A change of non-astrophysical origin in the detector state or in the statistical nature of data while an interferometer is in lock reflects an abnormality. The change can manifest itself in many forms: transients, drifts in noise power spectral density, change in cross correlation between channels, etc. We advance the idea of an algorithm for detecting such change points whose design goal is reliable performance, i.e. a known false alarm rate, even when statistically unmodelled data such as those from the physical environmental monitors are included, Reliability is important since following up on such change points could be fairly labour intensive. Such an algorithm need not be simply a collection of isolated independent monitors running in parallel. We present the first design steps towards building this detector characterization robot along with some preliminary results and outline some possibilities for the future
展开▼
