Erlangen 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

GR: Fachverband Gravitation, Relativistische Astrophysik und Kosmologie

GR 12: Relativistic Astrophysics III / Gravitational waves III

GR 12.6: Talk

Thursday, March 19, 2026, 15:00–15:15, KH 01.016

RealTime Seismic Waveform Prediction Using Low-Latency Transformer-Based Models — •Kyrill Emanuel Blümer, Alexander Kappes, and Waleed Esmail for the Einstein Telescope collaboration — Institut für Kernphysik Uni Münster

The Einstein Telescope (ET) is a planned third-generation gravitational-wave detector, that will be built underground. It is designed to improve the detection sensitivity by up to an order of magnitude relative to existing detectors, especially at low frequencies. Because gravitational waves generates an extremely small strains, seismic and Newtonian noise becomes a limiting factor for the low-frequency sensitivity of the ET. Transformer-based deep learning models are well suited for learning long-range temporal and spatial dependencies in seismic waveform data and can provide accurate short-term forecasts of the 3D ground motion. However, for longer prediction horizons, waveform prediction quality degrades due to the error accumulation, when predictions are produced autoregressively. This talk will explore architectural and algorithmic improvements aimed at achieving stable, real-time, low-latency seismic waveform prediction.

Keywords: Neural network; seismic noise