Particle swarming of sensor correction filters

TL;DR

This paper demonstrates that particle swarm optimization can design more effective sensor correction filters for seismic noise reduction in gravitational wave detectors, improving their operational stability during seismic disturbances.

Contribution

It introduces a particle swarm algorithm for designing sensor correction filters, outperforming existing filters in seismic noise mitigation for LIGO's active platform control.

Findings

Particle swarm-optimized filters outperform current filters in most frequency ranges.

The method successfully reduced platform differential RMS velocity.

Filters were validated with real LIGO seismic data and implemented at Hanford.

Abstract

Reducing the impact of seismic activity on the motion of suspended optics is essential for the operation of ground-based gravitational wave detectors. During periods of increased seismic activity, low-frequency ground translation and tilt cause the Advanced LIGO observatories to lose `lock', reducing their duty cycles. This paper applies modern global-optimisation algorithms to aid in the design of the `sensor correction' filter, used in the control of the active platforms. It is shown that a particle swarm algorithm that minimises a cost-function approximating the differential RMS velocity between platforms can produce control filters that perform better across most frequencies in the control bandwidth than those currently installed. These tests were conducted using training data from the LIGO Hanford Observatory seismic instruments and simulations of the HAM-ISI (Horizontal Access Module Internal Seismic Isolation) platforms. These results show that new methods of producing control filters are ready for use at LIGO. The filters were implemented at LIGO's Hanford Observatory, and use the resulting data to refine the cost function.