Embedded Conservatism

Posted Fri May 13, 2005 in

After an intense meeting and a short night, I am ambulatory this morning. The coffee is ready and I think I am a quart low. I am preparing myself for another intense day. We have a lot of work to get done.

Our work on Texas hydrology is coming together. Yesterday, we were looking at the time to peak of the unitgraph we have developed from Texas data and comparing it with the “standard” approach to estimating the time to peak. We see that our estimates are significantly longer than those from the standard method. This means there is conservatism built into the standard method, at least so much as our dataset is concerned. Therefore, analysts using the standard method generate higher design discharges for drainage design than the data we have would indicate.

We have seen layer upon layer of hidden conservatism built into the standard methods. While I can understand the need for conservatism (also known as a safety factor), my professional opinion is that an analyst should design for a specific load and then apply a safety factor based on his or her judgement depending on the consequences of failure. This is what structural and geotechnical engineers do, they compute the design load and then apply a safety factor, usually a multiplier in the 1.5 to 2.0 range. This is sufficient to ensure that the structure will function as designed and covers uncertainties in material properties and load estimates.

While we have not quantified the degree of conservatism present in standard methods (yet), my suspicion is that is it significantly greater than the standard 1.5 to 2.0 range. Maybe this is appropriate for some hydraulic designs. But I suspect that it is unnecessary for many hydraulic structures and that we (taxpayers) are paying for unnecessary risk of failure reduction when the consequences of failure do not justify the embedded conservatism.

When we finish our research, we will have produced a set of methods that provide the hydraulic engineer with our best estimate of the design discharge without hidden conservatism. That will put the choice of safety factor squarely in the lap of the designer, where it belongs. It will mean some retraining and I bet that there will be some resistance to the changes we are suggesting. It will be interesting to see what happens.