Pages

Monday, April 07, 2008

A retraction. A MnDOT sponsored 2001 bridge study on the I35W bridge apparently did not consider approach spans.

Earlier today, here, I published about MnDOT studies available online, and quoted the abstracts of three items started and finished during the time Elwyn Tinklenberg headed MnDOT.

From the abstracts, and because I lack an engineering background as I noted, I believe I reached an improper conclusion.

I want to qualify that nothing I wrote was factually incorrect, as to what was said online. My opinion was stated and argued. Upon learning more, I have altered that opinion. Fairness requires that I explain why I now am not willing to cast any blame upon Elwyn Tinklenberg based on how I now understand the facts.

I mentioned three studies. One was MnDOT Final Report 2000-16, 296 pages in length, which in early pages indicated it dealt with a totally different kind of bridge. This possibility I had mentioned, and as best as I understand things now, this report is not relevant to the I35W bridge, MnDOT bridge 9340.

A second report, MnDOT Final Report 2002-06, was a general study of instrumentation and monitoring possibilities, 111 pages long, and may be indirectly relevant. But I cannot see it leading to any negative conclusions.

The key item -- MnDOT Final Study 2001-10.

This item considered the actual bridge that fell. It noted that fatigue crack expectations based on design load presumptions and not on actual load conditions drew a worse picture than study measurements demonstrated.

Most importantly, the central span, alone, was studied and the approach spans dismissed in initial pages of the report. Upon reading that, I am posting this retraction.

The report is 89 pages long including introductory pages. The executive summary opens, at p. ix, stating:

Bridge 9340 is a deck truss with steel multi-girder approach spans built in 1967 across the Mississippi River just east of downtown Minneapolis. The approach spans have exhibited several fatigue problems; primarily due to unanticipated out-of-plane distortion of the girders. Although fatigue cracking has not occurred in the deck truss, it has many poor fatigue details on the main truss and floor truss systems. Concern about fatigue cracking in the deck truss is heightened by a lack of redundancy in the main truss system. The detailed fatigue assessment in this report shows that fatigue cracking of the deck truss is not likely. Therefore, replacement of this bridge, and the associated very high cost, may be deferred.

Strain gages were installed on both the main trusses and the floor truss to measure the live-load stress ranges The strain gages were monitored while trucks with known axle weights crossed the bridge and under normal traffic. Two- and three-dimensional finite-element models of the bridge were developed and calibrated based on the measured stress ranges. These finite element models were used to calculate the stress ranges throughout the deck truss.

The peak stress ranges are less than the fatigue thresholds at all details. Therefore, fatigue cracking is not expected during the remaining useful life of the bridge. The most critical details, i.e. the details with the greatest ratios of peak stress range to the fatigue threshold, were in the floor trusses. Therefore, if fatigue problems were to develop due to a future increase in loading, the cracking would manifest in a floor truss first. Cracks in the floor trusses should be readily detectable since the floor trusses are easy to inspect from the catwalk. In the event that the cracks propagate undetected, the bridge could most likely tolerate the loss of a floor truss without collapse, whereas the failure of one of the two main trusses would be more critical.


Then, the report again at its pages 1-3, indirectly mentions the approach spans and problematic concerns about them, but not in a way that raises a major red flag notice:

PROBLEM STATEMENT

Bridge 9340 supports four lanes in each direction (eight lanes total) of I-35W across the Mississippi River just east of downtown Minneapolis. The Average Daily Traffic (ADT) is given as 15,000 in each direction, with ten percent trucks. Bridge 9340 consists of a deck truss and steel multi-girder approach spans built in 1967. The deck tmss, shown in Figure 1, has a center span of 139 meters, north and south spans of 80.8 meters and cantilever spans of 11.6 and 10.9 meters. The bridge was designed using the 1961 American Association of State Highway Officials (AASHO) Standard Specifications [I]. At that time, unconservative fatigue design provisions were used. The American Association of State Highway and Transportation Officials, (AASHTO) fatigue design rules were substantially improved as a result of research at Lehigh University in the 1970's [2,3].

The approach spans have exhibited several fatigue problems; primarily due to unanticipated out-of-plane distortion of the girders. Although fatigue cracking has not occurred in the deck truss, it has many poor fatigue details on the main truss and floor truss systems.

Stress ranges calculated using the lane load as live load are greater than fatigue thresholds for many of the details. The poor fatigue details in the deck truss include intermittent fillet welds, welded longitudinal stiffeners and welded attachments at diaphragms inside tension members. These details are classified as Category D and E with threshold stress ranges 48 and 31 MPa, respectively.

The design analysis, using the AASHTO lane load in all lanes, shows design-live-load stress ranges in the truss members much higher than these thresholds. Design-live-load stress ranges were greatest, up to 138 MPa, in members that experience load reversal as trucks pass from the outside spans onto the center span. The predicted average life at that stress range is between 20,000 and 40,000 cycles. With 15,000 trucks per day crossing the bridge in each direction, these details should have cracked soon after opening if the stress ranges were really this high.

The actual stress ranges can be determined by instrumenting the bridge with strain gages and monitoring strains under both a known load and open traffic. Fortunately, the actual stress ranges are much lower than these design live-load stress ranges. Consequently, the fatigue life is far longer than would be predicted based on the design-live-load stress ranges. The difference between actual and predicted stress ranges is the result of conservative assumptions made in the design process. The primary reason is that the traffic on the bridge is 90 percent cars and weighs a lot less than the lane loading, (9.34 kN/m). The lane loading is approximately equivalent to maximum legal 356 kN trucks spaced at about 38 meters apart.

The lane load may be appropriate for a few occurrences during the life when there are bumper-to-bumper trucks in all lanes, and the bridge should be designed to have sufficient strength to withstand this load. However, a few occurrences of loading of this magnitude would not have a significant effect on fatigue cracking. In fact, it has been shown that essentially infinite fatigue life is achieved in tests when fewer than 0.01 percent of stress ranges exceed the fatigue threshold [4]. Therefore, only loads that occur more frequently than 3.01 percent of the time have an effect on fatigue. If there are 15,000 significant load cycles (trucks) pcr day, any load that happens less frequently than daily is irrelevant as far as fatigue is concerned. In observing this bridge closely over the period of more than a year, the authors have never seen a condition where there were closely spaced trucks in each lane.


That is vague and inconclusive language to me. The remainder of the study appears more dismissive of worry than cautionary. The ultimate conclusions at the end of the study are not extreme, but precautionary. Much of the report explains that some conservative design presumptions were not bourne out in actual loading experience during the study.

In summary, the most recent forensic work indicated it was a failure in one of the approach spans that ultimately caused a total failure and collapse. That consideration appears to have been entirely outside of the scope of this study, which is focused on the main span.

From the report, as best as I can read and understand it, Elwyn Tinklenberg would have had no direct notice of cause to worry about the bridge possibly failing in the way it appears, so far, to have failed.

Again I disclaim engineering expertise. However, as best as I can read the report it did not give notice of cause to worry as much as cause to monitor and inspect carefully.

The issue of insufficiently thick gussets, noted after the collapse, was there to be seen by these professionals doing the study - and their not noting it is an indication that until there was the failure and the forensic attention was heightened, there was not notice of the problem at MnDOT management levels unless and until someone in the field would have noticed the error - something that went unnoticed and undetected during building and since the span was built in 1967.

The study begs a host of questions, mainly, why were the two outer appraoch spans not part of the study. Also, if strain guages were placed on the structure, what would they be measuring except differences between the bridge in an unloaded static state and under usage loading - so that if the structure were inadequate [gusset thickness considerations], or overloaded in a static, at rest state [repaving as an increase in static loading], would those things have shown up in the strain guage measurements of differences between static and dynamic loadings? Those are my immediate questions of the entire study. However, I see no critical error nor major cause to fault Elwyn Tinklenberg arising from the text of this particular I35W span study at this point. It was done as it was, and focused as it was for reasons I do not know and cannot find fault with unless I knew other things which might contradict the impression that the report gave no red-flag notice.

Coming into the job years after the triple-span was built, and knowing what this report said, any reasonable person in Elwyn Tinklenberg's position heading MnDOT and being as he was without an engineering background (hence, having to rely on staff expertise), would likely have found this report more reassuring that the situation was under control than suggestive of a threat that the bridge was in imminent danger.

That is my sincere belief now, having read much of the study and understanding it as best as I do, and not being an engineer either.

Others more skilled than I am might read the report differently, or have a wider factual background for decision, and might see the report as more suggestive of a need for caution. That is why I believe it should be fully studied and questioned.

However, as best as I can read things, I at present stand by what this post says - I see no notice given Elwyn Tinklenberg at the time he headed MnDOT, based on what this study says, that he faced any imminent threat of that bridge falling during his watch, or within five years of when he left MnDOT.

That is all I can say, for now.