Results of Radically New Anisotropic Microseismic Inversion Algorithm that Successfully Avoids Many Local Minima (With Partial Matlab Codes)

Local minima is a very serious issue in microseismic/seismic inversion. It is almost equivalent to failure in inversion.

To understand how my new location algorithm performs, with different random initial values, I did 60 more synthetic tests of simultaneous microseismic/seismicity location and tomography in VTI media with old and my new algorithms. For the acquisition geometry below, the success rate is about 90% (still improving) for my new algorithm. That means I can invert noise-free data 90% times exactly. The second picture is a typical example of a successful inversion with noise. My implementation of the old algorithm of Vladimir Grechka has a success rate only around 65% (Ref 1). That means I can invert noise-free data around 65% times exactly with the old method. The time unit is around 0.01 second and the length unit is around 50 meter. Stiffness tensor is within 30% of their true values and initial location is within about 42 meters of their true values. With tighter bounds, the new algorithm's success rate is almost 100%. No implementation of relative algorithm such as double-difference algorithm. I can successfully invert noise-free data exactly around 90% times (still improving), different from recent papers (Ref 2) authored by professors from Hohai University, China University of Petroleum, and J. Carcione, whose numerical example is incapable of exact inversion of noise-free data. In other words, my new algorithm successfully avoids many local minima encountered by others for inverting noise-free data.

The new algorithm should perform better in other acquisition geometries. Check my open-sourced codes for part of the algorithm details (https://github.com/zhourui2058/layered_VTI_ray_tracing). Do not just believe what I said above without testing. You can test my inversion capabilities by forward modeling with my open-sourced codes and let me do the inversion.

Please talk to me if you want to know more! Please also give feedback and comments below, both positive and negative are welcome!

Thank you very much!

Ref: 1, Grechka, V. and Yaskevich, S., 2013. Inversion of microseismic data for triclinic velocity models. Geophysical Prospecting, 61, pp.1159-1170.
2, Huang, G., Ba, J., Du, Q. and Carcione, J.M., 2019. Simultaneous inversion for velocity model and microseismic sources in layered anisotropic media. Journal of Petroleum Science and Engineering, 173, pp.1453-1463.