A question using Example of Seismic Inversion

I’m a researcher on Seismology and recently I found this promising SimPEG and instantly attracted to it. I installed the environment and had a try at once, I noticed the Seismic Example in the documentation which is about the velocity tomography problem. I ran the example and got the exactly same results, very cool. Then, I tried to test more. I modified the sources and receivers in the example to simulate another scenario like this: np.random.seed(0) src_x = (np.random.rand(50)-0.5)*200 src_y = (np.random.rand(50)-0.5)*200 src_lct = np.c_[src_x, src_y]

rcv_x = np.r_[range(-100,100,20), [100]*9, range(100,-100,-20), [-100]*10] rcv_y = np.r_[[-100]*10, range(-100,100,20), [100]*10, range(100,-80,-20)] rcv_lct = np.c_[rcv_x, rcv_y] From forward simulation, I got the right arrival times data of all the receivers from all the sources successfully. Then I tried the inversion process. I set the initial model a uniform model with 2500m/s (3000m/s in the example), and left the other settings without change. Then the iteration began. The total iteration number soon reached 100 which is the maxIter and unfortunately I got a fairly wrong result like this: I tried to modify the parameters in regularization, optimization and inversion processes but I just couldn’t get the point. I think my modification is simple and there should be a right way to get a good recovery of the true model. I hope someone can help me with this issue by setting the appropriate parameters and I’ll very appreciate with that.

[chrisweird7]

@ckohnke I’ve figure it out. I don’t know what to say. How careless I am. I did all things right except to copy the code line np.random.seed(0). It means my sources were changing all the time. Haha, my problem becomes a joke. Anyway, now I have the right result and yes, SimPEG is good.

[chrisweird7]

@thast Sure.