Machine learning for early warning detection of gravitational wave signals

Alfaidi, Reem (2026) Machine learning for early warning detection of gravitational wave signals. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2026AlfaidiPhD.pdf] PDF
Download (23MB)

Abstract

The direct detection of gravitational waves has enabled detailed observations of compact binary coalescences, including binary black hole, binary neutron star, and neutron star-black hole systems. A central challenge in gravitational wave astronomy is the development of low-latency detection methods capable of identifying signals before merger, enabling rapid multi-messenger follow-up observations.
This thesis investigates the use of long short-term memory neural networks for early-warning detection of gravitational wave signals. The objective is to determine whether a data-driven model can reliably identify signals during the inspiral phase and provide early-warning of coalescence.
This work studies simulated datasets for binary black hole and binary neutron star systems and evaluates model performance using simulated signals embedded in both Gaussian and realistic detector noise. The long short-term memory model processes time-series strain data and produces per-timestep detection probabilities, enabling continuous monitoring of signal evolution.
These results demonstrate that long short-term memory-based models provide an effective and computationally efficient approach for earlywarning detection of gravitational wave signals, establishing that reliable detection during the inspiral phase is achievable on physically meaningful timescales. The model can detect binary black hole signals several seconds before merger and binary neutron star signals up to approximately one minute before coalescence, depending on the signal strength. When applied to realistic detector noise following preprocessing and conditioning, the model continues to exhibit responses to injected signals. However, the results also highlight the challenges associated with detector non-stationarity and the mismatch between idealised Gaussian training conditions and real interferometer noise.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Additional Information: Supported by funding from the Kingdom of Saudi Arabia.
Colleges/Schools: College of Science and Engineering > School of Physics and Astronomy
Funder's Name: Kingdom of Saudi Arabia
Supervisor's Name: Messenger, Dr. Chris
Date of Award: 2026
Depositing User: Theses Team
Unique ID: glathesis:2026-86366
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 18 Nov 2026 11:22
Last Modified: 18 Nov 2026 16:45
Thesis DOI: 10.5525/gla.thesis.86366
URI: https://theses.gla.ac.uk/id/eprint/86366
Related URLs:

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year