Data field application in removing large P-phase arrival picking errors and relocating a mine microseismic event

Abstract

Seismic source location is a key parameter in the microseismic (MS) monitoring technology, and its accuracy is correlated with the calculation of some other event source parameters (e.g., event magnitude and focal mechanism). The conventional location methods usually take advantages of objective functions based on all arrival-time dataset and solve them through optimization algorithms, while a local optimization may be obtained due to the influence of initial iteration point. Even if a global grid search algorithm is applied, it is likely to obtain a poor location result when there are large P-phase arrival picking errors. Therefore, this study proposed a method to remove large picking errors from P-phase arrival time dataset and relocate the seismic source based on bootstrap sampling selected sub datasets and data field theory. Its basic principles are shown as follows: by using the time difference (TD2) method through the simplex algorithm, seismic sources are located based on sub datasets of P-phase arrival time from multiple bootstrap sampling, and the location mean value of points with the 50 largest potential values in the data field is taken as the approximate location result of a MS event. Furthermore, according to the approximate location, the difference between theoretical travel time of P waves at each sensor and travel time based on the observed arrival time is calculated, and then a threshold value is set to remove large picking errors. Furthermore, by repeating the above steps of bootstrap sampling and data field based location, an accurate relocation is obtained. The synthetic tests and application were based on sensor locations of the Institute of Mine Seismology (IMS) acquisition system settled in the Yongshaba mine (China). Two typical synthetic events were separately set inside and outside the sensor array, and eight blasting events with known locations were treated as test data. These events were located and studied by utilizing the TD2 method, data field based TD2 method (DF1–TD2) and data field based TD2 method with large picking errors removed (DF2–TD2). Results demonstrate that the TD2 method shows a poor location stability when large picking errors are present, while location errors obtained by the DF1–TD2 and DF2–TD2 methods are obviously smaller than that of the TD2 method. Moreover, location results obtained by the DF2–TD2 method with large picking errors removed are superior to those without removing large picking errors. Furthermore, the DF2–TD2 method was applied to locate 300 MS events, the location elevations are well consistent with the stope distribution, showing that the method has a good application prospect. Finally, the data field theory was combined with other location objective functions and P-phase arrival time picked by the Akaike information criterion (AIC) method, and we obtained good location results. In addition, the data field based location results obtained through a grid search algorithm and the simplex algorithm are similar. Therefore, the combinations of data field with different objective functions, automatic P-phase arrival picking methods and iteration algorithms show broad prospects.