Publications Search

The detection of anomalous water quality events has become an increased priority for distribution systems, both for quality of service and security reasons. Because of the high cost associated with false detections, both missed events and false alarms, algorithms which aim to provide event detection aid need to be evaluated and configured properly. CANARY has been developed to provide both real-time, and off-line analysis tools to aid in the development of these algorithms, allowing algorithm developers to focus on the algorithms themselves, rather than on how to read in data and drive the algorithms. Among the features to be discussed and demonstrated are: 1) use of a standard data exchange format for input and output of water quality and operations data streams; 2) the ability to "plug in" various water quality change detection algorithms, both in MATLAB® and compiled library formats for testing and evaluation by using a well defined interface; 3) an "operations mode" to simulate what a utility operator will receive; 4) side-by-side comparison tools for different evaluation metrics, including ROC curves, time to detect, and false alarm rates. Results will be shown using three algorithms previously developed (Klise and McKenna, 2006; McKenna, et al., 2006) using test and real-life data sets. © 2007 ASCE.

More Details

TYPE Conference YEAR 2007

Scopus OSTI

Predicting building contamination using machine learning

Proceedings - 6th International Conference on Machine Learning and Applications, ICMLA 2007

Martin, Shawn; Mckenna, Sean A.

Potential events involving biological or chemical contamination of buildings are of major concern in the area of homeland security. Tools are needed to provide rapid, onsite predictions of contaminant levels given only approximate measurements in limited locations throughout a building. In principal, such tools could use calculations based on physical process models to provide accurate predictions. In practice, however, physical process models are too complex and computationally costly to be used in a real-time scenario. In this paper, we investigate the feasibility of using machine learning to provide easily computed but approximate models that would be applicable in the field. We develop a machine learning method based on Support Vector Machine regression and classification. We apply our method to problems of estimating contamination levels and contaminant source location. © 2007 IEEE.

More Details

TYPE Conference YEAR 2007

OSTI Scopus

Testing and Analysis of Geostatistical Characterization Techniques using Full-Coverage Magnetometer Data Sets

Roberts, Barry L.; Mckenna, Sean A.

Abstract not provided.

More Details

TYPE Presentation YEAR 2007

OSTI

Dispersion analysis using particle tracking simulations through heterogeneity based on outcrop lidar imagery

Tidwell, Vincent C.; Mckenna, Sean A.

Solute plumes are believed to disperse in a non-Fickian manner due to small-scale heterogeneity and variable velocities that create preferential pathways. In order to accurately predict dispersion in naturally complex geologic media, the connection between heterogeneity and dispersion must be better understood. Since aquifer properties can not be measured at every location, it is common to simulate small-scale heterogeneity with random field generators based on a two-point covariance (e.g., through use of sequential simulation algorithms). While these random fields can produce preferential flow pathways, it is unknown how well the results simulate solute dispersion through natural heterogeneous media. To evaluate the influence that complex heterogeneity has on dispersion, we utilize high-resolution terrestrial lidar to identify and model lithofacies from outcrop for application in particle tracking solute transport simulations using RWHet. The lidar scan data are used to produce a lab (meter) scale two-dimensional model that captures 2-8 mm scale natural heterogeneity. Numerical simulations utilize various methods to populate the outcrop structure captured by the lidar-based image with reasonable hydraulic conductivity values. The particle tracking simulations result in residence time distributions used to evaluate the nature of dispersion through complex media. Particle tracking simulations through conductivity fields produced from the lidar images are then compared to particle tracking simulations through hydraulic conductivity fields produced from sequential simulation algorithms. Based on this comparison, the study aims to quantify the difference in dispersion when using realistic and simplified representations of aquifer heterogeneity.

More Details

TYPE Conference YEAR 2007

OSTI