Award Abstract # 1762862
Fusing Data Analytics with Hydraulics in a Hydroinformatics Approach for Water Distribution System Monitoring

NSF Org: CMMI
Division of Civil, Mechanical, and Manufacturing Innovation
Recipient: UNIVERSITY OF ARIZONA
Initial Amendment Date: June 15, 2018
Latest Amendment Date: November 10, 2021
Award Number: 1762862
Award Instrument: Standard Grant
Program Manager: Yueyue Fan
CMMI
 Division of Civil, Mechanical, and Manufacturing Innovation
ENG
 Directorate for Engineering
Start Date: August 1, 2018
End Date: July 31, 2022 (Estimated)
Total Intended Award Amount: $498,949.00
Total Awarded Amount to Date: $502,364.00
Funds Obligated to Date: FY 2018 = $498,949.00
FY 2022 = $3,415.00
History of Investigator:
  • Kevin Lansey (Principal Investigator)
    lansey@email.arizona.edu
  • Jian Liu (Co-Principal Investigator)
Recipient Sponsored Research Office: University of Arizona
845 N PARK AVE RM 538
TUCSON
AZ  US  85721
(520)626-6000
Sponsor Congressional District: 07
Primary Place of Performance: University of Arizona
888 N Euclid Ave
TUCSON
AZ  US  85719-4824
Primary Place of Performance
Congressional District:
07
Unique Entity Identifier (UEI): ED44Y3W6P7B9
Parent UEI:
NSF Program(s): CIS-Civil Infrastructure Syst
Primary Program Source: 01002223DB NSF RESEARCH & RELATED ACTIVIT
01001819DB NSF RESEARCH & RELATED ACTIVIT
Program Reference Code(s): 062Z, 036E, 029E, 5948, 1057, CVIS
Program Element Code(s): 163100
Award Agency Code: 4900
Fund Agency Code: 4900
Assistance Listing Number(s): 47.041

ABSTRACT

This project will develop a novel scheme that seeks to provide full-scale monitoring from individual households to the entire water distribution system (WDS) by inexpensively combining various kinds of measurements from real-time metering of individual premise use to system flow and pressure conditions over time. The monitoring system will provide predictions on hydraulic conditions across the WDS over time. Comparing these predictions with field measurements capabilities in smart ways will increase burst detection accuracy, reduce detection time and improve accuracy of locating bursts resulting in shorter repair times, and lower costs and damages. Field demonstration and validation of the approaches within this project will demonstrate the active approach to burst detection/location. The project will promote inter-disciplinary graduate and undergraduate education and training, enhance research opportunities that promote female and underrepresented minority groups, and attract K-12 students to engineering programs.

The objective of this project is to develop a new generation of WDS monitoring systems within a new interdisciplinary hydroinformatics paradigm by fusing hydraulic knowledge with data-driven spatio-temporal analytics. The real-time burst detection algorithm will employ likelihood ratio test statistics to compare predicted and measured pressures and flows in conjunction with multivariate control charts to maximize event detection rates and avoid false alarms. After a burst is determined to be occurring, the approach returns to the WDS hydraulics to estimate responses to bursts at various locations and correlate measurement deviations to predicted conditions. Given the relatively low implementation costs, the concepts and technologies developed in this research will have significant commercial potential in a sizable international market. The project will demonstrate a platform for water utilities to rapidly assimilate new sensing technology and reap broader benefits as opportunities to incorporate this expanded data resources in WDS control systems are clear. The spatio-temporal sensor fusion mechanism will enhance data analytics for solving problems in other complex non-stationary systems where real-time modeling, monitoring and decision-making are of interest. The integrated systems and civil engineering research collaboration and the link to practice through field studies offer unique opportunities for establishing an inter-disciplinary education program, for multidisciplinary training of graduate and undergraduate students, and for student research opportunities. Broad dissemination will be ensured by maintaining data and case libraries, publishing papers in conferences and refereed journals, and collaborating with industry partners.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH

Note:  When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

Jun, Sanghoon and Arbesser-Rastburg, Georg and Fuchs-Hanusch, Daniela and Lansey, Kevin "Response Surfaces for Water Distribution System Pipe Roughness Calibration" Journal of Water Resources Planning and Management , v.148 , 2022 https://doi.org/10.1061/(ASCE)WR.1943-5452.0001518 Citation Details
Jun, Sanghoon and Jung, Donghwi and Lansey, Kevin E. "Comparison of Imputation Methods for End-User Demands in Water Distribution Systems" Journal of Water Resources Planning and Management , v.147 , 2021 https://doi.org/10.1061/(ASCE)WR.1943-5452.0001477 Citation Details
Zhang, Y. "Detecting Burst in Water Distribution Systems via Penalized Functional Decomposition" INFORMS 2019 , 2019 Citation Details
Zhang, Yinwei and Lansey, Kevin and Liu, Jian "Detecting Bursts in Water Distribution System via Penalized Functional Decomposition" 2020 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) , 2020 https://doi.org/10.1109/IEEM45057.2020.9309770 Citation Details

PROJECT OUTCOMES REPORT

Disclaimer

This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.

Leaks and bursts occur on a daily basis in many water distribution networks (WDN) like the system show in Figure 1.  They result in lost water that has been treated and pumped into the network.  If undetected, they will grow and can cause damage to other infrastructure including underground utilities and roadways.  Water utilities are moving to employ data collection systems and advanced methods to detect and locate leaks early in their development. 

Most often, residential and commercial consumer water demands are met by pumping water from a source through a WDN to withdrawal points.  As flow passes through pipes, energy is lost and water pressure drops.  The amount of pressure drop is related to the pipe flow rates and the pipe layout and sizes.  In earliest attempts to identify leaks, utilities install meters to measure pressure and pipe flow at selected location.  These measurements are compared to historical values and deviations indicate an anomaly (leak).  However, the measurement network tends to be sparse and the impact of a leak that has not made its way to the surface relatively small.  As a result, many leaks can not be identified until they breach the ground surface. 

This project proposed two major improvements to enhance leak detection.  First, a new data collection system is proposed that measures consumer demands at their service meter. These systems are in place primarily to help identify leaks after the meter in the household.  They have had limited use in detecting failures within the WDN.  A further enhancement to this system is to simultaneously measure pressure at the service meter.  This can be done at little cost and, at least, one meter manufacturer now offers this capability. 

The second improvement is incorporating this information in leak detection methodologies. Rather than be limited to historical data, the new data provides a significant opportunity to improve detection and location of leaks.  Three general approaches were developed to use this data.  First, a so-called mass balance approach was applied that compares the total flow withdrawn by customers to the flow that is provided to the WDN.  Since the customer meters do not measure flow lost from cracks in pipes or pipe joints, the difference between the two flow rates can be attributed to leaks.  This approach was very effective for small networks and during the night for most systems when the demand is low.  When demand is high, measurement uncertainties in the high demands can be significant and be larger than the leak.  Further, although this method provides information on if the leak is occurring, it does not help determine where it is located.

To improve over the mass balance approach, we linked hydraulic model and domain knowledge with optimization and machine learning/artificial intelligence (ML/AI) methods.  Most utilities construct hydraulic models to simulate the pressure drops and pipe flows based on the WDN physical characteristics and water withdrawals.  With measured withdrawals, pressure conditions without a burst can be estimated.  These predictions can be prepared with measured pressure values.  The differences are caused by the leak and larger differences provide information on the leak location.  Figure 2 shows a spatial representation of the pressure differences.  In the upper figures show without leak conditions (a and b), the errors are random due to measurement errors.  The lower figures (c and d) show conditions with a leak and the concentration of significant differences located around the circled leak location. 

Pictures like these can be assessed by artificial intelligence techniques that focus on pattern recognition.  It is complicated by having a series of snapshots representing conditions over time.  We demonstrated the value of considering temporal changes compared to a single snapshot in improving detection likelihood and confidence in the identification.  Figure 1 shows the detection probability (DP) for leaks of varying sizes and different locations in the system.  We also applied an optimization scheme to solve this problem.  Determining rules to state is a leak is occurring is possible but appears to be system dependent.  Lack of good rules can result in false alarms or delayed detection and WDN operators losing confidence in the leak detection algorithm.  ML/AI is able to generate detection rules without modeler intervention but may need to be retrained as the WDN changes. 

Comparing the methods above to data available from current data collection systems and several WDNs demonstrated significant improvements in leak detection with all and, in particular, small leaks, the time to detect a leak and the number of false alarms.   


Last Modified: 12/29/2022
Modified by: Kevin E Lansey

Please report errors in award information by writing to: awardsearch@nsf.gov.

Print this page

Back to Top of page