Reliability, Uncertainty, and Geo-Statistics
5.1.1 Introduction – Why Not Just Pick the Best Number?
When I started studying soil mechanics in the 1960s, we had plenty of theories and equations available to develop engineering problem solutions. The primary dilemma facing young geotechnical engineers was selection of the input “numbers” to use when solving a given problem. Fellow engineers told me at the time that with experience and a few basic field and laboratory tests, I could select the correct numbers to use to solve the client's problems. If I had any doubts about the soil parameters required for a particular problem, I could ask my mentor Neil Mason, who was a student of Karl Terzaghi, the father of soil mechanics himself!
As you might expect, I was considered to be a bit of a troublemaker by my co-workers. If I counted 12 blows per foot and the driller counted 11, which number did I use? If I observed the driller lifting the hammer 28 inches per blow rather than 30 inches per blow, what correction factor should I apply? If I ran a PI test and got 15, and the senior lab tech ran the same sample and got 18, do we each run a couple more samples and average the batch? How many PI tests should we run to get the right answer? I was continually asking questions about how much testing was enough testing.
These kinds of problems got me thinking, so I decided to pull out my college statistics text books (told you I was considered a troublemaker). The first thing that struck ...