Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


[CRITIS2014] Critical Infrastructure Online Fault Detection: Application in Water Supply Systems

No description

Constantinos Heracleous

on 18 April 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of [CRITIS2014] Critical Infrastructure Online Fault Detection: Application in Water Supply Systems

Critical Infrastructure Online Fault Detection: Application in Water Supply Systems
Water System Testbed
Online Fault Detection
Constantinos Heracleous
Experimental Results
Conclusions and Future Work
City's main Water Tanks.
Receives water from a Pumping Station
Water Tanks for Residential Area 2
They receive water from the main Tanks with gravity
Industrial Area Water Tank
Receives water from the main Tanks using a pump.
Water Leaks
Physical faults that can happen to this water supply system:
Testbed Objectives:
able to emulate the normal operation of Highlake's water supply system:
following a daily water demand curve
filling the tanks as needed to avoid any water shortages etc.
monitored and controlled by SCADA system
able to emulate physical faults:
pumps faults, valves faults and leaks
Tanks for Residential Area 2
(2x10L max water level 0.35m)
Manual valves emulating leaks

Reservoir (125L)
Three water pumps emulating the pumping station
Solenoid Valves emulating water demand (Res. Area 1)
Supply Valve
Tank 2 to Tank 4
PLC (Programmable Logic Controller)
Tanks for Residential Area 1
(2x25L max. water level 0.4m)
Tank for Industrial Area
(5L used [total 25L]
max. water level used 0.2m)
Water Level Sensor
SCADA HMI (Human-Machine Interface)
mass flow rate
fluid flow rate entering
and exiting the tank
liquid level in
cylindrical tank
Applying forward Euler discretization with time step
Online Fault Detection Algorithm Main Objectives:
Detect accurate online physical faults of the testbed:
Pump Failures, Valve Failures and Leaks.
Use the SCADA system database for online data through a network.

Fault Detection Scheme
Nonlinear uncertain
discrete-time dynamic system
the system uncertainty can be bounded by some known positive function
threshold error
state estimation error
nominal healthy dynamics
model uncertainty (unmodelled dynamics, parameter uncertainty, etc.)
*Ferrari, R.M.G., Parisini, T., Polycarpou, M.: A fault detection and isolation scheme for nonlinear uncertain discrete-time sytems. In: Decision and Control, 2007 46th IEEE Conference on. pp. 1009-1014 (2007)
Online Fault Detection Algorithm
General Scenario for Experiments
Testbed follows a predefine daily water demand curve where 24 hours are scaled down to 360sec (6 min.), i.e., the behavior of the real system in 1 hour is emulated by the testbed in 15 seconds.
In each area the demand valves open and close at specific time instances.
The testbed automatically starts and stops the pumps and opens and closes the supply valves do avoid water shortages.
Sensor measurements and control input values are obtained from the SCADA database every 1sec (
=1s) , i.e., every 4 minutes for the actual system.
A testbed was developed that emulates the operation of a critical infrastructure and some of its faults.
An online fault detection algorithm for critical infrastructures was implemented and tested.
Some experimental results that illustrate the effectiveness of this approach with a testbed.
Critical Infrastructure Online Fault Detection: Application in Water Supply Systems
Constantinos Heracleous
Estefania Etcheves Miciolino, Roberto Setola,
Federica Pascucci, Demetrios G. Eliades, Georgios Ellinas,
Christos G. Panayiotou, and Marios M. Polycarpou
KIOS Research Center for Intelligent Systems and Networks, and
Department of Electrical and Computer Engineering, University of Cyprus
This work was partially supported by the Prevention, Preparedness and Consequence Management of Terrorism and other Security-related Risks Programme European Commission - Directorate - General Home Affairs under the FACIES project, and by the European Research Council Advanced Grant FAULT-ADAPTIVE (ERC-AdG-291508).
Distribution networks in each area supply water to the consumers
We address the problem of Fault Detection in Critical Infrastructures.
SCADA system monitors and controls the city's water system
Pump Failures
Valve Failures
(pipe failures, tank leaks)
Pump Fault
At 225s (11:00pm) a fault introduced shutting down Pump 1 while in operation.
Valve Fault
At 200s (9:20pm) a fault introduced, closing the supply valve between Tank 2 and Tank 4
Leak Fault
At 180s (8:00pm) a leak fault (est. 6% of q3) was introduced through the water demand manifold of Tank 3.
We plan to implement, apply, and test fault isolation and accommodation approaches suitable for critical infrastructures.
Compare, with the help of the testbed, various fault diagnosis architectures.
Also explore ways for cyber attack detection.
water system description
technical drawing v.XX
Online identification of
ailure and
ttack on interdependent
Structural characteristics
tanks size and volume,
pipe lengths and connections,
pump flow rates,
Operational characteristics
water supply to the city - tank levels,
daily water demand patterns etc.
(w/ uncertainty)
flow rate through
an orifice:
- discharge coefficient
S -
cross-section area of the restriction
*Borutzky, W., Barnard, B., Thoma, J.: An orifice flow model for laminar and turbulent conditions. Simulation Modelling Practice and Theory 10(3-4), 141-152 (2002)
using (1) and (2):
state vector
valve control signal vector
input vector
pump supply flow rates
Testbed Model
Tank 1:
Tank 2:
Tank 3:
Tank 4:
Tank 5:
Testbed Model:
The fault is detected by the Tank 1 FD component at 227s (11:08pm).
For testbed
2 sec.
and for actual system
8 min.
after the fault.
The fault is detected by:
Tank 4 FD comp. at 204s (9:36pm), testbed
4 sec.
and actual system
16 min.
after the fault.
Tank 2 FD comp. at 210s (10:00pm), testbed
10 sec.
and actual system
40 min.
after the fault.
The fault is detected by the Tank 3 FD component at 186s (8:24pm)
For testbed
6 sec.
and for actual system
24 min.
after the fault.
For leak detection we need a very accurate water demand estimations.
Knowledge is POWER…but only once it’s applied!
Thank you for your attention!
Full transcript