Spatiotemporal Anomaly Detection in Gas Monitoring Sensor Networks
In this paper, we use Bayesian Networks as a means for unsupervised learning and anomaly (event) detection in gas monitoring sensor networks for underground coal mines. We show that the Bayesian Network model can learn cyclical baselines for gas concentra
- PDF / 974,245 Bytes
- 16 Pages / 430 x 660 pts Page_size
- 88 Downloads / 222 Views
2
CSIRO ICT Centre, Locked Bag 17, North Ryde, NSW 1670, Australia [email protected] School of Information Technologies, The University of Sydney, NSW 2006, Australia
Abstract. In this paper1 , we use Bayesian Networks as a means for unsupervised learning and anomaly (event) detection in gas monitoring sensor networks for underground coal mines. We show that the Bayesian Network model can learn cyclical baselines for gas concentrations, thus reducing false alarms usually caused by flatline thresholds. Further, we show that the system can learn dependencies between changes of concentration in different gases and at multiple locations. We define and identify new types of events that can occur in a sensor network. In particular, we analyse joint events in a group of sensors based on learning the Bayesian model of the system, contrasting these events with merely aggregating single events. We demonstrate that anomalous events in individual gas data might be explained if considered jointly with the changes in other gases. Vice versa, a network-wide spatiotemporal anomaly may be detected even if individual sensor readings were within their thresholds. The presented Bayesian approach to spatiotemporal anomaly detection is applicable to a wide range of sensor networks.
1 Introduction Since the 1980s, electronic gas monitoring sensor networks have been introduced in the underground coal mining industry. However, no current system can provide site specific anomaly detection. This means monitoring systems often give false alarms, which can be costly to the mining operation. The periodic variation in the gas concentration also increases the number of false alarms in these flat line threshold based systems. Further, current systems ignore the spatial relations between data gathered at different sensor network nodes. These spatial relationships between data could identify anomalies missed by individual sensors. Conversely, the spatial relationships could explain away the anomalies identified by the individual gas sensors, thus avoiding false alarms. Currently, the existing system integrates and interprets incoming data in accordance with a pre-determined set of rules, produces a risk profile, and autonomously initiates a response to a breach of these rules. A problem with this approach is that no clear-cut definitions of abnormal situations with respect to the concentration of different gases exist, so that it is difficult to produce a good set of rules. 1
Corresponding author. The authors list after the lead author is in alphabetical order.
R. Verdone (Ed.): EWSN 2008, LNCS 4913, pp. 90–105, 2008. c Springer-Verlag Berlin Heidelberg 2008
Spatiotemporal Anomaly Detection in Gas Monitoring Sensor Networks
91
The underground coal mining industry has been struggling with the issues of sitebased moving threshold levels for critical gases since the introduction of electronic gas monitoring systems in the 1980s. No satisfactory, scientifically validated methodology is in existence that can provide a mine with its own
Data Loading...
