Long-Term Pavement Performance Ohio SPS-1 and SPS-2 Dynamic Load Response Data Processing

10. CONCLUSIONS

The DLR study team reinterpreted 4,290 Ohio SPS-1 DLR raw traces (table 3) and 9,240 Ohio SPS-2 DLR raw traces (table 4), correcting the data issues identified by DAOFRs Ecomplex-75-77 and the technical memorandum, Investigation of Ohio DLR data in LTPP Database, for SDR 22.0, including trace peak time lag shift, incorrect sensor locations, and wheelpath offsets. (See references 2-5.) Using the methodology in chapter 5 of this report, the DLR study team calibrated and smoothed the SPS-1 and SPS-2 raw traces before categorizing those traces into three categories: good, maybe, and not good. For the SPS-1 data, the trace categorization QC results for smoothed and raw traces are listed in table 13 and table 14, respectively. Approximately 24 percent of strain gauge traces, 55 percent of LVDT traces, and 99 percent of PC traces were concluded to be good. For the SPS-2 data, due to significant noise in the raw traces, only smoothed traces were categorized, and the QC results are listed in table 17. Approximately 61 percent of strain gauge traces and 15 percent of LVDT traces were concluded to be good. Only good traces were used for further extraction of trace peaks and valleys for the upcoming SDR 27.0. In addition, the sensor locations and the corresponding wheelpath offsets were corrected using the approach in chapter 5. Overall, the newly created DLR data in SDR 27.0 appear to match the DLR raw traces, as demonstrated by the plots in chapter 9.

Moreover, the QC results from the categorization were manually checked, and the sensor status from visit to visit and run to run for all SPS-1 and SPS-2 tests for smoothed and raw traces were verified. In addition, the first peak value extracted for good traces was compared with the data from OU, which indicated that the values were very close for most of the sensors for all test sections (see table 18 and table 21 ).

Appendices A-E show the sensor layouts in the Ohio SPS-1 and SPS-2 DLR test sections as well as of the 23 Ohio SPS-1 DLR tests and the 24 Ohio SPS-2 DLR tests to aid future DLR data users in identifying the layout and status of each sensor from one test visitorrun to another.

In the remainder of this chapter, the data issues identified in the DLR raw traces are enumerated.

OHIO SPS-1 DATA ISSUES

Ohio SPS-1 data issues are as follows:

• Some tests in the DLR data did not have any test files, and some files did not have information pertaining to sensor location, truck pass, and truck peak in Ohio data. These test were not considered for processing. As a result, only 23 out of 34 tests were considered for DLR data processing.
  
  
• Strain gauge sensors Dyn10 and Dyn11 for tests J2A, J2C, J2D, J2E, J2F, and J2G in test section 390102 showed a flat trace pattern.
  
  
• All of the LVDTs were buried deep into the subgrade or close to the interface between the subgrade and the base layer in the test sections. Thus, the LVDT traces should not contain any trace valleys (no tensile strains) but only peaks (compressive strains). The LVDT3 sensor for tests J2A, J2C, J2D, J2E, J2F, and J2G (test section 390102), however, showed a trace pattern similar to a longitudinal strain gauge trace that contains trace valleys.
  
  
• Longitudinal strain gauges are expected to assume trace valleys, whereas transverse strain gauges are not. The longitudinal strain gauge sensor Dyn17 for tests J8A, J8D, J8E, and J8G (test section 390108), however, showed a trace pattern similar to a transverse strain gauge trace that assumed no valleys.
  
  
• As indicated by table 9, a significant difference between the extracted peaks and the Ohio data peak for some sensors was observed. This could be due to the sensor locations reported in the OU data (Truckpeak.txt) being approximately two to three inches off the actual sensor locations as measured from the southernmost deep LVDT.
  
  
• In table 25, inconsistent Z-coordinates (depth of the sensor from the pavement surface) for strain gauges Dyn16 to Dyn18 were found between test section390108 profile view and EmbeddedSensor.txt. Tests J8A, J8D, J8E, and J8G were conducted in test section 390108. The tests consisted of three AC layers (2, 2, and 3 inches), one permeable asphalt treated base (PATB) (4 inches), one dense graded aggregate base (8 inches), and subgrade. The test section 390108 profile view showed that Dyn16 to Dyn18 were buried at the bottom (Z = 11 inches from the pavement surface) of the top PATB base layer (4 inches), which is below the three AC layers. In contrast, EmbeddedSensor.txt showed that Dyn16 to Dyn18 were buried at Z = 7 inches from the pavement surface and the Layer was "bottom," referring to the bottom of the lowest of the three AC layers. Thus, the DLR study team recommends changing the Z-coordinate of test section 390108 Dyn16 to Dyn18 from 7 inches to 11 inches and changing the layer to "base PATB" or "base" from "bottom" for this case.
  
  
• The construction plan from one of the original DLR documents showed that two AC layers were planned for test sections 390102, 390104, 390108, and 390110. The section profile views of these sections, however, showed three AC layers. Based on the construction plan, the SECTION_LAYER_STRUCTURE table in SDR 26.0 had two AC layers, with the bottom AC layer combining the two bottom AC layers (2 and 3 inches) shown in the profile views into a 5-inch AC layer while keeping the top 2-inch AC layer as the top AC layer. The DLR study team recommends revising the SECTION_LAYER_STRUCTURE table in order to show the three (instead of two) AC layers.

Table 25 . Sample data from Ohio data for test section 390108.

CL = Center lane.
*Indicates suspect data.

OHIO SPS-2 DATA ISSUES

Ohio SPS-2 data issues are as follows:

• Due the fact that test J12A1 was empty and test J12J10M1 was a partial repeat of J12J10M, the two test were not processed. As a result, only 24 out of 26 tests were considered for DLR data processing.
  
  
• Site visits (A, B, and C) of SPS-2 test sections 0201, 0205, 0208, and 0212 were inconsistent with the SPS-2 subseries (H, I, and J) of OU data. Since subseries A-G were already used for SPS-1 in the OU data, it was assumed that OU assigned H, I, and J instead of A, B, and C for SPS-2.
  
  
• In the DLR_STRAIN_TRACE_SUM_PCC table, Dyn8 strain gauge from test J5J1P runs 1-10 had significantly larger raw strain values compared to other SPS-2 strain gauge values, which were mostly less than 100 µe (see table 22). Thus, further investigation is needed for this strain gauge.
  
  
• SPS-2 DLR sensors LVDT5 and LVDT6 were unresponsive for all tests (the values were all zero).
  
  
• The ASCII files for SPS-2 tests J5J1M, J5J1N, J5J1O, J5J1P, J8S3M, J8S3N, J8S3O, J8S3P, J12J10M, J12J10N, J12J10O, and J12J10P had 32 LVDT sensors (LVDT1 through LVDT32). The other tests (J1A, J1B, J1C, J5A, J5B, J5C, J8A, J8B, J8C, J12A, J12B, and J12C) had only 16 LVDT sensors (LVDT1 through LVDT16). The DLR study team processed only the first 16 LVDTs (LVDT1 through LVDT16) based on information present in the EmbeddedSensor.txt file, which showed only the first 16 LVDTs.
  
  
• The first 500 trace data points were used to determine the gain adjustment factor for SPS-2 data. On average, each SPS-2 time history dataset contains close to 7,000 data points, whereas each SPS-1 time history dataset contains approximately 5,000 data points. Due to significant noise in the SPS-2 data, the first 500 data points may not be enough. For future research, the first 700 data points should be used to determine the gain adjustment factor for SPS-2 data; 700 is approximately 10 percent of each SPS-2 time history dataset.

OHIO SPS-1 AND 2 COMMON DATA ISSUES

Common data issues for Ohio SPS-1 and SPS-2 are as follows:

• The proposed REF_LOC_NO, the distance between the beginning of a test section and the southernmost first LVDT that serves as the origin of the sensor coordinate system, is not possible to determine. The section beginning was not used as a reference for sensor location, and sections have since been overlaid, making this measurement unattainable.
  
  
• As listed in table 18 (SPS-1) and table 21 (SPS-2), the peak data information contained in the Ohio data file was unclear. The data had multiple sensor location values for the same sensor and run.
  
  
• The beginning offset, ending offset, and range values for strain gauges, LVDTs, and PCs (the latter for SPS-1 only) obtained from the DLR raw traces did not match the beginning and ending offset and range values in SDR 22.0. Per the technical support service contractor's recommendations, the beginning offset, ending offset, and range columns were removed and will not be in the upcoming SDR 27.0.
  
  
• The information on channel number, record status, input card, card gain, post gain, gauge resolution, etc. in DLR_STRAIN_CONFIG_AC/PCC, DLR_LVDT_CONFIG_AC/PCC, and DLR_PRESSURE_CONFIG_AC, and run time in DLR_TEST_MATRIX presented in DLR tables in SDR was not found.
  
  
• The inconsistent wheelpath offset field in DLR_TEST_MATRIX was updated only for test sections using the truck pass data in the Ohio dataset (TruckPass.txt). The DLR_TEST_MATRIX table in SDR, however, had wheelpath offset records for both Ohio and North Carolina test sections. Since the wheelpath offset data for the North Carolina test sections was not available, the wheelpath offset records for the North Carolina test sections were not updated.
  
  
• The nine records in the DLR_TEST_MATRIX table in SDR 22.0 that did not have any source data to update were removed (see table 23 ). As a result, 54 traces that did not have a matching record in the DLR_TEST_MATRIX were removed from the 5 trace tables in the upcoming SDR 27.0.
  
  
• No information could be found in the DLR_TEST_MATRIX to interpret data in the MATRIX_INDEX column (distinct coded reference number for controlled truck testing used to aggregate the tests according to the type of truck, vehicle speed, and general time of testing). Thus, the DLR study team recommends removing the column.