Abstract
Inversion of strain from the velocity change at multiple angle sets was attempted using 1) direct search and 2) linear surrogate model (LSM).
• In the synthetic case with velocity data obtained at the given six angle sets, six strain values can be accurately inverted at reasonable time cost using
local search algorithms, but this can be computationally expensive for a scaled-up problem.
• The dependence of the directional velocity change on strains can be described using a LSM, and the inversions based on the LSM show a reasonable accuracy. It has the advantage of much lower time cost for a
scaled-up problem because the model coefficients can be readily determined for given angle sets.
• In the synthetic case with velocity data obtained at the given six angle sets, six strain values can be accurately inverted at reasonable time cost using
local search algorithms, but this can be computationally expensive for a scaled-up problem.
• The dependence of the directional velocity change on strains can be described using a LSM, and the inversions based on the LSM show a reasonable accuracy. It has the advantage of much lower time cost for a
scaled-up problem because the model coefficients can be readily determined for given angle sets.