On the surface, it's all about measuring. But watch out—in the near future it's going to be all about sampling and statistics.

Today tolerances are on the books for the flatness of formed (CIP) concrete walls. They're in Section 4.8.3 of ACI A117-06, “Specifications for Tolerances for Concrete Construction and Materials Commentary,” and essentially permit a specific variance over a 5-foot distance. These tolerances refer to the wall's form, which in this case is its flatness.

“Abrupt surface irregularities shall be measured within 1 inch (25 mm) of the irregularity. Gradual surface irregularities shall be measured by determining the gap between concrete and near surface of a 5-foot (1.5-meter) straightedge, measured between contact points.”

The standard goes on to give the acceptable irregularities for four surface classes, A to D, from +1/8 inch to +1 inch.

Although this is a great improvement over the way the issue was addressed in the previous edition of the standard—where it was in section 4.5.4—it's difficult to confirm that an entire wall is within the tolerance. So representative checks are made with a straightedge, and the inspector and the contractor come to an agreement that the wall either does or does not meet the specified tolerance.

Although the current language is improved, it remains open to interpretation and dispute.

But Colin Milberg's current use of 3-D laser scanning to confirm as-built field conditions is bringing a whole new approach to answering the question of whether a wall surface falls within the flatness tolerance.

Laser scanning technology—which results in spatial imaging—goes far beyond checking a few wall sections with a straightedge. As explained in previous installments of this column (see below for links to articles by Milberg) this technology provides data on a grid covering the entire wall surface, then calculates each point's location relative to the ideal flat plane.

By applying standard mathematics to each data point, and the assembled mass of information, you now can know with much more certainty whether a wall surface is within tolerance. You also can identify exactly how much it varies—overall and point by point—from the plan.

That variation can be reported as the mean and the standard deviation. The mean tells the average variation between the locations of all the measured points and the plane they were supposed to be on. The standard deviation provides a weighted version of that information. If most of the measured points are close to the plane they were supposed to be on, the standard deviation is small. But when more points are farther from that plane, the standard deviation is larger.

This is where the principles of statistics are being introduced. Milberg started his data acquisition efforts collecting as much data as he could, knowing it was more than he needed. Now he has settled on taking measurements every 1 inch, rather than every ½ inch. The key was determining how much data was needed to obtain a statistically valid sample.

This depends on two things: the minimum area evaluated and what acceptable flatness tolerance to use as the baseline. Milberg has assumed a one-floor column, meaning an area of about 8 square feet from a column that is 1 foot wide and 8 feet tall, as generally the smallest sample of interest. He also assumed an acceptable variation of +/-¼ inch over a 5-foot distance. Then he looked at how to measure the data points.

Milberg's goal is “to be 99.7% sure that all points on the surface are within ¼ inch of the ideal flat plane defining that surface.” Although that percentage may sound a little arbitrary, it's not. That's simply the confidence level one gets when all the measured points fall within three standard deviations of the mean. Put another way, when three times the standard deviation of the location of the measured points is less than ¼ inch, you can be 99.7% certain that all your points are within that ¼ inch.

What does that have to do with concrete walls? Basically, scanning a concrete wall and comparing the standard deviation of the point location data to an allowable standard deviation will tell you instantaneously—and with 99.7% confidence—whether the wall flatness is within tolerance.

That raises the far more interesting question of the significance of these tolerances, which will be addressed in months to come.