ML_AE_relocation

Use machine learning (ML) methods to relocate acoustic emission (AE) events on a laboratory fault surface.

Reference:

Zhao, Q., Glaser, S.D. Relocating Acoustic Emission in Rocks with Unknown Velocity Structure with Machine Learning. Rock Mech Rock Eng (2019) doi:10.1007/s00603-019-02028-8

File description

Data files:

• AE_test_arrivals.mat - P-wave arrival pickings of 96 AE events recorded during the slip test.

• AE_train.mat - Locations (x,z) of pencil break events in the training data and their relative P-wave arrival pickings.

• AErelocNet_2D_Deploy.mat - ANNs trained to output AE source location on the laboratory fault (x,z).

Code files:

• AErelocNet_train_ANN.m - Train the ANN model

• AErelocNet_train_ANN_picking_quality_test.m - Check sensitivity of the ANN model to arrival picking quality.

• AErelocNet_train_ANN_with_Xvalid.m - ANN model accuracy estimation with ten-fold cross-validation.

• AEreloc_ANN.m - Apply the ANN model to the deployed ANN model for AE relocation.

• AEreloc_SVM_picking_quality_test.m - Check sensitivity of the SVM models to arrival picking quality.

• AEreloc_single_target_SVM.m - Train and apply SVM models for AE relocation.

• plotonfault.m - Function for plotting the AE events on the fault surface.

Image files:

• fault_surf_impose.jpg - Relocated AE locations plotted on top of the image of the laboratory fault after slip test.

• sample_after_slip.jpg - Raw image of the laboratory fault after the slip test.

• training_data_on_surf.pdf - Training data on laboratory fault surface with event IDs.

Additional data

Some additional data for the experiment. 12 sensors are used (11 sensors for the work in Zhao & Glaser (2019)). These data are not necessary for reproducing Zhao & Glaser (2019).

• AE_sensor_loc.mat - Locations of AE sensors in 3D.

• AE_signal_data.mat - Raw data for traning AE signals, locations and arrival pickings.

• disp_signal_and_picking.m - Code for plotting AE signals and arrivals.

• sensors_on_block.pdf - AE sensors plotted with the rock block in 3D with sensor IDs (sensor 12 not used in Zhao & Glaser (2019)).

Requirement

The ML methods are realized using MATLAB R2018a. The MATLAB neural network Toolbox and Statistics and Machine Learning Toolbox are required.