A look at the article: Dávalos, H., & Miranda, E. (2019). Evaluation of bias on the probability of collapse from amplitude scaling using spectral‐shape‐matched records. Earthquake Engineering & Structural Dynamics, 48(8), 970-986.
Kind of an intro.
Anyone doing a time history assessment deals with ground motion records. To many SE people, these records are nothing but a mere loading protocol, with a pinch of trust in people preparing them. However, (hopefully, if we are well-informed) we cannot just go and take an earthquake record from somewhere in California and apply it to Peru. We need to “select” appropriate records (from sites with similar seismicity) and then “modify” them to be very similar to what we expect for a possible earthquake at the site.
While different modification approaches exist, as the unwritten rule of life reminds us, the easiest way is also the most popular one. In this case, the easiest method is called “amplitude scaling,” which is a fancy way to say that the acceleration record is multiplied by a single number (AKA scale factor).
The choice of scale factors has been a long-standing debate (how to choose? should we limit it?). Still, until recently, the scaling was considered to lead to unbiased structural estimation (meaning that the results would be the same from scaled and unscaled records). Some folks raised the issue that the scaled and unscaled records would have different frequency content, which means the structural analysis results will be biased. Very recently, some authors argue that the bias is due to the matching spectral shape of the record ( spectral shape of a record shows spectral acceleration at different periods and is considered a surrogate for frequency content of the record)
What is this paper addressing?
What happens to our estimation of structural collapse if we use scaled records that have accurate spectral shapes?
Quickly, wrap up the results. I have 2 minutes tops
The authors showed that even with spectral shape matched records, the scaling introduces bias in collapse estimation! The bias increases by the scale factor and leads to collapse probability overestimation. More interestingly, the authors argue that this is because of the difference in input energy and distribution of energy between record pulses.
What I like about the results
A lot. For a long time, matching to spectral shape was deemed the most state-of-art way to avoid bias. The practice of doing spectral matching AND limiting scale factor still might provide some safeguards, but apparently it does not solve the issue. Shifting the seismic thinking into an energy-based framework.
I’m hooked. Give me some more related papers
Luco, N., & Bazzurro, P. (2007). Does amplitude scaling of ground motion records result in biased nonlinear structural drift responses?. Earthquake Engineering & Structural Dynamics, 36(13), 1813-1835.
Huang, Y. N., Whittaker, A. S., Luco, N., & Hamburger, R. O. (2011). Scaling earthquake ground motions for performance-based assessment of buildings. Journal of Structural Engineering, 137(3), 311-321.
Zacharenaki, A., Fragiadakis, M., Assimaki, D., & Papadrakakis, M. (2014). Bias assessment in incremental dynamic analysis due to record scaling. Soil Dynamics and Earthquake Engineering, 67, 158-168.
2 thoughts on “SE Literature Café: Wait, I cannot just multiply my earthquake records by 10?”
Hi, really nice initiative! Thank you!
Just adding one more reference that I read recently and tries ti tackle this problem (by reducing the ammount of scaling) is Cloud to IDA paper of Miano et al (https://doi.org/10.1002/eqe.3009 ).
LikeLiked by 1 person
Thank you Andreia! Glad you liked the initiative and awesome suggestion. Dr. Miano had done some interesting stuff on bridging cloud and IDA analysis and the one you mentioned is quite relevant to this topic. Personally, I’m more of a cloud guy than IDA because of limited scaling, if any.