In October 2015 a slab experiment got under way in North Carolina. A concrete contractor owned land with five identical sheds. Each was long and narrow with a dirt floor. The contractor placed a concrete slab in each shed, taking five different approaches to the control of cracking and curling. We reported on the initial findings in an article in April.
In the year since the experiment began, the slabs have been monitored for cracking, shrinkage, and curling. As so often seems to happen when you investigate concrete, the results provide a few answers but also raise new questions.
The Five Slabs
Slabs 1 through 3 were cast on 1 October 2015, and slabs 4 and 5 were cast 29 days later. Weather conditions were similar on both days. All five slabs were the same size and the same mix was used. This mix design has been used on other industrial floors, with generally good results.
Here are the details that differ from slab to slab.
- Slab 1: Type II straight steel fibers, 1 inch long, dosage of 44.4 pcy; no joints.
- Slab 2: Type I steel fibers with hooked ends, 1-3/8 inches long, dosage of 44.0 pcy; no joints.
- Slab 3: macrosynthetic fibers made of copolymer and polypropylene, up to 2-1/4 inches long, dosage of 7.5 pcy; no joints. Because these fibers made the concrete stiffer, the mix had extra water and extra superplasticizer.
- Slab 4: no fibers nor any other reinforcement; crack-control joints were cut with an early-entry saw, 1-1/2 inches deep, on a grid of 15 by 17 feet. The east half of this slab was treated with colloidal silica and did not get the dissipating-resin curing compound used everywhere else.
- Slab 5: no fibers, nor any other reinforcement; no joints.
The slabs spent their first year protected from rain, sun, and wind, but exposed to temperature changes since the sheds are neither heated nor air conditioned. According to weather records for the nearest city, the ambient temperature ranged from 15° to 98°F over the course of the year.
Mixing, Placing, and Finishing
The three fiber types proved easy to batch and mix into the concrete. Only two changes had to be made to accommodate the addition of fibers, and both involved the macrosynthetic fibers in slab 3. At the behest of the fiber supplier, the concrete for slab 3 contained 2 gallons more water, and the superplasticizer dosage went up from 9 to 14 oz.
The fibers differed in their effect on pumping. The original plan was to place all concrete by pump, using a 4-in. line. This worked fine for slabs 1 and 2, with steel fibers. It also worked well on slabs 4 and 5, which did not contain fibers. However, a problem arose on slab 3, with macrosynthetic fibers. After the pump clogged twice, the workers abandoned it and placed the rest of slab 3 by power buggy. One of the workers, when asked about the mix with macrosynthetic fibers, told me "workability is not there", but felt it would be usable on a job where concrete could be placed straight from the ready-mix trucks.
Slab 3 developed more bleed water than the other four, so much so that it was pulled off with squeegees. It seems likely the bleeding was a result of the extra water in the mix rather than a direct effect of the macrosynthetic fibers.
All five slabs got a smooth, burnished finish. The presence of fibers did not seem to affect the quality of the finish. The same worker who complained about workability said "finish is the same."
Just about every slab made with fibers ends up with some fibers visible at the floor surface. Fiber exposure can be a big problem, a small problem, or no problem at all -- depending not only on the number of fibers showing, but also on floor usage and the owner's expectations.
Slabs 1, 2, and 3 all have exposed fibers. Slabs 1 and 2, with steel fibers, look similar. Both have a few visible fibers, as well as a few fossils where fibers have popped loose, leaving shallow holes. Photos 3 and 4 show the worst spots in slabs 1 and 2. Slabs 1 and 2 easily meet the limits on exposed fibers specified in the only official standard that sets such limits, TV 204 from Belgium. The tightest tolerance in TV 204, applied to the slab size we have here, allows up to 857 exposed fibers on each slab. Slabs 1 and 2 don't come anywhere near that limit.
Slab 3, with macrosynthetic fibers, has far more exposed fibers, and they take two forms. Over most of slab 3, the fibers are well embedded and hard to see unless you look closely. But along the slab edges, where the surface was hand-troweled, the fibers at first stuck up in large numbers (Photo 5). Over time many of the protruding fibers wore off, so that by October 2016 they were far less obvious.
So which performed better with regard to fiber exposure: the steel fibers in slabs 1 and 2 or the macrosynthetics in slab 3? It's hard to say, because fiber exposure is a matter of esthetics, not function. If you count the fibers, steel comes out ahead. But if you consider the visual impact of each exposed fiber on its own, you might prefer macrosynthetic.
The slabs were examined weekly for cracks, at first. After several months passed with no cracks appearing, apart from the planned cracks beneath the sawcuts in slab 4, the inspection schedule became less frequent. During the summer of 2016 three small cracks appeared, but we don't know exactly when.
Here is the status after one year:
- Slabs 1 and 2 never cracked.
- Slab 3 has two cracks, located close together at one end of the slab (Photo 6). They run roughly parallel to the slab's long axis. One crack, which goes all the way to the end of the slab, is 4 ft long. The other, which does not reach a slab edge, is 8 ft long. Maximum crack width is 0.015 in.
- Slab 4 cracked under every sawcut, but nowhere else.
- Slab 5 has one crack that is barely detectable. I don't believe I would have found it on my own, but someone with a sharper eye pointed it out. It runs perpendicular to the slab's long axis and is located midway between the slab ends, but it does not go all the way across the slab. Crack width is about 0.001 in.
I monitored slab shrinkage by measuring the distance between bronze monuments embedded in the concrete. Monuments were installed near both ends of each slab, with a third at the midpoint. Lines were scribed on each monument. The end monuments were located 98 ft apart on slabs 1, 2, and 3, and 97 ft apart on slabs 4 and 5.
As a second check on shrinkage, I measured the width of the construction joint where each slab meets an older connector slab (Photo 8). This measurement is less precise, partly because the joints don't have perfectly straight edges, and also because we cannot be sure the connector slab is stationary.
Table 1 below shows the results of both tests.
Slabs 1 and 2 shrank the most, at a little over 300 microstrains. Slab 5 did better, with 15% less shrinkage than slab 1. Slab 3 shrank less still, with only two thirds the shrinkage of slab 1.
The odd man out is slab 4, with practically no shrinkage, but there is a good explanation for that. Unlike the others, slab 4 contains sawcut joints, all of which activated and all of which opened up noticeably. The shrinkage showed up at those joints while the slab ends stayed put. When I add up the openings at all the transverse joints (current joint width minus the width of the original sawcut), I get total shrinkage of 0.694 inch, which amounts to 596 microstrains. So depending on how you measure it, slab 4 shrank either hardly at all, or about twice as much as the other four slabs. Graph 1 (bottom of article) shows shrinkage for all five slabs, including both ways of measuring slab 4.
I checked for curling two ways. First, I measured the floor profile down the length of each slab on the day of construction, and again about a year later. I examined the profiles for changes and signs of curling. The unjointed slabs -- slabs 1, 2, 3, and 5 -- all performed the same. Their profiles did not change significantly, and there were no clear signs of curling. Graph 2 shows the before and after profiles on slab 1, which is typical of the unjointed slabs. In contrast, the sole slab with joints -- slab 4 -- clearly did curl, as shown on Graph 3. (See graphs at bottom of article.)
Second, I measured FF flatness numbers on all five slabs, on the day of construction and again about a year later. A reduction in FF numbers is strong evidence, if not absolute proof, that a slab has curled. Table 2 shows the results.
The FF results confirm the evidence of the profile graphs. The unjointed slabs -- slabs 1, 2, 3, and 5 -- were stable, with changes in FF that range from +2.8% to -7.7%. Such changes are well within the normal margin of error for such testing. In contrast, slab 4 showed a 29.5% reduction in flatness. Curling is the only plausible explanation for such a change.
Answers...and more questions
The chief goal of this experiment was to see how different fiber types affect cracking and curling in slabs with extended joint spacings. Regrettably, it failed to shed much light on that question, since all three fiber types performed well, as did slab 5 with no fibers at all. It seems that under these conditions, we didn't need fibers to control cracking and curling.
In some ways the slabs without fibers, slabs 4 and 5, teach us more than the three slabs that contained fibers. Slab 4 shows that if you saw closely spaced joints into a plain concrete slab. you can make it crack and curl. Slab 5 shows that if you take a similar slab and leave out the joints, it can remain substantially free of cracks and curling. Slab 5 breaks all the rules for joint spacing in an unreinforced slab, and yet it works. Slab 5's success doesn't mean the rules are wrong, but it does raise questions about whether they should apply in every case, without regard to other factors such as slipsheets and elimination of restraints to shrinkage.
The shrinkage measurements served up a surprise: markedly less shrinkage on slab 3, with macrosynthetic fibers. No one on the project predicted that, and it's hard to say what it implies. In principle, less shrinkage should mean less cracking. But slab 3 actually developed more cracks -- albeit just two of them, both short -- than the other unjointed slabs.
While the experiment didn't tell us conclusively which fiber type did best at controlling cracks and curling, it did reveal some differences with respect to workability and fiber exposure. There was no practical difference between the two steel fiber types, but differences did arise between the steel fibers and the macrosynthetics. When it came to workability, the steel fibers came out ahead. They required no adjustment to the mix design and they passed through the pump without trouble. In contrast, the macrosynthetic fibers required extra water and extra superplasticizer, and even so the mix proved impossible to pump with the equipment at hand. When it came to fiber exposure, the fiber types performed differently, but with no obvious winner. The steel fibers appeared at the surface in lower numbers. But when they did appear, the effect of each exposed fiber was greater.
The experiment highlighted several important questions that remain unanswered:
- Are there conditions under which we would see a substantial difference in performance between steel fibers, macrosynthetic fibers, and no fibers?
- Are there conditions under which we would see a substantial difference in performance between the two types of steel fibers?
- How important is the slipsheet under an extended-joint slab? Would the slabs in our experiment have performed as well without a slipsheet?
Clearly, we need more experiments.
While the main experiment focused on the effects of fibers, a second experiment on slab 4 looked at something completely different: a liquid finishing aid based on colloidal silica. The manufacturer claims that this product reduces curling and improves wear resistance.
The experiment on slab 4 was set up to test both claims. The west half of the slab got no colloidal silica and was cured with a dissipating-resin compound, the same as the other four slabs. The east half got no curing compound but was treated with the colloidal silica, sprayed on in two steps. The first coat went down right after the concrete was struck off, and the second coat a few minutes after the final trowel pass. A manufacturer's representative directed the work and actually performed some of it himself.
The results do not support the manufacturer's claims.
Curling was similar on both halves of slab 4. Look at Graph 3, which shows the control surface on the left side and the treated surface on the right. Curling is evident at every joint, on both sides of the graph. The FF results tell much the same story. On the control surface, over a year, FF went down by 26.0%. On the surface treated with colloidal silica, FF went down by 33.0%.
When we turn to wear resistance, the difference between the claim and the test results is even wider. Contrary to every expectation, the colloidal silica seems to have reduced the floor's resistance to wear. Graph 4 shows the results of Chaplin abrasion tests on both halves of slab 4. In this test, low numbers are good, and high numbers are bad. When testing at 31 days showed poor results on the treated surface, the manufacturer suggested the colloidal silica needed more time to work. Sure enough, a retest at 106 days gave better results, but the control surface improved, too. Even after 106 days, the surface treated with colloidal silica showed, on average, over six times as much wear as the control surface.
Colloidal silica may work on other slabs, under different conditions. But it did not work here.