Seismic signal processing for near-field source localization

The goal of the research performed in this dissertation is to correctly estimate the location of a seismic source from acceleration records collected via a network of seismic stations. Specifically, multi-axial seismic accelerometer stations coupled to the surface of the ground collect accelerations generated by seismic sources; the sensor network consists of many such accelerometer stations spaced less than 50 meters apart from each other; and, the seismic sources are located in the near-field of sensor arrays. Near-field scenarios refer to cases where the distances between sensors and the source are less than a few wavelengths of the signal in question. Typical seismic signals comprise wavelengths ranging between 0.05-15 km. Thus, near-field scenarios are pertinent to sensor-source distances less than 30 to 50 km. Near-field seismic signals exhibit characteristics different from the well understood and frequently utilized far-field signals, and pose unique challenges for source localization.; Two main approaches for seismic source localization are developed in this dissertation. The first approach involves estimating the Direction-Of-Arrival (DOA) of the source from collected sensor data first, and subsequently locating the source using the estimated DOA's. Three DOA estimation techniques, namely, the Covariance Matrix Analysis (CMA), Surface Wave Analysis (SWA), and Modified Kirlin Method (MKM) were investigated. CMA is an existing method developed for far-field data. Here, the effectiveness of CMA for near-field data is evaluated. The other two techniques---i.e., SWA and MKM---are novel, and were developed based on the unique characteristics of near-field seismic signals. SWA exploits the characteristics of Rayleigh waves; whereas MKM is based on a noise equalization estimate of the frequency domain signal-part covariance matrix of the near-field signal.; The second approach is a maximum likelihood optimization method that was developed based on a theoretical model of the near-field seismic signal. This, seismic Approximate Maximum Likelihood, algorithm is also a two-step process in which the experiment site parameters are calculated first, then source location is estimated directly from site parameters and collected sensor data.; All of the techniques developed for seismic source localization are validated using data collected in a number of field experiments. These experiments involve stationary impact sources, as well as a moving vehicle.