Inertial Profiler Certification for Evaluation of International Roughness Index (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

12 This chapter presents background information about inertial profilers and profile data col- lection. The topics covered in this chapter include a description of inertial profilers, the type of height sensors used in profilers, IRI, effect of texture on data collected by inertial profilers that are equipped with different height-sensor types, reference profilers, certification of inertial profilers, network-level data collection, operational procedures for collecting profile data, and resources for information related to pavement smoothness. Inertial Profilers Inertial profilers record the true profile of a pavement surface that affects ride quality. The principal components of an inertial profiler are the height sensors, accelerometers, the distance-measuring instrument (DMI), and a computer. The height sensor records the height to the pavement surface from the sensor. A profiler typically has two height sensors mounted on the vehicle such that each sensor collects data along a wheelpath. An accelerometer is located on top of each height sensor to record the vertical acceleration. A computer program is used to convert the acceleration to vertical displacement, and then the data from the height sensor are combined with the vertical displacement computed from the accelerometer data to determine the distance to the pavement surface relative to an inertial reference frame. The DMI consists of an encoder fitted to the rear wheel of the vehicle and keeps track of the distance with respect to a reference starting point. Instead of using an encoder, a manufacturer now offers a distance measuring system that is based on using data from a GPS to determine distance. A computer program computes the profile at each data-recording point using the data recorded by the height sensor and the accelerometer. The computed profile data are recorded in the computer and can be used to compute a roughness index. An inertial profiler that collects data at highway speeds is referred to as a high-speed profiler, with the profiling system housed on a van or a truck. High-speed profilers can be classified as profilers where the equipment is permanently fixed to the vehicle or as portable profilers. Figure 1 shows a high-speed profiler, where the profiling equipment is permanently fixed to the vehicle. In this profiler, the sensors that collect profile data are housed inside a sensor bar that is attached to the front of the van. The portable profiling equipment consists of a bar that houses the height sensors, accelerometers, and a signal processing unit. The portable system can be mounted on the trailer hitch of a van or a truck or on the front of a vehicle that is specially configured to mount the system. An encoder is attached to the rear wheel of the host vehicle, and the output from the encoder is sent to the signal processing unit located on the bar. The signal processing unit is then connected to a laptop computer located inside the vehicle by an Ethernet cable. The portable system eliminates the C H A P T E R 2 Background Information on Inertial Profilers and Profile Data Collection

Background Information on Inertial Profilers and Profile Data Collection 13 need to have a dedicated vehicle to house the profiler, as this system can be installed in any suit- able vehicle. Portable profiling systems have been popular with contractors, as they do not need a dedicated vehicle to house the profiler. When not needed, the profiling system can be taken off the vehicle, and the vehicle can be used for other purposes. Such a system can also be shipped to any location and then fitted onto a van or a truck. The term “lightweight profiler” is used to refer to a light utility vehicle that houses a profiling system, as shown in Figure 2. Typically, the total weight of a lightweight profiler is about 950 lb without the operator. Lightweight profilers were developed primarily because such equipment could collect profile data on PCC pavements as soon as the pavement could support the weight of the utility vehicle and the operator. A high-speed profiler that is housed on a van or a truck would have to wait several days before profiling a new PCC pavement until it gained sufficient strength to carry the weight of the vehicle. A lightweight profiler has all the functionality of a high-speed profiler except that the highest speed that it can collect data is limited to the maxi- mum speed of the utility vehicle, which is typically 15 to 20 mph. Although lightweight profilers Figure 1. High-speed profiler. Source: R.W. Perera. Figure 2. Lightweight profiler. Source: Pennsylvania DOT.

14 Inertial Profiler Certification for Evaluation of International Roughness Index were primarily developed to measure the profile of newly placed PCC pavements, they can be used to profile both AC and PCC pavements. AASHTO Standard M 328-14, Standard Specification for Inertial Profiler (AASHTO 2017B), defines the attributes required for an inertial profiling system. This specification covers the follow- ing items related to an inertial profiler: • Requirements for the height sensor, accelerometer, and the DMI; • Calibration of components in the profiler; • Verification of proper functioning of components; • Triggering mechanisms for initiating data collection; • Wavelength range for the collected data; • Data-recording interval; • Software requirements; • Operating speed; • Data storage; • Calculation of roughness indices; and • Operating temperature and humidity ranges. ASTM Standard E 950-09, Standard Test Method for Measuring the Longitudinal Profile of Traveled Surfaces with an Accelerometer Established Inertial Profiling Reference (ASTM 2017C), also covers measurement and recording of profile data. This standard covers the following items: • Components of the profiler; • Requirements for the height sensor, accelerometer, and DMI; • Longitudinal data sampling interval for various classes of profilers; • Data-recording and storage medium; • Calibration procedures for accelerometer, height sensor, and DMI; • Calibration checks; • Measuring speed; • Data filtering; • Data acquisition; • Data evaluation for correctness; and • Procedure to determine the precision and bias of the profiling system. As state DOTs need other information about their highway network, many state DOTs use multifunction data-collection vehicles to collect a variety of data in addition to pavement roughness. Figure 3 shows a multifunction data collection vehicle. A typical multifunction data collection vehicle contains a variety of subsystems, such as • Cameras to collect images of the pavement surface for identifying pavement distress; • Additional cameras to collect the right-of-way view; • Road profiling system to collect longitudinal profile data; • Transverse profile data collection system to obtain rut depths; • Inertial motion system to measure cross fall, grade, and radius of curves; • GPS to collect latitude and longitude data; and • Special laser to collect macrotexture data. Height Sensors Used in Profilers Laser height sensors are currently used in profilers to collect the height of the pavement from the sensor, which is needed for computation of the profile. The laser height-sensor types used in profilers today are either single-spot (SS) lasers, wide-spot (WS) lasers, or line lasers (LL).

Background Information on Inertial Profilers and Profile Data Collection 15 Figure 4 shows the footprint projected by these laser types, where the footprint of the SS laser, WS laser, and the LL are shown from left to right. An SS laser projects a small dot that is typically 0.03 to 0.12 in. in diameter onto the pavement surface. Research studies have shown that SS laser sensors cannot collect accurate profile data on pavements that have a longitudinal texture, such as PCC pavements with longitudinal tining, PCC pavements with longitudinal grooves, or longitudinally diamond-ground PCC pavements (Karamihas and Gillespie 2003; Perera et al. 2009; Fernando and Harrison 2013; de León Izeppi 2013). An SS laser sensor can obtain measurements on the land area between the longitudinal tines (or grooves), as well as on the trough of the tine (or groove) due to lateral wander as the profiler moves along a pavement. These measurements are misinterpreted as roughness and can result in a significant upward bias in the roughness of the pavement. Some profilers use a WS laser that projects a line that is about 0.75 in. long and 0.04 in. wide on the pavement surface. However, there is no published information to show how they will perform on longitudinally textured surfaces, and it is expected this sensor will function similar to a SS laser. LLs that typically project a 4-in.-wide beam on the pavement can overcome the issue that SS laser sensors face on such surfaces. An LL obtains a series of data points along the projected Figure 3. A multifunction data collection vehicle. Source: R.W. Perera. Figure 4. Footprint projected by an SS laser, WS laser, and an LL. Source: Florida DOT.

16 Inertial Profiler Certification for Evaluation of International Roughness Index laser line and then computes a single-bridged data point from this data. This strategy eliminates the problem that SS laser sensors have on longitudinally textured pavements. Therefore, when collecting profile data on pavements that have a longitudinal texture such as longitudinal tining, diamond grinding, or longitudinal grooving, a profiler that is equipped with an LL must be used to obtain repeatable and accurate IRI data. The LL is typically mounted on the profiler such that the projected line is perpendicular to the travel direction, but some manufacturers mount the sensor at an angle to the travel direction. LMI Technologies is the only sensor manufacturer that offers LLs in the U.S. The first LL offered by this manufacturer in the U.S. was the RoLine LL. LMI Technologies discontinued the RoLine sensor in December 2013 and offered an LL called Gocator to replace the RoLine sensor. All profiler manufacturers in the U.S. that offer profilers equipped with LLs currently use the Gocator LL. Profiler manufacturers can use the bridging method offered by LMI Technologies to obtain the single-bridged height data point from the series of data points obtained from the LL or use their own method to obtain the bridged data point using the series of data points collected by the LL. Some manufacturers have developed procedures to obtain profile data from the three- dimensional data collected by camera systems that are used to collect distress data on pavements. An inertial measurement unit (IMU) is attached to the camera in these systems in lieu of an accelerometer to record the vertical acceleration. The height data along the wheelpaths are extracted from the three-dimensional data collected by the camera system and used together with the data from the IMU to compute the profile along the two wheelpaths. International Roughness Index (IRI) The IRI was developed based on the work performed for an experiment in Brazil that was sponsored by the World Bank (Sayers et al. 1986). The computation of IRI is based on a math- ematical model called a quarter-car model. The mathematical model calculates the suspension deflection of a simulated mechanical system with a response similar to a passenger car when simulated along a profile. The simulated suspension motion is accumulated and then divided by the distance traveled to give an index with units of slope (e.g., in./mi). The mathematical simulation that is carried out by the computer program is shown schematically in Figure 5. The quarter-car model used in the IRI algorithm is what the name implies, which is a model of one IRI Measured Profile Figure 5. Illustration of computer algorithm used to compute IRI (Sayers and Karamihas 1998).

Background Information on Inertial Profilers and Profile Data Collection 17 corner (a quarter) of a car. As shown in Figure 5, the quarter car is modeled as one tire that is represented with a vertical spring, the mass of the axle supported by the tire, a suspension spring and damper, and the mass of the body supported by the suspension for that tire. The car having the properties that are used in the computation of IRI is referred to as the “Golden Car” (Sayers and Karamihas 1998). The IRI was mainly developed to match the responses of passenger cars, but subsequent research has shown good correlation with light trucks and heavy trucks (Sayers and Karamihas 1998). The IRI has become recognized as a general-purpose roughness index that is strongly correlated to most kinds of vehicle responses that are of interest. The IRI is highly correlated to three vehicle response variables that are of interest: road meter response (for historical continuity), vertical passenger acceleration (for ride quality), and tire load (for vehicle controllability and safety) (Sayers and Karamihas 1998). The response of the IRI quarter-car filter to different wavelengths is shown in Figure 6. The amplitude of the output sinusoid is the amplitude of the input, multiplied by the gain shown in the figure, which is dimensionless. The IRI is influenced by wavelengths ranging from 4 to 100 ft. However, there is still some response for wavelengths outside this range. The IRI filter has maxi- mum sensitivity to sinusoids with a wavelength of 7.9 and 50.5 ft. The computation of IRI from profile data is performed by a computer program. The ASTM Standard E 1926, Standard Practice for Computing International Roughness Index of Roads from Longitudinal Profile Measurements, presents the computer program to be used to compute IRI (ASTM 2017A). The IRI is calculated for a single profile. The specific steps that are taken in the computer program to compute IRI are as follows (Sayers and Karamihas 1998): 1. The profile is filtered with a moving average having a 9.8-in. base length. The moving aver- age is a low-pass filter (i.e., it attenuates short wavelengths) that smoothens the profile. This moving average filter should be omitted if the profile has already been filtered by a moving average or with an anti-aliasing filter whose cut-off attenuates wavelengths shorter than 2 ft. 2. Quarter-car simulation is performed on the profile. The parameters of the quarter car are defined in the IRI program. The quarter-car simulation on the profile is performed at a simu- lated speed of 50 mph, and the suspension motions of the quarter car are accumulated. 3. The absolute values of the suspension motion that are obtained from the simulation are summed and then divided by the profile length to obtain the average suspension motion over the simulated length. The value that is computed is the IRI and has units of slope (e.g., in./mi). Figure 6. Response of IRI filter (Karamihas 2003).

18 Inertial Profiler Certification for Evaluation of International Roughness Index Profilers collect profile data along two wheelpaths (i.e., left and right wheelpaths), and therefore, IRI values are obtained for the two wheelpaths. The average of the IRI values obtained along the two wheelpaths is referred to as the MIRI and is used to denote the roughness of a roadway. The HRI is the roughness index that is obtained when the IRI algorithm is applied to the average of the left and right wheelpath profiles. The HRI shows a high correlation to the MIRI (Sayers and Karamihas 1998). A regression analysis indicated the relationship between HRI and MIRI to be HRI = 0.89 MIRI (Sayers and Karamihas 1998). Effect of Surface Texture on Data Collected by Laser Height Sensors This section presents information on comparison of IRI values obtained on various texture types from profile data collected with an SS laser and an LL. In all of these comparisons, the LL data were collected with the LL mounted such that the projected laser line was perpendicular to the travel direction, unless noted otherwise. Information regarding IRI values obtained from an SS laser and an LL on the following surface/texture types are presented in this section: • Dense-graded AC, • Stone matrix asphalt (SMA), • Open-graded friction course (OGFC), • Chip seal, • Transversely tined PCC, • Longitudinally tined PCC, • Diamond-ground PCC, and • Longitudinally grooved PCC. Dense-Graded Asphalt Concrete Perera and Karamihas (2010) reported IRI values obtained on two dense-graded AC surfaces by a profiler built by the University of Michigan Transportation Research Institute (UMTRI), which was equipped with an SS laser and a RoLine LL. These two sensors were mounted on the host vehicle such that they collected data along the same path. This profiler collected data at two 200-ft-long test sections located at the National Center for Asphalt Technology (NCAT) test track in Opelika, Alabama. Table 1 shows the IRI values computed from the data collected by the two sensors at the two test sections. The IRI corresponding to the SS laser was higher than that corresponding to the RoLine laser at both sections, with the difference in IRI at the two sections being 2.3 and 1.7 in./mi. The percentage difference in IRI values at the two sections with respect to the IRI corresponding to the RoLine laser were 5.3% and 2.8%. Parameter Section 1 Section 2 IRI SS (in./mi) 45.6 62.4 IRI RoLine (in./mi) 43.3 60.7 Difference in IRI (in./mi) 2.3 1.7 Percentage difference in IRI 5.3 2.8 Table 1. IRI values corresponding to SS laser and a RoLine LL at two dense-graded AC sections (Perera and Karamihas 2010).

Background Information on Inertial Profilers and Profile Data Collection 19 Fernando and Walker (2013) reported a comparison of IRI values obtained from an SS laser and a RoLine laser for data collected on two types of dense-graded AC surfaces (Type C and Type D) in Texas in 2009. The Type C AC mix is a coarser mix than the Type D AC mix, with the maximum aggregate size for the Type C and Type D mix being 1 in. and ¾ in., respectively. Both sensors were mounted on the vehicle such that they collected data along the same path. Data were collected on two projects for each mix type. The total length of the two projects with the Type C mix was 14.4 mi, while the total length of the two projects with the Type D mix was 7.8 mi. The researchers compared the IRI values computed at 0.1-mi intervals from data collected by the two sensors statistically for all projects. Their analysis indicated the IRI corresponding to the SS laser was higher than that corresponding to the RoLine laser for all projects. They reported that the 95% confidence interval for the difference in IRI for the two projects with the Type C mix was 0.9 to 1.1 in./mi and 3.5 to 3.9 in./mi, while this value for the two projects with the Type D mix was 1.9 to 2.1 in./mi and 4.4 to 4.7 in./mi. Fernando and Walker (2013) compared IRI values obtained from data collected by another profiler that was equipped with an SS laser and a RoLine laser that collected data along the same path. Data were collected at test sections located on nine projects in Texas with the length of a test section typically ranging from 0.5 to 1 mi. The researchers computed IRI values at 0.1-mi intervals and compared the difference in IRI values from the two lasers statistically. They reported the average difference in IRI from the SS laser and RoLine laser (i.e., single-point IRI minus RoLine IRI) for these projects were 0.96, 0.67, 0.47, 0.67,-0.84, 0.64, 0.15, 0.66, and 0.77 in./mi. The researchers reported that the 95% confidence interval for the difference in IRI from the SS laser and RoLine laser (i.e., SS IRI minus RoLine IRI) was positive for seven of the eight projects. Hence, for seven of the eight projects tested, the IRI from the SS laser was slightly greater than the IRI from the RoLine laser. Fernando and Walker (2013) also reported results from a study where data were collected by a profiler equipped with an SS laser and a RoLine laser that collected data along the same path. Data were collected at two test sections located at the Texas A&M University System Riverside campus. One test section was a smooth section, and the other section was a medium-smooth section. The profiler obtained three repeat runs at both sections, and the authors compared the average IRI obtained from the collected data. The researchers reported that the IRI corresponding to the SS laser data was higher than that corresponding to the RoLine laser data at both sections. The researchers reported that the difference in IRI for the smooth and medium-smooth sections was 3.6 and 3.2 in./mi, respectively. The results presented above show that on a dense-graded AC surface, the IRI corresponding to the data collected by an SS laser was higher than that corresponding to an LL. The LL obtains a bridged elevation value from a series of transverse data points. Because of the bridging that is performed, the profile obtained from an LL is expected to be slightly smoother than the profile obtained from an SS laser on the same surface. The slightly lower IRI noted for the data from the LL is attributed to this reason. The surface texture of a dense-graded AC surface can vary according to the nominal maximum aggregate size and the gradation of the aggregates used for the AC mix. Therefore, the difference in IRI noted for these two sensor types on a dense-graded AC surface could vary depending on the macrotexture of the pavement surface. Stone Matrix Asphalt SMA is a gap-graded AC and is designed to minimize rutting by providing stone-on-stone contact. Most of the SMA pavements placed in the U.S. have a 0.5- or a 0.75-in. nominal maximum aggregate size (Brown and Cooley 1999). An SMA pavement will have a higher macrotexture than a dense-graded AC mix because of the coarse nature of the mix.

20 Inertial Profiler Certification for Evaluation of International Roughness Index Perera and Karamihas (2010) reported IRI values obtained on an SMA surface by an UMTRI- built profiler that was equipped with an SS laser and a RoLine laser. The two sensors were mounted on the host vehicle such that they collected data along the same path. This profiler collected data at a 200-ft-long SMA test section at the NCAT test track in Opelika, Alabama. The researchers reported that the IRI values corresponding to the data collected by the SS laser and the RoLine laser were 74.7 and 71.2 in./mi, respectively. The IRI corresponding to the SS laser data was 3.5 in./mi higher than that for the data corresponding to the RoLine laser. Fernando and Walker (2013) reported a comparison of IRI values obtained from data collected by an SS laser and a RoLine laser on two projects in Texas that had an SMA surface. The two sensors were mounted on the profiler such that they collected data along the same path. Data were collected on two projects totaling 5.5 mi, and IRI values were computed at 0.1 mi intervals. The researchers compared the IRI values corresponding to the two sensors statistically. Most of the IRI values for the 0.1-mi sections were between 50 and 100 in./mi. Their analysis indicated the IRI corresponding to the data from the SS laser to be higher than that corresponding to the RoLine laser, with the 95% confidence interval for this difference for the two projects being 2.6 to 3.0 in./mi and 1.8 to 2.4 in./mi. The results presented above show that on an SMA surface, the IRI corresponding to data from an SS laser was higher than that corresponding to an LL. As described for the dense-graded AC surface, the lower IRI corresponding to the LL data is attributed to the bridging performed by the LL, which will result in a smoother profile compared to the profile obtained by an SS laser. The surface texture of an SMA surface can vary according to the nominal maximum aggregate size and the gradation of the aggregate used for the SMA. Therefore, the difference in IRI obtained for data collected by the two sensors on an SMA surface could vary depending on the macrotexture of the pavement surface. Open-Graded Friction Course An OGFC has a high macrotexture compared to a dense-graded AC surface and provides a high level of skid resistance, particularly in wet weather. The OGFC is placed on top of a dense- graded AC or SMA surface. An OGFC is porous and has interconnected voids, thus providing high permeability to remove water from the pavement surface (Huber 2000). According to Huber (2000), the normal thickness of OGFC surfaces used in the U.S. is 0.75 in., with the nominal maximum aggregate size being 0.375 in. and the air voids content in the mix typically being 15%. Perera and Karamihas (2010) reported IRI values obtained on four OGFC surfaces by an UMTRI-built profiler equipped with an SS laser and a RoLine laser. The two sensors in the profiler were mounted on the host vehicle such that they collected data along the same path. This profiler collected data at four 200-ft-long test sections that had an OGFC at the NCAT test track in Opelika, Alabama. The IRI values obtained from the data collected by the two sensors at the four test sections are shown in Table 2. The IRI corresponding to the data from the SS laser was Parameter Section Number 1 2 3 4 IRI SS (in./mi) 68.0 69.3 65.9 69.7 IRI RoLine (in./mi) 60.5 68.1 62.4 63.0 Difference in IRI (in./mi) 7.5 1.2 3.5 6.7 Percentage difference in IRI 12.4 1.8 5.6 10.6 Note: Percentage difference in IRI with respect to RoLine IRI. Table 2. IRI values corresponding to an SS laser and a RoLine laser on OGFC surfaces (Perera and Karamihas 2010).

Background Information on Inertial Profilers and Profile Data Collection 21 higher than that for the RoLine laser at all four sections, with the difference in IRI ranging from 1.2 to 7.5 in./mi and the average difference being 4.7 in./mi. Fernando and Walker (2013) reported a comparison of IRI values obtained from an SS laser and a RoLine laser for data collected on permeable friction course (PFC) surfaces in Texas. A PFC is similar to an OGFC. Both lasers were mounted on the same vehicle such that they collected data along the same path. Data were collected on three projects totaling 21 lane miles. The researchers computed IRI values for the collected data at 0.1-mi intervals and compared the IRI values corresponding to the two sensors statistically. The IRI values for most of the 0.1-mi intervals sections were between 30 and 60 in./mi. The researchers reported that the analysis indicated the IRI corresponding to the SS laser data to be higher than that corresponding to RoLine laser data, with the 95% confidence interval for this difference for the three projects being 6.2 to 6.7 in./mi, 5.8 to 6.5 in./mi, and 10.3 to 10.7 in./mi. Fernando and Walker (2013) also reported IRI values computed from data collected by a profiler that was equipped with an SS laser and a RoLine laser that collected data along the same path on PFC test sections located on six projects in Texas. The length of a test section on a project typically ranged from 0.5 to 1 mi. IRI values were computed from the collected data at 0.1-mi intervals, and the IRI values corresponding to data from the two sensors were compared statistically. The researchers reported the following average IRI differences for the two lasers (i.e., IRI from SS laser minus IRI from RoLine laser) used for the six projects, with the 95% confidence interval for the difference in IRI indicated in parenthesis: 3.21 in./mi (2.50 to 3.92 in./mi), 3.73 in./mi (3.03 to 4.43 in./mi), 6.77 in./mi (6.31 to 7.23 in./mi), 5.27 in./mi (4.69 to 5.84 in./mi), 3.72 in./mi (3.28 to 4.17 in./mi), and 3.64 in./mi (3.11 to 4.18 in./mi). The results presented above show that on OGFC surfaces, the IRI corresponding to data collected by an SS laser was higher than that corresponding to an LL. As described for the dense-graded AC surface, the lower IRI corresponding to the LL data is attributed to the bridging performed by the LL, which will result in a smoother profile compared to the profile obtained by an SS laser. The macrotexture of an OGFC surface can vary depending on the nominal maximum size of the aggregate used in the mix. Therefore, the difference in IRI obtained for the two sensor types on an OGFC surface may vary depending on the macrotexture of the pavement surface. Chip Seal A single chip seal consists of a layer of asphalt binder covered by one-stone-thick aggregate. Chip seal is a preventive maintenance strategy, with the primary purpose of sealing fine cracks in the pavement surface to prevent intrusion of water into the base and the subgrade. The aggregate used in the chip seal improves the skid resistance of the existing pavement by increasing the macro- texture. Most agencies use a nominal maximum aggregate size that ranges from 0.375 to 0.5 in. for a single chip seal, with the most common size being 0.375 in. (Gransberg and James 2005). The ideal gradation for an aggregate used as a chip seal is one where all of the stone particles are close to one size (Gransberg and James 2005). Fernando and Walker (2013) reported a comparison of IRI values obtained on chip seal surfaces from data collected by a profiler equipped with an SS laser and a RoLine laser. The two sensors were mounted on the vehicle such that they collected data along the same path. Data were collected on a project that had a Grade 3 chip seal (total mileage of 13.6 mi) and on a project that had a Grade 4 chip seal (total mileage 24.3 mi). The Texas construction specifications indicated the maximum stone size for a Grade 3 and a Grade 4 chip seal to be 3⁄4 and 5⁄8 in., respectively. The researchers computed IRI values at 0.1-mi intervals for data collected by both sensors and compared the IRI values corresponding to the data collected by the two sensors statistically. The researchers reported the IRI corresponding to the SS laser to be higher than that corresponding

22 Inertial Profiler Certification for Evaluation of International Roughness Index to the RoLine laser, with the 95% confidence interval for this difference for the Grade 3 chip seal being 3.0 to 3.3 in./mi and for the Grade 4 chip seal being 5.8 to 6.1 in./mi. Briggs (2009) reported IRI values obtained on chip seal surfaces in Texas by a profiler that was equipped with an SS laser and a RoLine laser that collected data along the same path. Data were collected on a 4.8-mi-long section that had a Grade 3 chip seal and a 5.2-mi-long section that had a Grade 4 chip seal, with data being collected in both directions. The Grade 3 chip seal had stones that were slightly larger than the Grade 4 chip seal. The authors reported the Grade 3 chip seal appeared to be older, and the aggregates were more embedded into the asphalt when com- pared to the Grade 4 chip seal. Briggs (2009) compared IRI values computed at 0.1-mi intervals. On the Grade 3 chip seal, the majority of the 0.1-mi-long segments had IRI values between 80 and 130 in./mi, while for the Grade 4 chip seal the majority of the 0.1-mi-long segments had IRI values between 60 and 100 in./mi. Briggs (2009) reported that for the Grade 3 chip seal (in one direction) a slight systemic difference in IRI was noted, with the IRI corresponding to the SS laser data being higher. In the other direction, a systemic difference in IRI was also noted, with the IRI corresponding to the SS laser data being higher by about 5 in./mi. However, this value diminished with increasing IRI. Briggs (2009) reported that for the Grade 4 chip seal, a systemic difference in IRI was noted for the data from two lasers in both directions, with the IRI corresponding to the SS laser data being higher. Briggs (2009) reported that on average, this difference was about 6 to 7 in./mi. Briggs (2009) also reported that the difference between the SS laser data IRI and RoLine laser data IRI was higher on the Grade 4 chip seal when compared to Grade 3 chip seal, although the size of stones used for the Grade 4 chip seal was smaller. Briggs (2009) reported that the stones on the Grade 3 chip seal had a greater depth of embedment than the stones on the Grade 4 chip seal, and there was more bleeding on the Grade 3 chip seal. Briggs (2009) indicated that these two factors appear to have reduced the macrotexture depth on the Grade 3 chip seal, and this is likely why the IRI for the data from two sensors were closer to each other on the Grade 3 chip seal than the Grade 4 chip seal. The results presented above show that on a chip seal, the IRI corresponding to SS laser data was higher than that corresponding to data from an LL. As described for the dense-graded AC surface, the lower IRI corresponding to the LL data is attributed to the bridging performed by the LL, which will result in a smoother profile compared to the profile obtained by an SS laser. The magnitude of the difference in IRI for data from the two sensor types is expected to vary depending on the macrotexture of the surface. Factors that can affect the macrotexture of a chip seal surface are the size of the stone used for the chip seal, the depth of embedment of the stone, and the amount of bleeding. Transversely Tined PCC Fernando and Walker (2013) compared IRI values computed from data collected by a pro- filer that was equipped with an SS laser and a RoLine laser on continuously reinforced concrete (CRC) pavements that had transverse tining. The two sensors were mounted on the profiler such that they collected data along the same path. The profiler collected data at test sections located on six projects in Texas that had a tine spacing of 1 in. The length of a test section typically ranged from 0.5 to 1 mi. The researchers computed IRI values from the collected data for both sensors at 0.1 mi intervals and then compared the IRI values statistically. The following are the average IRI differences for the two lasers (i.e., IRI from SS laser minus IRI from RoLine laser) for the six projects, with the 95% confidence interval for the difference in IRI reported by the researchers indicated in parenthesis: 0.90 in./mi (0.46 to 1.40 in./mi), 0.11 in./mi (-0.40 to 0.60 in./mi), 0.68 in./mi (0.02 to 1.33 in./mi), 0.83 in./mi (0.20 to 1.46 in./mi), -1.4 in./mi (-2.3 to -0.54 in./mi), and -0.58 in./mi (-1.21 to 0.03 in./mi). The average difference indicated the IRI corresponding to the RoLine laser was slightly higher than the IRI corresponding to the SS laser for four projects, with the reverse occurring for the other two projects.

Background Information on Inertial Profilers and Profile Data Collection 23 Fernando and Walker (2013) presented IRI values reported by a profiler manufacturer who collected data on a transversely tined PCC section in Texas with a profiler that was equipped with both an SS laser and a RoLine laser. The sensors were configured such that they collected data along the same path, with the RoLine laser being at a 30° angle from the direction perpendicular to the travel direction. The majority of the 0.1-mi-long segments on the roadway had IRI values between 40 and 60 in./mi. The researchers reported that the data from the RoLine laser seemed to produce marginally lower IRI values when compared to data from the SS laser, but the differences were not regarded to be significant. Fernando and Walker (2013) also presented results from a study that compared IRI values collected on transversely tined PCC pavements with a RoLine laser oriented at different angles. The following three orientations were studied: laser mounted such that the projected line was perpendicular to travel direction (0°), projected line at a 30° angle from the direction perpendicular to travel direction (30°), and projected line at a 45° angle from the direction perpendicular to the travel direction (45°). Testing was performed on three projects. The researchers reported that the data indicated the 45° position resulted in the lowest IRI on all three projects. The average differ- ence in IRI between the 0° and 45° positions for the three projects were 1.5, 3.4, and 6.2 in./mi. The information presented above indicates that the IRI obtained from an SS laser on a transversely tined PCC section could be close to the IRI obtained from a RoLine laser. The orientation of an LL with respect to the travel direction appears to influence the IRI values obtained on transversely tined PCC surface. The difference in magnitude of the IRI between the two sensor types could depend on the depth of the tines, tine spacing, width of the tines, and the orientation of the LL. Longitudinally Tined PCC Fernando and Harrison (2013) reported the IRI values obtained on a CRC pavement with longitudinal tining that was profiled with a profiler equipped with a RoLine laser and two profilers equipped with SS lasers. This pavement had a longitudinal tine spacing of 1 in. The authors reported that the IRI values computed from data collected by the two profilers with SS lasers were higher than those computed for data collected by the profiler with the RoLine laser for both wheelpaths. The difference in IRI between the average IRI from the two profilers that had SS lasers and the profiler with the RoLine laser was 42 and 43 in./mi for the left and the right wheelpaths, respectively. Karamihas and Gillespie (2003) indicated that the slow drift of a height sensor with a narrow footprint (i.e., SS laser) into and out of the tines on longitudinally tined concrete (LTC) intro- duces significant content into the profile that would be misinterpreted as roughness that affects ride quality. This is the reason for the upward bias in IRI that is noted for data collected on longitudinally tined PCC pavements by an SS laser sensor. The information presented above indicates that the difference in IRI corresponding to data obtained from an SS laser and an LL on a longitudinally tined PCC surface could be very large. The difference in the magnitude of IRI corresponding to the two laser types may depend on the tine spacing, width of the tine, and the depth of tining. Diamond-Ground Concrete De León Izeppi (2013) presented IRI values obtained from data collected by seven profilers equipped with SS lasers and a profiler equipped with a RoLine laser on a diamond-ground pave- ment located at the Smart Road in Blacksburg, Virginia. The author of this synthesis averaged and compared the IRI values obtained from the data collected by the seven SS profilers to the

24 Inertial Profiler Certification for Evaluation of International Roughness Index IRI from the profiler that was equipped with RoLine laser. The average IRI for the data from the seven profilers that had SS lasers for the left and right wheelpaths was 129 and 108 in./mi, respectively. The IRI for the data collected by the RoLine laser for the left and the right wheelpaths was 108 and 92 in./mi. Overall, these data indicated that on this surface, the IRI corresponding to SS laser data was higher than that corresponding to RoLine data by 21 and 16 in./mi for the left and the right wheelpaths, respectively. Perera et al. (2009) reported IRI values obtained on two diamond-ground pavements in Florida for data collected by an SS laser and a WS laser, which were mounted on a profiler such that both lasers collected data along the same path. The WS laser had a spot size that was 0.67 in. wide, and the laser was installed on the vehicle such that the 0.67-in. side was perpendicular to the direction of travel. The profiler collected data on two diamond-ground projects: I-275 and I-75. The profiled length for the I-275 project was 2.1 mi, while that for the I-75 project was 3.1 mi. The overall IRI for the I-275 project for data from the SS laser and the WS laser was 62 and 38 in./mi, respectively. The overall IRI for the I-75 project for the data collected by the SS laser and the WS laser was 62 and 54 in./mi, respectively. The IRI difference for the data collected by the two laser types for the I-275 and I-275 projects was 24 and 8 in./mi, respectively. The grooves from the diamond-grinding operation were reported to be deeper on the I-275 project when compared to the I-75 project. The difference in IRI corresponding to the two sensor types was higher on the I-275 project, where the diamond grinding created deeper grooves. Although a RoLine laser was not used in these two projects, the data presented show how the laser spot size affects the IRI values on diamond-ground surfaces. The reason for the upward bias in IRI for data collected by an SS laser on a diamond-ground pavement is similar to the reason that is causing an upward bias in longitudinally tined PCC described previously. The drift of the SS laser into and out of the grooves in the ground pavement introduces significant content into the profile that would be misinterpreted as roughness that affects ride quality. The information presented above indicates that the IRI obtained from data collected by an SS laser on a diamond-ground PCC surface was higher than that for data from a RoLine and a WS laser. The difference in the magnitude of IRI from two laser types on a diamond-ground surface could depend on blade width of cutters on the grinder, spacer spacing on the grinding head, and the depth of the grooves on the ground surface. Longitudinally Ground and Grooved Concrete De León Izeppi (2013) presented IRI values obtained from data collected by seven profilers equipped with SS lasers and a profiler equipped with RoLine lasers on a longitudinally ground and grooved pavement section located at the Smart Road in Blacksburg, Virginia. The author of this synthesis averaged and compared the IRI values obtained from the data collected by the seven SS profilers to the IRI from the profiler that was equipped with RoLine lasers. The average IRI obtained by the seven profilers with SS sensors for the left and right wheelpaths was 135 and 122 in./mi, respectively, while the IRI from the RoLine laser for the left and the right wheelpaths was 38 and 29 in./mi, respectively. These data indicated that on this surface, the IRI corresponding to the SS laser data was higher than the IRI corresponding to the data from the RoLine laser by 97 and 93 in./mi for the left and the right wheelpaths, respectively. The reason for the upward bias in IRI for data collected by a SS laser on grooved pavements is similar to the reason that is causing an upward bias in longitudinally tined PCC that was described previously. The drift of the SS laser into and out of the grooves in the pavement introduces significant content into the profile that would be misinterpreted as roughness that affects ride quality. The information presented previously indicates that the difference in IRI obtained from data collected by an SS laser and an LL on a longitudinally ground and grooved

Background Information on Inertial Profilers and Profile Data Collection 25 PCC surface can be very large. The difference in the magnitude of IRI for the data collected by the two laser types on such a surface could depend on groove spacing, groove width, and the groove depth. Implications of Sensor Type on Network-Level Data and Smoothness Acceptance Testing The type of sensor that was used to collect the profile data for the computation of IRI on PCC pavements that have longitudinal tining, longitudinal grooving, or diamond grinding is an issue that will be faced at the national level if comparisons are made for IRI of PCC pavements among the states. If some state DOTs use SS lasers and others use LLs, then the comparison of IRI levels among states will not be valid for these PCC pavements. A state DOT that has PCC pavements with longitudinal tining, longitudinal grooving, or diamond grinding needs to make a decision about whether it wants to require LLs to be used for network-level data collection. If a profiler equipped with an SS laser is used to collect data on such surfaces, the obtained IRI value is expected to have an upward bias. If the percentage of pavements with such surfaces in the state DOTs’ network is small, the upward bias in rough- ness on such sections may not have much influence on the overall network-level roughness. However, if the state DOT has a significant amount of such pavements in its network, the overall network-level roughness could be affected. Also, the IRI values obtained by an SS laser on such surfaces can vastly overestimate the roughness of such sections and may create issues when using the Pavement Management System to compare IRI values among pavement segments. It should be noted that if a state DOT decides to use LLs for network-level data collection, then that choice can also affect the IRI obtained for other types of pavements when compared to IRI corresponding to data collected by an SS laser. A noticeable reduction in IRI may occur for OGFC surfaces, while a slight reduction in IRI is expected on dense-graded AC, SMA, and chip seal surfaces. The information in the literature indicates that when collecting data for smoothness accep- tance on the final paved surface of a roadway, LLs need to be used for PCC pavements that have longitudinal tining, longitudinal grooving, or diamond grinding to obtain accurate IRI values. Data collected with an SS laser can vastly overestimate the IRI on such surfaces. As indicated by Karamihas and Gillespie (2003), the upward bias in IRI is caused by the drift of the SS laser into and out of the longitudinal texture that introduces significant content into the profile that would be misinterpreted as roughness that affects ride quality. As described previously, using an LL to collect data for smoothness acceptance can also have an effect on OGFC, SMA, and dense-graded AC, where data collected with an LL can result in lower IRI when compared to an SS laser. A small change in IRI can make a difference between getting a positive pay adjustment or getting full pay, getting a higher positive pay adjustment, or getting full pay or having a negative pay adjustment. Therefore, using a profiler equipped with LLs on these surfaces will definitely be an advantage for the contractor. Reference Profilers To evaluate the accuracy of an inertial profiler, the data collected by an inertial profiler has to be compared to a reference profile. Equipment that is used to collect a reference profile is referred to as a reference profiling device. Measurements made with a rod and level were historically used to evaluate the data collected by an inertial profiler. ASTM Standard E 1364-95, Standard Test Method for Measuring Road Roughness by Static Level Method (ASTM 2017D), describes the procedures to be followed to collect

26 Inertial Profiler Certification for Evaluation of International Roughness Index rod-and-level data at a test section. The most important factor when collecting rod-and-level data is to make sure that the resolution of the level meets the requirements outlined in the ASTM standard. Levels that are used for standard surveying work do not meet the resolution require- ments indicated in ASTM Standard E 1364-95. In the 1990s, the Dipstick and the Australian Road Research Board (ARRB) Walking Profiler were commonly used to measure reference profiles. The Dipstick is a device that is walked along the pavement to collect data, while the ARRB Walking Profiler is pushed along the pavement to collect data. The Dipstick recorded data at 12-in. intervals, while the ARRB Walking Profiler recorded data at 9.5-in. intervals. With advances in technology, inertial profilers were capable of recording data at 1-in. intervals by the late 1990s. Therefore, to evaluate the data collected by such devices, a need arose for reference devices that were capable of collecting data at 1-in. intervals. Although the rod and level could be used to obtain data at 1-in. intervals, the time needed to perform such a survey would be excessive. In the early 2000s, push-along devices that were capable of collecting data at 1-in. intervals were developed. Currently, there are three manufacturers in the U.S. that offer reference devices that are capable of collecting data at 1-in. intervals. These manufacturers and the devices they market are • International Cybernetics Corporation (ICC): SurPRO, • Surface Systems and Instruments (SSI): Walking Profiler, and • ARRB Group: Walking Profiler G3. All of these devices are pushed along the pavement at a walking speed to collect data. The ARRB Walking Profiler G3 is different from the earlier version of the ARRB Walking Profiler that recorded data at 9.5-in. intervals, as this device can record data at 1-in. intervals. SurPRO, which is manufactured by ICC, is shown in Figure 7. This device has two wheels and is pushed along the pavement to obtain elevation measurements. It is capable of recording elevation data at user-selectable intervals that range from 0.5 to 12 in. SurPRO can be operated at speeds up to 2.5 mph, but the manufacturer recommends a lower speed for obtaining best results on rough or coarse-textured surfaces. Figure 8 shows the SSI Walking Profiler, which can record elevation data at 1-in. intervals. This device has three wheels on one side and two wheels on the other side. It is pushed with Figure 7. SurPRO. Source: R.W. Perera.

Background Information on Inertial Profilers and Profile Data Collection 27 the three wheels traversing the path where measurements are needed. A Panasonic Toughbook computer placed on a stand attached to the handle of the device is used to record the collected data. The SSI Walking Profiler can be operated at speeds up to 3 mph, but the manufacturer recommends a slower speed for best results on rough or coarse-textured surfaces. Figure 9 shows the G3 Walking Profiler. Manufactured by the ARRB Group, it is capable of recording data at 1-in. intervals. This device is equipped with three wheels: one in the front and two at the back. It features two additional small wheels inside the equipment that are in line Figure 8. SSI CS 8800 Walking Profiler. Source: SSI. Figure 9. ARRB G3 Walking Profiler. Source: R.W. Perera.

28 Inertial Profiler Certification for Evaluation of International Roughness Index with each other longitudinally and in line with the front wheel. These two wheels are in contact with the pavement surface, and the movement of these wheels is used to obtain measurements. The G3 Walking Profiler also has a laser sensor. The data collected are recorded on a tablet. The device has a pointer on top that can be rotated in a horizontal plane, which can be used as a guide to follow any pre-placed longitudinal mark to collect data at a specified offset to that mark. The FHWA sponsored reference profiler evaluation studies in 2009, 2010, 2013, and 2015. Reference profiler manufacturers were invited to bring their devices to these evaluations, where data collected by the reference devices at several test sections having diverse texture types were compared with data collected with a benchmark profiler that was developed by UMTRI. The benchmark profiler was developed using funding from the FHWA (Winkler et al. 2016) and was deemed to be collecting the ground truth data to evaluate the reference profilers that participated in the evaluations. Results from the reference profiler evaluations performed in 2009 and 2010 are documented by Karamihas (2011). Karamihas and Perera (2014) and Perera and Karamihas (2017) described the results obtained from the evaluation performed in 2013 and 2015, respectively. These reports document the repeatability and the accuracy of the reference devices that participated in these studies. Certification of Inertial Profilers Certification of an inertial profiler involves confirming that the profiler can collect repeatable and accurate data based on a specified set of criteria. Profiler certification involves performing the following steps: 1. Lay out test sections. 2. Perform profile measurements longitudinally along the two paths that will be traversed by the sensors in the profiler using a reference profile device. 3. Perform repeat measurements on each test section with the profiler. 4. Use the collected data to evaluate the repeatability and the accuracy of the profiler. The test sections should be selected such that they encompass the types of surface textures on which the profiler will collect data. The test sections should also have varying roughness levels to evaluate the performance of the profiler on smooth as well as rough pavements, depending on the purpose for which the profiler is used. The repeatability and the accuracy of data collected by each sensor in the profiler must be evaluated separately. The distance between the two wheelpaths where data are being collected should be specified in the state DOT document addressing profiler certification. Before collect- ing data with the profiler at the certification sections, the spacing between the two sensors in the profiler must be checked to ensure the sensor spacing will match the distance between the two wheelpaths that were laid out at the test sections to collect the reference profile data. Perera (2014) summarized the profiler certification procedures used by several states. Standards for Certifying Profilers The following ASTM standard and AASHTO standard describe procedures that can be used for certifying a profiler: • ASTM E950-09, Standard Test Method for Measuring the Longitudinal Profile of Traveled Surfaces with an Accelerometer Established Inertial Profiling System (ASTM 2017C). • AASHTO R 56-14, Standard Practice for Certification of Inertial Profiling Systems (AASHTO 2017C).

Background Information on Inertial Profilers and Profile Data Collection 29 ASTM Standard E950-09 This standard was first published in 1983. The standard contains a procedure for evaluating the repeatability and accuracy of a profiler by determining the precision and the bias of a profiler. The standard indicates that a test section that is 1,056-ft long should be established, and at least 10 repeat measurements from the profiler must be obtained at the test section. Profile data at 1-ft intervals are used to compute the precision of the profiler. The precision is evaluated by performing the following steps: 1. The standard deviation of the elevation data at each data-recording interval of the profiler is computed using the data obtained from the 10 runs. 2. These standard deviation values are averaged to obtain a mean standard deviation that is used to judge the precision of the profiler. The standard presents the precision requirements for three equipment classifications (i.e., Class 1 through Class 3). Profile data and reference data at 1-ft intervals are used to compute the bias of the profiler with the following steps: 1. Reference data are filtered using the identical filter that was applied to the data from the inertial profiler. 2. At each data-recording interval of the profiler, the elevation values obtained from repeat profiler measurements are averaged to obtain a mean elevation value. 3. The difference between the filtered reference device elevation and the mean elevation from the profiler measurements at each data location are computed. 4. The absolute values of the difference are summed and divided by the total number of data points to obtain the bias of the profiler. The standard presents bias requirements for three equipment classifications (i.e., Class 1 through Class 3). This standard indicates that the data collected by a profiler should be filtered such that no attenuation or amplification occurs for wavelengths up to 200 ft. Typically, an upper wavelength cut-off filter of 300 ft is applied to data collected by an inertial profiler. Agreement among profiles that have been subjected to a 300-ft upper wavelength cut-off filter is dominated by the agree- ment in long wavelengths. This phenomenon presents a shortcoming in the method presented in this standard for computing the precision and bias of a profiler. It is possible for the profile data to have significant differences in shorter wavelengths that affect ride quality but yet for the profiler to meet the precision requirement. Also, it is possible to have significant differ- ences in shorter wavelengths between the inertial profiler data and the reference profile data that will have an impact on the ride quality and yet for the inertial profile data to meet the bias criterion. AASHTO Standard R 56-14 This standard describes a certification procedure for inertial profilers and provides guidance on procedures to certify profiler operators. The data analysis procedures and threshold values recommended in this standard were established to ensure adequate determination of the IRI and the profile features that affect it. This specification covers the following items related to certification: • Block check to determine the accuracy of the height sensors; • Bounce test to determine if the accelerometers are canceling out the motion of the vehicle and if the signals from the height sensor and the accelerometer are properly combined to compute the profile; • Check of the distance measured by the device to evaluate the accuracy of the DMI; • Test sections to be used for dynamic certification testing;

30 Inertial Profiler Certification for Evaluation of International Roughness Index • Procedures to perform dynamic certification testing; • Procedures to analyze the data to determine equipment repeatability and accuracy using the cross-correlation technique; • Procedures to verify the computed ride statistics (i.e., IRI) from the systems software and procedures to report results from testing; and • Procedures to test if the profile operator is qualified to operate an inertial profiler. This standard indicates that the profiler must first pass the block check and the bounce test to verify that the height sensor and the accelerometer are functioning properly before the profiler is allowed to collect data at the test sections. The standard also indicates that the DMI of the device must pass a DMI check before collecting data at the test sections. The standard recommends using a section that is at least 1,000-ft long to check the DMI. The standard recommends perform- ing three runs each at the lowest and highest speeds of the profiler at the DMI section using the auto-trigger in the profiler to initiate and terminate data collection. The distances recorded by the profiler are then used to compute the absolute difference between the DMI reading and the known distance of the test section. The standard indicates that the average of the absolute dif- ferences must be less than 0.15% of the distance of the test section for the profiler to pass the DMI test. AASHTO R 56-14 specifies the use of two test sections: a smooth section where the IRI is between 30 and 75 in./mi and a medium-smooth section where the IRI is between 95 and 135 in./mi for certifying profilers that collect data for construction acceptance. The standard indicates that an agency may elect to perform certification only on the smooth section for pro- filers that are used to perform testing for construction acceptance. For certifying network-level profilers, a test section having an IRI up to 200 in./mi is recommended to be used in addition to the smooth and medium-smooth section. The standard recommends all surface types on which the profiler is expected to obtain measurements be included in the certification process. The standard indicates that each test section should be at least 528-ft long. A profiler is required to collect 10 repeat runs at each test section, with five runs being collected at the lowest desired certification speed and the other five runs collected at the highest desired certification speed. The standard indicates that reference data at the test sections should be collected with a device having a data-recording interval of at least 1 in. AASHTO R 56-14 specifies the use of the cross-correlation method to evaluate the repeatability and accuracy of the data collected by the profiler. The repeatability and accuracy are evaluated separately for the data collected by each sensor of the profiler. The cross-correlation method included in AASHTO R 56-14 is an objective method of assessing profile agreement, which was applied to profile data by Karamihas (2004). This procedure is based on the cross-correlation function described by Bendant and Piersol (1971). The cross-correlation method can be used to rate agreement between profiles in a given waveband or to rate agreement between devices for any given roughness index, including the IRI. This procedure provides a rating agreement ranging from 0 to 1 that describes how well two profiles correlate with each other, with 1 indicating perfect agreement. The rating agreement provided by this procedure represents repeatability when it is applied to two measurements of the same section by the same device, reproducibility when it is applied to two measurements of the same section by different devices, and accuracy when a measurement from one of the devices is deemed correct. Cross-correlation is superior to direct comparison of IRI values because it compares the spatial distribution of roughness between two profiles. The IRI algorithm is often thought of as an analysis procedure that takes a profile as input and produces an index value as output. However, it is actually a linear filter that takes a profile as input (in units of elevation) and produces a modified signal in units of slope as output at each data-recording interval in the profile (Karamihas 2005). The modified signal is produced

Background Information on Inertial Profilers and Profile Data Collection 31 by simulating the motion of a quarter car on the profile. The filter used in the IRI algorithm to obtain the modified signal (in units of slope) is referred to as the IRI filter. Figure 10 shows an example of an IRI-filtered profile for data collected by an inertial profiler that recorded data at 1-in. intervals at a 528-ft-long section prepared by the author of the synthesis. Although Figure 10 appears to show a continuous plot, this plot contains individual data points at 1-in. intervals. The IRI reported for a segment is the average rectified value of the output signal. The average rectified value of the data points shown in Figure 10 is 78 in./mi, which denotes the overall IRI of the 528-ft-long section. The following procedure is specified in AASHTO R 56-14 for evaluating the ability of the profiler to collect repeatable IRI data at a test section: 1. Apply the IRI filter to all repeat profiles collected by the profiler. 2. Calculate the cross-correlation between two profiles that have been subjected to the IRI filter for all possible run combinations (i.e., for 10 repeat runs, there are 45 possible run combinations). When performing cross-correlation between two profiles, keep one profile fixed and shift the other profile 3 ft in either direction from the start, calculating the cross- correlation value at every possible offset. The maximum cross-correlation value found over the 6-ft range is the cross-correlation value between the two profiles. 3. Compute the average cross-correlation (i.e., average the 45 cross-correlation values), which is referred to as the repeatability score for the profiler. The standard indicates that the repeatability score must be 0.92 or greater for a profiler to pass the repeatability criterion at a test section. The standard indicates that a lower score may be acceptable for surfaces having an IRI greater than 150 in./mi; surfaces having an IRI greater than 150 in./mi are not used to evaluate profilers that collect data for construction acceptance. The standard indicates that a repeatability score of 0.92 suggests that the IRI values of the two profiles agree within 5% of each, with a 95% confidence level. The following procedure is specified in AASHTO R 56-14 for evaluating the accuracy of a profiler at a test section: 1. Apply the IRI filter to the reference profile and to all inertial profiler runs. 2. If the data-recording interval for the inertial profiler and the reference profile are different, interpolate the inertial profiler profile to the recording interval of the reference profile. 3. Calculate the cross-correlation between the IRI-filtered reference profile and the IRI-filtered profile from the inertial profiler for all runs of the inertial profiler. For the 10 repeat runs of the profiler, 10 cross-correlation values will be computed by comparing each repeat run to the reference profile. When performing cross-correlation between a profile from the inertial profiler and the reference device, fix the reference profile and shift the profile from the inertial Figure 10. Example of an IRI-filtered profile.

32 Inertial Profiler Certification for Evaluation of International Roughness Index profiler 3 ft in either direction from the start, calculating the cross-correlation value at every possible offset. The maximum cross-correlation value found over the 6-ft range is the cross- correlation value between the reference profile and the inertial profile. 4. Compute the average cross-correlation by averaging the 10 cross-correlation values, which is referred to as the accuracy score for the profiler. The standard indicates that the accuracy score must be 0.90 or greater for a profiler to pass the accuracy criterion at a test section. The standard indicates that a lower score may be acceptable for surfaces having an IRI greater than 150 in./mi; surfaces having an IRI greater than 150 in./mi are not used to evaluate profilers that collect data for construction acceptance. The standard indi- cates that an accuracy score of 0.90 suggests that the IRI values of the two profiles agree within 5% of each, with a 95% confidence level. The FHWA has developed a software called ProVAL (Profile Viewing and Analysis) for analyzing profile data (ProVAL 2017). It is free and can be downloaded from the web. ProVAL has the capability of computing the repeatability and accuracy scores for certifying profilers that follow the AASHTO R 56-14 procedures. State-Specific Procedures for Certifying Profilers A profiler certification method can be based on the overall IRI values obtained at a test section by using the data collected from an inertial profiler and a reference device at that section. The following procedures are used in such a process to evaluate the repeatability and the accuracy of data collected by each sensor in a profiler at a test section: • Repeatability: The standard deviation of IRI is computed using the IRI values obtained for the repeat runs of the profiler at the test section. The standard deviation of IRI must be less than a specified threshold for the profiler to meet the repeatability criterion. The coefficient of variation (CV) of IRI (i.e., standard deviation of IRI divided by the average IRI and expressing this value as percentage) is another measure that can be used to evaluate repeatability. • Accuracy: An average IRI value is computed for the profiler by averaging the IRI values obtained for the repeat runs, and then the average is compared to the IRI obtained for the reference device data. The difference between the average profiler IRI and the IRI from the reference device data must be within a specified threshold for the profiler to pass the accuracy criterion. When using the CV to evaluate repeatability, the standard deviation of IRI is normalized by the IRI of the section. This can provide lower CV values on rougher sections than for smoother sections, which seems to imply that the profiler is more repeatable on rougher sections than on smoother sections. This is not actually the case, and the lower CV values on rougher surfaces occur simply because of the normalizing effect of the IRI of the section. The Michigan Department of Transportation (Michigan DOT) uses the aforementioned procedure to certify profilers with the repeatability evaluated using the standard deviation of IRI (Michigan DOT 2015). A profiler must first pass a DMI check at a 528-ft-long section where the DMI in the profiler must measure the length of the section within 0.1% of the actual distance. Thereafter, the profiler is required to collect 10 repeat runs at the test section used for certifica- tion. The profiler must satisfy the following two criteria for data collected by each sensor for the profiler to be certified: • Repeatability Criterion: The standard deviation of IRI computed from the IRI obtained for the 10 profiler runs must not exceed 2 in./mi. • Accuracy Criterion: The average IRI from the 10 profiler runs must be within 5 in./mi of the IRI calculated from the reference device data.

Background Information on Inertial Profilers and Profile Data Collection 33 Some state DOTs have developed their own certification procedures that include the cross- correlation method specified in AASHTO R 56-14. The Minnesota Department of Transportation (Minnesota DOT) is an example of a state that has adopted such a procedure. Minnesota DOT uses an AC-surfaced and a PCC-surfaced section for certifying profilers, with each section being approximately 528-ft long (Minnesota DOT 2017). A profiler is required to obtain five repeat runs at each test section, and the profiler must satisfy the following three criteria at both test sections to be certified: • The average IRI from the five runs corresponding to each sensor must be within 5% of the reference IRI value at both test sections. In addition, when individually compared against the reference profile, all five profiles must correlate at a level of 85% or higher. The average of the five correlations must be at least 90%. • The standard deviation of IRI for the five profiles corresponding to each sensor must be no larger than 3% of the average IRI of the five passes. • All five passes on each test section must be within 0.2% of the actual length of the test section. The Texas Department of Transportation (Texas DOT) has developed their own procedure for certifying profilers, and this procedure does not contain the cross-correlation method described in AASHTO R 56-14 (Texas DOT 2017). This procedure contains a repeatability and an accuracy criterion based on IRI similar to the Michigan DOT procedure for certifica- tion of profilers. In addition, the procedure contains additional requirements for evaluating the repeatability and accuracy of profile data that are similar to the precision and bias method described in ASTM E950. A profiler is required to obtain 10 repeat runs at a test section. The reference device is used to obtain three repeat runs. The Texas certification procedure requires the profiler to satisfy the following requirements at each test section for data collected by each sensor in the profiler: • Profile Repeatability: The variance of the profile data from the 10 repeat runs at each data- recording interval is computed for each wheelpath. Thereafter, the average of the variance is computed. The square root of the average variance must not exceed 35 mils for the profiler to pass the profile repeatability criterion. • Profile Accuracy: The average profile from the 10 repeat runs of the profiler is computed at each data-recording interval of the profiler. The reference data are filtered using the filter applied to the profile data by the software in the profiler, and the filtered reference data from the three repeat runs are used to compute an average reference value for each data-recording interval. The difference between the average profiler data and average filtered reference data is then computed at each data-recording interval. To pass the profile accuracy criterion, the average of the point-to-point differences must be within ±15 mils, and the average of the absolute differences must not be greater than 50 mils. • IRI Repeatability: The standard deviation of IRI computed from the 10 runs must not exceed 2 in./mi to pass the IRI repeatability criterion. • IRI Accuracy: The difference between the average profiler IRI and the reference device IRI must not exceed 6 in./mi for the profiler to pass the IRI accuracy criterion. Differences Between Certification Methods Based on Cross-Correlation and Overall IRI When the overall IRI along a wheelpath at a test section is used to evaluate repeatability, a pro- filer may show good repeatability because of compensating effects. A profiler may have signifi- cant differences in spatial distribution of IRI within a section for repeat runs but still have good overall agreement in IRI for the runs, thereby indicating that the profiler is obtaining repeatable IRI values. Similarly, a profiler and a reference device may have significant differences in spatial

34 Inertial Profiler Certification for Evaluation of International Roughness Index distribution of IRI within a section but still have good overall agreement in IRI. This effect is illustrated in the following example: Continuous IRI plots (based on a 25-ft base length) for two runs made by a profiler at a 528-ft- long section prepared by the author of this synthesis are shown in Figure 11. A continuous IRI plot shows how the IRI is distributed within a section. The IRI shown at a specific location in such a plot is the average IRI over the specified base length centered at that location. The base length used for the continuous IRI plot shown in Figure 11 is 25 ft. For example, the IRI shown in the plot at a distance of 100 ft is the average IRI from 87.5 ft (12.5 ft before 100 ft) to 112.5 ft (12.5 ft after 100 ft), where the base length of 25 ft was used to average the IRI from 87.5 ft to 112.5 ft. The overall IRI value for this section represents the average IRI of the section. However, the overall IRI does not provide any information on how the IRI is distributed within the section. A continuous IRI plot will show how the roughness is distributed within a section. Figure 11 shows that there are significant differences in the spatial distribution of IRI for the two runs. However, the overall IRI values for Runs 1 and 2 on this 528-ft-long section are 78 and 79 in./mi, respectively. If one looks at these overall IRI values, it appears that the two profile runs are very repeatable. However, the continuous IRI plots show there are significant differences in spatial distribution of IRI between the two runs. The agreement in overall IRI for the two runs occurred because of compensating effects (i.e., high and low values cancel each other out). As shown above, when using the overall IRI of a section to evaluate the repeatability, a profiler may not collect spatially repeatable IRI data but yet may meet the specified repeatability crite- rion and accuracy criterion because of compensating effects. The profiler may not be obtaining spatially repeatable data because of an equipment issue. Therefore, determining the repeatability and the accuracy of a profiler based on the overall IRI of the section has shortcomings. The cross-correlation method specified in AASHTO Standard R 56-14 evaluates the repeat- ability and accuracy of a profiler based on spatial distribution of IRI. In this method, a profiler will not be able to pass the repeatability criterion if it does not obtain spatially repeatable IRI values within a specified tolerance. The profiler will also not pass the accuracy criterion if the spatial distribution of IRI from the profiler data does not agree with the spatial distribution of IRI for data from the reference device within a specified tolerance. Methods Used in Australia Austroads is the Association of Australian and New Zealand Road Transport and Traffic Authorities. It publishes two tests methods to validate inertial profilers. In one method, the IRI Figure 11. Continuous IRI plot of two profile runs.

Background Information on Inertial Profilers and Profile Data Collection 35 from the profiler-recorded data is compared to the IRI from the data recorded by a reference device. In the other method, the IRI from inertial profiler–recorded data is compared to data from another inertial profiler, which is considered to be the reference. Austroads Test Method AG:AM/T002 Austroads Test Method AG:AM/T002, Validation of an Inertial Profilometer for Measuring Pavement Roughness (Reference Device Method), describes a procedure for validating the pavement roughness determined by an inertial profiler by comparing measurements from the inertial pro- filer with measurements made by a reference device (Austroads 2016A). The IRI is the roughness index used to perform this evaluation. The standard indicates that five 1,640-ft-long test sections with the following characteristics should be established for the evaluation: • At least one test section must have an IRI between 57 and 101 in./mi. • At least one section must have an IRI between 120 and 197 in./mi. • At least one section must have an IRI greater than 216 in./mi. • The remaining two sections must have an IRI greater than 63 in./mi and less than 444 in./mi. • At least two of the total 25 individual 328-ft segments that make up the five test sections (i.e., each 1,640-ft section has five 328-ft-long sections) must have an IRI of 285 in./mi or greater. The standard indicates that the test sections should be selected such that the surface charac- teristics (materials, texture, and so forth) are representative of the road network to be surveyed. Each profiler is required to collect data at three test speeds: a speed near the bottom, mid-range, and top of the inertial profilers specified operating speed range. Each profiler is required to obtain five repeat measurements at each test section at each test speed. The test method indicates reference measurements at the test section are to be obtained using a rod and level or using the ARRB Walking Profiler. The standard describes lane IRI as the average IRI of the left and right wheelpaths. The lane IRI is computed from the data collected by the inertial profiler, as well as the reference device at 328-ft intervals at each test section. The profiler must pass two criteria to be validated. For the first criterion, the IRI values obtained at a single speed are analyzed separately. For each test speed of the profiler, 125 records are available for 328-ft interval IRI values (i.e., one speed × five sections × five 328-ft-long sections × five repeat runs). This data are used with the reference device IRI values at 328-ft intervals to develop a regression equation having the fol- lowing form: IRI (Base) = A × IRI (Profiler) + B where IRI (Base) = lane IRI from reference device, in./mi; IRI (Profiler) = lane IRI from inertial profiler, in./mi; A = slope of regression equation; and B = intercept of the regression equation. For the profiler to pass the first criterion, A and B must fall within a specified range, and the coefficient of determination for the regression equation must be equal to or greater than 0.95 for the analysis performed at each test speed. For the second criterion, the IRI values obtained at all three speeds are analyzed as one data set. A total of 375 records are available for 328-ft interval IRI values for this analysis (i.e., three speeds × five sections × five 328-ft-long sections × five repeat runs). The data from all three

36 Inertial Profiler Certification for Evaluation of International Roughness Index speeds are used to perform a regression analysis similar to the analysis performed for the single speed data set. For the profiler to satisfy the second criterion, A and B must fall within a speci- fied range, and the coefficient of determination for the regression equation must be equal to or greater than 0.95. Austroads Test Method AG:AM/T003 Austroads Test Method AG:AM/T003, Validation of an Inertial Profilometer for Measuring Pavement Roughness (Loop Method), describes a procedure for validating the pavement roughness determined by an inertial profiler. The process compares the roughness obtained by an inertial profiler with roughness obtained from another inertial profiler that is considered to be the reference (Austroads 2016B). A test loop is used to obtain measurements for this procedure. The test method indicates that a 35-km-long roughness calibration loop maintained by the Roads and Maritime Services New South Wales and located approximately 75 km north of Sydney can be used to evaluate profilers. The test method indicates that this loop exhibits a wide range of roughness and that other sections that are at least 10-km long may be adequate to perform this test, but the section must contain a wide range of roughness. It is stated in the test method that this test method is used by many road agencies in Australia to accredit inertial profilers for project- and network-level roughness surveys. The test method was developed to • Ensure that the profiler is properly calibrated, • Assess the ability of the driver to track a consistent path, and • Assess the ability of the operator to accurately correlate road condition data with the physical location of the road. The reference IRI of the loop is obtained from measurements made by an inertial profiler generating data that are considered to be accurate. This profiler makes five repeat runs, and the lane IRI (i.e., average IRI of the left and the right wheelpath) is computed at 328-ft intervals for each run. The lane IRI obtained at each 328-ft interval for each run are averaged to obtain an average lane IRI for each 328-ft-long segment. The profiler being evaluated is required to obtain five repeat measurements along the test loop. The obtained data are used to compute lane IRI at 328-ft intervals, and the lane IRI values obtained from the five runs are used to compute an average lane IRI for each 328-ft length. The profiler being evaluated must pass two criteria to be validated. The first criterion is based on the correlation between IRI from the profiler being evaluated and the IRI from the reference inertial profiler. The second criterion is based on an allowable percentage difference in IRI between the profiler being evaluated and the reference inertial profiler. For evaluating the first criterion, the average lane IRI values obtained at 328-ft intervals for the profiler being evaluated and the reference inertial profiler are used to develop a regression equation having the following form: IRI (Profiler) = IRI (Reference) + B where IRI (Profiler) = average lane IRI at 328-ft interval for profiler being evaluated, in./mi, and IRI (Reference) = average lane IRI at 328-ft interval for the reference profiler, in./mi. The regression equation must have a coefficient of determination of at least 0.95 for the profiler to pass the first criterion.

Background Information on Inertial Profilers and Profile Data Collection 37 For evaluating the second criterion, the overall average of the percentage difference for each 328-ft section between the profiler being evaluated and the reference inertial profiler are computed, using the following formula: ∑ ( ) ( )( )= − = 100 1 Average Percent Difference n IRI Prof IRI Ref IRI Ref i i ii n where IRI (Prof) = lane IRI at 328-ft intervals from profiler being evaluated, in./mi; IRI (Ref) = lane IRI at 328-ft intervals from the reference inertial profiler, in./mi; and n = total number of 328-ft-long segments used for analysis. The average percentage difference must be less than or equal to ± 5% for the evaluated profiler to pass the second criterion. Method Used in the United Kingdom SCANNER (Surface Condition Assessment for the National Network of Roads) surveys were developed by the United Kingdom Roads Board to provide a consistent method for measuring the surface condition of roads in the UK using automated survey vehicles (United Kingdom Roads Board 2011). Among the data gathered during the automated surveys were longitudinal profile along the wheelpaths, transverse profile of the road, texture of the roadway, condition at the edge of the road, and cracking. The equipment used for SCANNER surveys is required to be accredited before data can be collected. The accreditation is valid for 14 months after the initial accreditation, and thereafter it is valid for 12 months during subsequent accreditations. In the UK, the ride quality is measured based on the longitudinal profile variance over 9.8 and 32.8 ft and is reported for the left and the right wheelpaths. As the profile variance is not always successful in identifying large local bumps, severe bumps in each 32.8-ft length are also identified from the profile data. During accreditation, the equipment is evaluated to determine the minimum speed and the maximum level of deceleration under which the surveys can be performed to obtain valid mea- surements for computation of the required parameters. The site for testing is selected by the accreditation tester. It is divided into sections, where a reflective post is placed at the start and the end of each section. The selected site may contain straight and curved sections but must not contain any extremes of geometry. The profiler is required to obtain data along both wheelpaths. The reference profile is measured using the ARRB Walking Profiler. The accreditation tester indicates the speeds at which measurements are to be made. Some runs are carried out under deceleration, and the tester marks locations where braking should start and end. The following procedure is used to analyze the data collected along each wheelpath: 1. Two filtered profiles each are obtained for the measured profile and reference profile by applying a moving average filter to attenuate wavelengths in excess of 9.8 ft and 32.8 ft. 2. If required, the measured profile is normalized to remove any constant or linear offset between the reference and measured profiles. The differences between the filtered reference profile and the filtered measured profile are computed by subtracting the filtered reference profile from the filtered measured profile. 3. The 9.8-ft and 32.8-ft longitudinal profile variances are computed over 32.8-ft lengths for the profiler data and the reference device data. The variances obtained from the reference data are subtracted from the variances obtained for the profiler data for both the 9.8-ft and 32.8-ft variances.

38 Inertial Profiler Certification for Evaluation of International Roughness Index For tests conducted at a constant speed, the profiler must meet the following criteria for data collected along each wheelpath to pass the accreditation test: • Difference between measured longitudinal profile and reference profile: 95% of the differences must fall within a specified range, where separate ranges are provided for the 9.8-ft and 32.8-ft wavelength data. • Cross-correlation coefficient between the measured profile and reference profile: 95% of the cross-correlation coefficient values must equal or exceed a specified value, where separate values are provided for the 9.8-ft and 32.8-ft wavelength data. • Error between longitudinal profile variance computed from measured profile and reference profile: 65% of the errors for the 9.8-ft wavelength must fall within a specified range, and 95% of the errors for the 32.8-ft wavelength must fall between a specified range. The performance of the equipment at each level of deceleration is compared with the above requirements. The accreditation tester records the level of deceleration that in the tester’s opinion does not provide measurements to an acceptable level of accuracy. This level of deceleration is the limit of deceleration beyond which the equipment does not provide accurate data. Method Used in Ontario, Canada The procedure for certifying an inertial profiler by the Ministry of Transportation in Ontario is described in Test Method LS-296, Method of Test for Calibrating, Correlating, and Conduct­ ing Surface Smoothness Measurements Using an Inertial Profiler (Ministry of Transportation, Ontario 2017). A profiler must be certified using the procedures outlined in this standard for the profiler to collect data on ministry contracts. The profiler certification is carried out at established correlation sites that are at least 1,312-ft long. Two parallel reference profiles are established at each site that are 69 in. apart. The profiler obtains five repeat measurements at each site at a speed of 44 mph. The profiler must pass a repeatability criterion and an accuracy criterion to be certified. The criteria are similar to the repeatability criterion and accuracy criterion presented in AASHTO R 56-14 (AASHTO 2017C), which is a repeatability value of at least 0.92 and an accuracy value of at least 0.90. The profiler operator must demonstrate the ability to compute IRI and find locations of localized roughness sites using the ProVAL software (ProVAL 2017). The localized roughness sites are found using a continuous IRI plot created with a 25-ft base length, with the threshold for the localized roughness being an IRI of 152 in./mi. The operator is required to compute the IRI of each wheelpath from the collected data and identify the limits of the localized roughness at each site. The operator is required to fill out a form included in the standard with the computed IRI values and locations of localized roughness, which is evaluated by a ministry representative. Network-Level Data Collection The AASHTO Standard R 43-13, Standard Practice for Quantifying Roughness of Pavements (AASHTO 2017D), describes standard procedures for measuring a longitudinal profile and calculating the IRI for highway pavement surfaces to produce consistent estimation of IRI for network-level pavement management. This standard covers the following items: • Transverse spacing between the two sensors that collect data, • Calculation of IRI, • Reporting requirements, and • Guidelines for developing a QA plan.

Background Information on Inertial Profilers and Profile Data Collection 39 This standard indicates the following requirements for data collection: • Height sensors used for data collection must be separated by approximately 65 to 71 in. • Longitudinal profile data points used for calculation of IRI must have a longitudinal spacing not greater than 2 in. • A long wavelength filter should be applied to the profile data to remove wavelengths exceed- ing 300 ft. This standard references ASTM Standard E 1926 (ASTM 2017A) for the computer code that is used to compute the IRI. The standard indicates that the IRI should be computed at 0.1-mi intervals for each wheelpath, and the wheelpath IRI values should be averaged to obtain MIRI at 0.1-mi intervals. This standard indicates that each agency should develop their own QA plan, which should include survey personnel certification training records, accuracy and repeatability of equipment, daily QC procedures, and periodic and ongoing QC activities. The standard offers the following guidelines for developing such a plan: • Certification and Training: Agencies should implement a certification program to certify profil- ers at least annually to ensure repeatability and accuracy of the equipment. Agencies are respon- sible for training and/or certifying their data collection personnel and contractors for proficiency in operating the equipment according to standard practice and applicable agency procedures. • Equipment Calibration: Calibration of the components in the profiler (i.e., accelerometer, height sensors, and the DMI) must be performed according to the manufacturer’s recommendations. The equipment must be operated according to the manufacturer’s specifications. A maintenance and a testing program for the equipment must be developed according to the manufacturer’s recommendations. • Verification Sections: Verification sections are sections with known IRI values along the two wheelpaths. Such sections are to be measured on a regular basis, and the collected data pro- vide information about the accuracy of measurements and give insights into issues with the equipment. • Quality Checks: Data quality checks can be made by comparing the most recent IRI values with previously obtained values. Additional data checks can be performed at locations where there is a large difference in IRI. AASHTO Standard R 57-14 (AASHTO 2017A) also describes the procedure for establishing control sections (i.e., verification sections) for network-level profiling. This standard indicates that several test sections having a range of IRI up to 200 in./mi can be selected as control sections. The standard indicates that the IRI of the control section should be established by using an inertial profiler certified within the past 90 days to make at least five profile measurements at the section. The average IRI from these measurements is used to determine the IRI of the control section. Operational Procedures for Collecting Profile Data Correct operational procedures must be followed to obtain accurate data from an inertial profiler. AASHTO Standard R 57-14 (AASHTO 2017A), Standard Practice for Operating Inertial Profiling Systems, describes the procedure for operating and verifying the calibration of an iner- tial profiling system. This standard describes procedures that are applicable to project-level data collection (i.e., collecting data for construction acceptance), as well as for network-level data collection. Among the items covered in this standard are • Procedures to verify whether the DMI is within the specified accuracy requirement; • Procedures to verify whether the height sensors are obtaining readings within the specified accuracy requirement;

40 Inertial Profiler Certification for Evaluation of International Roughness Index • Procedure to perform the bounce test, which is an overall integrity test that checks whether the accelerometers and the height sensors in the profiler are working correctly; • Procedures to be followed during data collection, such as speed requirements, lead-in distance, and longitudinal path to profile; and • Guidelines on establishing verification sections. The operational procedures used for data collection with an inertial profiler can be divided into pre-operational tests/checks and procedures to follow during data collection. Pre-Operational Checks The three main components of an inertial profiler are the height sensor, accelerometer, and the DMI. The operator must perform pre-operational checks/tests to ensure that all of these components are functioning properly before collecting data. The following tests/checks are performed to ensure these components are functioning properly: • Vertical verification of calibration that is performed to ensure that the height sensor and the accelerometer in the profiler are functioning properly, and • Longitudinal verification of calibration that is performed to ensure that the DMI on the profiler is measuring distance accurately. AASHTO R 57-14 describes the procedure for vertical verification of calibration (AASHTO 2017A). The vertical verification of calibration consists of performing two procedures: the block check and the bounce test. In the block check, blocks of known height are placed below the height sensor, and the height of the block is measured using the software in the profiler. AASHTO R 57-14 indicates that the height of the block measured by the profiling system must be within 0.01 in. of the actual height of the block. The bounce test is performed to determine whether the motions of the vehicle are canceled by the accelerometer and is an overall integrity check on the proper functioning of the accelerometer and the height sensor. This test is performed by imparting a bouncing motion on the vehicle while the vehicle is stationary. The procedure for performing this test—and the criteria for passing this test—are described in AASHTO Standard R 57-14 (AASHTO 2017A). AASHTO R 57-14 indicates that the block check and the bounce test must be performed prior to collecting profile data for construction acceptance. For network-level profiling, AASHTO R 57-14 indicates that the block check must be performed monthly and the bounce test must be performed daily before data collection. The DMI of the profiler must be calibrated at intervals recommended by the manufacturer. AASHTO R 57-14 (AASHTO 2017A) describes the procedures for verifying the accuracy of the DMI and indicates that the accuracy of the DMI must be verified prior to using the profiler to collect data for construction acceptance. For network-level data collection, AASHTO R 57-14 (AASHTO 2017A) indicates that the accuracy of the DMI must be verified monthly. The accuracy of the DMI is verified by driving the profiler over a section of known length and comparing the dis- tance measured by the DMI of the profiler with the actual length of the section. AASHTO R 57-14 (AASHTO 2017A) indicates that the distance measured by the profiler must be within 0.15% of the actual length of the section. The accuracy of the DMI is extremely important when col- lecting data for construction acceptance, as grinding limits for localized roughness locations are determined from this data. If there is an error in the DMI, problems will be encountered when identifying these locations in the field. Procedures to Follow During Data Collection AASHTO R 57-14 (AASHTO 2017A) describes the procedures to follow during data collection when collecting data for project-level (i.e., construction acceptance) as well as for network-level

Background Information on Inertial Profilers and Profile Data Collection 41 data collection. Attention must be paid to the following items when collecting data with an inertial profiler: • Sensor spacing, • Operational speed, • Lead-in length, • Use of the auto-trigger, • Wet pavement, and • Debris on the pavement. Sensor Spacing Sensor spacing refers to the distance between the left and the right sensor in the profiler. The left and the right sensors must be located equidistant from the center of the vehicle. When collecting data for construction acceptance, the sensor spacing in the profiler must match the spacing stated in the smoothness specification, and the profiler must be driven along the center of the lane. For network-level data collection, the sensor spacing must be consistent with the agency’s practice and the vehicle must be driven along the center of the lane. Operational Speed The profiler manufacturer will specify the speed range for which the profiler will collect valid data. Collecting data outside these limits will result in erroneous data. Coming to a stop during data collection can introduce a major error to the collected data (Karamihas et al. 1999). Profiler manufacturers will typically mark the data that are collected outside the operational speed of the profiler and not use these data in the computation of roughness statistics. The profile data collection must be performed at a constant speed to ensure quality data. Rapid acceleration and sudden braking should be avoided during data collection, as these can introduce errors to the data (Karamihas et al. 1999). Lead-In Length A profiler must have the software operational and travel a certain distance before the start location for data collection to stabilize the filters that are applied to the collected data to compute the profile. The lead-in length is the distance the profiler must travel before the start location of data collection. This issue is particularly important when collecting data for construction acceptance because if the lead-in distance is insufficient, the IRI computed for the first 0.1 mi can be erroneous. Therefore, the operator must ensure that the lead-in length satisfies the manufac- turer’s recommended values when collecting data. Using the Auto-Trigger Profilers are equipped with an auto-trigger that can automatically start and stop data collection when the auto-trigger detects a reflective tape placed on the pavement surface or on a cone that is placed on the side of the road. Using the auto-trigger is important when collecting data for construction acceptance to ensure that the data collection is initiated at the correct location. The data collected for construction acceptance are used to detect locations of localized roughness; if data collection is not initiated at the correct location, difficulties will be encountered when marking these locations on the pavement for correction. Wet Pavements Collecting data on wet pavements should be avoided as spray from the water can affect the measurements obtained by the height sensor, generating erroneous profile data that will result

42 Inertial Profiler Certification for Evaluation of International Roughness Index in an upward bias to the IRI values. Collecting data on damp pavement without any standing water has not shown to result in erroneous data (Evans and Eltahan 2000). Surface Contaminants Any debris on the pavement can affect the collected data and result in an upward bias in the computed roughness indices (Karamihas et al. 1999). When collecting data on the final paved surface for construction acceptance, the surface must be swept before collecting data if any surface contaminants are present on the pavement. Locations for Profiler Certification A difficulty faced by many state DOTs in certifying profilers is to find suitable locations to establish test sections. Establishing test sections on in-service roads will require traffic con- trol to perform reference profile measurements. The profile of a jointed PCC pavement can change due to slab curling. Therefore, when certifying profilers on such test sections, reference profile measurements must be obtained either immediately before or after data collection is performed by an inertial profiler. This means if such a section is established on an in-service road, traffic control is needed to obtain reference profile measurements every time a profiler is certified. Some state DOTs have used existing test tracks—such as a test track used by the police, a drag racing facility, and so on—to set up test sections to certify profilers. At least one state DOT has established a test section on a taxiway of an airport. Another state DOT has established the test section on the center turn lane of a low-volume road. Some state DOTs use a low-volume road, such as a frontage road, to establish test sections for profiler certification. Some state DOTs have access to facilities that are not open to public traffic that can be used to establish test sections for certifying profilers. MnROAD is a pavement test track owned and operated by Minnesota DOT. It is located near Albertville, Minnesota, where pavement-related research activities are performed. Minnesota DOT uses two test sections—one surfaced with AC and the other surfaced with PCC—that are located on the low-volume loop of MnROAD to certify profilers. The low-volume loop of MnROAD is not open to public traffic. Alabama DOT uses test sections located at the NCAT test track in Opelika to certify their profilers. This profiler certification program is administered by NCAT. A few state DOTs—such as California DOT, New Jersey DOT, and Texas DOT—have constructed a facility specifically for certifying profilers. California DOT constructed its certifi- cation facility at a regional transit metro parking lot. The New Jersey DOT constructed its certi- fication facility at an abandoned rest area off a highway. The Texas DOT constructed its facility at the Texas A&M University System Riverside campus, and the profiler certification program is administered by the Texas A&M Transportation Institute (TTI). A profiler certification program administered by a state DOT is aimed at certifying profilers that operate within the state to collect profile data. Such profilers can include profilers owned by the state DOT, profilers owned by a contractor or a testing company that collects data on the final paved surface for construction acceptance (if the state DOT allows such data collection), and profilers owned by a vendor that collects network-level data. The profiler certification programs administered by TTI and NCAT offer certification to any profiler, regardless of whether the profiler collects data within the state or out of the state. A description of the test sections available at these two facilities and the cost associated with certification are presented below.

Background Information on Inertial Profilers and Profile Data Collection 43 Texas A&M Transportation Institute (TTI) Certification Program TTI is affiliated with Texas A&M University. The TTI facility has the following test sections that can be used for certifying profilers: • Smooth and medium-smooth dense-graded AC sections, • Open-graded AC section, • Smooth and medium-smooth transversely tined PCC sections, and • Longitudinally tined PCC section. TTI offers the following certification levels: • HMA: certification at the two dense-graded AC sections and at the open-graded AC section; • PCC2: certification at the two transversely tined concrete sections; • PCC1: certification at the two transversely tined PCC sections and at the longitudinally tined PCC section; • HMA/PCC2: certification at the HMA and PCC2 sections previously indicated; and • HMA/PCC1: certification at the HMA, PCC2, and PCC1 sections previously indicated. The profiler owner can select the level of certification, and TTI has different charges for each level. The profiler owner can elect that the profiler be certified according to the methods specified in Texas DOT Standard Tex-1001-S (Texas DOT 2017) or according to the AASHTO R 56-14 procedure. The following are the charges assessed by TTI: PCC1 or PCC 2—$2,500; HMA—$3,000; and HMA/PCC1 or HMA/PCC2—$4,000. TTI charges a fee of $400 for certifying an operator. National Center for Asphalt Technology (NCAT) Certification Program NCAT is affiliated with Auburn University in Alabama. NCAT operates a 1.7-mile-long oval test track located in Opelika, Alabama. The outside lane of this test track consists of 200-ft-long test sections. These test sections are subjected to accelerated loading using truck traffic to study the performance of various AC mixes. The research cycle for these test sections is 3 years. The inside lane of the test track serves as a haul route and work platform when the test sections in the outside lane are reconstructed every 3 years. The inside lane is a perpetual pavement, which has a thick AC surface and is not subjected to accelerated loading as is the outside lane. Hence, there is very little change in the condition of the inside lane with time. NCAT has established four test sections for certifying profilers on the inside lane of the test track. Three of the test sections have a dense-graded AC surface, and the IRI of these test sections are approximately 55, 100, and 165 in./mi. The other test section has an OGFC and an approximate IRI of 55 in./mi. The profiler certification program is administered by NCAT personnel and certifies profilers according to the procedures described in AASHTO Standard R 56-14 (AASHTO 2017C). NCAT charges a fee of $1,500 for certifying a profiler and any number of operators associated with that profiler. Resources Related to Pavement Smoothness Over the past several years, the FHWA has undertaken several endeavors to improve the understating of smoothness and roughness issues, create documents related to pavement smoothness, develop software for analyzing profile data, and establish training courses for profiler operators. This section describes a document, software, and a training course that have been implemented through FHWA initiatives.

44 Inertial Profiler Certification for Evaluation of International Roughness Index The Little Book of Profiling The Little Book of Profiling (Sayers and Karamihas 1998) presents basic information about measuring and interpreting road profiles. This document describes how profilers work, what can be done with measurements obtained by profilers, and procedures to follow to eliminate errors during profile data collection. ProVAL Software Through a pooled-fund effort, the FHWA has developed a computer program called ProVAL for evaluation and analysis of profile data collected by inertial profilers and reference devices (ProVAL 2017). This software is useful for state DOTs that plan to implement or are already implementing a ride-quality specification based on data collected by inertial profilers. ProVAL has a variety of features that can be used to view profile data, as well as to compute various rough- ness indices such as IRI, HRI, and Ride Number. ProVAL also detects localized rough spots and performs a grinding simulation to correct the pavement if a bump is causing the rough spot. This option can be used to determine and evaluate how the roughness of the pavement will change with different grinding scenarios. ProVAL also has a module for profiler certification, where cross-correlation values for evaluating repeatability and accuracy of profilers can be computed. National Highway Institute (NHI) Course The FHWA has developed a course titled “Pavement Smoothness: Use of Inertial Profilers for Construction Quality Control”, which is offered through the National Highway Institute (NHI) (NHI 2017). This course provides information on procedures to be followed for collecting accurate profile data with inertial profilers and determines how to analyze the collected data. The course is beneficial to state DOTs that are currently using or planning to use a ride-quality specification based on IRI. The course is also useful to persons involved in network-level profile data collection.