Research Activities

Design and development of an integrated CPHS middleware for coupling information sensors, simulations, geospatial data and human input for community water systems.

In AquaSCALE, IoT sensing data from water infrastructures can track the changes of the network in a timely manner, and reflect a certain level of failures in the network. Pipe breaks or bursts often reduce pressure heads and increase flow rates at failure points [ref]. However, these measurements are highly cor- related with each other and also rely on other factors such as elevations and pipe properties. It is then hard to isolate the damaged pipes by these data itself, but aggregating with external sources will be helpful. For example, hu- man reports related to leak events can narrow down the region of potential damage. Extremely cold temperatures that are likely to cause pressure increasing along with pipe freezing indicate a pattern of changes in pressure heads (increasing first due to pipe freeze and decreasing due to pipe leak). This pattern can then be used as an additional information for event detection.

A novel two phase process for managing water workflows at multiple levels of observation and control.

AquaSCALE supports a multiphase approach where robust simulation using an enhanced version of a hydraulic simulator EPANET (with added support for IoT sensors and failure modeling) are used offline to generate profiles of anomalous events. The offline robust simulations and fault profiles are used for rapid coarse fault isolation that further informs statistical approaches for fine-grained localization of failures.

A plug-and-play analytic engine that enables the selection/integration of machine learning based techniques for fault isolation.

AquaSCALE incorporates a plug-and-play approach allows for a range of machine learning methods to be used for training models in Phase 1. These models are then combined with external data sources in Phase 2 to improve the accuracy and speed of the online event detection. The trained model incorporating additional observations can rapidly narrow down failure zones and help rapid identification of the problems in near real-time.

Enabling fault predictions of higher level events and logical actuations to prevent cascading impact.

AquaSCALE bridges the gap between the application- level concepts at service layer and the raw sensor data at base layer. By executing a logical observe-analyze-adapt loop at its core, it transforms input sensor data streams to higher level semantic streams that capture application level concepts and entities (e.g. locations of broken pipes and region impacted by flooding due to pipe breaks). This allows different classes of users to leverage their interests. AquaSCALE enables - city planners to design IoT instrumentation and management by identifying the vulnerable spots in water infrastructures; water agency operators to quickly locate and fix the problems and prevent cascading failures by event detection and prediction; community citizens to participant and obtain meaningful information, such as leak and flood awareness, via decision makers.

A prototype implementation of AquaSCALE and extensive evaluation of the proposed approach under diverse failure conditions.

The AquaSCALE prototype builds upon multiple tools developed by domain experts - these include the industrial strength hydraulic simulation platform EPANET (from the Environmental Protection Agency) for simulating the behavior of water networks, G-WADI, a hydrological remote sensing platform that supplies information from geophysical data sources and a novel adaptive Tweet Acquisition System (TAS) enables adaptive acquisition of tweets based on the situation. Using the flood model BREZO (originally designed to study the impact of high levels of precipitation), we incorporate the hydrodynamics of flood propagation to predict and study the cascading impact of leak events. These along with a range of ML approaches has enabled us to develop novel techniques for identifying leakage in multiple real failure-scenarios in water networks.

This web page was last updated on Monday, June 6, 2016.
Webmaster: Qing Han <qhan3 (-AT-) uci (-DOT-) edu>