Making the most of credible evidence10 November 1998
Professor Duron calls for the setting up of a national database for the seismic and stability analysis of dams, and argues that such a database will enable the industry to make informed decisions,
Members of the dam community will no doubt play significant roles in ensuring dam safety in the next decade. With numerous US dams exceeding their original design lives it is a sobering fact that many which were built in the early 1900s were not designed for seismic events, yet they are routinely analysed using predicted maximum credible earthquakes.
As we continue to ratchet up our safety requirements, the potential for conflicting results between predicted analytical behaviour and the known operating history of dams (especially the older ones) will only increase. Many of us who have performed seismic and stability analyses have had to address comments and suggestions from reviewers which have, at times, not coincided with our intuitive ‘feel’ for the dam’s structure or for its surrounding region.
Reconciling these various opinions can lead to some lively discussions. What is required is some baseline knowledge which has credible evidence relevant to the manner in which we currently evaluate the seismic and stability behaviour of dams.
This baseline knowledge should take the form of a database from which response characteristics for a dam site can be extracted as required.
There have traditionally been two main obstacles to the idea of creating a database — the cost and the implementation. The perception in the dam community is that the gathering of data from dam sites is costly and time consuming. If the benefits of any proposed effort are not clearly identified, any expenditure will seem unjustified. The reality is that the typical engineering firm does not have the expertise to conduct field testing in a cost-effective manner, and high cost estimates are often the result. As a researcher, however, I have conducted 17 field tests at 10 dam sites since 1985 and have watched the price of conducting each test fall as advances in technology have allowed us to develop more efficient testing and analysis techniques. Cost, I argue, is really not the major obstacle it appears to be. Convincing the owner that the benefit warrants the expense, however, is.
A valuable database
What is the value of such a database? Ultimately, its value will be demonstrated by our ability as engineers to use the information for condition assessment. This includes using the data for numerical model verification and calibration, and for the development of new and improved field testing and analytical procedures.
Some examples of ways in which a database could be used are:
•The analysis of large concrete multiple arch dams. For such dams, particularly where there are large numerical models, the availability of dynamic response characteristics allows the engineer to explore more efficient modelling and analysis options. For instance, a truncated model of the dam can alleviate computer requirements leading to a more practical symmetric analysis approach. In such a case, validation of the model and an analysis approach can lead to enhanced confidence in the predicted stress results. Specifically, the model can be validated by matching response characteristics derived from the analysis (eg resonant frequencies and response shapes) with actual observed behaviour.
•Using the database can improve the researcher’s confidence in the selection of appropriate material properties for the dam concrete and foundation rock. Dynamic characteristics in the form of frequency response functions in the dam, foundation and in the reservoir provide a more detailed picture of the overall dam behaviour than one obtained from discrete core sample testing.
•The existence of a database of known response behaviour can also be of value in the development of accurate diagnostic field test procedures. At a particular dam site, the engineer can explore or refine test approaches which could lead to health monitoring on a routine basis.
The response signal shown (previous page) is offered as one example of what can be derived from diagnostic testing. The signal is the hydrodynamic pressure taken on the upstream face of a small concrete arch dam, subjected to a sudden burst of compressed air released 20ft away from the dam in the reservoir. The pressure is characterised as the pre-event ambient condition followed by the arrival of the pressure wave, decay and return to pre-event conditions.
From a diagnostic standpoint the question can be asked — does the response shown contain useful information about the dam's seismic characteristics? The answer lies in the subsequent analysis and interpretation. It turns out that the accuracy of the characteristics derived from the measured response actually depends on the analytical technique employed.
The figure on the previous page contains two power spectral densities based on fourier (dashed line) and entropy (solid line) estimation methods. The two estimates, while obtained from the same record, are clearly not in agreement over the frequency range shown. Assessing the applicability of the test approach and determining which analysis result is correct requires knowledge of the dam's response characteristics for comparison. The database for this dam contains measurements from a forced vibration test which reveal the entropy result to be in excellent agreement with the known 12Hz fundamental resonance and 8% critical damping estimate.
•Finally, the database can be used to guide efforts for improving current analysis procedures. Presently there is much concern surrounding the modelling of reservoir boundary absorption effects. Research efforts are planned in which both analytical and experimental approaches will be implemented to define the parameters required accurately to represent these effects. Results from these efforts will be compared with the ability of the analytical models to reproduce measured dam and reservoir characteristics. Researchers attending a recently convened workshop sponsored by the US Army Corps of Engineers have agreed that a database of measured response characteristics is vital to the evaluation of the proposed efforts.
Considering the contents
Although the specifics of what the national database should contain have not been addressed here, for the purposes of seismic and stability evaluations the type of data needed would probably include measurements of acceleration on the dam crest, adjacent foundation, across joints, and along cantilevers, as well as hydrodynamic pressures in the reservoir. Other data types could also be included (eg piezometers and seepage gauges).
The next steps
What needs to be done now is to encourage the industry to begin thinking about this issue on a national level. While there are groups currently working to develop databases in California, I would like to see a unified approach so that we can address issues of data quality, integrity, data management and use.
Those of us with substantial data and observational evidence of dam response behaviour should feel almost compelled to share this information with community. The long term health of our dams and the decision-making processes for continued operation and repairs will certainly benefit from the establishment and use of a national database.