Hazarding a guess

Engineering techniques can point to sources of uncertainty

What's the scientific likelihood of a steep slope crumbling onto the highway below? Encountering boulders when you're excavating a tunnel? Or having soils settle beneath your brand new skyscraper?

Uncertainties like these are now being addressed in a methodical manner with reliability-based engineering. The result is a clearer estimation of overall risk - and a better allocation of construction resou-rces.

Better than guestimates

Soil properties have traditionally been estimated using field and laboratory measurements, as well as the judgment of seasoned engineers. Variations in the measured values and applied loads, plus limitations of the analytical model, are then combined into a single "safety factor." Safety factors are evaluated to determine acceptability based on commonly accepted levels of safety and engineering practice.

The problem with this approach is that the natural variation of the soils' properties, which can be significant, are not considered. But when material property variations are taken into account, statistical methods can be used to render a quantitative evaluation of the risk of failure.

With the advent of computers on engineers' desks, risk estimation can now be easily quantified. A range of input values as well as distribution patterns can be plugged into analytical models. With reliability-based analysis, a safety factor can be calculated using most-likely values. In addition, the impact of variable conditions can be evaluated.

For instance, a report on slope stability might conclude that a particular safety factor is acceptable. But reliability-based analyses might indicate that there is a probability of slope failure. In other words, the slope had better be stabilized in a hurry.

Bottom-line benefits

Quantifying the unknown can save money in several ways. First, estimating the variation of the safety factor could reduce direct costs for owners, who can make more informed decisions in the allocation of construction dollars. Reliability-based analyses identify the major sources of uncertainty, allowing design budgets to focus on characterizing the nature of these impacts. Does it make more sense to study the compressibility of a soil layer so footings could be small and shallow? Or should the footings be constructed larger and deeper to minimize settlement? Since all risks can never be addressed, trade-offs must often be made. Reliability-based analysis provides the tools to make (and justify) logical decisions in the face of the unknowable.

Furthermore, reliability-based analyses can reduce costly change orders. When bidders must guess about uncertain soil or water conditions, they're often forced to protect themselves with conservative estimates. For instance, encountering boulders in a tunnel construction operation usually results in project delays and change orders. If the number of boulders that might be encountered can be estimated, the owner's risk can be reduced.

Boulders beneath Seattle

Shannon & Wilson, a Seattle-based geotechnical consultant, frequently works on tunneling projects. One of our most common challenges, therefore, is determining the amount of boulders that could potentially be encountered during the tunnel boring process.

At first glance, the situation seems like guessing the number of needles in a haystack by poking in a broomstick. Fortun-ately, the number of boulders that can be reasonably expected in a tunneling can be estimated by performing a statistical analysis of boulders encountered in geotechnical subsurface investigations, as well as from comparable jobs nearby.

Using this type of analysis for the downtown Seattle bus tunnel, Shannon & Wilson predicted a 50 percent chance of hitting between 12 and 31 boulders greater than two feet in diameter. During tunnel mining, 21 of these boulders were found - well within the estimated interval - and a gratifying demonstration of the technique's usefulness.

Managing Magnolia's mudslide

Restoring the mammoth landslide above the on-ramp to the Magnolia Bridge has also required reliability-based engineering. The city of Seattle needed to repair the slide area quickly - and safely - so the bridge ramp could be reopened for traffic. But first, our engineers needed to consider the variations in groundwater levels and engineering properties in designing the slope stabilization systems.

Shannon & Wilson's staff used reliability-based techniques on the project and found that the probability of slope failure was low and, therefore, not a significant worry. With this back-up data in hand, city engineers could go ahead with their original plans and shore up the slope with much greater confidence.

Forensic fodder

Another application of reliability-based engineering can be found in the courtroom. Who is at fault when the best-laid plans go awry and part of a project fails? Establishing responsibility (and financial liability) is often done with reliability-based engineering. Forensic hydrogeologists can use statistical modeling to describe the likelihood of a pipeline bursting, a construction site flooding or a hazardous discharge occurring.

Often it's just as important to determine how an incident most likely occurred, so that a responsible party can be named and coverage determined. A reliability-based estimate can also help evaluate liability and estimate appropriate damages.

Reliability-based analysis will become increasingly popular as engineers discover its usefulness as a tool to quantify likely risks. The goal of engineering, after all, is to ensure that structures will last. And until crystal ball technology evolves to let us forecast the future with real conviction, any tool that allows us to better quantify probable risk is of tremendous value.

---

djc home | top | special issues index