Can structural health of bridges be determined solely by acceleration measurement? (no strain gages)

[Rob Stanovski]

Well, this article claims that the structural health of a bridge can be evaluated based only on acceleration (vibration) measurement.

This would definitely solve a lot of cost issues about monitoring a large number of existing ageing bridges. But the key question here is if it is actually possible to determine the health based solely on acceleration measurement with a low-noise triaxial accelerometer. Sure, operational modal analysis can be done. Modal shapes obtained, natural frequency determined. Velocity and displacement of oscillation can be integrated and observes. But to get stiffness, you would have to know what is the actual load that gives you certain displacement and the load (excitation) is difficult to measure (varying traffic, wind etc.).

I surely think it is good to gather the data, but can engineers with more experience in the field elaborate on the feasibility of the idea? If we could get rid of strain gauges, fbgs etc, it would be a dream come true for structural monitoring engineers.

What do you think?

 

[BridgeSmith]

It may be possible to detect major steel section loss in the way they describe, but to me, that would seem to be the extent of the usefulness of the method. It's doubtful that many of the the sudden, catastrophic bridge failures we've seen would have been detected by this method. The I-35W bridge in Minneapolis is a perfect example of this. The stiffness of the bridge, which is what the proposed method would measure, would not have changed until seconds before collapse. Brittle failures are typically not preceded by any changes in stiffness, and ductile failures typically present visual cues of imminent collapse. Major section loss of structural steel members would also typically be visible to inspectors.

Monitoring a bridge in this manner seems dangerous to me, as it would likely lead to a decrease in visual inspection.

 

[3DDave]

I believe the failure of the Italian bridge was essentially predicted using this method (a look at the article shows that's the lead example.) If I recall right an Italian university did a vibration survey indicating a large shift in the natural frequency responses that lead to looking into contracting repairs. However, no one saw it as sufficiently urgent as to close the bridge before those repairs were made. Vibration analysis will detect cases where there is a progressive failure of structural materials, such as loss of section by corrosion or the breaking of individual reinforcing elements. It might also be helpful if there is a prediction made of natural frequency for given loading conditions to confirm the calculations.

However, if there's no base to compare with, doing a frequency analysis probably won't predict much. So - wire up a bridge to begin with with strain gages and accelerometers and then keep the accelerometers which are enviromentally tougher and need no localized calibration and cheaper to operate to monitor for hidden damage.

The only other bridge that might have benefited was the Hyatt skywalk in Kansas city. By changing the structure from the original design it would have been obvious the stiffness was significantly altered and might have flagged a second look; though they didn't seem to have done a first look so it's less likely that the collapse would have been avoided.

 

[canwesteng]

I think the Hyatt is an example of where this doesn't work, since the dseign was change to basically double the load on a connection. Because connecting elements like bolts don't really affect the stiffness of the structure, you'll have no warning until failure basically.

 

[3DDave]

That would be incorrect. The builders of the Hyatt had made using the bridge off-limits because it bounced too much when moving construction materials. The section of tube was much lower stiffness than the solid rod that was supposed to connect the lower bridge to the roof. As I said, if they had done a frequency response analysis, the radical change was detectable.

 

[JStephen]

I was thinking on the Hyatt bridge, they changed a rod that passed through the bridge to one that was interrupted at the bridge, which doubled the connection loads at that point but didn't change the load on the rod.

 

[3DDave]

It changed from a solid rod to including a hollow-tube spring in the middle. The tube was far more elastic than the rod.

 

[BridgeSmith]

In the case of broken prestressing or post-tensioning strands, the stiffness would change. However, the cases where a significant stiffness change doesn't lead to immediate collapse or noticeable movement, displacement, or deformation of the structure, would be a small percentage.

The Ponte Morandi was detailed poorly, with numerous maintenance headaches and corrosion-prone connections that were nearly impossible to inspect, and more difficult to repair, which is why, despite knowing that the bridge was in danger of failure, it had not been repaired.

As far as the Hyatt Regency skywalk, I don't see how there would have been a detectable stiffness change before collapse. The failure occurred at a detail that did not change in any significant way until it failed. When it did, catastrophic collapse immediately followed.

As I said, it may be possible in some limited circumstances for this method to identify potential failure, but it would likely to do more harm than good by leading to reduced visual inspections, which have a far better chance of identifying issues in the majority of bridges.

 

[Ron]

I have done quite a bit of work with accelerometers and strain gages. For bridge ratings or evaluations, I have used strain gages and direct measurements. For more significantly dynamic structures I have used accelerometers and strain gages.

For bridge evaluations, I would not use accelerometers alone. Supplemental direct strain information is needed.

 

[3DDave]

The Hyatt stiffness didn't change from as-built. It was just not what an analysis of the original design should have predicted.

If I calculate the lowest natural frequency should be 30Hz and the as-built measure is 10Hz, it doesn't matter that it stays 10Hz until it fails. It matters that it is 10Hz at all.

The point of the article is not for acceptance, but long term monitoring; which was used by an Italian university to predict the failure of the bridge in Genoa.

 

[BridgeSmith]

At the Hyatt Regency, the only thing that changed from the design was the support point of the vertical threaded rods that supported the lower walkway. The configuration (and therefore the stiffness) didn't change significantly from the design, certainly not enough to affect the natural frequency of the system.

3DDave, you seem to misunderstand what actually failed in Hyatt skywalk. It had nothing to do with a tube connecting the skywalk to the roof. What changed from the design wasn't anything that would have changed the natural frequency of the structure (assuming it could be calculated with any accuracy in the first place). You can read up on the failure Here.

 

[271828]

Having measured vibrations on many structures, I am very skeptical of this.

 

[bridgebuster]

I don't have any significant experience with accelerometers and strain gages but I agree with HotRod10; it will lead to a false sense of security. I've inspected 100's of bridges and have seen things don't fail when you think they should and things that fail when you think they shouldn't.

 

[Rob Stanovski]

I don't think the purpose of the proposed system in the article is to get rid of the visual inspection. IMO the writer is corect when he claims there is not enough long-term data acquired with the present maintenance system. The system presented in the article is much cheaper to deploy on a wide scale than strain gauges, fbgs etc., because it simply requires bolting the devices on the bridge and connecting them with a network cable.

Today structural engineers definitely struggle in pinpointing which bridges should be repaired. There si this huge number of deficient bridges, but which of those are really the most critical? Ponte Morandi is a good case: everyone knew it is deficient, but the priority of maintenance went to the other bridges and then it was too late. More data should definitely be collected in a big, nation-wide data acquisition system.

Apparently the accelerometers presented in the article can also measure static angular deflection, basically act as inclinometers. Therefore the system could also indicate the absolute displacement of the parts of the structure. Do you think that information would be meaningful next to the vibration data?

 

[BridgeSmith]

"I don't think the purpose of the proposed system in the article is to get rid of the visual inspection."

Don't be naive. That may not be the original purpose, but that's exactly what the end result would be.

I still don't buy into the idea that this technology will be able to reliably predict imminent failure in the majority of common bridge types. However, it will tempt oversight agencies to use technology as a cheap substitute for thorough visual inspections at the necessary intervals on those bridges, anyway.

 

[charliealphabravo]

The success of some of these applications might depend on how early in the life of the structure the monitoring system was installed and how the data was analyzed and calibration against design loading thresholds.

For example on the I-35W, an acceleration-based system could have detected the change in response due to the over-loading of the spans. Even if a brittle connection failure was not detectable wouldn't there be some value in a system that kept track of and could flag load cycles, load duration, and mobilized percent of design load capacity?

 

[BridgeSmith]

The I-35W bridge was thought to be within its design capacity. It wasn't until after the buckling/shear failure of the gusset plates that the design error was discovered. It actually took several years for the real cause of the collapse to be discovered - the thickness of the gusset plates (and all the steel components) were reduced in size by half when the steel strength was doubled. However, while the original 1" thick gusset plates were not subject to buckling, the 1/2" gusset plates in the revised design were, so the 1/2", 100ksi gusset plates did not have the same capacity as the 1", 50ksi gussets. A visual inspection identified partial buckling of one of the undercapacity gusset plates, but it was assumed that it wasn't a problem.

 

[JoshPlumSE]

When I was in college one of my professors was doing a lot of research on this subject. Using dynamic characteristics (and accelerometers) to detect damage in a truss structure. I was one of his research aides and spent many an hour working on the tests. I read a few papers or drafts of papers on the subject as well.

What we did was pretty successful in detecting changes in stiffness or mass (anything that would change the mode shapes or frequencies). So, this could be very useful in detecting certain types of damage. If certain members are rusting significantly, this could probably show it over time. Not when you compare from Jan to Feb, but more likely when you compare data from 2000 to 2019. What's nice is that it can also point you in the direction of where the problems are.

That being said, as some others have suggested, I don't know that it could predict an imminent connection failure like we saw in that bride in Minnesota. Maybe it would have shown some loss of stiffness due to corrosion. But, it probably wouldn't have been significant enough to shut down until it had really started to rupture. It definitely couldn't have shown anything with the Hyatt Regency.

 

[charliealphabravo]

the thickness of the gusset plates (and all the steel components) were reduced in size by half when the steel strength was doubled. However, while the original 1" thick gusset plates were not subject to buckling, the 1/2" gusset plates in the revised design were, so the 1/2", 100ksi gusset plates did not have the same capacity as the 1", 50ksi gussets.

I'm just reading through the I-35W NTSB report and I don't think this is true. If I understand the report correctly, the failed gusset plates at nodes U10 & L11 were not made of the high strength T1 steel and the design was unchanged from the original preliminary designs which showed 1/2-inch A441 steel. It looks like the preliminary calcs considered the pass through forces but not the forces from the diagonals and ended up being severely under-designed against buckling and side shifting instability of the diagonal.

The Safety Board found no evidence that the gusset plates
at the U10(′) and L11(′) nodes had ever been intended to be fabricated from T-1
steel or that their specifications (0.5-inch-thick A441 steel) had changed from the
earliest design documents through fabrication and installation.

As best I can tell, most of the T1 steel proposed early in the design was swapped out with A441. Many gusset plates were consequently increased in thickness to 1-inch but not the gussets at U10 & L11.

Final Design, March 1964–March 1965
Because of the decision regarding T-1 steel and the lower allowable stress
levels for the required A242/A441 steels (27,000 psi for A242/A441 steels versus
45,000 psi for T-1), Sverdrup & Parcel had to redesign all the truss members and
gusset plates for which T-1 steel had initially been proposed. For example, the
thickness of all tension members would have to be increased by about 67 percent
to maintain the same tensile stress capability. These thicker members increased
the dead load on the bridge, which required additional design computations.

 

[BridgeSmith]

CAB, well after that report was written, they finally located the original design. Those gusset plates were originally designed and detailed as 1" A242 plates. The design was revised to use 1/2" T-1 plates, apparently without any buckling stability checks, which were not required for the 1" plates in the original design.