Evolution of geotechnical risk analysis in North American practice - Risk and Reliability in Geotechnical Engineering - page 474

12.9.3 Dynamic risk analysis and management
The emergence of real-time sensing combined with web-based data communication and visu-
alization has led to risk-monitoring capabilities to enhance risk management that has great
potential and, at one time, was only speculated about. Modern sensor systems have allowed
geotechnical performance data to be measured in real time and communicated to central
databases where even more powerful computer systems are used to model field conditions as
they change. As noted by Marr (2011), who terms the approach Active Risk Management ™,
this is principally an extension of Peck's concept of the observational method (Peck 1969)
by using modern tools, of which risk modeling and decision science are central components.
Real-time monitoring is used during construction and operation to identify emerging risks
as soon as possible so that steps can be taken to reduce the probability of the risk event,
reduce the consequence, or do both.
12.10 ten unreSolVeD queStIonS
While a great deal of progress has been made, certain questions still remain that deserve
continued thought. The authors recently suggested a short list of 10 unresolved questions in
geotechnical risk and reliability analysis and that list may bear repeating for those without
easy access to ASCE conference proceedings (Christian and Baecher 2011). The 10 issues
range from technical mechanics problems to matters of communication. The list is certainly
not exhaustive, but we propose it to challenge the reliability community. It will be interest-
ing to return to these issues 10 years hence to see what progress we have made.
1. Why are failures less frequent than predicted ? Typical coefficients of variation for soil-
engineering properties are reported to be in the order of 20-30%. Presuming a mean
factor of safety of 1.5, corresponding reliability indices (β) are about 1.67, implying
probabilities of failure of about 0.05. These are an order of magnitude larger than the
observed frequency of adverse performance and two orders of magnitude larger than
the frequency of all-modes failures of earth dams. Why the difference?
2. What is the actual variability of soil and rock properties ? Several studies have been
published on the variability of soil properties, but more work needs to be done, espe-
cially on how to improve estimates at a particular site. Variations in soil-engineering
data involve at least two things: (1) actual variability from one point to another, and
(2) noise. Some of that noise arises because the measurements are index properties and
only loosely correlated to engineering properties (e.g., the field vane with its Bjeruum
correction factor) introduced by methods of measurement. In addition, however, there
are at least two bias errors that creep into assessments: (3) statistical error due to lim-
ited numbers of observations, and (4) model error due to the approximate nature of
our mathematical descriptions of soil behavior.
3. What are the effects of spatial correlation ? Geological materials arrive at their pres-
ent configurations by a geologic process that follows physical principles. Therefore,
their physical properties exhibit spatial correlation. While there have been successes
in describing spatial correlation statistically and in modeling spatially correlated
variables, the techniques for dealing with spatial correlation are difficult to imple-
ment, and they are poorly understood in practice; thus, their consequences are often
ignored.
4. How should we account for scale effects ? Much of geotechnical engineering involves
scaling properties from laboratory or field tests on limited volumes. We know that