Risk Assessment in Geotechnical Engineering
Soils and rocks in their natural state are among the most variable of all engineering materials, and geotechnical engineers must often “make do” with materials that present themselves at a particular site. In a perfect world with no economic constraints, we would drill numerous boreholes and take multiple samples back to the laboratory for measurement of standard soil properties such as permeability, compressibility, and shear strength. Armed with all this information, we could then perform our design of a seepage problem, foundation, or slope and be very confident of our predictions. In reality we must usually deal with very limited site investigation data, and the traditional approach for dealing with this uncertainty in geotechnical design has been through the use of characteristic values of the soil properties coupled with a generous factor of safety.
If we were to plot the multitude of data from the hypothetical site investigation as a histogram for one of the properties, we would likely see a broad range of values in the form of a bell-shaped curve. The most likely values of the property would be somewhere in the middle, but a significant number of samples would display higher and lower values too. This variability inherent in soils and rocks suggests that geotechnical systems are highly amenable to a statistical interpretation. This is quite a different philosophy from the traditional approach mentioned above. In the probabilistic approach, we input soil properties characterized in terms of their means and variances (first and second moments) leading to estimates of the probability of failure or reliability of a design. Specific examples might involve estimation of the reliability of a slope design, the probability of excessive foundation settlement, or the probability of excessive leakage from a reservoir. When probabilities are
coupled with consequences of design failure, we can then
assess the risk associated with the design.