Aggregate Image Measurement System 2 (AIMS2): Final Report
Project Questions and Responses
HIGHWAYS FOR LIFE FINAL REPORT
Grant Number DTFH61-08-G-00003
Aggregate Image Measurement System (AIMS2)
Pine Instrument Company
Grove City PA 16127
Phone: (724) 458-6391
Fax: (724) 458-6418
This section is intended to respond to questions or comments that may arise from readers of this report. It will be appended as needed to provide clarification. Readers may submit these inquiries to Pine Instrument Company per the contact information above.
- The AIMS2 values are forced to fit the AIMS1 values in order to allow continued usage of the data collected with the previous AIMS1 device.
- What is the value of shifting the AIMS2 data to the AIMS1 data? If the AIMS1 data have ambient light effects and less effective/accurate optics, why do we want to preserve these historic values?
AIMS1 to AIMS2 2D-Form, Angularity, and Dimensional ratios required no adjustments. The AIMS2 texture value required adjustment to match the AIMS1 texture. A linear fit was determined to be an appropriate correlation correction and a multiplier was determined which provided a reasonable fit. This AIMS Texture shift was discussed in detail in the Phase I report. A brief discussion is provided below to help explain the reason this adjustment was selected.
Dr. Eyad Masad developed the AIMS technology and is a recognized leader in pavement materials research. Therefore, his research team at Texas A&M was selected as the primary laboratory to validate the AIMS2 design. Dr. Masad provided expert advice regarding the AIMS technology throughout the Highways for LIFE project work.
In 2005, Manjula Bathina, working under the direction of Dr. Masad evaluated the repeatability and reproducibility of the AIMS1 system at the Texas Transportation Institute (TTI), which is the same system used by Dr. Masad at Texas A&M. Bathina reported in her thesis Quality Analysis of the Aggregate Imaging System (AIMS) Measurements, "AIMS has been found to have excellent repeatability and reproducibility for all measured parameters when compared with many other test methods."
Enad Mahmoud, also working under the direction of Dr. Masad, compared the outputs of two AIMS1 systems, one at TTI and one at the Texas Department of Transportation. He reported in his thesis work titled Development of Experimental Methods for the Evaluation of Aggregate Resistance to Polishing, Abrasion, and Breakage (December 2005), "Measurements using two AIMS units and two Micro-Deval machines were used to assess the variability. There was no statistical difference between the measurements of the two AIMS units or between the measurements of the two Micro-Deval units."
This same AIMS1 unit had also been reported to produce repeatable and reproducible results in "Simulation, Imaging, and Mechanics of Asphalt Pavements FHWA DTFH61-03-X-00026" by Eyad Masad. This AIMS1 system has been located in the same laboratory since it was acquired by TTI; therefore, the results from this system have not been influenced by variation in ambient light, as was confirmed by the high repeatability and reproducibility of this system by Bathina (2005). No other direct system comparisons, other than factory configuration data, were known to be available for consideration.
Based on this information, it was believed the system at TTI (located at Texas A&M) provided a reasonable representation of the AIMS1 system outputs. These reports provided support for selecting this AIMS1 system as the definitive baseline for comparison and calibration of the AIMS2 system in the Highways for LIFE project.
It was the understanding of Pine Instrument Company that the intent of the Highways for LIFE program was to leverage existing AIMS historical data, which had shown promising results. The AIMS1 system at this location had been used in the majority of the published research work utilizing AIMS technology. Matching the results of this AIMS1 system was believed to be appropriate and concurred with the advice from Dr. Masad.
- As we continue to attempt to develop pavement performance relationships with the AIMS2 values, would it be more valuable to shift the AIMS1 data to the AIMS2 data and "draw a line in the sand" to continue forward with AIMS2 unshifted or uncorrected values based on improved lighting and optics for performance relationships?
Dr. Masad recommended a shift of the AIMS2 data to fit the established AIMS1 system scale. The texture scale was also extended to be from 0 to 1000 instead of 0 to 800. This was simply an extension to a scale of 10 rather than 8 with no scaling of the actual values.
Each AIMS1 system was set up at the factory under controlled lighting conditions to achieve reproducible results. The repeatability work done at Texas A&M suggests that setup was successful. Therefore, based on the information available, it was believed this texture shift was appropriate to match the existing AIMS1 systems.
Top lighting problems had been reported by users of the AIMS1 systems, and the system at TTI was configured with additional lighting to account for this problem. The AIMS2 system's top-lighting arrangement was changed to address the cost and reliability issues associated with the ring light system used on the AIMS1. The camera technology was also updated from the outdated camera used on the AIMS1 system. These changes were believed to be the cause of the shift in AIMS2 texture from the AIMS1 data and a linear multiplier was determined to be the best fit.
The ruggedness testing performed in Phase I showed the texture output to be extremely sensitive to ambient lighting making the AIMS1 system output sensitive to laboratory conditions. This sensitivity is addressed in the AIMS2 system by creating an enclosed image acquisition chamber. The correlation of the results between the Texas Department of Transportation and TTI AIMS1 systems suggests that the ambient light conditions were similar, but no specific information on ambient lighting at the time the data were collected is available to corroborate this assumption.
Since the texture value output is sensitive to ambient light conditions, the only way to understand how a specific AIMS1 data set matches up with the TTI system, and therefore the AIMS2 systems, is to compare the results of a specific AIMS1 system directly to an AIMS2 system. A correlation between the specific AIMS1 system in question and the AIMS2 systems can then be established with reasonable certainty. The AIMS2 units have proven to be reproducible in multiple laboratory environments, so the goal would be to establish the texture relationship of any given AIMS1 system to the AIMS2 platform.
It is also important to consider the realized performance of any specific AIMS1 system. System field operation may have differed from these factory settings (aperture setting, ambient lighting, etc.), which may influence the results. Also, some AIMS1 users have reported poor focusing (blurry texture images) and partial particle images, which may impact the texture values. As a research platform, the AIMS1 system did not include controls on many of these variables.
- The reference scale used in AIMS1 versus AIMS2 is not clear.
- Example (Angularity) Descriptions:
- AIMS1 (Angular, Sub-angular, Sub-rounded, and Rounded)
- AIMS2 (High, Medium, and Low)
- Why were the descriptions changed? Which descriptions are to be adopted and why?
The angularity scale was not changed. The AIMS2 system angularity output compared favorably with the AIMS1 system outputs over multiple aggregate sources.
The change to Low, Medium, and High for each of the AIMS characterizations was specifically requested by Dr. Masad. This change was intended to be an improvement to clarify and simplify the information. The Low, Medium, and High classifications are used consistently throughout the AIMS shape characterizations. In the specific case of angularity, the three categories of Low, Medium, and High are believed to be less ambiguous than sub-rounded and sub-angular categories. These changes were motivated by input from the aggregate industry in response to presentations that were made by Dr. Masad in various conferences, including especially, the annual meeting of the Transportation Research Board and the annual meeting of the International Center for Aggregate Research.
- In relating the AIMS shape properties to performance data, I think we would prefer to use the AIMS1 descriptions. The AIMS1 descriptions are also easier for people to relate visually to aggregate shape.
The AIMS Angularity values and scale have not changed. The Low, Medium, and High category descriptions were presented to the FHWA for review in September of 2008. The AASHTO provisional specifications were balloted and accepted using the Low, Medium, and High categories. Given the variety of AASHTO committee members' comments on the terminology used in the specification, changing the descriptions is expected to raise additional concerns, comments, and negative votes. The break point value for each category will likely require supporting documentation for the AASHTO ballot process.
- AIMS1 quantifies angularity using two methods (gradient and radius). The gradient method is only utilized in AIMS2 and the AASHTO specification.
- Why is the radius method not included or used? The report should discuss why the radius method is not included or adopted.
The use of multiple angularity values was a cause of confusion. From a practical point of view, it is preferred to provide one method for measuring each of the physical characteristics (shape, angularity, and texture). The work that was conducted as part of NCHRP 4-30 showed that the gradient method is better than the radius method for differentiating between aggregates with different angularity characteristics. In several presentations to the industry, Dr. Masad received feedback that providing two methods to measure the same property is confusing and poses a challenge for developing one set of specifications for angularity. Therefore, the Radius Angularity is not provided in the output.
- How can we use the angularity values obtained using the radius method?
To obtain the Radius Angularity values, a separate analysis of the existing images must be performed
- AIMS1 reports coarse aggregate 2D Form. It is not mentioned in the report why AIMS2 and the AASHTO specification are not using the 2D Form index for coarse aggregate particles.
- What happened to the coarse aggregate 2D Form and why was it removed?
The 2D Form is a two-dimensional parameter. The system provides a three-dimensional measurement in millimeters for each coarse aggregate particle; short, intermediate, and long dimensions. The AIMS Sphericity, F&E and ForE ratios are calculated from these dimensions. This three-dimensional information is not available for fine aggregates; therefore, the two-dimensional form is provided as a means to characterize fine aggregate particle form.
The 2D Form for coarse aggregate particles is still available on the data sheet in each AIMS report workbook (raw values and cumulative distribution). These data can be evaluated if desired. These output data are presented in Excel workbooks specifically for the purpose of customization for specific user requirements. As discussed in the response to Question 3 above, only one method to measure each of the physical characteristic is desired. Therefore, the three-dimensional shape measurements over the two-dimensional (form index) measurements of coarse aggregates is provided. Current AIMS2 AASHTO specifications do not include 2D Form for coarse aggregate sizes.
- Does the coarse aggregate 3D Form also address the 2D Form? Descriptions should be given if the coarse aggregate Sphericity (3D Form) is thought to address the coarse aggregate 2D form.
AIMS2 Sphericity addresses the three dimensional form of coarse aggregate particles. There is correlation between coarse aggregate 3D and 2D Form. However, the 3D form is the preferred method since it considers the three dimensions of particles instead of the two dimensions. The industry has used three-dimensional measurements of shape for many years through the use of the flat and elongated percentage of particles. As such, the use of the 3D Form is more accurate and more consistent with the industry experience.
- The aggregate texture shift factor (AIMS1/AIMS2) is reported to be 2.4563. This value was obtained using 32 aggregate samples.
- Certainly the aggregate texture index depends on the illumination and gray-scale intensity of aggregate surface image. In order to obtain reasonable texture shift factors (if we even want to use shift factors, see comment 1 above), wouldn't the study need a larger number of aggregate types and specimens with different type combinations including RAP aggregates to ensure a precise shift factor?
While this concern might apply if the AIMS1 system had wide scale usage, the AIMS1 platform has been utilized in a very limited manner, by limited users. The 32 aggregate samples are believed to provide a reasonable representation of a broad spectrum of materials tested in the AIMS1 system. The 2.4563 value provides a reasonable fit for the materials tested as shown in Figure 1.5 of Appendix A. Because the AIMS1 has only been used in limited research applications, this fit provides a reasonable link to the research data, while essentially drawing a "line in the sand" as the AIMS2 platform moves forward.
- The report states that additional work is needed to improve the performance of the system for the sieve size 0.075 mm (#200) due to multiple particles (connected/touching) being analyzed as a single particle.
- When is this work going to occur on CHRP values and touching particle analysis in order to allow the imaging and analysis of the #200 material and improve the variability?
- The AASHTO specification still states the procedure is applicable to the #200 material when the statement above and not utilizing the #200 material in the precision statement means that it is not applicable to #200 material.
At the time the draft procedure was submitted to AASHTO, it was expected that the #200 work would be completed before the specification was accepted. However, time constraints delayed further investigation into the selection of the appropriate CHRP value for #200. In addition, the specifications were approved rather quickly.
It is appropriate to revise the precision statement in the AASHTO specification to include a separate line for the #200 retained material. This will allow AIMS2 data to be collected at that size while providing the user with the proper variability information. As the system performance is improved, this precision value can be adjusted accordingly.
- The method of edge detection followed by image segmentation may possibly be used to effectively delineate the overlapping or toughing aggregate particles of size 0.075 mm. We can discuss in more detail to see if this is worth pursuing.
There were many methods investigated for solving the issue of defining particle edges with several showing promise. It is certainly possible that there are more effective ways to address the touching particles in fine aggregate images. The CHRP was selected as it appeared to be reasonably effective at removing touching particles while not impacting the angularity values.
- The precision statements are a result of the combination of all sieve sizes (except the #200 material).
- Why was the precision of all aggregates combined versus separate statements for individual sieve sizes? If some sizes are more variable than other sizes would it be valuable to differentiate the higher variability sizes from the others?
Early in the Interlaboratory Study, labs reported higher variability in the #200 results. This variability in the #200 ILS data was determined to be linked to touching particles in the images. This indicated the CHRP value used for #100 and larger is not the optimum value for the #200 retained. The need for a different CHRP variable is likely due to particle size and image resolution interactions. The number of touching particles in a specific image is also related to how well the operator distributed the material over the tray surface. The system is currently capable of collecting #200 shape characteristics but with slightly higher variability than the other sizes.
The ILS data, shown in Appendix B and analyzed by Gates et. al. (Texas A&M University 2009) with results shown in Table 3.13 and Table 3.16, clearly shows similar variability in all sizes with statistically insignificant differences, except #200. Therefore, for simplicity and practicality, one precision statement value based on the geometric average variability is reasonable for all sizes excluding #200 retained. As noted in the response to question 6, the AASHTO specification precision statement should be revised. This revision will be introduced at the next AASHTO Sub-Committee on Materials ballot opportunity along with the appropriate information to correct the error.
- The AASHTO M 323 (i.e., Standard Specification for Superpave Volumetric Mix Design) test protocol requires an aggregate particle to be flat or elongated if the longest particle dimension (i.e., length) divided by smallest particle dimension (i.e., thickness) exceeds five. In addition, the percentage of particles for this ratio is limited to a maximum of 10%.
- AIMS2 present this information using Figure 1. This graphic is busy and not clear for the average user who is looking to see if their material meets the M323 specifications. For the average user to determine pass or fail requirements the software should include an additional simplified figure (example Figure 2) for the stockpile blend and sieve sizes. This is how we present the MAMTL reported data.
Figure 1. Comparison of Aggregate Sphericity and Superpave Flat and Elongated Limits; Aggregate Source CG-1 (After KS0882 project).
Figure 2. AIMS Aggregate Flat & Elongation Index (After KS0882 project).
Figure 1: Elongation vs. Flatness as provided by the AIMS2 output file is certainly a "busy" chart. The cumulative distribution graphs for the F&E and for the ForE ratios were added (see figure 1A below) specifically to address the need to understand how the material fits within the M323 specification. These cumulative distribution charts are consistent with the other AIMS2 characteristic charts and provide a clear means to determine if an aggregate material meets F&E requirements.
Changes to the current F&E and ForE specifications have been discussed within the research community, so it is believed this cumulative distribution graph, which provides the full spectrum of ratio data, not just integer break points, was the best configuration for this information. This distribution chart is not limited to providing integer F&E ratios. The scatter plot represented in figure 1 was retained in the output file so that users familiar with the F&E concept utilized in M323 could understand how it relates to the concept of AIMS2 Sphericity, introduced by Dr. Masad in the AIMS shape characterizations. Sphericity includes all three dimensional measurements of the coarse particle.
Figure 1A below shows the AIMS2 F&E cumulative distribution chart. The limit line in this figure was added to depict how specification limits might be shown for multiple ratio categories. In this case, arbitrary limits of Maximum 20% >3:1 and Maximum 10% >5:1 are shown. All materials meet the Max 10% >5:1 criteria, but two materials fail the Max 20% >3:1. These red-line limits are not currently included on the graphs. It is up to the user to understand the information and apply the appropriate specification limits.
The bar graph representation had been casually suggested by one AIMS1 user. It is respectfully suggested that the cumulative distribution graph (figure 1A) presents the data in a reasonable form and is consistent with the other AIMS2 data.
Figure 1A: Cumulative F&E Distribution with shaded specification limits added.
However, if a specific user wishes to present the information in a specific way, all the AIMS2 shape characterizations are provided in an Excel workbook. Each material's scan data are provided on a separate worksheet. These AIMS2 values are accessible to the user, permitting custom charts within the Excel format to be created. These custom formats can also be saved as the working template within the AIMS2 system allowing future scans to be loaded directly into the desired format. The Excel format was chosen specifically to permit customization of the report formats for specific reporting requirements.
The AASHTO TP81 specification was approved with the F&E cumulative distribution chart. Any changes to the reporting will need to be submitted to the AASHTO Sub-Committee on Materials for approval.
Center for Accelerating Innovation