the fixer work to; achieved cover is what the concrete did. Be careful with the word nominal, because the two main code families do not use it the same way. In Eurocode 2, nominal cover is minimum cover plus an allowance for deviation, and it is the designer's drawing value, not a workshop number. ACI 318 and NSCP call the drawing value the specified cover. Say which document you are quoting before arguing about which number was missed. The third of the trio is a distribution, not a number: it scatters with spacer height, mat weight, foot traffic and form movement, so measuring it is a statistical exercise. One low reading is not a defect and one good reading is not compliance. A disposition needs a mean, a scatter and a rule for the tail.
In the ACI 318 and NSCP cover tables the ladder runs one way: concrete cast against and permanently in contact with ground carries the largest tabulated cover, concrete exposed to weather or earth after casting comes next, and members never exposed to weather or ground carry the least. Aggressive chloride exposure does not sit on top of that same ladder. It is handled separately, through the exposure class and the mix requirements that come with it, and it can drive the specified value higher still. The numbers themselves come from your own project specification and the code edition that specification references, not from memory and not from this article.
Specify by capacity, not appearance, because a spacer must survive being stepped on. Cementitious spacers matching the parent concrete in strength and permeability suit aggressive exposures. Roughly 500 to 1000 mm each way on bottom mats is common practice, not a code mandate. On a top mat the design load is a person and a hose.
The beam is 400 mm wide and 12 m long, specified cover 50 mm, 20 mm main bars, 10 mm stirrups, assumed overall depth 600 mm. A soffit survey returns 30 readings: mean 41 mm, standard deviation 7 mm, three below 30 mm, lowest 22 mm. Both the specified 50 mm and every survey reading are taken to the outside of the stirrup.
What follows is a screening, not a durability design, and it rests on inputs that were assumed rather than measured. Each of these would move the answer, and two of them decide it.
1.645 × 7 = 11.515 mm, and 41 - 11.515 = 29.485 mm, say 29.4 mm. Round a lower bound down, never up.
On a normal assumption that is the five percent lower fractile. Be clear about what that is: no mainstream concrete code judges as-built cover by a five percent fractile with a factor of 1.645. It is a screening convention used here to put a number on the tail. Cover compliance is normally judged against minimum cover plus the construction tolerance your specification states, and that check, not the fractile, is what the non-conformance will be argued on, so run both and report both.
Thirty readings is also a thin basis for normality. 22 mm sits at (22 - 41) ÷ 7 = -2.714 standard deviations, about 0.33 percent of a normal population, so 0.33 percent × 30 = 0.10 readings expected that low, and one was found. That is not by itself evidence against normality: under the same model, the chance of at least one reading this deep somewhere in thirty is 1 - 0.99668 raised to the thirtieth power, about 0.095, or roughly one survey in ten. The companion figure points the same way: three readings below 30 mm against an expected 1.74 has a probability of about 0.25 under the same model. Thirty readings cannot resolve the tail either way, so do not lean on the fitted fractile alone.
Mean: 50 - 41 = 9 mm, and 9 ÷ 50 = 0.180, so 18.0 percent.
Characteristic value: 50 - 29.485 = 20.515 mm, and 20.515 ÷ 50 = 0.4103, so 41.0 percent.
Lowest reading: 50 - 22 = 28 mm, and 28 ÷ 50 = 0.560, so 56.0 percent.
The mean matters before the tail: 18 percent light on average is systematic, not scatter.
A widely used first pass model advances the chloride front with the square root of time, x = K √t, with x in mm, t in years and K a site and mix coefficient in mm per square root of a year. Rearranged, t = (x ÷ K)². Take K = 8. That value is illustrative only, and the entire service life screening moves with it: every year figure below is a direct function of K, so halving or doubling K changes each of them by a factor of four. It must be replaced by a coefficient measured on your own concrete, or published for a comparable mix in a comparable exposure zone, before anyone quotes a year to a client.
To reach 50 mm: 50 ÷ 8 = 6.25, and 6.25² = 39.06 years.
To reach 41 mm: 41 ÷ 8 = 5.125, and 5.125² = 26.27 years.
To reach 29.485 mm: 29.485 ÷ 8 = 3.6856, and 3.6856² = 13.58 years.
To reach 22 mm: 22 ÷ 8 = 2.75, and 2.75² = 7.56 years.
Loss at the 22 mm spot: 39.06 - 7.56 = 31.50 years, and 31.50 ÷ 39.06 = 0.806, so 80.6 percent of the initiation period is gone. At the mean, 39.06 - 26.27 = 12.80 years, and 12.80 ÷ 39.06 = 0.328, or 32.8 percent. At the characteristic value, 39.06 - 13.58 = 25.48 years, and 25.48 ÷ 39.06 = 0.652, or 65.2 percent.
So 56 percent of the cover lost costs 80.6 percent of the initiation period, while 18 percent lost costs 32.8 percent. The relationship is quadratic, so cover losses cost service life faster than they look on a tape measure. State the limits: it ignores concrete quality, binder, curing, cracking and the real surface chloride level, any of which can move the answer more than cover does, and it models only initiation. It ranks options; it is not a durability design.
Effective depth is overall depth less cover, stirrup diameter and half the bar diameter. At specified cover, 600 - 50 - 10 - 10 = 530 mm; at the survey mean, 600 - 41 - 10 - 10 = 539 mm. That is 539 - 530 = +9 mm, and 9 ÷ 530 = 0.0170, an increase of 1.70 percent.
Now flexure, sagging case, with an illustrative four 20 mm bars, 415 MPa steel and 28 MPa concrete, using the rectangular stress block of the ACI 318 and NSCP family with its 0.85 factor. Eurocode 2 uses a different block, but the ratio at the end of this step barely moves, because the steel force cancels out of it. Steel area = 4 × 314.16 = 1256.6 mm². Stress block depth = 1256.6 × 415 ÷ (0.85 × 28 × 400) = 521,489 ÷ 9520 = 54.78 mm, so half of it is 27.39 mm, and that does not move with cover. The lever arm does: 530 - 27.39 = 502.61 mm at specified cover, 539 - 27.39 = 511.61 mm at the mean, and 511.61 ÷ 502.61 = 1.0179, so nominal moment capacity rises by 1.79 percent.
Losing cover made the sagging section marginally stronger. That is why strength is the wrong argument to have about low cover, and why "the beam still checks out" is not a disposition. Two exceptions: the same arithmetic applies to top steel in a hogging region, and excess cover does cut capacity and widen surface cracks.
The member wide question is settled by the mean: 41 mm against 50 mm is an 18 percent systematic shortfall costing roughly a third of the screened initiation period. That is a member wide non-conformance if the minus tolerance in your specification is smaller than the 9 mm gap, and confirming that tolerance is the check to run before the word is written; where the specification allows more, the mean is inside tolerance and the whole argument shifts to the low readings. Either way the fix is upstream, in the spacer schedule and the pre-pour hold point.
The three low readings are a local defect only if they cluster in one zone with an identifiable cause; scattered lows are the tail of a bad process, already covered above. Re-survey that zone on a tighter grid to fix its extent, 150 mm being a common working choice rather than a requirement. Typical repair practice is to break out behind the bar, commonly of the order of 20 to 25 mm, then clean the steel to bare metal, prime, and reinstate with a mortar of at least parent strength and no greater permeability. The break-out depth and the reinstatement material are specification items: take both from the project repair specification, or from the designer in writing, not from this paragraph. Skimming adds cover on the drawing only.
All rates here are illustrative, as are the patch sizes, the spacer arrangement and the board quantities set out in the assumptions above; use your own records. Repairing three 300 mm by 400 mm patches: area = 3 × 0.30 × 0.40 = 0.36 m², and at ₱4,800 per m², 0.36 × 4,800 = ₱1,728. Access and supervision, ₱8,000. Re-survey and close out, 4 hours at ₱450, so 4 × 450 = ₱1,800. Total = 1,728 + 8,000 + 1,800 = ₱11,528.
Prevention on that 12 m length: bottom spacers at 500 mm centres, 12 ÷ 0.5 + 1 = 25 per line, two lines gives 2 × 25 = 50, plus say 40 side spacers, so 90 spacers at ₱12 each is 90 × 12 = ₱1,080, and 11,528 ÷ 1,080 = 10.7. Walking boards run the same way: 30 metres at ₱320 per metre is 30 × 320 = ₱9,600, and over eight reuses, 9,600 ÷ 8 = ₱1,200 per pour.
To run this arithmetic on site rather than in the spreadsheet you left in the office, the calculators on the RHCES web tools page cover the routine section checks behind a disposition like this.
Usually, but not always. The confinement term uses the smaller of the distance from the bar centre to the nearest concrete surface and half the centre-to-centre spacing of the bars being developed, and it is capped, so where bars are close the spacing governs and a cover reduction may not change the calculated length. That does not rescue the member: the splitting mode and the durability loss are real either way.
Sometimes, with the designer's written agreement, as a compensating measure. A barrier or sealer slows chloride ingress but has a finite life and a reapplication interval that must enter the maintenance manual. It is a management commitment, not a repair, and does nothing for fire or splitting risk.