M20 cube results that pass individually but fail as a group
A ready-mix supplier is supplying M20 concrete for the columns of a two-storey school building. Six consecutive samples, each the average of three 150 mm cubes cured in the site tank and tested at 28 days, give: 18.0, 21.5, 23.0, 19.0, 22.5 and 17.5 N/mm². The mix was designed for a water-cement ratio of 0.50 and a slump of 40 mm. Site records show that on at least three deliveries the gang added about 15 litres of water per cubic metre at the hopper because the concrete 'would not flow' in the pump, taking the slump to 85 mm. The site engineer argues that no single result falls more than 4 N/mm² below the characteristic strength, so the concrete is acceptable; the owner's representative disagrees. The laboratory has a rebound hammer and a core cutter. The next pour is due in two days, and the ground-floor columns have already been cast with this concrete. The site has no established standard deviation.
1. Apply the acceptance criteria for M20 and decide who is right.
Read the worked answer
Criterion 1 - individual results. fck − 4 = 20 − 4 = 16 N/mm². Lowest result is 17.5 N/mm², so every individual result passes. The site engineer is correct on this point only.
Criterion 2 - mean of any four consecutive samples. With no established standard deviation the requirement is fck + 4 = 24 N/mm².
- (18.0 + 21.5 + 23.0 + 19.0)/4 = 20.4
- (21.5 + 23.0 + 19.0 + 22.5)/4 = 21.5
- (23.0 + 19.0 + 22.5 + 17.5)/4 = 20.5
All three are well below 24 N/mm².
Verdict. The concrete fails acceptance. Individual passing results do not rescue a group whose mean has no margin above the characteristic strength. The overall average is about 20.3 N/mm² against a required characteristic of 20 N/mm² - that means roughly half the concrete in the structure could be below 20 N/mm², which is the opposite of what a characteristic strength promises. The scatter (range 5.5 N/mm²) also shows the batching is not under control.
2. Explain quantitatively how the site practice produced these numbers, and state the corrective action before the next pour.
Read the worked answer
The water addition. 15 litres per m³ against a cement content of roughly 320–340 kg/m³ at w/c 0.50 raises the water-cement ratio from about 0.50 to about 0.55–0.58. The Abrams relationship is steep in this range: a rise of 0.08 in w/c typically costs 15–25% of strength. Applied to a mix that would have given about 20–21 N/mm², that predicts roughly 15–17 N/mm² - which is precisely where the low results (17.5, 18.0, 19.0) sit. The three deliveries where water was added should be checked against the sample dates; the correlation is the evidence.
Also note. Extra water raises the slump to 85 mm as recorded, increases bleeding, shrinkage and permeability, and lowers durability in a tropical climate - three problems that no later test corrects.
Corrective actions before the next pour:
- Issue a written instruction that no water is added beyond the approved w/c; post the w/c and slump on the hopper.
- Use a mid-range water reducer or plasticiser to get pumpable concrete at 0.50.
- Move to weigh batching with moisture correction on aggregates; check the batching plant printout per truck.
- Fix the slump acceptance band and send back non-conforming trucks.
- Review the mix design for the required target mean strength (roughly fck + 1.65σ, around 26 N/mm²), not just fck.
3. What do you do about the columns already cast, and how do you make sure the problem is the concrete and not the testing?
Read the worked answer
First, eliminate testing error, because condemning good concrete is as serious as passing bad concrete:
- Check cube moulds for size and squareness, compaction (number of layers and blows), curing tank temperature and whether cubes dried before testing.
- Check the testing machine calibration certificate and the rate of loading.
- Check that the cubes were actually taken from the deliveries in question - traceability of sample to truck is essential.
Second, assess the structure in place:
- Rebound hammer survey across the columns, taking a grid of readings per column and comparing columns cast from different deliveries - the hammer gives a relative picture, not a strength.
- Extract 3 cores, preferably from a non-critical zone (with the designer's approval), and accept the concrete only if the average of cores is at least about 85% of fck and no single core falls below about 75%. Cores cut from the top of a column are not comparable with cubes; drill carefully and avoid reinforcement.
- If cores fail, refer to the designer for a load test or strengthening decision; do not accept on NDT alone.
Third, hold the next pour until the mix is corrected and at least three consecutive samples meet the group criterion. Recover the cost of any remedial work from the supplier's retention, since the water addition was directed by site staff - record who instructed it.