Page 180 - MY STORY

When the final report came my way for approval and

signature, I almost immediately had a problem. The
linear regression algorithm was making predictions that
somehow seemed to contradict everything I knew about

creep behavior of metals, and in fact, seemed to
contradict “laws of nature.” The final report, of course,
was extolling the evolution of our wonderful linear

regression predictive tool, and I understood that even the
NASA engineer overseeing this program thought it was
one of the best things since God invented “sliced bread.” I

took the report home, examined the experimental data,
and replotted that data by hand on semi- logarithm graph

paper.

The experimental work was so well performed that
almost no data scatter existed on my hand- drawn plots

and the plotted data matched predicted empirical creep
performance curves that dated back many years. The data

clearly showed that cyclic stress had virtually no bearing
on the deformation characteristics, with sustained load
creep being the predominant strain mechanism.

The real problem was that experimental data had little to
do with the linear regression predictive tool being

evolved under the NASA contract. We had spent a lot of