U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information

Publication Number: FHWA-RD-03-089
Date: September 2005

Dynamic Bridge Substructure Evaluation and Monitoring

Previous | Table of Contents | Next

Chapter 9. Conclusions and Recommendations

Field Modal Vibration Dynamic Testing for Damage

The results of flexibility TF measurements of Trinity Relief Bridge bent 12 (piles only) and bent 2 (piles plus footing pile cap) are discussed in chapter 5 for intact, excavated (simulated flood scour damage), and broken (simulated earthquake damage) pile cases. The TF results showed distinct and clear increases in flexibility and lower natural first vertical mode fundamental resonant frequencies with increasing damage. Because flexibility is the inverse of stiffness, increases in flexibility of more than 50 percent correspond to greater than 50 percent reductions in stiffness.

The flexibility results show promise for helping directly identify substructure damage if a baseline of the sound response of a bridge has been determined above all critical members. The damage will be indicated by a comparatively small decrease in natural frequency and a larger increase in flexibility (corresponding decrease in stiffness) using traditional modal testing and analysis techniques. This result is not surprising—it is theoretically predicted, and in a sense a flexibility modal TF test is a dynamic load test. It is expected that more complicated substructures where loading cannot be close to the damaged member would not show as much of an effect in the flexibility results. The drawbacks of this flexibility TF approach are the need to put sensors on critical members and test before damage to obtain a basis of comparison by which to judge and identify the extent of damage in future tests. This method does seem to show real promise for determination of damage if before-and-after modal vibration measurements can be performed.

Field Modal Vibration Dynamic Testing for Shallow Versus Deep Foundation Determination

As discussed in chapter 5, dynamic tests were conducted on the Woodville Road Bridge (shallow footing foundations) and Old Reliance Road Bridge (footing pile cap on steel BP piles) to investigate whether modal vibration testing could differentiate between shallow and deep foundation systems for otherwise almost identical bridges. Although the two bridges had similar modal response signatures, their modal resonant frequencies were substantially different; however, the first vertical mode resonant peak for Old Reliance Bridge with the steel piles was not at a significantly higher frequency (9 Hz) than the first vertical mode resonant peak for Woodville Road Bridge (10 Hz). This result was contrary to the expected results—that the Old Reliance Road Bridge with the steel piles plus pile cap would show greater stiffness, and therefore a higher frequency for the peak of the first vertical mode.

It is interesting that the flexibilities, and inversely, the stiffness magnitudes, were similar for the two bridges. The first three resonant peaks for Old Reliance Road Bridge were all at lower frequencies than the corresponding resonant peaks for the Woodville Road Bridge. Considering these initial results, it appears doubtful that modal testing could uniquely and clearly determine shallow versus deep foundations without having otherwise identical bridge structures situated in similar soils. It further appears that at least one of the foundations tested should have a known shallow or deep foundation system to serve as the basis against which unknown foundation systems could be judged. It is extremely unlikely that modal testing could determine unknown depths of foundations for scour safety evaluation.

Structural Modeling and Parameter Estimation

As discussed in chapter 6, considerable effort was devoted to determining changes in foundation stiffness through structural modeling and parameter estimation. Unfortunately, the structural parameter estimation approach did not find a consistent, correct solution to the intact versus excavated versus broken pile cases. The modal vibration data quality was as high as could reasonably be expected, but the parameter estimation ultimately was unable to resolve the basic case of the first vertical mode of vibration. Considering the effort required and complexity of this approach, it does not seem suitable for bridge substructure damage detection at this time. The fundamental problem may be that the effects of the foundation damage are too small for parameter estimation, and they are buried in the overall modal response of the structure.

Hilbert-Huang Transform for Evaluation of Bridge Substructure Damage

The completion of this research was delayed in part to evaluate the potential of the HHT to indicate the presence of damaged substructure foundations. Specifically, HHT analysis was used to identify, by decreases in frequencies, nonlinear responses that indicate damage. HHT analysis shows promise in being able to identify damage comparatively in conjunction with modal vibration testing without a structure having been tested before the damage occurs. The HHT results and comparisons with the traditional modal and wavelet analysis method suggest the following findings:

The following novel signature recognition technique based on modal vibration testing with HHT analysis is proposed for structural health monitoring and damage detection:

  1. Two or more similar structural members such as two of four bridge piles with the same size, cross-section, and materials are selected for a nondestructive vibration test. A sensor (an accelerometer) is mounted on each pile. The piles are subjected to a dynamic excitation acting at a location close to, but not on, the piles. A computer data acquisition system digitizes and stores the sensors’ response to the excitation. This results in two sets of data and either modal or spectral data from which to determine resonant frequencies with traditional methods to aid in interpreting the HHT results.
  2. For the HHT analysis, the two data sets are decomposed in the EMD process into a number of IMF components, which are used to compute the Hilbert spectrum. The driving frequency, natural frequency, and other frequency contents can be identified partially, if not completely, by analyzing the Hilbert spectra and the IMF components.
  3. If the natural frequencies of the two members are different, the member with the lower frequency has damage while the other is comparatively sound, or the former has more severe damage than the latter. On the other hand, if the frequencies are the same, both members are either undamaged or damaged to the same degree. For the result of identical frequencies, the testing will be repeated with a third member involved. It is always better to have a priori data from testing the same structure before a damaging event (such as flooding or an earthquake) to use as a baseline.
  4. After identifying the damaged member and if access permits, the excitation is exerted at different locations on the damaged member and a series of vibration data sets are recorded at one sensor location.
  5. The data from step 4 are analyzed as in step 2.
  6. If one of the data sets shows the lowest natural frequency of the member, the location of the pertinent excitation is at or close to the damaged location.

The first three steps were explained in chapter 7 in the analysis of the in situ data from the bents of the Trinity River Relief Bridge. The reasons for the last three steps are given in this paragraph. When the excitation is acting directly on the damaged location, the effect on the structure at that location will be the strongest. Even if the load is small and well within the design range, the damaged member could have nonlinear responses, that is, responses characterized by lower natural frequencies resulting from local damage. On the other hand, if the excitation acts neither directly on nor close to the damaged spot, the vibration responses of the structure could remain elastic and linear. Because HHT analysis can identify the instantaneous frequency of data, the damaged location can be found by identifying the lowest natural frequency of vibration from the series of data sets.

Implementing Dynamic Testing Results in Bridge Management Systems

As discussed in chapter 8, dynamic bridge substructure evaluation and monitoring provides important opportunities for improving a BMS and enhancing the current state of the art in bridge evaluation and monitoring. Specifically, measures of dynamic bridge foundation vertical stiffness or HHT results, or both, that identify downward frequency shifts indicating damage show considerable promise for the following uses:

In conclusion, recent research on the role of NDE in BMSs suggests the desirability of integrating dynamic testing results, including HHT results, with visual ratings data.

Recommendations for Research

In the research described in this paper, modal vibration testing and the HHT approach appear to have successfully identified the effects of damage to concrete pile and concrete pile plus pile cap deep foundations; however, particularly for the HHT method, it must be emphasized that only limited studies have been performed to date and further research is necessary to understand, validate, and apply the combined approach of modal or spectral vibration testing (or both) and the HHT method. Because it appears capable of identifying local, short-duration, lower frequency responses indicative of nonlinear and nonstationary damage to local structural members, the HHT method has real potential for detecting damage to bridge structures and substructures after catastrophic events such as floods, earthquakes, and collisions. Future research also should address the issue of how the vibration and HHT approach could be implemented in a BMS.


Previous | Table of Contents | Next