A circumbinary planet is a planet that orbits two stars in a binary star system. These systems are relatively hard to detect, and currently there are only 14 confirmed transiting circumbinary systems. We conducted our own search for new circumbinary planets using data from NASA's Transiting Exoplanet Survey Satellite (TESS) mission. We looked at about 1,600,000 monthly light curves of 222,903 stars and used machine learning techniques to find eclipsing binaries. In particular, we used an unsupervised learning K-means cluster algorithm and a Convolutional Neural Network. Tests show these methods have about 91% accuracy in identifying eclipsing binaries. About 5000 eclipsing binaries were found, including one previously known circumbinary planet system. At this time, no new circumbinary planet systems were found.In the second part of this thesis, we studied the previously known circumbinary planet system KIC 10753734. We analyzed all of the available Kepler data, ground-based optical observations, and new data from TESS to find a comprehensive model for the system. In this case, the analysis of this system is complicated by the presence of variability in the light curve caused by star spots on both stars. We therefore developed a new technique to remove the effect of the star spots from the light curve. Updated values for the component masses and radii and updated values for the orbital elements are presented.
展开▼
