Fatigue Loading and Design Methodology for High-Mast Lighting Towers (2012)

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.

34 3.1 Results of Field Tests Results from field test data including dynamic properties, wind stochastic data, stress-range histogram data, and effects of mitigation are presented and discussed in the following sections. 3.1.1 Dynamic Properties of HMLTs A summary of modal frequency and damping ratio values for the HMLTs studied in this report is provided in Table 3.1. The damping ratio values listed are a composite of the two calculation methods described in Chapter 2. For the HMLTs included in the long-term monitoring study, values for the first mode are an average of pluck data using the log-decrement method and ambient data using the half-power bandwidth method. Ambient data were given priority for the higher modes. Damping ratio values for HMLTs not included in the long-term monitoring study were calculated using only pluck data. Modal frequency values are plotted in Figure 3.1 for all poles. The same data are presented in Figure 3.2 plotted against HMLT height. Note the tendency of the modal frequency values to decrease with increasing height, as expected. Damping ratio values are plotted in Figure 3.3. Note the tendency of the damping ratios to decrease with increasing mode number. This decrease may be explained by the tendency of the HMLT to respond to different forms of excitation. For example, since Modes 2 and 3 are more likely to be excited by vortex shedding, the effects of aerodynamic damping would be expected to be more prevalent for these modes. This would be consistent with the observed behavior in the field (i.e., vortex shedding occurring in Modes 2 and 3). Statistical analysis was used to determine confidence limits for damping ratios to be used for evaluation and design of HMLTs. In addition to the data presented in Table 3.1, damping ratio values from the report Field Instrumentation, Testing, and Long-Term Monitoring of High-Mast Lighting Towers in the State of Iowa (Connor and Hodgson, 2006) were included in the data set. Histograms and frequency distribution curves for Modes 1 through 3 are shown in Figure 3.4. Each figure includes a Weibull distribution scaled to fit the shape of the corresponding histogram. The Weibull distribution was chosen because of its application in a domain from zero to infinity and suitability in failure analysis. Table 3.2 lists the damping ratios at the 50 percent, 80 percent, and 95 percent exceedance levels for each mode. For Mode 1, the damping ratio at 95 percent confidence equals 0.5 percent, a match with the damping ratio specified in the current edition of AASHTO Signs. The damping ratio at 80 percent confidence equals 0.8 percent, a reasonable match with the Canadian Specification (2006). Both are conservative with varying degrees of confidence. However, for Modes 2 and 3, the C h a p t e r 3 Findings and Applications

ID FREQUENCY (Hz) DAMPING RATIOS (%) 1st 2nd 3rd 4th 1st 2nd 3rd 4th CA 0.37 1.73 4.72 9.45 1.3 0.4 0.4 - IA-N 0.32 1.33 3.41 6.53 1.2 0.4 0.4 0.3 IA-S 0.35 1.41 3.26 6.24 1.6 1.0 0.5 - KS 0.36 1.66 4.45 9.21 0.8 0.3 0.1 0.1 ND 0.31 1.28 3.15 5.94 1.3 0.5 0.2 - ND (LL) 0.34 1.40 3.38 6.29 0.8 0.3 0.3 0.4 ND-8 3 0.30 1.12 2.98 5.84 1.5 0.4 0.1 0.4 ND-9 4 0.28 1.12 2.95 5.64 1.9 0.9 0.5 0.6 ND-EXP 0.25 1.09 2.91 5.54 2.2 0.6 0.4 0.7 ND-MEM 0.31 1.58 4.18 7.60 1.2 0.4 0.4 0.3 ND-SUN 0.31 1.34 3.36 6.40 1.4 0.4 0.6 0.7 OK-E 0.28 1.18 3.16 6.03 1.4 0.4 0.2 0.5 OK-NE 0.26 1.03 2.83 5.15 1.9 0.7 0.3 0.3 OK-SE 0.26 1.03 2.85 5.28 0.8 1.3 1.5 - OK-SW 0.26 1.20 3.25 6.91 1.4 0.6 0.1 0.3 PA 0.39 1.68 4.60 9.37 1.2 1.1 - - PA-AD 0.38 1.93 4.77 9.52 0.6 2.3 1.7 1.1 SD 0.29 1.17 2.98 5.81 0.8 0.3 0.1 0.3 SD-1E 0.33 1.33 3.44 6.50 1.4 0.6 0.4 0.8 SD-1W 0.33 1.32 3.46 6.43 0.6 0.2 0.1 0.5 SD-42E 0.38 1.76 4.74 8.92 1.0 0.8 0.8 2.2 SD-42W 0.44 2.33 6.28 13.37 2.1 0.9 0.8 0.6 WY-219E 0.58 2.80 6.84 13.99 1.4 1.3 0.3 0.1 WY-219W 0.39 1.77 4.57 9.00 0.9 - 0.8 - WY-228W 0.39 1.56 3.96 7.52 0.7 0.4 0.2 0.6 WY-CJE 0.35 1.50 3.85 7.54 1.2 0.6 0.1 0.2 WY-CJE(LL) 0.47 1.78 4.25 7.89 0.8 0.2 0.5 0.7 WY-CJW 0.35 1.53 3.89 7.56 1.1 0.6 0.1 0.4 WY-CJW(LL) “LL” refers to pluck tests where the luminaire was lowered. 0.48 1.78 4.24 8.26 0.8 1.3 0.3 0.4 Table 3.1. Modal frequency and damping ratio summary. Figure 3.1. Modal frequency versus mode number.

36 Fatigue Loading and Design Methodology for high-Mast Lighting towers Figure 3.2. Modal frequency versus HMLT height. Figure 3.4. Frequency distribution of damping ratios. Mode 1 Mode 2 Mode 3 Figure 3.3. Damping ratio versus mode number.

Findings and applications 37 damping ratios are significantly less. Since AASHTO Signs only assumes one mode of vibration, no comparison can be made for the higher modes. In summary, the data suggest that a lower bound estimate of the damping for each mode could be used for design if a method for calculating dynamic load effects is required. A similar conclusion was reached in the previously cited study by Connor and Hodgson (2006). 3.1.2 Wind Data Mean wind speeds, peak 10-minute average wind speeds, and maximum recorded wind speeds for each site are given in Table 3.3. Data were collected using an anemometer at each site mounted near the standard 33-foot (10-meter) height. Average wind speed and maximum wind speed were recorded at 10-minute intervals for the duration of the study. The mean wind speed listed in the table is simply the mean value of all 10-minute averages recorded during the entire monitoring period for the given pole. The peak 10-minute wind speed average is the highest 10-minute average recorded and is indicative of the highest sustained wind speed. The maximum wind speed listed is the highest instantaneous wind speed sampled. All values in Table 3.3 are Table 3.2. Damping ratios at selected confidence limits. CONFIDENCE MODE (% EXCEEDENCE) 1 2 3 50 (mean) 1.3% 0.6% 0.4% 80 0.8% 0.3% 0.1% 95 0.5% 0.1% 0.02% Table 3.3. Measured wind speed data (mph). ID SPEED (mph) MEAN PEAK MAX 10-MIN AVG CA 8.7 37.8 55.4 IA-N 11.5 46.4 56.5 IA-S 10.3 38.3 53.4 KS 9.4 45.9 64.6 ND 6.7 34.0 55.4 OK-NE 6.8 31.4 53.3 OK-SW 8.8 41.8 65.5 PA 3.5 21.5 45.3 SD 7.8 37.0 60.0 CJE (FR) 12.8 40.9 78.0 CJE (MT) 14.3 49.2 63.8 CJW (FR) 14.0 45.9 64.6 CJW (MT) 12.4 40.8 60.4 FR—Cumulative data without strakes (free) MT—Cumulative data with double strakes (mitigated)

38 Fatigue Loading and Design Methodology for high-Mast Lighting towers independent of wind direction; Appendix D (available on the TRB website) includes wind rosettes for percent occurrence and average wind direction. Mean wind speed data are used later in this report to establish the proposed fatigue wind load. 3.1.3 Stress-Range Histogram Data Stress-range histogram data for most of the long-term monitored HMLTs are plotted together in Figure 3.5. Using the normalized parameters, static pressure-range and cycle frequency, a direct comparison of the data may be made. Cycle frequency is presented in terms of cycles per day by dividing the total number of cycles by the true monitoring period (i.e., excluding any time the data logger may not have been recording). All data plotted are for periods without rope strakes, or, in other words, unmitigated. Data for the IA-N HMLT are not included since a strake was present for the entire duration. A general trend of decreasing cycle frequency with increasing load effect can be seen. As expected, cycles of high magnitude stress range occur less often. Also apparent is the increase in variation of frequency with increasing load effect. This variation can be explained, in part, by observed variation in the wind speed. Data for selected HMLTs are re-plotted in Figure 3.6 along with best-fit lines illustrating the increase in the frequency of higher load cycles with mean wind speed. For example, a load effect corresponding to a stress range of 4 psf occurred with a frequency of about once every hundred days at the Erie, Pennsylvania, HMLT (mean wind speed of 3.5 mph), while the same load effect occurred about a hundred times a day in Creston Junction, Wyoming (mean wind speed of 14.0 mph). Hence, the greater the mean wind speed, the greater the fatigue-limit-state load that would be expected. Figure 3.5. Normalized pressure-range histogram data for all sites.

Findings and applications 39 Values for the fatigue-limit-state stress range, static pressure-range, and velocity-range are tabulated in Table 3.4 along with cycle counts. The histograms used to determine the fatigue- limit-state stress range were not truncated—in other words, all the data were considered in determining the best fit for the distribution. In addition, the upper-limit of all stress-range bins was used to return the most conservative result. For example, a bin counting all stress ranges between 3.0 ksi and 3.5 ksi would use 3.5 ksi to represent the bin during the distribution fit. This ensures all cycles contained in any given bin would be less than the stress range used to represent that bin, which is a conservative approach appropriate for design. The highlighted values for static pressure-range are used to formulate the proposed fatigue design load for new poles. Note, in every case, the number of cycles per day for the across-wind direction exceeds those for the along-wind direction, which indicates the effect of vortex shedding due to the vibration at a higher mode. Also, note the significant reduction in cycle counts between the free and mitigated conditions for the Wyoming HMLTs. “Mitigated” is used to describe data collected while double strakes were placed on the pole. The unmitigated, or free, condition is noted in tables and figures as “FR”, and the mitigated condition is noted as “MT”. Values for constant-amplitude effective stress range, static pressure-range, and velocity range are tabulated in Table 3.5 along with cycle counts. Values are presented for two different levels of truncation, one above 0.5 ksi and the other above 1.0 ksi. Truncating the lower bins of a histogram is common practice in a fatigue analysis. This is typically done so the effective stress range is not falsely “pulled down” by the high number of very small stress range cycles. The two truncation levels correspond to about one-third and one-half the constant-amplitude fatigue threshold for Category E′, respectively. The histograms used to determine the constant-amplitude effective Figure 3.6. Pressure-range histogram data for selected sites illustrating variation due to wind speed.

40 Fatigue Loading and Design Methodology for high-Mast Lighting towers ID STRAIN GAGE SRfls pfls Vfls Ntot Days N/Day (ksi) (psf) (mph) CA-A CH_3 5.76 4.60 42.4 9,214,499 602.3 15,300 CA-X CH_5 4.51 3.60 37.5 9,962,977 602.3 16,543 IAN-A (MT) CH_9 6.17 4.46 41.7 1,851,318 130.7 14,160 IAN-X (MT) CH_12 5.21 3.76 38.3 2,321,984 130.7 17,760 IAS-A CH_2 3.24 6.07 45.1 473,141 168.7 2,805 IAS-X CH_1 2.87 5.36 42.4 585,046 168.7 3,468 KS-A CH_2 7.17 4.96 44.0 15,125,738 457.3 33,079 KS-X CH_6 7.73 5.34 45.7 13,549,087 345.5 39,217 ND-A CH_1 3.54 3.96 39.3 9,592,205 593.8 16,154 ND-X CH_5 3.87 4.32 41.1 10,713,385 593.8 18,043 OKNE-A CH_3 4.58 3.68 37.9 5,599,228 242.4 23,096 OKNE-X CH_5 4.19 3.36 36.3 6,331,421 242.4 26,116 OKSW-A CH_8 4.45 2.87 33.5 9,979,764 251.9 39,613 OKSW-X CH_6 4.61 2.97 34.1 17,245,973 360.8 47,798 PA-A CH_6 2.32 2.60 31.9 394,474 139.3 2,833 PA-X CH_1 2.23 2.50 31.3 669,765 139.3 4,809 SD-A CH_6 3.40 2.83 33.3 18,971,633 593.2 31,979 SD-X CH_8 3.74 3.12 34.9 20,623,451 593.2 34,764 CH_8 4.06 4.87 43.6 17,778,845 317.2 56,050 CH_6 4.68 5.61 46.8 28,777,530 317.2 90,725 CH_4 4.57 5.48 46.3 4,549,210 301.6 15,084 CH_6 4.67 5.60 46.8 5,992,470 301.6 19,869 CH_8 5.10 6.12 48.9 23,030,758 341.6 67,413 CH_6 5.45 6.54 50.5 27,628,271 341.6 80,870 CJW-A (MT) CJE-A (FR) CJE-X (FR) CJE-A (MT) CJE-X (MT) CH_1 3.97 4.76 43.1 1,248,053 74.5 16,743 CJW-X (MT) CJW-A (FR) CJW-X (FR) CH_3 3.95 4.73 43.0 1,320,528 74.5 17,715 A—Along-wind direction X—Across-wind direction FR—Cumulative data without strakes (free) MT—Cumulative data with double strakes (mitigated) Table 3.4. Summary of fatigue-limit-state data. stress-range are based on the average bin stress-range values. Average bin values are appropriate for evaluation because they are an approximation of the center of the bin. The highlighted values for static pressure-range are used to formulate the proposed fatigue evaluation load for existing poles. Note that in terms of cycle counts, the effect of vortex shedding diminishes with increased truncation, indicating the cycles produced by vortex shedding are concentrated in the lower bins. 3.1.4 Vortex Shedding Mitigation Results of the mitigation testing using various rope strake configurations are presented here along with a discussion of the load effect associated with the mitigation strategy.

Findings and applications 41 3.1.4.1 Results of Rope Strake Method The full-length double-wrapped rope strakes described in Chapter 2 were effective at reduc- ing the number of fatigue cycles accumulated and at reducing the occurrences of across-wind excitation, both of which are indicators of vortex shedding. Similar results were obtained at the three HMLTs where full-length strakes were installed. Experiments with the single-wrapped strake and the partial-length double strake were effective at reducing the effect, however, not to the same extent as the full-length double-wrap strake. 3.1.4.1.1 Creston Junction, Wyoming. To analyze the effectiveness of the single-wrap and double-wrap rope strake configurations at minimizing vortex shedding, a test was performed on Table 3.5. Summary of effective constant-amplitude fatigue data. TRUNCATION LEVEL > 0.5 ksi > 1.0 ksi ID STRAIN GAGE SReff peff Veff N/Day SReff peff Veff N/Day (ksi) (psf) (mph) (ksi) (psf) (mph) CA-A CH_3 1.28 1.02 20.0 5,820 1.80 1.44 23.7 1,793 CA-X CH_5 1.12 0.89 18.7 5,016 1.63 1.30 22.6 1,234 IAN-A (MT) CH_9 1.36 0.98 19.6 5,927 1.94 1.40 23.4 1,788 IAN-X (MT) CH_12 1.19 0.86 18.3 7,173 1.70 1.23 21.9 2,016 IAS-A CH_2 0.92 1.47 24.0 2,805 1.47 2.36 30.4 356 IAS-X CH_1 0.87 1.40 23.4 3,468 1.41 2.26 29.7 350 KS-A CH_2 1.55 1.07 20.5 12,730 2.12 1.46 23.9 4,622 KS-X CH_6 1.64 1.13 21.1 14,359 2.20 1.52 24.4 5,593 ND-A CH_1 0.92 1.02 20.0 4,547 1.46 1.64 25.3 579 ND-X CH_5 0.97 1.08 20.5 6,170 1.46 1.63 25.3 1,100 OKNE-A CH_3 1.11 0.89 18.6 8,294 1.64 1.31 22.7 1,942 OKNE-X CH_5 1.04 0.83 18.0 8,872 1.55 1.25 22.1 1,845 OKSW-A CH_8 1.08 0.70 16.5 13,997 1.61 1.04 20.1 3,165 OKSW-X CH_6 1.05 0.68 16.3 16,832 1.55 1.00 19.7 3,856 PA-A CH_6 0.81 0.91 18.8 294 1.35 1.51 24.3 16 PA-X CH_1 0.83 0.94 19.1 441 1.36 1.52 24.4 33 SD-A CH_6 0.93 0.77 17.4 11,515 1.51 1.26 22.2 1,453 SD-X CH_8 0.98 0.82 17.9 12,750 1.60 1.33 22.8 1,827 CJE-A (FR) CH_8 1.02 1.22 21.9 18,693 1.57 1.88 27.1 3,472 CJE-X (FR) CH_6 1.08 1.29 22.4 35,437 1.58 1.90 27.2 8,254 CJE-A (MT) CH_4 1.08 1.30 22.5 6,037 1.62 1.94 27.5 1,345 CJE-X (MT) CH_6 1.10 1.32 22.7 7,598 1.62 1.94 27.5 1,800 CJW-A (FR) CH_8 1.06 1.27 22.3 28,228 1.61 1.93 27.5 5,721 CJW-X (FR) CH_6 1.13 1.36 23.0 36,382 1.65 1.98 27.8 9,083 CJW-A (MT) CH_1 1.03 1.23 22.0 6,688 1.59 1.90 27.3 1,252 CJW-X (MT) CH_3 1.02 1.23 21.9 6,934 1.59 1.90 27.3 1,258 A—Along-wind direction X—Across-wind direction FR—Cumulative data without strakes (free) MT—Cumulative data with double strakes (mitigated)

42 Fatigue Loading and Design Methodology for high-Mast Lighting towers the twin Creston Junction HMLTs. Three different strake configurations were tested: no strake, single strake, and double strake. WY-CJW was used as the experimental group where strakes were installed. WY-CJE was used as the control group, where no strakes were installed. Rainflow cycle counting was used to create stress-range histograms for both structures over all periods. The total number of cycles in the histogram was divided by the monitoring duration for each strake period to obtain the equivalent number of cycles per day. The data in Figure 3.7 and Figure 3.8 present the results of this experiment. As seen in the figures, the strakes are successfully reducing the number of cycles in WY-CJW as compared to WY-CJE. Before discussing these results, there are a few important notes to consider: 1. The no-strake, single-strake, and double-strake labels shown in the legend refer to monitoring periods that are the same calendar periods for each HMLT and lasted approximately the same duration (e.g., the no-strake period refers to a monitoring period where strakes were not installed on either pole). Figure 3.7 presents the data for a given “control” pole (i.e., the pole for which there was no strake). Figure 3.8 presents the data from the same time for the corresponding pole where there was either no strake, a single strake, or a double strake installed. 2. The histogram data used to create the figures were truncated to stress-ranges above 1.0 ksi. Truncating the lower bins of a histogram is common in fatigue analysis, and is done here to keep the numbers manageable. 3. The strain gage labels or “channels” for WY-CJW have been shifted to account for the true orientation of the pole. The data presented for WY-CJW starts with Channel 8 followed by Channels 1 through 7. This shift aligns the channels in the figures with those having similar cardinal directions in the field. A more direct comparison between structures can be made in this fashion, as the wind would be coming from approximately the same direction for both HMLTs. Figure 3.7. Cycle counts for WY-CJE (control group—no strakes installed). Figure 3.8. Cycle counts for WY-CJW (experimental group—strakes installed).

Findings and applications 43 4. Wind data for the Creston Junction HMLTs indicate the wind is primarily out of the west. Thus, it would be expected that vortex shedding would cause vibrations mainly in the north- south direction, the across-wind direction, and strain gages placed on the northern and southern faces would have the greatest numbers of cycles. By examining the instrumentation plans in Appendix E (available on the TRB website), WY-CJE has Channels 2 and 6 placed in the north-south direction and Channels 4 and 8 placed in the east-west direction. For WY-CJW, Channels 1 and 5 are in the north-south direction and Channels 3 and 7 are in the east-west direction. Begin by comparing the no-strake period to the single-strake period for WY-CJW, the experi- mental group shown in Figure 3.8. A significant reduction in cycles is noted in all channels with a reduction to less than 1,000 cycles per day in Channels 3, 4, 7, and presumably 8. Unfortunately, the data for Channel 8 are unavailable for the single-strake period. These channels represent activity in the along-wind direction. The remaining channels represent activity in the across-wind direction, where vortex shedding is expected to be more prevalent. Extending the comparison to the double-strake period, there again is a significant reduction in the number of cycles per day. However, cycles for the double-strake period are reduced by approximately the same value for all channels, to less than 1,000 per day in each case. The cycle reduction for all channels is presumed to be because the double wrapping disrupts the vortices formed by wind from any direction. The sketches in Figure 2.18 best illustrate this reduction. Note that regardless of direction, there is always a strake on any windward face. This contrasts the single strake where there are lengths of pole where there is no strake on a downwind face. Providing a disruption at any given point on the pole significantly increases the likelihood of disrupting the flow that leads to vortex shedding. After establishing the effectiveness of the double-wrapped strake, data continued to be collected for WY-CJW with the strake in place. The strake was then removed and installed at WY-CJE to replicate the results. Cycle counts for the free and mitigated conditions are presented in Figure 3.9 and Figure 3.10. These figures illustrate the effectiveness of the double-wrap strake for both HMLTs. Note that time periods and wind conditions were not the same for each; however, a reasonable comparison may be made considering the aggregate effect in the long-term. It is also important to review the effective stress-range for any notable differences. One major concern of helical strakes is the increased area and corresponding drag they add to a structure. From the fatigue analysis there is no evidence that the addition of strakes results in the generation of larger stress ranges. The difference in the effective stress range from WY-CJW compared to WY-CJE is negligible between the free and unmitigated data sets. This is illustrated in Table 3.6 and Table 3.7. Histogram data used to calculate the stress ranges in the tables were truncated in Figure 3.9. Free (FR) and mitigated (MT) cycle counts for WY-CJE.

44 Fatigue Loading and Design Methodology for high-Mast Lighting towers WY-CJE CHANNEL 1 2 3 4 5 6 7 8 SReff (ksi) FR 1.68 1.57 1.44 1.60 1.69 1.58 1.49 1.57 MT 1.62 1.61 1.58 1.62 1.62 1.62 1.59 1.60 Table 3.6. Free (FR) and mitigated (MT) effective stress-range values for WY-CJE. the same manner as the data used in the previous figures. For all practical purposes, the effective stress range is the same for both HMLTs given the number of other varying factors between the structures such as local terrain effects, location of handhole, etc. A discussion of the load effect is presented later in this section and in 3.1.4.2. In addition to cycle counts, the effectiveness of the double-wrap strake can be measured by examining the occurrences of across-wind excitation. In contrast to cycle counts, this method excludes the effect of along-wind stress cycles and, theoretically, the effect of buffeting. By excluding the along-wind response, the effect of vortex shedding can be more clearly evaluated. Plots of across-wind excitation for the free and mitigated conditions are presented in Figure 3.11 and Figure 3.12. Although the histogram data have previously shown a reduction in the number of damaging stress-cycles, this data shows the reduction in damaging stress-range. 3.1.4.1.2 Clear Lake, Iowa. To analyze the effectiveness of strake coverage, an experiment was completed on the IA-N HMLT. Data were collected for a period of time where the HMLT was covered by a full-length double-rope strake, and another period where only the top third was covered. The HMLT has previously been monitored (Connor and Hodgson, 2006) and was known to be susceptible to vortex-induced vibration. Strain gages for this study were placed in the same location as strain gages from the previous study so a direct comparison of the two data sets could be made. No data for the free condition were collected during this study, since it could be incorporated from the previous research. Results of this experiment are presented in Figure 3.13 and Figure 3.14. Figure 3.10. Free (FR) and mitigated (MT) cycle counts for WY-CJW. Table 3.7. Free (FR) and mitigated (MT) effective stress-range values for WY-CJW. WY-CJW CHANNEL 8 1 2 3 4 5 6 7 SReff (ksi) FR 1.61 1.73 1.64 1.62 1.60 1.72 1.65 1.60 MT 1.55 1.59 1.59 1.59 1.57 1.57 1.60 1.57

Findings and applications 45 Figure 3.14. Free and mitigated occurrences of across-wind excitation for IA-N. Figure 3.13. Cycle counts for IA-N. Figure 3.12. Free and mitigated occurrences of across-wind excitation for WY-CJW. Figure 3.11. Free and mitigated occurrences of across-wind excitation for WY-CJE.

46 Fatigue Loading and Design Methodology for high-Mast Lighting towers Data for the free condition were only available for Channels 9 and 11; Channel 9 represents the along-wind direction, and Channel 11 represents the across-wind direction. Both figures show that the one-third strake was not as effective at mitigating vortex shedding as the full-length strake. Although the one-third strake did reduce the number of cycle counts as shown in the bar chart, it was not as effective as the full-length strake, indicating vortex shed- ding still occurred on the lower regions of the pole. Also, it is important to note the full-length strake on the Iowa pole is shown to produce results similar to those obtained in Wyoming. This further suggests the full-length strake is not an anomaly of one type of structure or one specific geographic location but rather something applicable to a variety of structures across the country. 3.1.4.2 Mitigated Fatigue Load Effect As stated earlier, based on the data collected herein, there is no evidence that the addition of strakes results in generation of larger stress ranges. Mitigation of vortex shedding appears to mainly affect the accumulation of stress cycles. This is illustrated in Figure 3.15, Figure 3.16, and Figure 3.17. Each figure plots normalized histogram data for both the free and mitigated conditions along with best-fit lines. Each histogram can be modeled according to an exponential equation of the type Equation 3.1: Exponential model for histogram data y Aebx= where A equals the y-intercept, a parameter representative of the cycle count, and b equals the slope of the line, a parameter related to static pressure-range. In each figure, the slope of the free (FR) and mitigated (MT) data is observed to be roughly parallel, while the mitigated data is clearly offset below the free data to a lower y-intercept. In addition to the histogram data presented above, the effect of mitigating vortex shedding can be observed by comparing values for the fatigue-limit-state pressure range, constant-amplitude effective pressure-range, and cycle frequency. The values are listed in Table 3.8, Table 3.9, and Table 3.10, along with the ratio of the free to mitigated condition, and a root-mean-square-error IA-N Figure 3.15. Free and mitigated histogram data for IA-N (“free” data from Connor & Hodgson).

Findings and applications 47 (RMSE) calculation using an expected value of unity. By comparing the RMSE values, it can be seen that the difference in variation between the load values and the cycle frequency values varies by an order of magnitude. In terms of fatigue load effect, there is little variation in the amplitude of the load applied, but considerable reduction in frequency. 3.2 Results of Aerodynamic Tests This section encompasses the results found in the aerodynamic studies. One of the interesting results was that the CFD Unsteady Reynolds Averaged Navier Stokes (URANS) Program was not calculating separation well, and change to a large eddy simulation (LES) would need to be pursued. WY-CJE Figure 3.16. Free and mitigated histogram data for WY-CJE. Figure 3.17. Free and mitigated histogram data for WY-CJW. WY-CJW

48 Fatigue Loading and Design Methodology for high-Mast Lighting towers FATIGUE-LIMIT-STATE PRESSURE RANGE (pfls) ID FR MT FR/MT E2 (psf) (psf) IAN-A n/a 4.46 - - IAN-X n/a 3.76 - - CJE-A 4.87 5.48 0.889 0.012 CJE-X 5.61 5.60 1.001 0.000 CJW-A 6.12 4.76 1.286 0.082 CJW-X 6.54 4.73 1.381 0.145 RMSE = 0.245 n/a data not available in Connor and Hodgson report E2 error calculated using an expected value of unity Table 3.8. Comparison of free and mitigated pfls values. CONSTANT-AMPLITUDE EFFECTIVE PRESSURE-RANGE (peff) ID FR MT FR/MT E2 (psf) (psf) IAN-A 0.81* 0.98 0.828 0.030 IAN-X 0.92* 0.86 1.068 0.005 CJE-A 1.22 1.30 0.943 0.003 CJE-X 1.29 1.32 0.980 0.000 CJW-A 1.27 1.23 1.028 0.001 CJW-X 1.36 1.23 1.106 0.011 RMSE = 0.091 * data from Connor and Hodgson E2 error calculated using an expected value of unity Table 3.9. Comparison of free and mitigated peff values. Table 3.10. Comparison of free and mitigated cycle frequency values. CYCLE FREQUENCY ID FR MT FR/MT E2 (N/day) (N/day) IAN-A 44,054* 5,927 7.43 41.4 IAN-X 66,620* 7,173 9.29 68.7 CJE-A 18,693 6,037 3.10 4.4 CJE-X 35,437 7,598 4.66 13.4 CJW-A 28,228 6,688 4.22 10.4 CJW-X 36,382 6,934 5.25 18.0 RMSE = 5.1 * data from Connor & Hodgson E2 error calculated using an expected value of unity

Findings and applications 49 From experiments, there were important findings. The corner upwind configuration is much more prone to lock-in than the face upwind configuration. Another finding was that the size of the vortex cells was much larger than expected, with some cells being 18 inches out of a 60-inch model, almost one-third of the total span-wise length. From the pressure side, the irregular geometry is creating artificial separation areas around corners making the flow separate at an upstream position far away from the expected cylinder case. These results are discussed in the following sections. 3.2.1 Pressure The results obtained from the pressure scanner show the mean pressure at each of the model faces. The wind tunnel model was rotated so that the pressure taps were at the required location to take the data presented here. This is illustrated in Figure 3.18. In this figure, the first test was taken when the pressure taps are set at the stagnation point, data were taken, then the tunnel was shut off and the model was rotated 45 degrees, data taken again, and so forth. When 180 degrees worth of data were taken, the tests were concluded and the data were compiled. The same procedure was done with the 12- and 16-sided models, but more tests were taken to ensure all face locations were tested at each configuration. Five tests were performed for the 8-sided model to cover 0 to 180 degrees in 45-degree increments. Seven tests were performed for the 12-sided model in 30-degree increments, and nine tests were performed for the 16-sided model in 22.5-degree increments. Two sets of data were collected for each of the models to check for repeatability. Only the 12-sided model with face upwind configuration will be shown in this section. This model was chosen because it was most prone to oscillate during testing. The other models’ data are shown in Appendix I (available on the TRB website). The data for the face upwind model shown in Figure 3.19 also has experimental pressure data and theoretical inviscid data for a circular cylinder. In Figure 3.19, the non-dimensionalized pressure is shown as a diagram with length-appropriate vectors. This shows the difference in Wind Direction Pressure Taps Test 1, 0 deg Test 2, 45 deg Test 3, 90 deg Test 4, 135 deg Test 5, 180 deg Figure 3.18. Pressure data experimental procedure.

50 Fatigue Loading and Design Methodology for high-Mast Lighting towers pressure between the 12-sided model and the experimental cylinder. The results for the 12-sided model show where separation occurs and how it is different from the circular cylinder case. Separation occurs when the non-dimensionalized pressure vectors start to look the same farther downstream. In this case, the 12-sided cylinder separates at 90 degrees while the circular cylinder separates closer to 60 degrees. The separation for the 12-sided cylinder occurs at a location farther upstream than seen on a constant-taper circular cylinder. This is due to the sharp corners on the model that promote separation at the corner location. (Note that for the real poles, “sharp” corners are not present since there is always some radius at the fold due to the actual fabrication process.) The circular cylinder does not have sharp corners, so the flow doesn’t have a discontinuous point that promotes an adverse pressure gradient. 3.2.2 Wake This section includes a look at some of the data acquired using the hot-wire and traverse system. The hot wires were placed in the wake to check for periodic flow velocity. 3.2.2.1 Data Taking and Signal Conditioning A multitude of waveforms were taken using the current hot-wire system. The data taking points were organized in a grid pattern as shown in Figure 3.20 and Figure 3.21. In these two figures, the distance between points in the span-wise direction is 9 inches, and in the cross-flow direction it is 1 inch. The black edges on the outside of Figure 3.20 represent the wind tunnel walls; flow is out of the page. The aforementioned testing pattern aides in checking the data for consistency; as the probe moves to different span-wise locations, different frequencies are captured. When the probe returns to a previous span-wise location (but a dif- ferent cross-wind location), the frequency should be the same as the data taken in the previous sweep. This proved to be true as long as the data were taken away from the near-model wake in the cross-wind direction (i.e., below the line traced by points 7, 8, 9 in the figures above). To estimate the sampling frequency, a generic value for the Strouhal number of 0.21 was used and Grey arrows: Dodecagon tapered pole Black arrows: Circular cylinder Inwards arrows: P > Patm Outwards arrows: P < Patm Wind direction Figure 3.19. Vector plot of pressure data for the 12-sided model and circular cylinder.

Findings and applications 51 the velocity in the tunnel was set to approximately 10 m/s (22.4 mph). This Strouhal number value came from Figure 3.22 for 103 < Re < 10 5 and the velocity from the geometry and Reynolds number that was expected. The results showed expected frequencies between 15 and 20 Hz for the range of diameters used on the models. A sample of one of the waveforms can be seen Figure 3.23. The FFTs of the signal were not very clear, so some conditioning was done on the data; the wave was windowed and the mean was subtracted from the value. An FFT before and after clean-up can be seen in Figure 3.24. The post-processing of data shows a sharper frequency peak as well as less noise overall in the rest of the data. All sets of data were processed this way. The windowing function and code can be seen in Appendix I (available on the TRB website). 3.2.2.2 Cross-Wind Study The results from the hot-wire grid pattern study referenced in the previous subsection are presented in this subsection. After the data were post-processed, the frequency peak on each FFT was plotted against cross-wind position in model diameters as shown in Figure 3.25. This study was done to check that the vortex shedding frequencies were consistent at different cross-wind locations. There are five different cross-wind points and three span-wise points as discussed at the beginning of the chapter. To represent the different span-wise locations, different symbols were used. Three locations were picked, the span-wise center of the model and 9-inch locations to either side of the span-wise center. 1 2 3 7 8 6 5 4 9 12 11 10 13 14 15 Figure 3.20. Hot-wire testing grid and order of data taking procedure. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 3.21. Side view of the testing grid–flow is to the right.

52 Fatigue Loading and Design Methodology for high-Mast Lighting towers 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 Mooring cables and Umbilicals Towed Cables Marine Structures Region in which eddy shedding frequency can be defined as the dominant frequency in the spectrum Region of turbulent vortex trail laminar boundary layer on cylinder R eg io n of d ev el op in g tu rb ul en ce in th e fre e vo rte x la ye r R eg io n of tu rb ul en t v or te x, Tu rb ul en t b ou nd ar y la ye r o n cy lin de r Un st ab le v or te x re gi on La m in ar v or te x sh ed di ng re gi on La m in ar to tu rb ul en t t ra ns itio n in B ou nd ar y La ye r o n Cy lin de r ST RO UH AL N UM BE R, S REYNOLDS NUMBER, RE Figure 3.22. Strouhal versus Reynolds number for a circular cylinder with no taper. Figure 3.23. Hot-wire signal before and after processing. 0 1 2 3 4 5 6 7 8 9 10 -0.02 -0.01 0 0.01 0.02 Data before post-processing Time (s) Vo lta ge (V ) 0 1 2 3 4 5 6 7 8 9 10 -0.02 -0.01 0 0.01 0.02 Data after post-processing Time (s) Vo lta ge (V )

Findings and applications 53 Figure 3.25 shows cross-wind frequency data from the face-upwind configuration of the 12-sided model. The same configuration from other models shows the trend shown in Figure 3.25 where the data points are close to 16 Hz as they approach one model diameter distance from the model but are more spread out. If the probe is located closer to the center of the model, the readings are not coherent vortex shedding. When taking data farther than one model diameter from the cross-wind center, the results are the same as those taken near the one model diameter mark. The difference between the data points denoted by the first three symbols in the key to Figure 3.24. Real value of FFTs before and after processing. Figure 3.25. Cross-wind frequencies for 12-sided model, face-upwind configuration.

54 Fatigue Loading and Design Methodology for high-Mast Lighting towers Figure 3.25 and the data points denoted by the last three symbols in the figure’s key is that the natural frequency of the system on the first three symbols is much higher than the shed vortices coming off the model. For the data points denoted by the figure’s last three symbols, the springs have a natural frequency close to the shedding frequency of the model. It is important to note that at a certain cross-wind distance from the model, some frequencies are very low. This is due to not getting a coherent signal; the data are from one of the recirculation zones on the wake. The hot-wire readings at these locations are not periodic vortex shedding. According to the Strouhal relation, if the wind velocity is kept constant and the Strouhal number is assumed constant, then as diameter changes, the frequency should change linearly. This is true for a circular cylindrical case as seen in the section on Vortex Shedding. This is readily apparent for this face-upwind configuration. However, this frequency change is not so readily seen in the vertex-upwind configuration shown in Figure 3.26. The vertex-upwind configuration shows the dominant frequency to be approximately the same for different span-wise configurations. This configuration exhibits “lock-in” for different diameters, showing the same frequency for different diameters. The only parts that show low frequencies are close to the trailing edge to the model inside the recirculation zones. These areas show pure noise and no real coherent signal. This was expected after seeing the results from the face-upwind configuration. The frequencies lining up at one frequency means that the vortices hitting the hot-wire probe have nearly the same frequency and confirms the existence of vortex cells. These cells are necessary to get a uniform excitation frequency on the structure. If the vortex cells’ frequency is near the natural frequency of the HMLT pole that means that a portion of the structure is resonating and this resonance can cause an increase in amplitude and a greater potential for fatigue and failure. 3.2.2.3 Signal Partitioning Another test was performed to check that the shedding frequency is constant in time. That is, a section of the model is not locking on to one frequency and, after a short time, is locking on to another shedding frequency. This would be a shift of the vortex cells through time and an Figure 3.26. Closer frequencies for 12-sided model at different diameters, or “lock-in.”

Findings and applications 55 important result to determine the size of the vortex cells. The results can be seen in Figure 3.27. A cross-wind scan (five 10-second waves taken at different cross-wind locations) was segmented into 2-second windows and the shedding frequencies for each of these 2-second segments was plotted as a “sample.” Each cross-wind location thus has five samples that are portrayed as the X-axis in Figure 3.27, the cross-wind location is shown in the legend, and the frequency for each segment corresponds to the data point’s location on the Y-axis. As Figure 3.27 shows, for the most part, the 2-second windows have approximately the same results for the waves that are farther away from the model. The data taken closer to the model’s center (with respect to the cross-wind direction) exhibit a large discrepancy, sometimes giving an incorrect frequency. This is due to the nature of the inner wake where the data are scattered and full of noise due to the recirculation zones on the trailing edge of the model. These results have confirmed that as long as the probe is far enough away from the inner wake of the model, the data acquired will be useful and repeatable for this study. This study means that the readings at each location, except the ones too close to the model, are repeatable and are not jumping between frequencies. 3.2.3 Smoke Wand The smoke wand study was used initially to find a good cross-wind location to take hot-wire data. Data were taken at this location, but later the cross-wind study used for Section 3.2.2, Wake, was done to check for consistency. The shear layer separates at approximately the 45-degree location, however the oscillations are at a frequency between 15 and 20 Hz and very low ampli- tude so they cannot be seen in any of the smoke wand pictures. This separation location appears to be the same as the pressure data suggest. 3.2.4 Computational Fluid Dynamics The computational fluid dynamics models were done as a way to check separation surfaces using a standard URANS on the program FLUENT, anything more complex than 2-D is suspect since there is a range of magnitudes that would require an extremely large amount of grid points. Figure 3.27. Two-second segments of signals.

56 Fatigue Loading and Design Methodology for high-Mast Lighting towers To put this in perspective, the boundary layer may be ½-inch thick (and you need 10 points in it to calculate separation) and the pole is up to 160 feet tall. The amount of points it would take would be a stretch; the whole experiment time would have to be devoted to this endeavor. However, a 2-D model can be used to check general separation locations. The Cfdesign software provided limited results. It showed certain low-pressure areas cor- rectly and seemed to predict an “updraft” (a span-wise flow due to the taper on the structure). The software used organic meshing, creating its own mesh with certain input parameters from the user. The FLUENT CFD model showed separation further downstream than was seen on the experimental model for every case. These models are URANS calculations and may not take into account surface roughness or other factors. Separation has traditionally been an issue in CFD modeling. In all, the CFD modeling could be done using large eddy simulation (LES) for greater turbulent accuracy, but this would have to be done as a future study. 3.3 Development of Proposed Specification The proposed specification changes developed herein perpetuate the method of infinite life design for HMLTs while redefining the fatigue design load. A “combined wind effect,” a new loading concept, is introduced as part of the proposed specification. This new load more realistically mimics the loading of an in-service structure by combining the effects of buffeting and vortex shedding into a single load. The proposed specification also considers variation of mean wind speed and gives designers the ability to select a design pressure based on local wind data. Changes to the fatigue importance categories also have been made in the proposed specification. Using the importance categories, consequence of failure is considered based on HMLT proximity to the roadway. 3.3.1 Fatigue Design Life The stress-range histogram data collected through this study diminish the uncertainty in the number of lifetime loading cycles for a given HMLT and have established that the number of loading cycles is well beyond the limiting number of cycles at the constant-amplitude fatigue limit (CAFL). Table 3.11 demonstrates the need to use infinite life design for HMLTs. For each channel considered, critical detail categories were selected for comparison such that the CAFL exceeds the fatigue-limit-state stress range (SRfls), which is the minimum stress-range used to achieve infinite life. Then, the limiting number of cycles (Nlim) was determined for the detail based on the appropriate S-N curve and compared to the total expected number of cycles (Ntotal) for a 50–year life. This process was carried out for three different truncation levels assuming the selected CAFL is the maximum anticipated stress-range regardless of truncation level. For most cases, infinite life design is required. Where an HMLT has not entered the infinite life regime, it does not necessarily imply finite life: the detail may be altered, thereby increasing the critical fatigue stress-range and decreasing the limiting number of cycles for infinite life. For the instances where infinite life is not required, the sites are observed to exhibit either low demand in terms of cycle counts, or a high fatigue resistance, each resulting in low SRfls values. For example, prior to monitoring, the IA-S HMLT was retrofitted with a 0.625-inch-wall tubular pole section that drastically reduced the observed stress-range data. (The retrofit was not part of this study but a strategy undertaken by Iowa DOT as part of a separate project.) This reduction results in a conservative critical detail category and has a significant effect on Ntotal with increased truncation. The ND HMLT has a similarly high fatigue resistance with a 0.438-inch wall. Together, the IA-S and ND HMLTs were the stiffest in

Findings and applications 57 the study. For the PA HMLT, two factors allow for finite life: low demand and a unique inner reinforcing sleeve. In addition, the IA-S and PA HMLTs have relatively short intervals of data collection, which may skew the results with regard to expected life. 3.3.2 Fatigue-Limit-State Pressure Range Consequence of failure is currently incorporated into the specification through the use of Fatigue Importance Categories. The concept of “importance” allows owners to adjust the level of structural reliability of sign, signal, and lighting structures based on individual design conditions. Currently, the specification commentary recommends most HMLTs be considered Category I, the most conservative importance category. However, conditions such as distance to roadway and installation of effective vibration mitigation devices can affect the consequence of failure and structural reliability, respectively. Although the mitigation strategy previously TRUNCATION LEVEL NONE* > 0.50 ksi > 1.0 ksi ID SRfls DETAIL Nlimit NTotal INF. LIFE NTotal INF. LIFE NTotal INF. LIFE (ksi) CAT. @ CAFL REQUIRED? REQUIRED? REQUIRED? CA-A 5.76 D 6.4E+06 2.8E+08 Yes 1.1E+08 Yes 3.3E+07 Yes CA-X 4.51 D 6.4E+06 3.0E+08 Yes 9.2E+07 Yes 2.3E+07 Yes IAN-A (MT) 6.17 D 6.4E+06 2.6E+08 Yes 1.1E+08 Yes 3.3E+07 Yes IAN-X (MT) 5.21 D 6.4E+06 3.2E+08 Yes 1.3E+08 Yes 3.7E+07 Yes IAS-A 3.24 E 1.2E+07 1.9E+08 Yes 5.1E+07 Yes 6.5E+06 No IAS-X 2.87 E 1.2E+07 2.5E+08 Yes 6.3E+07 Yes 6.4E+06 No KS-A 7.17 C 4.4E+06 6.0E+08 Yes 2.3E+08 Yes 8.4E+07 Yes KS-X 7.73 C 4.4E+06 7.2E+08 Yes 2.6E+08 Yes 1.0E+08 Yes ND-A 3.54 E 1.2E+07 2.9E+08 Yes 8.3E+07 Yes 1.1E+07 No ND-X 3.87 E 1.2E+07 3.3E+08 Yes 1.1E+08 Yes 2.0E+07 Yes OKNE-A 4.58 D 6.4E+06 4.2E+08 Yes 1.5E+08 Yes 3.5E+07 Yes OKNE-X 4.19 E 1.2E+07 4.8E+08 Yes 1.6E+08 Yes 3.4E+07 Yes OKSW-A 4.45 E 1.2E+07 7.2E+08 Yes 2.6E+08 Yes 5.8E+07 Yes OKSW-X 4.61 D 6.4E+06 8.7E+08 Yes 3.1E+08 Yes 7.0E+07 Yes PA-A 2.32 E' 2.2E+07 5.2E+07 Yes 5.4E+06 No 2.9E+05 No PA-X 2.23 E' 2.2E+07 8.8E+07 Yes 8.1E+06 No 6.1E+05 No SD-A 3.40 E 1.2E+07 5.8E+08 Yes 2.1E+08 Yes 2.7E+07 Yes SD-X 3.74 E 1.2E+07 6.3E+08 Yes 2.3E+08 Yes 3.3E+07 Yes CJE-A (FR) 4.06 E 1.2E+07 1.0E+09 Yes 3.4E+08 Yes 6.3E+07 Yes CJE-X (FR) 4.68 D 6.4E+06 1.7E+09 Yes 6.5E+08 Yes 1.5E+08 Yes CJE-A (MT) 4.57 D 6.4E+06 2.8E+08 Yes 1.1E+08 Yes 2.5E+07 Yes CJE-X (MT) 4.67 D 6.4E+06 3.6E+08 Yes 1.4E+08 Yes 3.3E+07 Yes CJW-A (FR) 5.10 D 6.4E+06 1.2E+09 Yes 5.2E+08 Yes 1.0E+08 Yes CJW-X (FR) 5.45 D 6.4E+06 1.5E+09 Yes 6.6E+08 Yes 1.7E+08 Yes CJW-A (MT) 3.97 E 1.2E+07 3.1E+08 Yes 1.2E+08 Yes 2.3E+07 Yes CJW-X (MT) 3.95 E 1.2E+07 3.2E+08 Yes 1.3E+08 Yes 2.3E+07 Yes NTotal values based on 50-year design life. *All stress-range data collected was automatically truncated at 0.25 ksi. Table 3.11. Determination of infinite life based on study data.

58 Fatigue Loading and Design Methodology for high-Mast Lighting towers discussed is effective and would result in a more reliable design, the research team believes that further study is needed to assess the effect of all potential mitigation devices prior to explicitly considering them for the design of new poles. Mitigation methods, for both disrupting vortex shedding and damping out vortex-induced vibrations, should be examined in more detail. For this reason, the proposed importance categories do not yet consider mitigation as a method of explicitly increased reliability. Decisions are based solely on consequence of failure. This study divides consequence of failure of HMLTs into the following two outcomes: 1. Low risk to traffic, where the distance from edge of the roadway to the HMLT is greater than the height of the HMLT and 2. High risk to traffic, where the distance from edge of roadway to HMLT is less than the height of the HMLT. It is recognized that application of this provision could result in different HMLTs being designed for different fatigue loads at the same interchange, depending on the consequence of failure. Although this may not seem to be worth the effort in design, it could result in more economical structures or, more importantly, encourage designers to place the structures in locations where the risk is less. Proposed importance categories for HMLTs are presented in Table 3.12. The proposed combined wind effect is directly related to the fatigue-limit-state pressure-range by the following expression: Equation 3.2: Combined wind effect P P CCW FLS d= where PCW is the combined wind effect pressure-range, PFLS is the fatigue-limit-state pressure range, and Cd is the AASHTO drag coefficient. Calculated fatigue-limit-state static pressure-ranges based on histogram data and corresponding measured mean wind speed values are tabulated in Table 3.13. Note, the measured mean wind speed values presented are independent of the velocity range values listed in previous tables, which are derived from stress-range histogram data. The highlighted values exceed the current equivalent static natural wind gust pressure- range set forth in Equation 11-5 of AASHTO Signs (2009), which equals 5.2 psf when Cd and IF are both equal to 1. The data are plotted in Figure 3.28 along with curves for Equations 11-5 and C11-5 (AASHTO, 2009). From this graph, it is apparent that the upper bound limit and the existing “adjustment” equation accounting for yearly mean wind velocity are unconservative in many cases. The pressure-range histogram data in Figure 3.6 are shown to vary with mean wind speed, and the same is true for the fatigue-limit-state pressure range values. To account for variation in wind speed, a linear model was fit to the data representing the mean pressure-range. That line was then offset two standard deviations to form an upper bound accounting for all other variation in HMLT geometry, HMLT details, topographic effects, etc. The upper-bound static pressure-range is plotted TRAFFIC IMPORTANCE RISK? CATEGORY HIGH I LOW II Table 3.12. Proposed fatigue importance categories for HMLTs.

Findings and applications 59 with the fatigue-limit-state data in Figure 3.29. To determine the proposed PFLS values, the variation in measured mean wind speed is divided into the following three outcomes: 1. Location-specific yearly mean wind velocity is less than national average, 2. Location-specific yearly mean wind velocity is greater than national average but less than the yearly mean wind velocity with 16 percent exceedance (one standard deviation), and 3. Location-specific yearly mean wind velocity is greater than 16 percent exceedance. ID WSmean Pfls (mph) (psf) CA-A 8.7 4.6 CA-X 8.7 3.6 IAS-A 10.3 6.1 IAS-X 10.3 5.4 KS-A 9.4 5.0 KS-X 9.0 5.3 ND-A 6.7 4.0 ND-X 6.7 4.3 OKNE-A 6.8 3.7 OKNE-X 6.8 3.4 OKSW-A 8.3 2.9 OKSW-X 8.8 3.0 PA-A 3.5 2.6 PA-X 3.5 2.5 SD-A 7.8 2.8 SD-X 7.8 3.1 CJE-A (FR) 12.8 4.9 CJE-X (FR) 12.8 5.6 CJW-A (FR) 14.0 6.1 CJW-X (FR) 14.0 6.5 Table 3.13. Fatigue-limit- state pressure-range and recorded mean wind speed. EXISTING Figure 3.28. Plot of fatigue-limit-state data against existing specification equations.

60 Fatigue Loading and Design Methodology for high-Mast Lighting towers The 16 percent exceedance level is the upper bound previously set by NCHRP Report 412 (1998) and corresponds to approximately 11 mph. By finding the intersection of the national average yearly mean wind velocity with the measured upper-bound pressure-range from this study, a proposed minimum design pressure-range of 5.8 psf was set. The national wind data used to determine pressure-range values from the upper bound are provided in the report “Comparative Climatic Data for the United States through 2010” published by the National Climatic Data Center, a division of NOAA (2010). This is a more current and comprehensive set of data than used in the work performed in the mid 1990s during the preparation of NCHRP Report 412 (1998). Annual mean wind velocities from 238 stations in the lower 48 states were considered. Data from Alaska, Hawaii, and U.S. territories were specifically excluded because of the tendency for higher annual wind speed at coastal and island stations. From the data set, the national average yearly mean wind velocity is 9.0 mph, the yearly mean wind velocity with 16 percent exceedance is 10.9 mph (11 mph for simplification), and the yearly mean wind velocity with 2 percent exceedance is 12.8 mph. A wind contour map from the National Climatic Data Center is shown in Figure 3.30. A design equation similar to equation C11-5 (AASHTO, 2009), which allows designers to vary static wind pressure based on the location-specific mean wind velocity, is not recommended for design of HMLTs. Three issues prevent recommending this approach for the proposed revisions. 1. The measured fatigue-limit-state data do not easily fit a parabolic curve through the origin like equation C11-5 suggests. Hence, the equation is inaccurate in “scaling” actual response to various mean wind speeds. 2. Considerable variation exists in the local measured wind speed data collected in this study when compared to wind velocity data provided by NOAA (2010). It is recognized that the NOAA data is from a more comprehensive study of wind across the country. As a result, the measured mean wind speed for a given HMLT in this study may vary significantly from those suggested in the contour map at the same location. Although they may be reasonable overall, adjusting the suggested wind speed solely on the contour map implies a level of accuracy not justified by this study. 3. Additional variation exists due to topographical effects and site-specific details encountered at the long-term monitoring sites. For example, the difference in mean wind speed noted at the two Oklahoma HMLTs is likely due to differences in elevation, location at the highway interchange, and/or blockages near the structure such as trees. Figure 3.29. Plot of fatigue-limit-state data and proposed upper boundary. PROPOSED

Findings and applications 61 Therefore, a means to adjust the design pressure-range based on yearly mean wind velocity is proposed using a three-tiered approach. Rather than scale entirely on the local wind speeds measured herein, it is proposed to scale on the readily available NOAA data, but in a much coarser fashion than the existing AASHTO Signs Commentary section provides. Although the local mean wind speed measured by this project does not always compare well with the data available from NOAA, histogram data do confirm that as the average wind speed increases, the fatigue- limit-state pressure range for a given HMLT also increases. Again, the level of accuracy suggested by the existing commentary, scaling directly on yearly mean wind velocity, is not justified. The proposed minimum fatigue-limit-state pressure range (PFLS) is based on the national average yearly mean wind velocity of 9 mph. This corresponds to a PFLS value of 5.8 psf. If the yearly mean wind velocity is greater than 9 mph, but less than 11 mph (i.e., within one standard deviation from the national mean wind speed), the PFLS is increased to 6.5 psf. If the yearly mean wind velocity is greater than 11 mph (i.e., greater than one standard deviation away from the mean), the PFLS is further increased to 7.2 psf. As stated, this is different than the existing method as it does not attempt to “split hairs” regarding the estimate of the location-specific yearly mean wind velocity. The three possible PFLS outcomes are presented in Table 3.14 according to the applicable yearly mean wind velocity. In summary, the proposed PFLS values of 5.8, 6.5, and 7.2 psf are based on yearly mean wind velocities of 9, 11, and 13 mph, respectively. However, designers should be cautioned on the effects of topography and site-specific wind effects when considering location-specific mean wind velocity in their design. For example, take the data measured at the Pennsylvania HMLT and compare it to what a designer would use based on the NOAA map and proposed static pressure-range values. The NOAA contour map suggests the Pennsylvania location would be subjected to a greater mean wind speed. However, Figure 3.30. NOAA annual mean wind speed contour map (2010). Source: NOAA National Climatic Data Center, http://www.ncdc.noaa.gov/oa/ncdc.html

62 Fatigue Loading and Design Methodology for high-Mast Lighting towers the HMLT in Pennsylvania was largely shielded by local terrain and trees, which is likely the reason the mean wind speed shown in Table 3.13 was much lower than the NOAA contour. In contrast, the Wyoming site recorded average wind speeds exceeding those suggested by the NOAA contour map. However, the topography around the Wyoming sites was open with little obstruction. Further, it is highly unlikely one of the 238 NOAA monitoring stations was located near Creston Junction; thus, it was smeared into the contours of the nearest recording stations. Additionally, it should be noted that it is not appropriate to use a value less than 5.8 psf where the yearly mean wind velocity is shown to be less based on the NOAA map. The minimum pressure is based on the mean wind speed and helps to account for any local effects unknown to the designer. A comparison of experimental data with the proposed method is presented in Table 3.15. Measured mean wind speed data are listed side by side with estimated yearly mean wind veloci- ties from the NOAA contour map. The highlighted values show where experimental values are unconservative compared with the map. Measured Pfls values are listed side by side with proposed YEARLY MEAN WIND VELOCITY, Vmean PROBABILITY BASED ON Vmean PFLS (psf) Vmean ≤ 9 mph 50% 5.8 9 mph < Vmean ≤ 11 mph 34% 6.5 Vmean > 11 mph 16% 7.2 Table 3.14. Proposed static pressure-range values. Table 3.15. Comparison of experimental data with proposed method. ID WSmean NOAA MEAS. Pfls PROP. PFLS (mph) (mph) (psf) (psf) CA-A 8.7 8-9 4.6 5.8 CA-X 8.7 3.6 IAS-A 10.3 9-10 6.1 6.5 IAS-X 10.3 5.4 KS-A 9.4 10-11 5.0 6.5 KS-X 9.0 5.3 ND-A 6.7 9-10 4.0 6.5 ND-X 6.7 4.3 OKNE-A 6.8 9-10 3.7 6.5 OKNE-X 6.8 3.4 OKSW-A 8.3 9-10 2.9 6.5 OKSW-X 8.8 3.0 PA-A 3.5 9-10 2.6 6.5 PA-X 3.5 2.5 SD-A 7.8 9-10 2.8 6.5 SD-X 7.8 3.1 CJE-A (FR) 12.8 9-10 4.9 6.5 CJE-X (FR) 12.8 5.6 CJW-A (FR) 14.0 9-10 6.1 6.5 CJW-X (FR) 14.0 6.5

Findings and applications 63 PFLS values. Even though the wind speed values don’t always match, a conservative design value is determined for each HMLT. The two possible outcomes for importance category and three possible outcomes for yearly mean wind velocity combine to make a total of six possible outcomes. To incorporate consequence of failure into the proposed design method, it is proposed that HMLTs presenting a high risk to traffic be restricted from using the lowest pressure-range value of 5.8 psf. A high risk to traffic was defined by the researchers as any HMLT that could actually fall into the path of traffic. Such a structure is proposed to be classified as Category I. It was felt that poles that could fall into the path of traffic should be designed with a higher probability of survival than is associated the mean wind speed (i.e., a 50-50 chance of exceedance). As a result, it was decided that for structures classified as Category I, increasing the PFLS by one standard deviation provided sufficient reliability. Hence, the intermediate PFLS value of 6.5 psf would then be required for HMLTs located where the mean wind speed is 9 mph or less, but classified as Category I. However, for poles located where the mean wind speed is greater than 9 mph, the research team decided that further increases in the design value for PFLS were not warranted. This decision is based both on engineering judgment and the approach contained in the fatigue loading provisions developed during the research for NCHRP Report 412. Rationale for the proposed approach follows. Both the Category I fatigue importance factor of 1.0, and the wind gust pressure-range set by NCHRP Report 412 are set at the same confidence limit and correspond to a yearly mean wind velocity of 11 mph. This velocity is consistent with wind that is the speed used to develop the PFLS of 6.5 psf. The reduced probability of failure associated with one and two standard deviations from the mean PFLS value was felt to be already conservatively set as they are based on the worst-case response of all poles instrumented. It is recognized that the above approach is somewhat arbitrary; it is felt to adequately address the risk associated with a majority of poles. Further adjustments as a function of ADT or ADTT do not appear warranted at this time as insufficient data exist to set such limits. Rather, the simple check of whether an HMLT can fall into the path of traffic is a straightforward and sufficient criterion. The PFLS values proposed for design are presented in Table 3.16. The proposed approach provides a reasonable and familiar method of accounting for local conditions and consequence of failure without implying a level of accuracy that is not justified. It is noted that fatigue importance factors are conspicuously missing. It is noted that the proposed method of accounting for consequence of failure is not based on a quantitative reliability analysis, which was outside the scope of this study, but from qualitative assessment of the current state of HMLT design. In other words, if the HMLT can fall into the path of traffic, the research team felt that it should meet a higher standard, and hence a higher load range is suggested. One final attractive option of the proposed approach is that state DOTs could easily specify which static pressure-range and/or importance factor should be used in their state, either regionally or for the entire state, to further simplify the approach. FATIGUE-LIMIT-STATE PRESSURE RANGE, PFLS (psf) YEARLY MEAN WIND IMPORTANCE CATEGORY VELOCITY, Vmean I II Vmean ≤ 9 mph 6.5 5.8 9 mph < Vmean ≤ 11 mph 6.5 6.5 Vmean > 11 mph 7.2 7.2 Table 3.16. Proposed PFLS values for design (psf).

64 Fatigue Loading and Design Methodology for high-Mast Lighting towers 3.4 Large-Amplitude Oscillation Over the course of this study, a YouTube video surfaced, which recorded the behavior of an HMLT outside of Watertown, South Dakota, during a late winter storm. A passing motorist shot the video, which shows the HMLT experiencing extreme displacements in the first mode of vibration. It can be viewed at http://www.youtube.com/watch?v=2wpc8qD6AtI. Wind speed that day was estimated to be 30 to 40 mph with higher gusts. Later investigation found the pole to be cracked near the base, and it was promptly taken out of service. Unfortunately, the HMLT in the video was not part of the long-term monitoring study, so no data are available for that particular event. However, the research team searched through collected trigger data, data that was specifically recorded during periods of high wind speed, for any instances of behavior similar to that shown in the video. Two instances were found that may exhibit similar behavior, one at the WY-CJW HMLT in Creston Junction, Wyoming, and the other at the SD HMLT in Rapid City, South Dakota. Both experienced sustained Mode 1 oscillation at stress ranges of approximately 10 ksi for a duration of 3 minutes or more. Both instances occurred during sustained winds of approximately 30 mph, and the movement appears to be across-wind. Data for the WY-CJW event are presented in the following figures. Figure 3.31 shows a stress-time plot for one of the channels and illustrates the sustained harmonic motion. Figure 3.32 is a close-up of the same plot and illustrates the Mode 1 behavior at 0.35 Hz and maximum stress of about 10 to 11 ksi. The estimated dynamic displacement for this magnitude of stress is 3 feet in one direction or about 6 feet of total travel at the luminaire. Figure 3.33 shows a plot of the wind speed for the event at a 3-second average, which is a common averaging time for wind gusts. The mean wind speed over the entire duration is about 32 mph. Also note that the HMLT did not have any strakes installed at the time that the large-amplitude oscillations were recorded. The WY-CJE did have strakes installed, but did not exhibit the same behavior; this may be a coincidence. With regard to fatigue, this phenomenon is a matter of concern; it may be responsible for low-cycle fatigue behavior leading to collapse or significant cracking. If the two instances of large-amplitude oscillation mentioned above truly match the behavior in the video, then it is encouraging to know that it is extremely rare, exceeding the 1:10,000 fatigue-limit-state threshold. Also, the two instances of large-amplitude oscillation are theoretically part of the loading spectrum used to create the proposed fatigue design loads. However, without fully understanding the phenomena responsible for this type of behavior and without stochastic data related to how often it may occur, it is impossible to safeguard against it in a specification. Figure 3.31. Large-amplitude oscillation at WY-CJW.

Findings and applications 65 At this time, the research team can only speculate on the phenomena responsible for this type of behavior. It certainly warrants future research. 3.5 Fatigue Life Evaluation of Existing HMLTs The intention of the proposed fatigue provisions is to provide for infinite life design. For new structures, this is appropriate and economical. However, recent failures and subsequent research shows that, for many existing structures, guidance on how to estimate the remaining fatigue life is needed. The procedures contained herein provide such guidance. These procedures are equally applicable for finite life design of new HMLTs in cases where such an approach is warranted. The first step in evaluating an existing HMLT is to determine whether infinite life is achieved using the appropriate proposed fatigue-limit-state pressure range (i.e., 5.8, 6.5, or 7.2 psf). Although an existing HMLT may have originally been designed for infinite life using the earlier AASHTO Signs provisions, the pole may not meet the proposed (NCHRP Project 10-74) provisions. There are three primary reasons this may be the case, as follows: 1. It is likely that the original design used an unconservative fatigue load. Assuming the proposed 6.5 psf fatigue-limit-state load will apply to most HMLTs, the 5.2 psf load in the current edition Figure 3.32. Close-up of same large-amplitude oscillation. Figure 3.33. Plot of 3-second wind speed during large-amplitude event.

66 Fatigue Loading and Design Methodology for high-Mast Lighting towers of AASHTO Signs (2009) is obviously less. Hence, depending on how “overdesigned” the pole was, it will likely not meet the proposed specification. 2. It has been observed that in many cases, older designs often assumed a higher fatigue category for a given detail, in particular the baseplate detail, than is actually the case. The wealth of information relating to cracking in HMLTs (Dexter, 2004; Connor and Hodgson, 2006) supports this. 3. The pole was designed and built before any fatigue provisions were included in AASHTO Signs. For infinite life design to be valid, applied loading cycles must not exceed the CAFL 99.99 percent of the time, or, in other words, the 1:10,000 confidence limit of the applied loading spectrum must be at or below the CAFL. This load is referred to as the fatigue-limit-state pressure range. To check for infinite life, the stress-range due to the fatigue-limit-state load, SRfls, must be less than the CAFL. In addition to the calculation described above, a qualitative approach also may be used to determine whether the HMLT is subject to a finite fatigue life. Through several research programs and field observations, certain criteria have been identified that may alert an owner that finite life is likely for an existing HMLT. An owner may wish to consider finite fatigue life for any of the following reasons: 1. A fillet-welded socket-type tube-to-base-plate connection, 2. Baseplate thickness less than 3 inches, 3. A history of loose anchor nuts, 4. Less than 6 anchor rods, 5. Tube wall less than or equal to 5⁄16-inch, or 6. Excessive corrosion of tube wall. HMLTs with just one of the factors listed above have been shown through experience (i.e., observed cracking) to have less than the intended fatigue life. After establishing that an HMLT cannot attain infinite life, the evaluation should proceed using the constant-amplitude effective fatigue pressure-range to calculate the effective stress-range, SReff. The number of cycles the HMLT can sustain can then be estimated from the appropriate stress-life (S-N) curve. Since it has already been shown that SRfls exceeds the CAFL, the CAFL should not be considered on the S-N curve. Because the constant-amplitude effective fatigue load is considerably less than the fatigue-limit-state load, SReff will most certainly be less than the CAFL. Therefore, only the sloping portion of the S-N curve need be considered for finite life evaluation. A straight- line extension of the S-N curve will typically be required to perform the assessment, as is commonly done in fatigue evaluation of highway bridges. 3.5.1 Constant-Amplitude Effective Pressure-Range Measured constant-amplitude effective static pressure-ranges and corresponding mean wind speed values are tabulated in Table 3.17 and plotted in Figure 3.34. The effective static pressure- ranges were arrived at by truncating the histogram data to stress-ranges above 0.5 ksi for all sites except IA-S. The IA-S HMLT was retrofitted by replacing the bottom tube portion with one 5⁄8-inch thick, which is 52 percent thicker than the next thickest pole. This additional stiffness causes the calculated pressure range associated with the histogram bins to be significantly larger than the other HMLTs in the study. As a consequence, cycle counts are lower. In other words, the lowest bin was effectively truncated on site because the base of the pole was so stiff. The values given in Table 3.18 are recommended for evaluation of existing HMLTs. The value for the constant-amplitude effective fatigue static pressure-range was determined in a manner similar to the fatigue-limit-state pressure range discussed above. It is noted that this is similar to the

Findings and applications 67 ID WSavg Peff CA-A 8.7 1.02 CA-X 8.7 0.89 IAS-A 10.3 1.01* IAS-X 10.3 0.95* KS-A 9.4 1.07 KS-X 9.0 1.13 ND-A 6.7 1.02 ND-X 6.7 1.08 OKNE-A 6.8 0.89 OKNE-X 6.8 0.83 OKSW-A 8.3 1.04 OKSW-X 8.8 1.00 PA-A 3.5 0.91 PA-X 3.5 0.94 SD-A 7.8 0.77 SD-X 7.8 0.82 CJE-A (FR) 12.8 1.22 CJE-X (FR) 12.8 1.29 CJW-A (FR) 14.0 1.27 CJW-X (FR) 14.0 1.36 *Peff values are truncated to stress ranges > 0.5 ksi except for IAS. Table 3.17. Constant-amplitude effective pressure-range and recorded mean wind speed. Figure 3.34. Plot of constant-amplitude effective pressure data and upper boundary.

68 Fatigue Loading and Design Methodology for high-Mast Lighting towers concept of the HS-15 fatigue truck (0.75 × HS-20), which is used for finite life and is intended to represent the cumulative amplitude truck loading spectrum (AASHTO LRFD, 2010). Similarly, the Peff is intended to represent the equivalent fatigue damage of the variable-amplitude wind pressure spectrum. Both values in Table 3.18 correspond to the national average mean wind speed, which would best represent the effect on HMLTs for evaluation purposes. In lieu of adequate local site data, an owner may wish to use the values provided in Table 3.16 for evaluation. However, this will likely result in an overestimate of the pressure ranges. It is recognized that the data shown in Table 3.16 could be utilized for the infinite life check (i.e., one could account for local mean wind speed effects with Table 3.16). The proposed evaluation procedure will provide commentary that allows an owner to use Table 3.16 if they wish. Overall, using the PFLS of 5.8 will provide a reasonable value for assessment of most HMLTs. It is also noted that little variation exists for Peff values over the range of yearly mean wind velocities set as the design limits. Peff values of 1.25, 1.33, and 1.41 psf are determined from yearly mean wind velocities of 9, 11, and 13 mph, respectively. As a result, a single value of 1.3 psf was selected to represent most locations. Using a single effective pressure-range simplifies the approach, as it does for the design of bridges. 3.5.2 Stress-Range Cycles for Evaluation As discussed above, measured histogram data show cumulative fatigue damage varies pro- portionally with wind speed, as would be expected. Most of this variation occurs in terms of cycle counts. Referring back to Table 3.5, note that the normalized values for constant-amplitude effective pressure-range and velocity range are surprisingly similar, while values for cycles per day vary greatly. The proposed evaluation method takes advantage of this variation and allows evaluating engineers to choose an appropriate cycle frequency (i.e., cycles per day), based on mean wind speed similar to the proposed fatigue loads. Mean wind speeds and effective cycle counts are listed in Table 3.19 and plotted in Figure 3.35. The cycle counts were arrived at by truncating FATIGUE LOADS FOR EVALUATION (psf) Fatigue-limit-state static pressure range, Pfls 5.8* Constant-amplitude effective static pressure-range, Peff 1.3 *Owners may wish to utilize the values for PFLS shown in Table 3.16. Table 3.18. Recommended fatigue load pressure-ranges for evaluation. ID WSavg (mph) N/Day CA-A 8.7 5,820 CA-X 8.7 5,016 IAS-A 10.3 10,167* IAS-X 10.3 13,770* KS-A 9.4 12,730 KS-X 9.0 14,359 ND-A 6.7 4,547 ND-X 6.7 6,170 OKNE-A 6.8 8,294 OKNE-X 6.8 8,872 OKSW-A 8.3 13,997 OKSW-X 8.8 16,832 PA-A 3.5 294 PA-X 3.5 441 SD-A 7.8 11,515 SD-X 7.8 12,750 CJE-A (FR) 12.8 18,693 CJE-X (FR) 12.8 35,437 CJW-A (FR) 14.0 28,228 CJW-X (FR) 14.0 36,382 *Histograms truncated to > 0.5 ksi except IAS. Table 3.19. Effective cycle frequencies and recorded mean wind speed. Figure 3.35. Plot of effective cycle frequencies.

Findings and applications 69 the first bin of histogram data for all sites except IA-S, same as the effective pressure-range data discussed above. The curve in Figure 3.35 is the best-fit parabola through the data crossing the Y-axis at the origin. The recommended stress-range cycles for evaluation were derived from this curve using the same confidence intervals for wind speed discussed in the proposed fatigue loads. However, instead of using the upper limit of the interval, the median value was used. To determine the lower-bound cycle frequency of 9,500/day, 8 mph was used, which fits in the range of wind speed between 7 and 9 mph (one standard deviation below the mean). To determine the intermediate cycle frequency of 15,000/day, 10 mph was used, which fits in the range of wind speed between 9 and 11 mph (one standard deviation above the mean). To determine the upper-bound cycle frequency of 23,000/day, 12 mph was used, which fits in the range of wind speed between 11 and 13 mph (one and two standard deviations above the mean). Also included in the table is the proposed number of stress-range cycles to be used with an HMLT effectively mitigated against vortex shedding. The value is determined from cycle counts from the three sites where full-length double strakes were installed (WY-CJE, WY-CJW, and IA-N). The same level of truncation was used and the mean value is given. This value is likely to be conservative for most HMLTs mitigated against vortex shedding since the three sites that formed the data set experienced mean wind speeds greater than the average. Recommended stress-range cycles for evaluation are summarized in Table 3.20. Although it is recognized that the above approach may be questionable to some, it also must be recognized that there is no rational method available to estimate “N” for a given HMLT short of conducting a field instrumentation study. Hence, at present it is offered as a reasonable method based on a significant amount of data to estimate the number of cycles to which a given HMLT may be subjected. STRESS-RANGE CYCLES FOR EVALUATION N/DAY Vmean ≤ 9 mph 9,500 9 mph < Vmean ≤ 11 mph 15,000 Vmean > 11 mph 23,000 Vortex Shedding Mitigated 7,000 Table 3.20. Recommended stress-range cycles for evaluation.