Particle Swarm Optimization and its Modification in Gravitational Wave Data Analysis

Pal, Tathagata (2019) Particle Swarm Optimization and its Modification in Gravitational Wave Data Analysis. Masters thesis, Indian Institute of Science Education and Research Kolkata.

PDF (MS dissertation of Tathagata Pal (14MS058)) 14MS058.pdf - Submitted Version Restricted to Repository staff only Download (8MB)

Abstract

Einstein predicted the existence of gravitational wave when he gave the General Theory of Relativity (GTR) in 1915. The 1st detection of gravitational wave was made on 14th September, 2015 and as a result of this, 2017 Nobel Prize in Physics was awarded to Rainer Weiss, Barry C. Barish, and Kip S. Thorne. The detection of GW has opened up a new window of observation into space as well as time. It brings with it never foreseen information! The space based detector, LISA promises to detect many more GW signals with lower frequencies. The GW detection on the other hand adds one more feather to the cap of success of GTR. When two massive bodies in space collide and merge together,they send out ripples in space-time in all directions. Such cataclysmic events include merger of two black holes (BH-BH), merger of a black hole and a neutron star (BH-NS), merger of two neutron stars (NS-NS) etc. The ripples are then detected as ’Gravitational Waves’ (GW) by the humongous interferometers (arm length of 4 km!) at LIGO Hanford, LIGO Livingstone and VIRGO.The Big Bang theory of the Creation also predicts the existence of primordial GW. But unfortunately the ripples are so weak that we do not have the technology at this moment to detect them. Noise from a number of sources make it very difficult in detection of these waves which generally have very low amplitude (which is of the order of 10−20, measured in terms of strain produced in matter as GW passes through them). The main source of noise with the ground based observatories are the seismic noise, thermal noise etc. Before doing any analysis it is thus very important to obtain the noise characteristic of the system. We have characterized the noise by obtaining the Power Spectral Density (PSD). The technique of matched filtering with a template bank is employed in order to detect any signal buried in the data. The current algorithm scans over all the possible templates in the bank and it is heavily time consuming. The Particle Swarm Optimization (PSO) algorithm minimizes this time of search greatly and still able to find the signal with some desired accuracy. We have implemented the PSO algorithm ourselves with the gravitational wave data and looked into its performance. In this work, we have also developed some new algorithms which are modifications over PSO and tested their efficiency adding different types of noise (Gaussian noise and O2 run noise)to the signal. At the end we have analyzed the results from different algorithms and compared with each other. This work is also going to lay the foundation of implementation of machine learning (ML) techniques in gravitational wave data analysis.

Actions (login required)

View Item