Plate steels are frequently subjected to through-thickness tension tests to assess their susceptibility to lamellar tearing. In the use of test data, however, the technologist is often hindered by a large variability in measured through-thickness reduction-of-area (TTRA) values. In particular, a group of reduction-of-area values from through-thickness specimens may exhibit a standard deviation 5 to I0 times that exhibited by a group from adjacent planar specimens. The technologist must either accept an inability to discern small differences in test results or increase the number of specimens in replicate tests.

Accordingly, an analysis was made of standard deviation values obtained in through-thickness tension testing of plate. In particular, six coupons were taken from each of 108 samples of various plate steels. Prolongs were stud-welded to both surfaces of each coupon, and 12.83-mm (0.500-in.)-diameter by 127-mm (5-in.)-long tension test specimens were strained to failure in a universal testing machine at a crosshead-separation speed of 0.1 mm/s ( 1 ∕ 4 in./min). Reduction of area was measured for each specimen and for each group of six specimens representing a sample, a mean value and a standard deviation value were calculated.

Regression analysis of the data produced the model

standard deviation, % = 4.614 + 1.136 (plate thickness, in.) - 0.004778 (TTRA% - 38)²

Variability among test results increased with increasing plate thickness and decreased as the mean TTRA value departed in either direction from 38 percent.

The overall standard deviation in TTRA values observed was about 5.1 percentage points. This figure indicates that mean TTRA values resulting from sextuplicate tests must differ by more than about 6.5 percentage points to be judged statistically different at the 97.5 percent confidence level. More tests would be required if finer discrimination were desired; fewer tests could be tolerated if lack of certainty about differences in TTRA values greater than 6.5 percentage points were tolerable.