Supercomputers to Aid With Concrete Mixes

Tue May 02, 2006 - National Edition
Pete Sigmund

Advanced software codes being developed with the aid of the world’s fourth-fastest supercomputer are expected to significantly speed the process of choosing the right concrete for a construction job, saving millions of dollars.

Researchers also expect the codes to significantly aid efforts to develop much more durable concrete, which may last more than half a century.

The new codes may be available on CD ROMs to ready-mix producers, cement companies, contractors and other users as early as 2009. Instead of evaluating perhaps several hundred batches over one week or longer, users employing the programs on their computers could find, and test, several “optimum” mixtures for their jobs in just a few hours.

High Hopes

A layman may see a concrete mixer on the street and figure there’s no big science involved. Actually, though the basic process is relatively simple — mixing perhaps 15 percent cement with 60 to 75 percent aggregates (small rocks), 15 to 20 percent water, and often other substances such as silicates, fly ash and slag — the science of finding the best mixtures for different concrete applications is mind-boggling.

“It’s a huge computational problem because concrete is getting much more complicated all the time,” said Edward Garboczi, leader of the Inorganic Materials Group at the National Institute of Standards and Technology (NIST) Building and Fire Research Laboratory in Gaithersburg, MD.

“Optimizing the mixture based on six or seven ingredients can be quite a task. What is the size-distribution of the rocks? What is the shape, flat or pointed? Our models try to take those into account. They try to characterize such elements as the cement, sand, and rock particle size distribution. Our goal is for people to use these software tools to choose the optimum mixtures for their projects, their environment, and their local materials. Up until now, concrete hasn’t been characterized that well and has sometimes been a hit-or-miss operation.”

One Million Hours on Supercomputer

As one of four winners of a national “grand challenge problem” competition, NIST has just received one million hours of computer time to work on such mixtures for optimum concrete using the Columbia supercomputer at the National Aeronautics and Space Administration’s Ames Research Center, Moffett Field, CA.

NIST has been working on a software package for optimizing concrete for 13 years. It formed an NIST/industry consortium to develop this software, called the Virtual Cement and Concrete Testing Laboratory (VCCTL). The supercomputer will allow further development, including the flow, dispersion and merging of materials for concrete under a variety of flow conditions.

The supercomputer, valued at approximately $120 million, includes 10,240 processors. Its speed is 51.87-trillion floating point operations per second.

Occupying approximately 15,000 sq. ft. at the Ames Center, Columbia (named for the space shuttle that exploded) offers approximately 1 PB (petabyte or 1,000,000,000,000,000 characters) in disk storage and a 10 PB-capacity tape archive.

The Space Agency said that a person making one multiplication or addition per second would take approximately 1.645 million years to complete the computations, which Columbia can do every second. It also observes: “It would take the entire 6.356 billion population of 8,160 earths working at one calculation per second to keep up with Columbia.”

That’s a lot of power, but the research program on concrete needs it. Columbia will be able to compare an astounding number of standard and specialty combinations of concrete mixtures, including many variables like different local materials and environment conditions.

Paul Tennis, consulting engineer of the PCA headquarters in Skokie, IL. explained the magnitude of the challenge (and the need for the supercomputer) this way: “We are dealing with billions and trillions of moving particles of all different sizes and shapes, including sand-sized particles only a couple of millimeters across, as well as aggregates perhaps a few inches across. You need a supercomputer because it’s extremely difficult to trace all these particles and all the changes that are occurring.”

The computer can calculate optimum tradeoffs while users will only have to make a handful of tests on the most likely combinations, instead of today’s time-consuming trial batches, to confirm the results.

Advanced Code in Three Years

A new version of VCCTL is released each year. An initial version of the software (Version 1.0) was released five years ago. Version 6.0 is about to be released.

“We operate in three-year phases,” said Garboczi, who is leading the consortium project. “We’re just finishing Phase 2. Three years from now, aided by the supercomputer, Phase 3 should offer something very usable by people in the concrete community.”

The CD ROM holding the code can be used on normal desktop computers. It is expected to offer many benefits to ready-mix suppliers, cement manufacturers and larger contractors.

“Any place that uses concrete should benefit from it,” Garboczi said. “That includes, of course, buildings and pavements, like highways. You wouldn’t dream of making a million-dollar building without some kind of finite-element package like Computer-Assisted Design (CAD). We feel the same thing should be done for new concrete mixtures, which right now are pretty empirical.”

The consortium includes such industrial members as the Portland Cement Association (PCA), the Ready Mixed Concrete Research Foundation (RMCRF) and the National Stone, Sand and Gravel Association (NSSGA).

Concrete Would Last Much Longer

Besides providing faster, more accurate preconstruction testing and evaluation, the code is regarded as a big step toward much longer-lasting concrete, potentially saving millions of dollars in maintenance and other costs.

“Durability is our main concern; once the models are in place, the industry can do a much better job of predicting concrete durability, and therefore make the concrete last longer,” said Tennis. “Users will be able to chemically predict what will happen in five, 10, 20 or more years, projecting long-term properties rather than waiting five years to actually physically test real specimens.

“The code will, for instance, allow people to factor in environmental effects, like sulfate-bearing soil of a certain concentration, or the local freeze-thaw environment. They will have a lot more confidence in specifying certain mixtures to be durable in certain environments.

“In hours, or days at most, we can predict 20 or 30-year properties. Computer models will be that sophisticated. At present, we have pavements that are 40 years old, but we don’t know what the initial properties of those pavements were, what the cement chemistry was, or what placement conditions existed.”

75-Year Pavement Life

“With proper care and material selection aided by the computer, I don’t think it’s unreasonable to expect bridge pavements to last 50 to 75 years,” Tennis said, “and there’s no reason not to expect most general-purpose highways to last that long, too. You might have to pay a little more money upfront to obtain the right mix designs, work longer on preparing the base cement, and pay attention to all the details.”

Tennis said the computer codes could help prevent early cracking of bridge surfaces.

“There have been some noted bridge failures, including some that have only lasted three to five years, because something went wrong early in the process, so that deterioration began much faster, with abnormal cracking in the first few months,” he said. “The advantage of the consortium’s research is that it provides an additional prediction tool to closely examine all the materials and their combinations.”

Could Save Millions of Dollars

Tennis said that “it is not unreasonable to expect the industry to save millions of dollars by using the computer models.

“Industry spends tens of millions of dollars on developing and testing thousands of batches every few years, typically testing for a strength profile over one, three, seven and 28 days,” he said.

“The benefit of the program is that it can look at all the characteristics of everything in the mix, and perform a complete hydraulic simulation to see exactly how they all interact, especially the early rheology [changes in form and flow]. The program allows you to simulate everything in 15 minutes, so that you can change parameters, like which aggregate is better for a certain application.”

Garboczi observed: “Doubling the surface life of concrete could save billions of dollars per year; that’s a big chunk of change.”

The United States used 127 million tons (115.2 million t) of cement in 2005, an increase of 5.9 percent from 2004. An increase of 3.5 percent is expected this year. U. S. cement production is expected to increase by 16.2 million tons (14.7 million t) by 2010.

An estimated 35.2 million tons (31.9 million t) of concrete are used on U.S. highways and streets each year.

In its simplest form, concrete is a mix of paste and aggregates. The paste, composed of portland cement and water, coats the surface of the fine and coarse aggregates. Through a chemical reaction called hydration, the paste hardens and gains strength to form a rock-like mass.

Concrete is plastic and malleable when newly mixed, strong and durable when hardened. The key to a strong, durable concrete is the careful proportioning and mixing of the ingredients, which makes the research of top importance. A mixture that doesn’t have enough paste to fill the voids between the aggregates will be difficult to place, with rough, honeycombed surfaces. CEG