Comparing Narrow Spectral Artifact Line Finders to Enable Continuous Wave Searches in LIGO

Continuous gravitational waves (CW) from sources such as a non-axisymmetric spinning neutron star have not yet been detected. If CW from neutron stars exist, the weak signals could be hidden within noise. The search for CW signals is impeded by the presence of narrow spectral artifacts (lines) caused by instrumentation or the environment at the Laser Interferometer Gravitational Wave Observatory (LIGO). Better identification of line noise would make a detection more likely, and a detection has the potential to expand our current knowledge of neutron stars. A non-machine learning (non-ML) line finding algorithm is currently used on a daily basis at the LIGO Hanford and Livingston Observatories. Studying and improving the existing non-ML FScan line finder across normalized and non-normalized data could enable accurate line-finding on non-normalized data such that researchers at LIGO would not need to switch between normalized and non-normalized views, and make it easier and quicker for LIGO data quality shift workers to spot changes in lines. A different approach to line finding utilizes a machine learning (ML) algorithm. My research uses both the ML and non-ML line finding methods, comparing the accuracy and efficiency of these two methods applied to various data sets. Expected results are that the non-ML method is currently more accurate and efficient from a computing resources perspective, but that the ML approach has the potential for high accuracy and adaptability, and will eventually be more efficient in human hours. The impact of my research is to implement more accurate and efficient line finding algorithms. This work is important because automated line finders at LIGO Hanford and LIGO Livingston save researchers hundreds of hours of work and could be set up to alert researchers on changes in noise behaviors.