|
Abstract:
|
Sensor systems, such as modern wind tunnels, require continual monitoring to validate their quality, as corrupted data will increase both experimental downtime and budget, and worse, lead to inconclusive scientific and engineering results. One approach to validate sensor quality is to monitor the distribution of individual sensor measurements. Although, in general settings, we do not know how correct measurements should be distributed, for each sensor system. Instead of monitoring sensors individually, our approach relies on monitoring the co-variation of the entire network of sensor measurements, both within and across sensor systems. That is, by monitoring how sensors behave, relative to each other, we are able to detect anomalies expeditiously. Previous monitoring methodologies, such as those based on Principal Component Analysis, can be heavily influenced by extremely outlying sensor anomalies. We propose a Bayesian mixture model that decomposes the anomalous and non-anomalous sensor readings into two parametric compartments. Specifically, we use a non-local, heavy-tailed Cauchy component for isolating the anomalous sensor readings, which we refer to as the Modified Cauchy Net.
|
