Click for next page ( 25

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 24
24 is typically determined using the "most appropriate" method- In general, sources of variability for pavement condition ology available, which is done by using trained and experi- data collection can be related to the equipment used, equip- enced raters for pavement surface distress and a calibrated ment operation (including rater/operator training and skills), or certified piece of equipment for sensor-based measure- processing of the data collected, environmental conditions, ments. Because the most appropriate methodology is some- and shape and condition of the pavement surface. All of these times difficult to determine, some agencies prefer to use potential sources have to be considered and, if possible, con- the term reference value instead of ground truth. Ongoing trolled because they will affect the quality of the data collected. studies may provide additional insights on this issue. For exam- Although the potential sources of variability are many, this sec- ple, the Strategic Highway Research Project 2 (SHRP 2), S03: tion summarizes the primary sources that can be controlled Roadway Measurement System Evaluation, is developing, during the data collection process. Calibration and/or valida- organizing, and conducting a roadway measurement accuracy tion before, during, and after data collection are necessary to evaluation of mobile road and pavement inventory services ensure accuracy, given the possible variations between dif- collected at highway speeds (53). The recent National Work- ferent devices and operators, as discussed in chapter four. shop on Highway Asset Inventory and Data Collection (54) compared automatic systems for inventorying roadway geom- etry and roadside element data. Surface Distress For surface distress, the control sections are typically eval- Sources of variability for manual and automated methods of uated using visual surveys or by independently analyzing the data collection are similar and generally involve distress iden- images captured by pavement evaluation equipment. Rutting tification and classification, as well as assigning distress sever- and faulting are typically determined using a straight edge and ity levels. McGhee (6) classified the automated distress data in rut wedge or a static inclinometer, and faulting using a straight distresses collected with sensors (e.g., smoothness, rutting, and edge and a ruler or the Georgia Faultmeter. Statistical meth- faulting) and distresses obtained from processing of pavement ods are typically used in conjunction with control site testing images (e.g., cracking). Because some of the sources of vari- to establish acceptable ranges for various data collection tech- ability for sensor-based (rutting) and image-based (cracking) niques. The ground truth smoothness is currently determined distresses are in general different, they are presented in two based on the profile measured using rod and level, the static separate sections. Typically, both manual and automated crack inclinometer, or "walking" profilers; however, research is detection methods display a noticeable bias toward detecting underway to find a more appropriate reference profiler (55). higher-severity distresses, while missing lower-severity dis- tresses. This is usually the result of higher-severity distresses To compare friction measurements obtained with differ- being more readily identifiable than medium- or lower-severity ent types of friction equipment, the Permanent International distresses (18). Association of Road Congresses (PIARC) has devised the IFI (56). ASTM currently uses the Dynamic Friction Tester (DFTester) and Circular Track Meter (CTMeter) as the ref- Cracking erence devices for the IFI. The index is composed of two Cracking is measured from a moving vehicle or "walking" numbers, the friction value at 60 km/h (F60) and the change the section. The evaluations from the moving vehicle can be of friction with speed (sp) (57 ). done manually ("windshield" evaluation) or automatically by processing images of the pavement collected by a pavement SOURCES OF VARIABILITY IN PAVEMENT distress data collection vehicle. The processing of the images CONDITION DATA COLLECTION is done manually in semi-automated data collection and auto- matically by the processing software in the fully automated One of the purposes of the quality control, and to some degree process. Sources of variability for automated, semi-automated, quality acceptance, processes is to reduce the variability in and manual distress include (59): the pavement condition measurements. Thus, it is important to understand the various sources of variability before design- Type of equipment/data collection method ing pavement control and acceptance programs. Because Image quality--for automated and semi-automated there is inherent variability in the pavement condition data surveys. that are collected, it is important that agencies understand the Type of technology used; for example, analog magnitude and sources of variability in the data being col- images, digital images, laser-based images. lected to be able to compare results and establish target control Resolution of the imaging equipment (detection of and acceptance level for quality management. For example, smaller cracks require higher resolution equipment). Stoffels et al. (58) proposed a process to identify acceptable Field of view (distresses may be missed if they do ranges for comparing results from two independent sources not cover entire lane). using standard variability control concepts with pavement Quality of the color contrast of the pavement data collection. image (color contrast between the crack and the

OCR for page 24
25 surrounding pavement is an important factor when Edge drop-off and/or narrow lanes--The far right distress software programs are evaluating the sever- sensor may pick up a drop-off or grass on the right ity of the distress). shoulder, which will affect the measurements. Lighting method. Equipment driver and/or operator--The experience, Rater's vision--in case of windshield surveys. training, and driving skills of the data collection per- Raters/equipment operator training--The experience sonnel will affect the measurements and thus the qual- and understanding of rating protocols is paramount to ity of the data collected. reduce the variability of the collected data. Rut depth calculation method is to be controlled to ensure Processing software--The algorithm used to detect and consistency from year to year or from service provider quantify the various types of cracks is critical in the case to service provider because there are different algorithms of automated surveys. For example, one common prob- for processing the transverse profile and calculating rut lem with an automatic cracking detection algorithm is depth. The main methods currently in use are the string- the classification of the pavement shoulder joint as a line or "wire" method [AASHTO PP-38-00 Standard longitudinal crack. Practice for Determining Maximum Rut Depth in Asphalt Measurement environment--The conditions under which Pavements (2005)] and the straight edge. The stringline the distress surveys are conducted will affect the detec- method allows an imaginary line to bend at the hump tion of the cracks. For example, cracks are typically between the wheel paths if the hump is higher than the more visible soon after rain because they may be filled outside and centerline of the road. The straight edge with water. The severity of the extent detected under method projects a straight line across from the inside to such circumstances may be different. the outside of the lane and results in lower rut depth cal- culations than the stringline method. Measurement environment Rutting Temperature, wind, humidity, and surface moisture affect the various types of sensor differently and can Most agencies are currently measuring rut depth (transverse add to the variability of measurements. profile) with some type of automatic data collection equip- Presence of pavement contaminants, such as sand, ment, including rut bars with multiple sensors or continuous gravel, etc. measurement systems. In general, the sources of variation Lighting conditions affect optical sensors. include the following: Surface texture--High-textured surfaces, such as open- graded friction courses and chip seals, can affect the Type of equipment sensor readings Type of sensors--Common sensors include point laser, ultrasonic, and continuous scanning lasers. It is generally accepted that rut bars with a greater number "Point" lasers are currently the most commonly used of sensors (or transversal measurements points in the contin- sensors. Continuous scanning lasers can cover a total uous systems) yield more accurate and consistent measure- width of up to 13 ft with a resolution of 1,280 points. ments. When changing from the older style rut bar to a new These sensors are very accurate; however, the user scanning laser, the Oklahoma DOT (ODOT) found that the needs to input the width of the lane to allow the soft- rut calculations were usually deeper but closer to manual ware to exclude curbs or edge drop-off. Ultrasonic measurements. Older rut bars could under-report rut depth sensors are still used in some systems, but they can because of a lack of full-lane-width coverage. be affected by temperature and moisture. Rut bar width (and lane coverage)--Even with a large Smoothness number of sensors, most older (and still used) rut bars will only cover a certain width of the lane. Further- Given the large variety of devices commercially available, more, even the extendible rut bars will almost never smoothness measurements of the same pavement segment can be fully extended for safety reasons. Typically, the rut show significant variation from device to device (effecting bar is only extended to 10 ft. Some rut bars have the reproducibility). However, measurements by the same device two most lateral sensors angled out to increase the are generally repeatable. The main factors that affect variabil- width of coverage. ity of smoothness measurements include the following (60): Number of sensors--The number of sensors can range from 3 to 37, to continuous, with continuous systems Type of profiler--The various profilers commercially becoming more common. available use different technologies, sensors, and signal Distance measuring system. processing techniques. Equipment operation Height sensor--Most current profilers use a laser Wheel path wander affects noncontinuous measure- sensor; however, some agencies still operate profilers ment because it affects the data points collected. with ultrasonic or infrared sensors. Important sensor

OCR for page 24
26 characteristics include sampling rate, resolution, foot- Wheel path measured--Although some agencies print, and range. report one of the wheel paths, others compute the Sensor footprint--Although traditional laser profil- average of both wheel paths. This is significant because ers use point lasers, agencies are starting to require measurements from the outer (right) wheel path are sensors that have a wide footprint. generally rougher than those in the inner wheel Accelerometer--The type (range) and location of the path (20). accelerometers may affect the measured profile and Bridges--There is lack of agreement on how to deal processed indices. with bridges included within the considered road seg- Distance measurement system. ment. These bridges are often localized areas of high Number of sensors, sensor location, and spacing-- roughness. Some agencies include bridges as part of when multiple sensors are used. the road segment because this better reflects the actual Profiler operation--the manner in which the profiler is user's perception for the overall road segments. Other driven. agencies do not include bridges to avoid artificially Profiler driver and operator--experience, training, and high estimates of the pavement IRI in the segments. driving skills. Measurement environment Lateral position--Wheel path wander is a big source Temperature, wind, humidity, and surface moisture of variation; thus, operators are trained to stay in the affect the various types of sensor differently and can center of the lane. add to the variability of measurements. Longitudinal positioning/triggering. The presence of pavement contaminants, such as Measurement speed--Although it has been hypothe- sand, gravel, etc. sized that measurement speed has an influence on the Lighting conditions affect optical sensors. measured smoothness (61), some of the latest research Surface shape--The condition and texture of the sur- indicates that most profilers produce measurements face affect the accuracy and repeatability of profilers. that are stable with respect to the measuring speed (55). Surface distresses have a major influence on trans- Lane measured--Although most agencies measure versal variations of the profile. only the outermost lane, others are starting to measure Daily and seasonal profile variations are caused by the profile on all lanes. curling of PCC slabs, moisture changes in the sub- Tire inflation pressure affects longitudinal distance grade, freeze and thaw cycles, etc. Differences as high measurements. Calibration of the various components of the as 0.4 m/km have been observed in some sites. equipment. Road geometrics--Cross slope, curves, hill, and grades Profile data interpretation and processing can affect the output of accelerometers, which are key Filters--Most profilers use filters to eliminate un- components of inertial profilers. wanted high and low frequencies in the measured profile; although some allow the user to select the fil- ters, others do not. Surface Friction Properties Profiler computation algorithm--The algorithm used to combine the output from the key component sen- Typical sources of variation in friction measurements include sors and determine the profile. the following. IRI calculation algorithm and procedure--for exam- ple, some profiler manufacturers automatically apply Equipment used--Most state DOTs use the locked- nonstandard filters to the profile. Other manufactur- wheel skid trailer (ASTM E274) for high-speed friction ers and states may choose to average the profiles measurements (62). The system includes a truck with a from the left and right wheel paths before applying water tank and a trailer system that can lock one of the the IRI algorithm, thus generating a half-car rough- wheels for measuring the friction coefficient between tire ness index instead of an IRI from a single wheel path and wet pavement. The wet pavement friction (or skid) or an average of two wheel paths of IRI. number is reported as 100 times the coefficient of friction Integration interval--The length of the segment over (63). Some of the equipment has incorporated laser- which the smoothness is reported is important because based systems to determine the surface macrotexture and the profile elevation data are aggregated. A relatively use it to estimate the gradient of friction with speed. large sum of elevation values can indicate a pavement Operation that is moderately rough over the entire segment or Type of tire--Some states (e.g., Virginia) use smooth very rough over a small section of the entire segment. tires and others (e.g., Florida) still use ribbed tires. Measurements over smaller segments tend to yield Testing speed--Some states have equations for cal- more useful results because short, rough areas are culating the friction number from skid testing at other detected and might be unnoticed if the segments speeds, but those correlations have been only locally were larger. verified (61).