Several diaries have mentioned the 2001 study (here)commissioned by MinnDOT on the I-35W bridge that collapsed yesterday. This study concluded that
Mn/DOT does not need to prematurely replace this bridge because of fatigue cracking, avoiding the high costs associated with such a large project.
This study was seriously flawed. Detailed analysis below
The study combined field tests and finite element analysis to reach its conclusion. There are three main parts, 1) controlled load test; 2) open traffic test; 3) finite element modeling. All three parts contain serious flaws that could have been discovered by a rigorous peer review process. Unfortunately, such research reports are usually only reviewed "internally," by friendly reviewers. These critical flaws were not discovered. Let me give a critique of each part of the research.
- Controlled load test. Curiously, the asymmetric tests were all conducted over the west truss. This was very significant because we will see below that abnormal behaviors were observed in open traffic tests on the east truss.
- Open traffic test. During the open traffic test, the effective stress range on the east truss was consistently higher than the west truss. Especially on the lower chord, the stress range on the east truss was nearly 60% higher than the west truss. To a naiive person like myself, such an uneven load would seem to be a yellow flag that at least should be understood. But this was never investigated. Even more disconcerting, is the following note on the top of page 53:
The gages in the east truss displayed excessive noise during triggered data collection and therefore are not included in the following discussion.
No explanations were given on what the source of this "excessive noise" might be. Compounding this problem, was the lack of data on the east truss during the controlled load test. This means that the entire test essentially skipped the east side of the bridge.
- Finite element modeling. Here was sloppy science at its worst. During the modeling, they found out that adding a concrete deck to the model would significantly reduce the predicted stress level (thus predicting a longer life for the bridge). However, after adding a concrete deck, they found that the simulated results did not agree with the test measurements. This in my opinion should have been enough for them to abandon the idea of modeling a concrete deck. But no, they didn't give up. Instead, they averaged the results with and without a concrete deck, and used that average to predict the stress! Imagine that, a bridge in a quantum mixed state, half way between having and not having a concrete deck! Using this prediction produced by mixing two calculations, they proclaimed that (page 84)
the remaining life ... is considered to be infinite.
The failure of this bridge is the result of failure on many fronts. The sloppy science might have contributed to it. This was an example of using science to serve the political need to cut costs. At the end, it ended up costing us much much more. Science should not be done this way.