Posted: April 2013

Introduction
Falling-weight deflectometers (FWDs) are used to collect important data that is used to characterize layer stiffness, and for planning maintenance and rehabilitation activities on highways and other major roads. Accurate data collection leads to a better prediction of maintenance and rehabilitation needs, which in turn leads to more effective planning and budgeting of these activities. Unfortunately, it is estimated that at this time only about half of the FWDs in use in the United States are regularly calibrated. This article will describe the primary sources of measurement error in FWD data collection and how the calibration protocol is used to effectively minimize these errors.

Sources of FWD Measurement Error
There are three primary sources of measurement error that must be accounted for when collecting FWD data: seating errors, random errors, and systematic errors. Let’s examine each of these sources of error in greater detail.

Seating Errors
Before testing, the FWD is driven into location, and the sensors and load plate are placed on the pavement surface. When initially put into place, the sensors and load plate may not be seated firmly on the pavement surface. To reduce measurement errors caused by this problem, the FWD mass is dropped at least twice without collecting data to ensure that the sensors and load plate are properly seated before testing begins.

Random Errors
Random errors (repeatability) are, as the name might suggest, unpredictable. They may occur for several reasons- variation in the electrical signal of the measurement equipment, environmental conditions, or many other unknown or unpredictable reasons. Because of random errors, we can expect that some of the measurements that we take will be higher than the true value, and some will be lower than the true value. Most random error follows a normal distribution, looking something like this:

(For more information on normal distributions, check out Metrology Musings: Measurement Uncertainty for Hillbillies.) To reduce the effects of random error on FWD measurements, several replicates of drop data are collected and averaged. By doing so, the measurements that are higher than the true value and those lower than the true value will cancel each other out. It is impossible to completely remove random error from FWD deflection data, but it can be reduced to somewhere around 0.04 to 0.08 mil.

Systematic Errors
Systematic error, or bias, occurs when there is a consistent offset of the measured values from the true value. This type of error is predictable, and can therefore be accounted for and corrected through calibration. It is reduced through comparing traceable reference standards to the FWD measurement equipment.

The Effect of Error on FWD Data
No measurement is perfect, and a small amount of error will always exist. The goal of calibration is to reduce this error to a level that will still allow us to use the FWD data in meaningful ways. Irwin, et al. (1998) found that “random error alone, without any additional bias error, can be enough to cause backcalculated modulus of the surface layer to be off by a factor of two or more. The effect on base and subgrade layer moduli was less dramatic.” The combined effects of random and systematic error could be detrimental to backcalculation data. It is therefore important to reduce the systematic portion of the error as much as reasonably possible in order to ensure that the FWD is providing meaningful test results.

How Low Can You Go?
So how much can we expect to reduce systematic error through calibration? Let’s take a closer look. Systematic error, in the case of FWD deflection data, is directly proportional to the measurement being made.

This means that the smaller the deflection, the lower the error. In cases where the deflection is small, the random error (usually between 0.04 to 0.08 mil) can actually be larger than systematic error. For these reasons, it is assumed that bias can only be reasonably detected at a level of around 0.06 mil or above.

Typically, deflection levels seen during the calibration procedure are around 20 mil. If systematic error can be reasonably detected at approximately 0.06 mil, then we can calculate the practical lower limit in error that FWD calibration can produce as follows:

(20 mil / 0.06 mil)* 100 = 0.3 percent

Therefore, the goal of the calibration procedure is to reduce systematic error (bias) to as close to 0.3 percent as possible. This ensures that the random error is greater than the systematic error in deflections up to 26 mil.

The Calibration Procedure
There are two primary components of the FWD calibration procedure in accordance with AASHTO R 32, Practice for Calibrating the Load Cell and Deflection Sensors for a Falling Weight Deflectometer: reference calibration and relative calibration. The procedure is completed using a software program called WinFWDCal, which was developed by the Cornell Local Roads Program (CLRP) as part of the pooled-fund study. It is recommended that all FWDs be calibrated in accordance with AASHTO R 32 annually.

Reference Calibration – Deflection Sensors
To perform the reference calibration of the deflection sensors, the load cell measurement data is compared to traceable reference standards. The reference equipment includes a reference accelerometer and a reference load cell. The accelerometer is sensitive to changes in the acceleration of Earth’s gravity. The acceleration data is double-integrated to obtain deflection and then compared to the deflection levels detected by the geophones or seismometers on the FWD.

Comparing FWD Deflection Sensor Data to Reference Accelerometer Data in WinFWDCal

Reference Calibration – Load Cell
To calibrate the FWD load cell, the load measured by the reference load cell is compared to the load measured by the FWD load cell. The reference load cell is calibrated annually in accordance with AASHTO R 33, Practice for Calibrating the Reference Load Cell Used for Reference Calibrations for Falling Weight Deflectometer. The calibration procedure in AASHTO R 33 is used to simulate the unique dynamic loading that an FWD load cell is subjected to. The end result of calibration is an adjustment of the gain factors for the deflection sensors and the FWD load cell, which are entered back into the FWD software.

Relative Calibration (Deflection Sensors Only)
The relative calibration procedure is completed by comparing the deflection sensors to one another. The relative calibration is done to further reduce the systematic error of the deflection data as close to 0.3% as possible.

Relative Calibration of FWD Deflection Sensors in WinFWDCal

Importance of Quality Assurance in Calibration
The FWD calibration procedure requires specialized equipment and a specialized facility that must be properly maintained in order to complete the calibration. In addition, the procedure requires highly-trained operators with an in-depth understanding of the calibration acceptance criteria. Ensuring uniformity in the way the calibration protocol is implemented from center to center is critical to collecting universally meaningful and accurate FWD data.

AASHTO’s certification program for FWD calibration center operators plays a critical role in ensuring quality in FWD calibration. A list of certified operators can be found on the AASHTO re:source website. Calibration centers and operators are evaluated annually by AASHTO re:source’s staff for conformance to AASHTO R 32.

What To Expect When Getting Your FWD Calibrated
The first step in getting your FWD calibrated is to contact the calibration center to schedule the calibration. Contact information for certified centers is found on the AASHTO re:source website. The calibration center will probably ask you to fill out a pre-calibration checklist. In order to ensure a smooth and hassle-free calibration it is important that you review the checklist carefully. Make sure that your FWD is in good working condition, and replace any parts that are damaged before you arrive. Most centers are unable to perform repairs on your FWD for you. If the FWD is not in good working order, it cannot be calibrated. Keep in mind that one of the reasons we calibrate FWDs is to detect equipment issues. If your FWD fails to meet the calibration criteria, you may have to take it to the equipment manufacturer for service before calibration can be performed.

Unless told otherwise, the owner/operator of the FWD must be available to operate the FWD during the entire calibration procedure. The person that accompanies the FWD for calibration must be familiar with the equipment and how to operate it. The procedure generally takes approximately 3 hours to complete, but if there are equipment or software issues, it may take longer. For more information on how to prepare your FWD for calibration, check out this presentation.

Monthly Calibration
In addition to annual calibration, AASHTO R 32 also includes a monthly calibration procedure. This procedure is generally performed by the FWD owner or operator. It can be used as a quick check to ensure that the FWD is working properly. The monthly calibration procedure is similar to the relative calibration procedure that is performed as part of annual calibration, and can be completed using WinFWDCal. A monthly calibration can identify any deflection sensors that need to be replaced. It can also be used to determine a temporary gain factor if a sensor has to be replaced unexpectedly. It is recommended that monthly calibration be incorporated into your FWD’s regular maintenance program.

More Information
In addition to the references below, you may also wish to check out this check out this video, produced by the Cornell Local Roads Program, which explains the calibration process in detail:http://www.fhwa.dot.gov/multimedia/research/infrastructure/calibration/index.cfm

Questions?
If you have any questions about FWD calibration, please contact Greg Uherek or Maria Knake.

References

• Irwin et al, “Falling Weight Deflectometer Calibration Center and Operational Improvements: Redevelopment of the Calibration Protocol an Equipment,” Report Number FHWA-HRT-07-040, Office of Infrastructure Research and Development, Federal Highway Administration, McClean, Virginia, October 2011.
• Knake, Maria, “What to Expect When Getting Your FWD Calibrated,” Falling Weight Deflectometer Users Group Meeting, PowerPoint Presentation, Embassy Suites Sacramento – Riverfront Promenade, Sacramento, California, October 15, 2012.
• Lutz, Bob, “Metrology Musings: Measurement Uncertainty for Hillbillies,” AMRL in Focus, Spring 2012, <http://www.amrl.net/amrlsitefinity/default/Resources/newsletter/Spring2012/5.aspx> (January 31, 2013).
• Norris, Rob, “The Skinny on Precision and Bias: Aiming or Sights on Precision,” AMRL in Focus, Spring 2012, <http://www.amrl.net/AmrlSitefinity/default/Resources/newsletter/Spring2012/4.aspx> (April 2, 2013).
• Price, Evan, “The Skinny on Precision and Bias: Aiming or Sights on Bias,” AMRL in Focus, Fall 2012, <http://www.amrl.net/AmrlSitefinity/default/Resources/newsletter/Fall2012/4.aspx> (April 2, 2013).