March 1, 2008
by Daniel Brown, Contributor
Nobody likes problems with aggregate quality or gradations, but they happen. Sampling errors can throw off test results. A crushing plant manager can strive to overproduce and cause gradations to go out of spec. A loader operator can make mistakes. The list goes on and on. So when problems develop, how quickly are you able to respond and pinpoint the cause?
FNF Construction, an Arizona-based highway contractor, ensures aggregate quality through various steps in its production process. It owns several quarries and gravel pits, and typically crushes aggregates as project needs dictate. Stacker belts can segregate the material, so stockpile management is important. The contractor works to keep the end of the stacker belt within 5 feet of the top of pile, to prevent segregation. And FNF builds smaller surge piles at the crusher for testing purposes, says quality control manager Tom Kennedy. The smaller piles produce more accurate samples.
For new asphalt mixes, Kennedy says FNF always paves a test strip with 500 to 700 tons of hot mix. Doing a test strip debugs the hot-mix plant, establishes a flow of materials, and limits the contractor’s liability in case of a problem, Kennedy says. Aggregate gradations are checked at a minimum of three places:
At the crusher, either off the belt or from the surge piles;
At the hot plant, after a pugmill blends the aggregates but before they go into the dryer drum; and.
In the hot mix, either sampled from the truck or behind the paver.
If a problem develops with material passing one spec sieve, Kennedy first checks for a testing or sampling error. Usually re-sampling and retesting once or twice will confirm or refute the first test.
Suppose the target is 80 percent passing the 0.375-inch sieve, but suddenly the mix sample shows 65 percent passing that sieve. Kennedy begins a logical tracking process to trace down the problem. “Say I check the cold feeds at the asphalt plant, and I get 65 percent on the cold feeds,” he says. “That tells me the plant is doing its job. Simultaneously, I check those surge piles at the crusher. If I check the 1-inch-minus pile and expect to find 30 percent passing the 0.375-inch sieve — but I get 5 percent passing the 0.375-inch, that’s the problem.”
Kennedy says if an entire stockpile was crushed out of spec, the problem should have been caught during the mix design process. “You may have to do a new mix design,” he says. “Sometimes you do a product change.”
Contractors use various statistical control methods and software to troubleshoot aggregate gradation problems. David A. Bramble, a Maryland asphalt contractor with five hot-mix plants, has programmed an Excel spreadsheet to flag aggregates that are out of spec. “We keep a three-dimensional database in a spreadsheet program, and I have every gradation test that I have taken since 1988,” says G. Marshall Klinefelter, bituminous quality control manager for Bramble.
|Microsoft’s Excel program has a feature called conditional formatting, shown here with the colored shading. David A. Bramble uses this feature to spot irregularities in aggregates. Percentages passing that are on target are shown in green; numbers within plus or minus one standard deviation from average are white; percentages beyond one standard deviation on the fine side are yellow; and percentages beyond one standard deviation on the coarse side are colored orange.|
The spreadsheet lists percentages passing every sieve for the following on a given aggregate blend: the current test; the maximum (the finest it’s ever been); the minimum (the coarsest it’s ever been); the average of all tests taken to date, the standard deviation, and the average of the last five tests.
“I can watch the gradations on any given sieve, and if it gets beyond one standard deviation on a certain sieve, it encourages me to call the aggregate supplier to find out what’s going on,” Klinefelter says. Typical factors that may impact aggregate shape include crusher downtime, speeds, and components, as well as equipment operator changes.
Klinefelter hopes to catch aggregate problems before the material reaches the hot-mix plant. “But sometimes we have to adjust on the fly,” he says. Bramble might add coarse material to a blend that is too fine, or compensate in other ways.
Microsoft’s Excel program has a feature called conditional formatting. On a spreadsheet of percentages passing all of the various sieves, conditional formatting will color in certain cells to show Klinefelter percentages passing that are either on target (green); within plus or minus one standard deviation from average (white); or beyond one standard deviation on the fine side (yellow); or beyond one standard deviation on the coarse side (orange). “It makes reading the data fast and accurate,” Klinefelter says.
One supplier delivers aggregates that are off-loaded from a barge by a clamshell bucket. “By the time it gets to me it’s a finer material,” Klinefelter says. “I compensate for that by asking for a material that is slightly coarser than I need it to be. We’ve developed a relationship over the years, so that my supplier can notify me of a problem before it’s off-loaded. If I know there is a problem upfront, I can deal with it. You have to stay in communication with your aggregate suppliers.”
The Cusum technique
Lafarge North America’s western region owns 12 quarries and 24 sand and gravel operations spread across the western United States. The company supplies 10 internal hot-mix asphalt plants, 45 ready-mix concrete plants, and several dozen external customers.
Statistical control of aggregate gradations gets a high priority. Says Heath Waddell, director of aggregate quality and product development: “We want our production ranges to be tighter than the specifications ranges — to ensure that we get paid and that our customers get paid.”
To look at trends toward finer aggregates or coarser ones, Lafarge uses a statistical device called the Cumulative Sum Technique (Cusum). This technique sums up and plots the differences between each new test and the mean of all tests. So if a series of gradation tests all run to the fine side, for example, the differences in percentages passing will all add to the positive side of the mean. But if some tests ran finer and some coarser, the cumulative deviations would cancel each other out and the plot line would be more flat. Because the Cusum is cumulative, it is more sensitive to changes in gradation than standard control charts that only show tests and standard deviations one at a time.
The Cusum technique can be used to predict production ranges in gradations, Waddell says. For example, if the mean of a number of tests is 28 percent passing the 0.375-inch screen, and the standard deviation is 3 percentage points, then the material can vary from 25 to 31 percent passing the 0.375-inch sieve and stay within one standard deviation. The Cusum totals the deviations, adding the deviation of each new test result as they come in — so you can see the trend over time.
“One standard deviation is too small to show up on a control chart,” Waddell says. “That’s why we look at a Cusum chart — to look at shifts in the tests smaller than a typical deviation would be. The Cusum measures trends, not consistency.”
Waddell says a variation of 4 to 8 says from the mean, on a given sieve, will cause Lafarge to resample and retest the material. If errant test results are found and validated on the third test, Lafarge will shut the plant down. “We stop production and make any adjustment we need to make,” Waddell says. “The problem could be crusher wear, or screen wear, or stockpile management. We can’t run the risk of continuing to produce until we figure out what’s wrong.”
Clarifying ‘Percent Within Limits’
The Federal Highway Administration is encouraging state transportation departments to use a statistical technique called “Percent Within Limits” (PWL) to pay contractors. And more and more states are beginning to use PWL.
“Generally, the highway industry does not use an accept/reject model for evaluating contractors’ work,” says Jim Walls of the FHWA. “Instead highway agencies accept what is produced and pay accordingly, using payment systems that have incentives and disincentives.”
The PWL model encourages highway contractors to produce consistent quality work and then rewards that work by tying payment to a statistically valid measure of quality. For hot-mix asphalt, payment might be based on measurements such as air voids content, pavement density, and/or smoothness of the pavement.
PWL is defined as the percentage of the lot falling above the lower specification limit and beneath the upper specification limit. This quality measure uses the sample mean and the sample standard deviation to estimate the percentage of the population (lot) that is within the specification limit. This is called the PWL method. In theory, the use of the PWL method assumes that the population being sampled is normally distributed — that it follows a normal bell curve. In practice, it has been found that statistical estimates of quality are reasonably accurate provided the sampled population is reasonably bell-shaped and not bimodal or highly skewed.
If a state requests it, the FHWA offers a one-day introductory workshop on PWL. The first such meeting was held Feb. 8, 2006, in Raleigh, N.C. More than 20 staffers from the North Carolina Department of Transportation attended the workshop, which offers an overview of quality measures in general, details on how PWL works, and specifics on computing PWL.
For more information on the FHWA workshop, contact Lee Gallivan, Office of Pavement Technology, at 317-226-7493; or Ewa Flom, also OPT, 202-366-2169; or Jim Walls, Resource Center, 410-962-4796