The Use of a Decentralized Wireless Sensor Network for

Document Sample
The Use of a Decentralized Wireless Sensor Network for Powered By Docstoc
					           The Use of a Decentralized Wireless Sensor Network
                    for CSO Abatement and Control
                       L. A. Montestruque1 and T. P. Ruggaber2
 EmNet, LLC, 12441 Beckley St., Suite 6, Granger, IN 46530; PH (574) 360-1093;
FAX (574) 968-0269; email:
 A.M. ASCE, EmNet, LLC, 12441 Beckley St., Suite 6, Granger, IN 46530; PH (574)
303-3031; FAX (574) 968-0269; email:


        EmNet, LLC, the City of South Bend, IN, the University of Notre Dame, and
Purdue University have developed CSOnet, an embedded sensor network that
provides decentralized real-time control (RTC) for CSO abatement. Unlike
conventional centralized RTC systems, CSOnet implements optimal control strategies
using a distributed network of microprocessors rather than a massive central
computer, which results in a system that is robust, incrementally implementable, cost-
effective, and easily integrated into any existing infrastructure. CSOnet was
demonstrated during a pilot demonstration, which increased the storage ability of a
retention basin by up to 110% and cost 55% less than conventional solutions.
Furthermore, in a series of computer simulations, CSOnet was able to reduce the
amount of overflow by up to 84% through interceptor optimization, to significantly
reduce the number of CSO events, and to prevent dry weather overflows.


         Each year in the United States, combined sewer overflow (CSO) events result
in the release of 3,217 billion liters (850 billion gallons) of untreated wastewater into
lakes and rivers, causing drinking water contamination, human illness, animal and
fish kills, and eutrophication (USEPA 2004). The United States Environmental
Protection Agency (USEPA) estimates that reducing this problem by 85% over the
next 15 years will cost over $50.6 billion using traditional abatement methods (sewer
separation, treatment plant expansion, etc.) An alternative to construction intensive
solutions such as sewer separation projects or waste water treatment plant (WWTP)
expansions is the use of real time control (RTC) systems. RTC solutions can make
use of underutilized in-line and off-line storage areas to precisely control the flows
within the sewer system. This allows the optimization of the transport system and
delivery to the WWTP effectively reducing overflows at a reduced cost. These
systems tend to be centrally controlled, requiring a powerful central computer to
solve thousands of complex equations every few minutes (Heinz and Schultz 2006,
Colas et al. 2001). Unfortunately, RTC systems have not been widely used due to the
cost of the components and the complexity of the control system. Furthermore,
existing RTC systems do not scale linearly, meaning that the system is proportionally
more expensive for smaller municipalities than large cities (Pleau et al. 2005).
        Many of these obstacles can be overcome through the use of a decentralized
RTC scheme. In this system, control decisions are made and implemented by a
network of microprocessors, which gather data from the local area and communicate
with each other to determine the optimal control strategy (see Figure 1). Such a
system is inherently redundant, robust, and easily installed (either incrementally or at
one time) into any existing infrastructure (Ruggaber et al. 2006). EmNet, LLC, the
City of South Bend, The University of Notre Dame, and Purdue University have
collaborated extensively during the past two years to create and implement just such a
system, called CSOnet.

CSOnet Description

        CSOnet is composed of a number of embedded real-time processors that
communicate over a multi-hop wireless communication network. Each node in
CSOnet consists of at least three basic subsystems; the sensor/actuation subsystem,
the processor module, and the wireless communication subsystem. The processor
module is either a micro-processor or embedded single board computer that is
interfaced to sensors and actuators through the sensor/actuation subsystem. The
processor module can store sensor information and can make control decisions based
on information it receives over its wireless communication subsystem.
        The novelty of CSOnet rests in two main aspects; information
dissemination using low power radio systems and a decentralized control for
regulating in-line and off-line water storage. The first aspect uses multi-hop mesh
networking techniques that allow the devices to communicate over large distances by
relaying data through other intermediary devices. This differs from traditional radio
modem approaches where each device communicates directly with a central
controller, usually located at the WWTP. For a large number of sensors (>50), these
centralized data-gathering systems require considerable bandwidth since one single
central location must communicate with each device simultaneously at each sampling
time. Radio modem installations can be expensive (around $10,000 per location)
whereas a CSOnet device costs approximately $2,000 per location.
        The second novelty of CSOnet rests in its use of decentralized control
strategies. Typically, a central computer gathers the information from all system
sensors using the radio modem connections. After all the information is received, the
central computer calculates a control strategy using a complex model of the sewer
system and an optimization algorithm. Traditionally, receding-horizon predictive
global optimization control techniques are used over a SCADA system. The weakness
of this centralized method is that it relies heavily on a-priori models of the existing
sewer system. Additionally, it requires powerful computational infrastructure to
obtain adequate set-points for the flow diverting structures in a timely manner,
especially if the amount of real-time information on the sewer network comes from a
limited number of field sensors.
        CSOnet, on the other hand, processes information in a decentralized manner.
Each processor in CSOnet sits in a manhole or retention basin of the sewer system.
Each CSOnet processor only gathers data from its adjacent neighbors to make
control decisions that asymptotically approach the optimal control achieved
using centralized methods. Moreover, CSOnet’s use of recursive feedback
mechanisms in controlling wastewater flows endows it with added robustness to
model uncertainty. CSOnet, therefore, has the ability to adapt to unforeseen
changes in the sewer infrastructure, something which is entirely impossible for
traditional centralized SCADA-based solutions. The technology implications of the
use of CSOnet or centralized SCADA-based solutions are multiple and have
orthogonal results (see Figure 2).
        The wireless nature of CSOnet allows its implementation into existing
distribution systems with only minor infrastructure modifications. No hardwired
communication media such as optic fiber, coaxial cable, or twisted pair is needed. If
such networks already exist, the nodes can easily access them. Some roadwork and
sewer work is required for the placement of Smart Valves (actuated valves) into the
sewer lines and retention basins, but this localized work can be done with minimal
inconvenience to the public. Smart Valves do require electricity, but battery or solar-
powered options are possible.
        The concluding differences between the use of the CSOnet solution or the
SCADA based solution are summarized in Table 1. These are powerful motivators for
municipalities with restricted economic and engineering resources that seek a
technologically efficient solution that will bring both EPA compliance and
environmental enhancement.
        TABLE 1. Comparison between CSOnet control solution and a SCADA-
                         based control solution.
                          CSOnet decentralized control      SCADA-based centralized
                                   solution                    control solution
                             Inexpensive: reusable and        Expensive: customization
 Control Algorithm
                            homogeneous components.              required for each
Implementation Cost
                             Inherent robustness due to         Highly dependent on:
                            localized decision, massive            hydraulic model,
    Robustness            spatial and temporal feedback,            communication
                                and communication             infrastructure, and central
                                    redundancy                        computer.
                                No communication             Expensive communication
                            infrastructure required, no         infrastructure and
  Hardware Cost             central computing element        computing systems due to
                                     required.                   large bandwidth
                           Fast and inexpensive, battery       Slow and expensive,
                          or solar operated devices form    requires implementation of
  Implementation           ad-hoc network and perform       computing center, design of
                               decentralized control.          control architecture,
                                                              electrical installations.

        The chief technological components in CSOnet were developed by EmNet
LLC, Purdue University, and the University of Notre Dame. The original concept for
CSOnet was developed by Drs. Jeff Talley and Michael Lemmon (University of
Notre Dame) as a method for monitoring and controlling CSO events. The realization
of that concept relied on hardware designs and middleware services developed at
EmNet LLC, as well as a novel antenna design developed by Dr. W. Chappell of
Purdue University. The remainder of this section briefly describes the existing
CSOnet technology.
        CSOnet consists of three main types of embedded nodes (see Figure 3), which
are refered to as the gateway (Gnode), the instrument node (Inode), and the routing
node (Rnode). The Inode and Rnode are both based on the Chasqui processor module
which was developed by EmNet LLC for this project. The difference between these
two node types is that the Inode has a sensor/actuation subsystem, whereas the Rnode
(which is only used to relay data between nodes) requires no such subsystem. The
Gnode is a single board embedded computer that can connect to the Internet and
which often has an actuator interface for controlling valves in the sewer network. In
practice, Inodes gather data about head level and flow rates within the sewer system,
Rnodes relay that sensor data to a Gnode, and the Gnode then computes the required
control scheme and controls valves within the network to adjust local flow rates
within the sewer system. Note that connectivity between the Gnodes do not
necessarily require accessibility to the internet or other network, although scalability
is guaranteed if the network can be partitioned in subnets with Gnodes connecting
with each other through a faster media.
        The main piece of technological hardware is the Chasqui processor module.
The Chasqui Wireless Sensor Node was developed by EmNet LLC to address a
number of real-life issues that were found in existing embedded network nodes
such as Crossbow’s MICA2 processor module. These issues concerned the limited
radio range of the MICA2 module and the need for specialized sensor/actuator
interfaces required in this application. To answer these questions, EmNet LLC started
with the original embedded node designs developed by U.C. Berkeley and modified
the radio subsystem and sensor/actuator interface subsystems. The Chasqui node uses
a MaxStream radio operating at 900MHz that uses Frequency Hopping Spread
Spectrum (FHSS) signaling to reduce interference and allow large range (TX power =
100mW with a range >700m in urban environments and >5km in line of sight). The
radio complies with the limits specified by the FCC for the use of license free ISM
spectrum. The Chasqui’s microprocessor is an ATmega128 running at 8MHz. This
microprocessor and support hardware enables the use of Berkeley’s TinyOS for
firmware coding. TinyOS is a programming tool widely used by academic institutions
for research in the area of sensor networks. The Chasqui node uses a highly efficient
switching power supply that generates 3.5V for the microprocessor, 5V for the radio,
and 12V for the sensor. The power supply can work with an input voltage between
4V and 7V allowing extended operation on battery power. The sensor interface
allows the connection of off-the-shelf sensors to the Chasqui Node. MOSFET
switches enable the microprocessor to turn off the sensors when not needed.
        The Gnodes are the main control nodes in CSOnet. These are Linux
embedded single board computers. These nodes collect data from Inodes and Rnodes
and then use that data to adjust Smart Valves to meet changing conditions. To
perform this, an actuator interface was designed providing two Pulse Width
Modulated outputs with 1A capacity and two 4mA-to-20mA current outputs. In
addition to controlling Smart Valves, Gnodes are responsible for keeping all of the
nodes in CSOnet synchronized. Synchronization ensures that all of the nodes are
awake and operating at the same time, allowing for real-time communication while
conserving power through a hibernation cycle. Gnodes also contain PC card slot and
an Ethernet connection. The PC card allows, for example, the use of a cellular card to
connect to the Internet via a wireless cellular connection to exchange information.
The wireless connection allows the Gnodes to post current conditions of the system
and the Smart Valves on the Internet for the WWTP. These functions require
considerable power, which requires the Gnodes to be linked into the AC power
supply for the Smart Valves. Because they have a continual source of power, Gnodes
do not undergo a power conservation cycle.
        The communication structure is a core part of CSOnet. It enables the use of
low power radios powered by batteries to cover large distances by multi-hopping
connections. Data is disseminated by means of an advanced routing algorithm called
Persistent Stateless Gradient-Based Routing. This algorithm was developed by
EmNet LLC and it enables the network to maintain connectivity in spite of poor
node-to-node reception while requiring low computational power. The result is robust
data communication over the network.
         The Inodes are usually placed within manholes and they communicate up to
street level. Initial tests showed that the Chasqui radio could only broadcast about 10
meters outside of a manhole. This distance could be greatly enhanced if the manhole
is converted into an antenna. The design of the manhole antenna was done by Dr. W.
Chappell (Purdue University). With this new antenna design the broadcast distance
increased to over 75 meters if the receiving node was placed 4.5 meters above the

The Ireland-Miami Pilot CSOnet

Pilot Site Background
        A pilot CSOnet was deployed and implemented in the CSO 22 service area in
South Bend, IN. South Bend currently has 36 CSO locations and a CSS that spans
50.1 km2 (Greeley and Hansen 2003). Each year, South Bend receives an average of
91.71 cm of rainfall from an average of 122 storms. A majority of the storms are
rather small, resulting from the climatic impact of nearby Lake Michigan (Greeley
and Hansen 1994).
        The CSO 22 service area spans 15.3 km2 (see Figure 4a) and is responsible for
17% of the city’s total overflow volume (Greeley and Hansen 1994). Due to its vast
size and the relatively small size of the interceptor line at that location, an overflow
will occur at the CSO 22 control structure if more than 2.5mm of rain falls in less
than 7 hrs (Greeley and Hansen 2003). The diameter of the trunkline at the outfall is
2.3 m (90 in), and the weir height is set at 1.42 m (56 in) above the trunkline invert.
        Within the CSO 22 service area, there is a 1.09 km2 section (65% commercial
and 35% residential) of separated sewers. The sanitary sewers are directly connected
to the CSS, but the storm sewers empty into a 47,540,000L retention basin, which
then drains into the CSS. Before CSOnet, the basin outflow was controlled by a 25
cm outlet pipe and a manual valve, which had been stuck half-open for several years.
Even prior to the valve being stuck, the basin was shown to be ineffective at storing
rainwater runoff during most storm events. The City of South Bend replaced the
manual valve with an actuated valve, which served as the prototype of the Smart

CSOnet Components
        This pilot CSOnet consists of 1 Gnode, 7 Rnodes, and 3 Inodes (see Figure
4b). The Gnode is directly connected to the Smart Valve at the basin, drawing its
power from the same source as the valve. It is mounted to a nearby antenna pole, and
from this position, the Gnode is able to have direct RF communication with the
surrounding nodes and connect to the Internet via a cellular connection. After every
time step, the Gnode uploads the data from the Inodes and the valve position via the
Internet to a specified website for the network administrator.
        Two Inodes are deployed in the basin itself, and both use pressure transducers
to determine the depth in the basin. The reason for this Inode redundancy is to further
minimize the risk of flooding in the areas surrounding the basin by providing a
failsafe. Both Inodes communicate directly with the Gnode via RF radios. Should
the two depth measurements ever differ beyond a given threshold, the Gnode selects
the higher measurement and notifies the administrator.
        The last Inode is deployed at the CSO 22 outfall, approximately 5.3 km away
from the Gnode. It monitors the depth of flow at the outfall using an existing level
sensor. Because distance between this Inode and the Gnode is too great for direct
communication, the Inode communicates with the Gnode via a series of Rnodes. The
Rnodes are mounted about 6m above ground on traffic signals along main roads,
which allows for very clear lines-of-sight. These clear lines-of-sight not only allow
the Rnodes to be placed farther apart, but they also enable a signal to make two hops
at once should one Rnode fail. When the outfall Inode tranmits a message, it is then
relayed by the Rnodes downgradient until the Gnode receives the message.

Control Scheme
        The overall goal of this pilot CSOnet is to maximize the storage ability of the
basin, thereby minimizing the amount of overflow at the outfall. Once the runoff is
stored in this basin, it must then be drained as quickly as possible once there is no
longer a threat of a CSO event in preparation for the next storm event and to prevent
anoxic conditions from occurring in the basin (USEPA 1999). CSOnet must also
ensure that depth in the basin does not exceed a predetermined level, which could
result in the flooding of the surrounding area. Should the depth in the basin ever
reach this level, the Gnode would begin draining the basin even though such an action
could result in a CSO event. A permitted overflow is deemed preferable to flooding
or property damage in the area surrounding the basin.
        Under normal (non-flooding) conditions, the Gnode begins releasing water
into the CSS at a constant rate of 56 L/s once the depth at the outfall falls below
0.76m. A constant, controlled discharge rate minimizes the impact of the additional
flow on the CSS and downstream structures (USEPA 1999) and allows the Gnode to
predict the effect of the additional flow on the flow depth at the outfall. To keep this
rate constant, the Gnode must continue opening the valve further throughout the basin
drainage process to account for the decreasing head in the basin. If the depth at the
outfall exceeds 1.02m, the Gnode closes the valve until the depth at the basin falls to
safe levels again. This usually happens when a rain event begins while the basin is
draining. Once the basin is empty, the Gnode closes the gate in preparation for the
next storm event.
        Each morning (if a storm event is not occurring), the Gnode opens and closes
the gate as part of a systems check.

Results and Discussion
       This pilot CSOnet was deployed during the summer of 2005 and has
functioned accurately during a number of storm events. Each time, water was stored
and discharged at the appropriate time in a controlled manner. The total cost of
implementing CSOnet, including the purchase and installation of a new actuated
valve, was approximately 55% less (US$26,000 compared to US$58,000) than the
estimated cost of providing the same control using a more traditional Programmable
Logic Controller. The long term maintenance cost of CSOnet has not been able to be
determined yet, but it is expected to be less than or equal to that of existing, more
traditional solutions.
         Figure 5 demonstrates how CSOnet functions during a typical storm. The
data shown is from a storm that occurred in South Bend on September 22, 2005, in
which CSOnet was responsible for storing 1,890,000L of runoff. However, a better
illustration of CSOnet’s storage improvement when compared with the passively
controlled system occurred during a November 1, 2005 storm. During this storm
event, 2.01 cm of rain fell during the span of 9.5 hours. In the CSOnet-controlled
basin, the depth in the basin reached 1.38m (see Figure 6), resulting in the storage of
6,020,000L of runoff. Moreover, no runoff entered the CSS while a CSO event was
occurring. A computer simulation was then run in which the same storm hit this
basin without CSOnet. In this scenario, the depth in the basin only reached 0.77m,
for a total storage of 2,870,000L. CSOnet increased the storage ability of the basin by
3,160,000L, or 110%. Furthermore, of the 3,160,000L that the passively controlled
basin released into the CSS, 3,060,000L was released while a CSO event was
occurring, causing an equal amount of combined wastewater to overflow at the
outfall. CSOnet prevented this additional overflow.
         The City of South Bend estimates that a liter of long-term storage potential is
worth US$0.80. With this in mind, CSOnet’s ability to improve a basin’s storage
potential more than pays for itself.

Interceptor Optimization

         Most sewer systems are composed by a number of basins and one or a few
interceptors. Each basin serves a specific area of the city. The flows in each basin are
aggregated into bigger pipes that connect each other in a tree structure. Typically, the
aggregated flow terminates in a single large trunk pipe that connects to the interceptor
line. The interceptor line collects the flows of all the basins as it runs towards the
wastewater treatment plant (WWTP). In the specific case of South Bend, the
interceptor pipe runs along the St. Joseph River. South Bend has 36 CSO outfalls,
located at the points where each basin connects to the interceptor (see Figure 7). The
outfall structure is usually simple, consisting of a weir structure and a throttle line that
connects the basin with the interceptor. The throttle line and weir maximum level
have been designed so to allow a fixed maximum flow into the interceptor line. If the
flow is greater than this maximum flow per basin, the hydraulic head level at the weir
will be higher than the weir itself causing a CSO overflow event. Since all the weir
levels and throttle line diameters are fixed, they had to be designed assuming the
worst case scenario of all CSO points at maximum flow. That is, rain is assumed to
fall uniformly over the entire city exactly at the same time. This is seldom the case.
Typically, rain storms sweep the city and therefore the different service areas are
affected at different times. Not only is precipitation time varying, but it is also non-
uniform. That is, the intensity of the storm varies greatly depending on the area. As a
result, extra capacity in the interceptor line is available for those basins more affected
by the storm event. The city has identified that the first step in reducing CSO
occurrences is to balance the load into the interceptor.
        For example, in March 2005 a heavy rain event that greatly affected the south
part of South Bend generated a flow that was slightly higher than the threshold value
for a period of 2 days in CSO22 (the storm event had duration of 4 hours). Even
though the interceptor line had capacity to convey the excess flow (the rain event had
stopped several hours earlier), CSO22 was still discharging into the river. If the
sensors downstream from CSO22 had determined that the interceptor had extra
capacity, CSOnet would have allowed the extra flow from CSO22 to enter the
interceptor instead of releasing it into the river. Currently, only 5 outfalls have some
kind of monitoring device to measure the amount of CSO flows directed to the St.
Joseph River.
        A more effective system than the current weir diversion system is to
dynamically adjust the flow into the interceptor at each outfall using CSOnet so that
the flow in the interceptor is optimized (i.e., when the flow in the interceptor is equal
to the maximum capacity of the WWTP). For such a system to be utilized, each
throttle line must be replaced with a larger line fitted with a Smart Valve. A Gnode
controls each Smart Valve and gathers flow depth data from the outfall and the
interceptor (see Figure 8). It then uses this data, as well as data from other Gnodes, to
determine the existing capacity in the interceptor line and then how much flow should
enter the interceptor at that outfall to optimize the overall interceptor performance.
At this point, CSOnet sets the maximum depth at each manhole equal to the highest
crown height of the pipes that enter the manhole. Soon, CSOnet will be able to adjust
this maximum height (i.e., allowing for surcharge) if this is deemed advantageous and
        Currently, the goal of control algorithm is only to minimize the total amount
of overflow during a given storm, without taking into account where the overflows
occur. However, CSOnet does possess the ability to control where the overflows
occur for most storms, should that be necessary to minimize the environmental
impact of the CSO events.

Results and Discussion
       A series of computer simulations have been run using the South Bend CSS to
determine the impact of CSOnet on minimizing the amount of overflow. Table 2
shows a list of the South Bend design storms used in this study. During the
simulations, it was assumed that the storms moved across the city from west to east at
33 km/hr.

  TABLE 2. Descriptions of the South Bend design storms used in this study
   Storm    Total Rainfall      Storm        Percentage of Storms with Total
  Number        (cm)         Duration (hr)     Rainfall Equal to or Less(%)
     1          0.61              11                       65
     2          1.24              11                       80
     3          2.03              13                       90
        The results for some key simulations are shown in Table 3. In addition to
decreasing the amount of overflow by 84%, CSOnet also reduced the number of CS0
events from 13 (which would occur with the existing system) to 1 in Simulation A.
In Simulation B, CSOnet reduced the number of CSO events from 11 to 1. CSOnet is
also very effective for larger storms, decreasing the amount of overflow by nearly
half and preventing 21,690,000L of untreated combined wastewater from overflowing
for a storm in the 90th percentile (Simulation C), and storms that affect the entire city
(as in Simulation D). This study focused on the impact of CSOnet on larger storms,
and based on this data, it is reasonable to assume that CSOnet will virtually eliminate
CSO events for smaller storms.

                TABLE 3. Results for the computer simulations.
                                  Existing Controlled Overflow
                                  System      System      Volume               Overflow
             Storm Description Overflow Overflow Decrease                      Decrease
                                         6           6           6               (%)
                                 (L x 10 )   (L x 10 )   (L x 10 )
              Storm 1 falls on
    A          areas north and     4.43        0.70        3.732                  84
                west of river
            Storm 2 falls on the
     B      western half of the    18.28       7.32        10.96                  60
            Storm 3 falls on the
     C                             47.21       25.52       21.69                  46
            western half of city
            Storm 2 falls on the
            western half of the
    D         city and Storm 1     40.63       29.39       11.24                  28
            falls on the eastern
               half of the city

        In addition to reducing the amount of overflow, CSOnet also fulfills five of
the EPA’s Nine Minimum Controls. In the process of maximizing flow to the
WWTP for treatment, CSOnet monitors each outfall to effectively characterize CSO
impacts and the efficacy of CSO controls. This also enables the City to properly and
accurately inform the public of CSO events. CSOnet automatically performs
preventative maintenance through the self-flushing effect of daily Smart Valve tests.
Lastly, CSOnet effectively prevents dry weather overflows by automatically detecting
and compensating for unexpected additional dry weather flows and by preventing the
throttle lines from clogging.
        Although interceptor optimization alone may not fulfill all of the EPA’s
standards (i.e., decreasing the amount of overflow by 85%) for most municipalities, it
can be an integral part of any Long Term Control Plan. In addition to interceptor
optimization, CSOnet can be used to maximize the storage capability of in-line and
off-line storage areas, thereby minimizing the flow in the trunklines. With the data
CSOnet provides, the municipality is also able to determine where additional storage
must be added or where sewer separation must occur with greater certainty,
eliminating any unnecessary construction and minimizing CSO abatement costs.


         EmNet, LLC, the City of South Bend, the University of Notre Dame, and
Purdue University have developed an embedded sensor network solution to the CSO
problem called CSOnet. CSOnet was proven to be an effective, robust, and cost-
effective RTC system during the pilot demonstration, which increased the storage
ability of a retention basin by up to 110% and cost 55% less than conventional
solutions. In a series of computer simulations, CSOnet was able to reduce the amount
of overflow by up to 84% through interceptor optimization and significantly reduce
the number of CSO events. The implementation of CSOnet at each outfall also
fulfills five of the NMCs, including the prevention of dry weather overflows. CSOnet
is designed to dovetail into any existing infrastructure and to work in conjunction
with other abatement measures to fulfill all of the EPA’s CSO standards in an
effective, timely, and inexpensive manner.


        We wish to thank the Indiana 21st Century Fund, which funded the research
and development of CSOnet. We also wish to thank the City of South Bend, IN, in
particular Mayor Steve Luecke, Gary Gilot, Jack Dillon, and Patrick Henthorn, for
their continual support and assistance. We thank Drs. Jeffrey Talley and Michael
Lemmon of the University of Notre Dame and Drs. William Chappell and Saurabh
Bagchi for their technical contributions. Finally, we thank Greeley & Hansen, LLC
for providing us with the trunkline flow data for the design storms used in the

Works Cited

Colas, H., Jolicoeur, N., Pleau, M., Marcoux, C. E., Fields, R., andStinson, M. (2001).
       The choice of a real time control strategy for combined sewer overflow
       control.” Proceedings of Novatech'2001, 4th International Conference on
       Innovative Technologies, Lyon-villeurbanne, France, June 25-27, 2001.

Greeley and Hansen, LLC. (1994). Combined sewer overflow control study, South
       Bend Department of Public Works, Division of Environmental Services,
       South Bend, IN.

Greeley and Hansen, LLC. (2003). Stream reach characterization and evaluation
       report, South Bend Department of Public Works, Division of Environmental
       Services, South Bend, IN.
Heinz, S., and Schultz, N. (2006). “Milwaukee case study in example evolution of
       sewer controls.” World Environmental and Water Resources Congress,
       Omaha, NE, May 21-25, 2006.

Lawson-Fisher Associates P.C. (2003). Stormwater management master plan, South
      Bend Department of Public Works, Division of Environmental Services,
      South Bend, IN.

Pleau, M., Colas, H., Lavallee, P., Pelletier, G., and Bonin, R. (2005). “Global
       optimal real-time control of the Quebec urban drainage system.” Env.
       Modeling and Software, 20(4), 401-413.

Ruggaber, T. P., Talley, J. W., and Montestruque, L. A. (2006). “Using embedded
      sensor networks to monitor, control, and reduce CSO events: A pilot study.”
      Environmental Engineering Science. Accepted.

United States Environmental Protection Agency (USEPA). (1999). Combined sewer
       overflow technology fact sheet: Retention basins, Office of Water,
       Washington, D.C.

United States Environmental Protection Agency (USEPA). (2004). Report to
       congress on impacts and control of combined sewer overflows and sanitary
       sewer overflows, Office of Water, Washington, D.C.
  Figure 1. CSOnet uses a widely distributed embedded network with specific
                      control points to control the CSS.

  CSOnet Decentralized Technology                     SCADA-Based Centralized Technology
 Multi-hop          No communication     Robust        Centralized     Dependent on comm.     Robust
mesh network          infrastructure   decentalized     network           infrastructure    decentalized
                                         control                                              control
                      Multi-hop                                            Large central
                     mesh network         Dense                              computer           Model to
 Low power                              temporal       Long range                           compensate for
   radio                                  spatial     radio SCADA                             lack of field
                       Low cost                                             Expensive
                                        feedback           RTU                                    data
                       hardware                                             hardware

   Battery              Low cost          Large       External power        Expensive           Small
  operated             installation    deployments        supply            installation     deployments

               Figure 2. Technology implications of the decentralized CSOnet control
                  system and the centralized SCADA-based control system.
    Figure 3. CSOnet components

(a)                              (b)
  Figure 4. CSOnet pilot deployment
                           September 22, 2005 Storm Event Results

     Depth at outfall

     Flow in trunkline (MGD)
     (from two sensors)

                     1                  2                       3        4
     Depth in basin (ft)
     Valve position (%)

     Rainfall (in)

Figure 5. Results from a typical storm event in the CSO 22 service area. In box
1, the Gnode opens the valve to release water stored during a prior storm. In
box 2, a storm event began during this draining, raining 0.41 cm in one hour.
This storm caused the flow depth at the outfall to increase. The Inode at the
outfall signaled the Gnode of this increase, and the Gnode closed the valve before
the CSO event began. In box 3, the storm event ended, and then the CSO event
ended. The Gnode then began releasing the stored water at a constant rate of 56
L/s, gradually opening the valve further to account for the decreased head in the
basin. In box 4, the Gnode opens and closes the gate as part of a system check.
Figure 6. Comparison of basin depth vs. time for a CSOnet controlled basin and
                         passively controlled basin.

               Figure 7. South Bend’s CSO basins and outfalls.

                                                         or Traffic Signal
Manhole Cover

                                                         Stilling         Line
                             Actuated Valve              Well
        Stilling                                                         Combined
        Well         Enlarged                            Weir            Sewer
                    Throttle Line                                        Trunkline

        Sensor                      Interceptor Line

  Figure 8. Configuration of a CSOnet controlled outfall.