the timber pile. Figure 13a shows a shift in signal strength measured 

 by the standard deviation from 0.2 to 2.5 mV for a constant 0% cross- 

 sectional area loss. These natural variations were verified when large 

 or significant fluctuations in signal strength were measured in different 

 quadrants of a completely solid pile (see Figure 14) . 



Figure 12. PILE 5X: Cross section divided into pie-shaped 

 quadrants. 



The relationship between the cross-sectional wood loss and the 

 calculated standard deviation of the ultrasonic signal over the 100-usec 

 time increment for the test piles (standard, 5x, 4x1, and 8335D) is in 

 Figure 15. As shown in Figure 13b the ultrasonic readings taken for 

 pile 5x (Figure 15b) were unlike the standard pile. The steady decrease 

 in standard deviation with an increase in cross-sectional wood loss of 

 pile 5x is not as evident. The largest discrepancy exists in the region 

 where 74% cross-sectional wood is lost. Standard deviation increases 

 where a decrease is expected. 



Again, readings vary significantly for a constant cross-sectional 

 area loss. The plots of standard deviation versus the percent of cross- 

 sectional wood loss for pile 4x1 and pile 8335D show a wide range of 

 values with no consistent or significant decrease in standard deviation 

 with cross-sectional wood loss. 



To clarify the extent of association between the percent of cross- 

 sectional wood loss and standard deviation of the ultrasonic signal, a 

 regression analysis was performed on the data collected for each pile. 

 Regression deals with the prediction of one variable from its correla- 

 tion with others. The two variables being correlated (percent cross- 

 sectional area loss and standard deviation of the ultrasonic signal) are 

 shown in graphic form in Figure 16. 



26 



