77)RJUDWHOKU(6'' PJFP

**Figure 7.** Average temperature of right and left trunk areas during the whole test.

From the plots, the different temperature trends are clearly visible for zone a, suggesting that the formation of dry bands is facilitated by the presence of the shed, shielding the wetting on this zone. Another clear indication from the average temperature of subareas is that subarea c always presents the lowest average temperature in each trunk. This suggests that the wetting is enhanced in the other two subareas, meanwhile in the area closer to the shed, the wetting is slightly reduced, creating better conditions for dry-band formation.

It is noted that the textured design tends to initiate dry bands on all the three areas, creating a more uniform distribution. This is not observed on the conventional design, exhibiting a more localized formation, as shown in the Appendix A in Figure A1. In fact, the middle trunk, area 3, does not show any significant increment, meanwhile the other two trunks show temperature increments in zone a. A rapid increase in temperature at around 5 to 6 min before the flashover event is observed.

## *3.3. Dry-Band Analysis*

Another useful indication is the localization and progression of any dry band during the test. The identification of dry bands using a fixed selected threshold cannot be detected correctly, as described in the temperature profile reported in Figure 3. In addition, using a variable level function only on the first IR frame is still not always applicable because of the cooling trend by the fog, as described previously.

The proposed algorithm of dry-band detection is based on identifying the points/area where the minimum average value is exceeded within the trunk zone. This permits to take into account any cooling e ffect in the initial period or overall variation not caused by a localized event. Each dry-band width is then evaluated calculating the average temperature of each left- and right-hand horizontal row of data given by the individual IR pixels. If these values exceed the threshold, the row is flagged and counted. The total dry band extension along the vertical axis for each frame is calculated converting the number of pixels in mm, given the distance between the terminals of the insulator; and in this case, it is equal to 175 mm.

The sum of all dry-band extensions along the insulator axis for all the duration of the test (Design CONV, ESDD 0.64 mg/cm2, fog rate 3 L/hr) is presented in Figure 8. An analogue test using a lower pollution level (ESDD 0.42 mg/cm2) is shown in the Appendix A as Figure A3.

**Figure 8.** Cumulative dry-band extension on all the trunks. Ramp-test: Design CONV, ESDD = 0.64 mg/cm2, and a fog rate of 3 L/hr.

In order to gain a deeper understanding of the growth of each dry band, the calculated dry-band location and duration were plotted on a single graph taking advantage of the contour facility introducing time as the x-axis value, the vertical position as the y-axis value and the temperature of the dry band as the colour level. In addition, the last frame of the IR image is automatically cropped according to the identified boundaries and rescaled along the vertical axis of symmetry, facilitating the user to localize the dry band on the insulator. The resulting plot is shown in Figure 9. This graph o ffers a valuable overview of the full test since each dry-band width is presented as function of time and as its dry-band temperature range. The graph allows the comparison of the dry-band temperature distributions for a specific design very readily. The individual dry-band extension function of time and the dry-band temperature range for the conventional design (Figure 9a) and textured TT4 insulator (Figure 9b) are also presented.

**Figure 9.** Dry-band extension as function of time and dry-band temperature range for the (**a**) conventional design and (**b**) textured TT4 insulator examples.
