Application of an extended Kalman filter for obtaining low-strain DST shear damping ratio estimates

1. Home
2. Publications
3. Online Library
4. Application of an extended Kalman filter for obtaining low-strain DST shear damping ratio estimates

The so-called FMDSMAA algorithm was developed for processing Downhole Seismic Testing (DST) datasets so that low-strain shear damping ratio (_s) estimates could be obtained. The low-strain DST _s values serve as reference values for laboratory test such as the Resonant Column Test (RCT). RCT can suffer from various disadvantages such as sample disturbance and sample preparation effects which can shift the estimated _s values. The RCT results can be adjusted so that the low strain RCT _s estimates agree with the low strain in-situ DST _s estimates. Low strain _s estimates are also very important for predicting and assessing ground amplification during earthquakes. The FMDSMAA algorithm takes into account source waves true raypath, geometric spreading, apparent attenuation (due to mode conversion, reflection-refraction at an interface) and material losses (intrinsic attenuation or absorption).  The FMDSMAA algorithm also addresses limitations of the spectral ratio technique such as inaccurate raypath assumptions and significant spectral ratio estimation sensitivities. An essential part of the FMDSMAA implementation is to identify seismic traces with either poor trace metrics or nonsensical Peak Particle Accelerations (PPAs) indicative of a nonconstant source energy output (e.g., plate slippage, poor trace quality and\or poor or variable hammer impacts).  The FMDSMAA algorithm was initially implemented where traces were iteratively dropped due to nonsensical PPAs values (e.g., not decreasing with depth) and large FMDSMAA residually errors. To address this requirement a new algorithm was developed which incorporates an Extended Kalman Filter (EKF) into the FMDSMAA algorithm. The EKF applies a multicomponent exponential best fit to all the measured and normalized PPAs of a DST profile (so-called EKFAE algorithm). A multicomponent exponential best fit is utilized to ensure that the PPAs decrease with depth in case of significant measurement errors. An EKF is required because the measurement equation is nonlinear. This paper outlines the associated mathematical governing equations of the EKFAA algorithm. In addition, a real data example is provided which demonstrates the effectiveness of the EKFAE and FMDMSAA algorithms when processing DST datasets.

5th International Symposium on Frontiers in Offshore Geotechnics (ISFOG2025)

5 - Data Analytics and Machine Learning