Water provider finds that a fixed network for leak monitoring—and meter reading—makes economic sense for a historic Pennsylvania city’s aging pipe network.
Two technologies—acoustic leak detection and automatic meter reading (AMR)—are helping water utility managers automate the routine tasks of monitoring their systems for leaks and collecting usage data for billing. As each technology gains greater acceptance, any remaining hesitance to adopt them has a predictable origin: capital investment cost. But the experience of American Water, which services the city of Connellsville in southwestern Pennsylvania and invested in a regional fixed-network acoustic leak-detection/AMR system, demonstrates that it’s possible to make a financial case for both detecting leaks and reading meters with minimal human intervention—in one capital outlay. The case of Connellsville also demonstrates that fixed-network acoustic leak detection may make the most economic sense for water utilities serving communities with very old water infrastructure. The water service provider for the community, Voorhees, NJ–based American Water, is the largest in North America and serves about 17 million people in 29 states and Canada. Its relationship with the city of about 8,500 goes all the way back to the early 1800s, when the city was established. Through the next few decades, Connellsville boomed due to its emergence by the end of the century as the world’s largest producer of coke, a solid residue of nearly pure carbon produced by burning off the sulfur, water, hydrocarbons, and other impurities from bituminous coal used to feed Pittsburgh’s steel mills.
According to Dave Hughes, infrastructure engineer with American Water, the company operated about 40 miles of underground water pipe in the prosperous boomtown populated by several millionaires by the turn of the century. “We haven’t replaced much of that original 40 miles, so you can imagine that we’ve got a very old system and to see that it is leak prone, given its age, is really not a surprise,” Hughes says. American Water also operates 18 miles of pipe in addition to the original infrastructure, which it is periodically replacing according to its condition. This incremental approach to replacing water mains has its economic advantages, particularly since American Water also needed to invest in some means of assessing the condition of various areas in such a large infrastructure.
 |
Photo: Itron |
| Leak detection technologies continue to advance. |
Without some way to monitor the condition of an entire underground pipe network—particularly one that is aging like Connellsville’s—water service providers such as American Water have to figure out a way to utilize their limited resources for detecting leaks in a given area of a network. Traditionally, those utilities that are proactive enough to use modern technologies for this task have first used listening devices to detect the general vicinity of leaks and then pinpointed them using correlators. But this approach is so impractical from the standpoint of labor utilization and safety that it seems to make as much sense to wait for leaks to manifest on the surface.
An Itron, formerly Flow Metrix, fixed-network leak detection system used in Connellsville is centered around an MLOG sensor, a waterproof, battery-powered data logger that is permanently installed near a water service meter and records vibrations, stores vibration data, and transmits the data via radio signals to a server that processes the data to reveal a noise pattern over days and months. American Water opted to use a fixed network to detect leaks that relies on these data loggers for data collection because the volume of water being lost in Connellsville was adversely affecting the financial viability of the provider’s service to the community and the company needed a way to continually monitor the entire system.
A Need to Do Something
“We don’t provide the treated water for Connellsville; we buy that from a local municipality,” Hughes points out. “We pay more or less a wholesale price for finished water, which is, I believe, in the range of $2 per thousand gallons—that can be as much as 10 times higher than the cost we experience when we produce water in other systems. There are other places where the water is fairly pristine, the treatment costs for pumping it out of the ground are fairly inexpensive, and it’s down at the $0.20 level.
“Clearly, there was a need to do something in Connellsville. In that particular system, we were running in excess of 25% non-revenue water.” Hughes adds that if the community needed 1.3 million gallons a day on average with a leaky system instead of 1 million gallons a day, the additional 300,000 gallons a day multiplied by 365 days a year was costing American Water several hundred-thousand dollars a year. “In Connellsville, it was a very clear economic case that we could make and the system itself is not atypical of Pennsylvania,” Hughes says.
American Water tracks the water lost in a network such as Connellsville’s as “non-revenue water” and attempts to limit it as much as possible. As a division of a publicly traded company, the RWE Group, the amount of non-revenue water lost in an American Water network versus the amount the company is able to retain for use—called “retail water”—impacts individual investors. Additionally, the company’s priorities include minimizing the number of customer complaints about low pressure and defer the replacement of entire underground networks and water treatment facilities for as long as possible, so long as regular maintenance results in economically viable networks. This latter motivation for maintaining existing networks is certainly a long-term consideration, but minimizing non-revenue water could be considered intangible capital savings in the short term, notes Itron’s (formerly Flow Metrix’s) Director of Business Operations Paul Lander. These savings include deferment of treatment plant expansion (as less water needs treatment) and underground pipe replacement, as well as lower repair costs resulting from early intervention.
Hughes describes American Water’s previously used method of detecting leaks as reactive. For some leak-prone water networks, considerable manpower would be devoted to conducting leak surveys, and in others, conditions causing leaks would force water to the surface in a short period of time—a truly reactive situation. For less leak-prone networks, leak surveys sometimes would be deemed not worth the cost despite losses of retail water. In some situations, Hughes says, it’s possible to detect abnormalities in network operation; in these situations, conducting leak surveys on an as-needed basis makes economic sense.
“Demand management is where you break your system into smaller pieces,” says Hughes. “Our plan is to use that methodology where we think it’s practical. There are systems where not only can we meter the water but we can also lower the pressure, and, obviously, lowering the pressure also lowers the flow that might escape from the pipe over time. It’s a mixed bag.
 |
Photo: Itron |
| Some systems’ leak detection programs lower the pipe’s water pressure in addition to metering the water. |
“If we had indications that the amount of water that we were metering versus the amount of water that we were pumping had wide variance, that would be sufficient because there are systems where water simply does not come to the surface or takes a very long time to do so—it has to be severe enough to actually percolate to the surface. We have systems with subpressure systems and we might be able to identify which subsystem is actually having the problem—maybe a pump is running much harder than it used to run, for example. One thing we do as a company is look very hard at the non-revenue numbers from place to place. We are very aware of what system is running 20% non-revenue water versus the ones that are reading 10%—that helps us to focus.”
Conducting leak surveys as needed was not an approach that made sense for Connellsville, however. The relatively high incidence of leaks in such an aging underground water network meant that leaks would be common. Hughes and American Water started looking at available fixed-network leak detection technologies. “We found that the acoustic monitoring was getting increasingly sophisticated and companies were getting much better at understanding the dynamics involved, how far the sound traveled, what frequencies to listen for, and also a visual aid—it’s not a huge leap to provide that in a graphical form that could be recorded,” says Hughes.
 |
Photo: Itron |
| A graphic interpretation of leak detection results |
Still, this would be the first use of more or less constant leak monitoring technology via a fixed network. “Originally, the MLOG was designed to be kind of a radio device that would send out data as you drove by,” Hughes says. “But it certainly looked logical to us and to Flow Metrix that if you had a fixed network, if we had the information available every day, it would really give us better insight and we might actually be even better able to respond to leaks and for the high cost of the leaks that we had in Connellsville, we might be able to get out there even faster. With this system, if we had a next-day leak, we would know it because the sound history would reveal that to us.”
Based on good results they had experienced in other communities, Hughes and American Water had also considered using AMR in Connellsville. AMR, coupled with the need for whole-network leak monitoring, made piggybacking leak-detection monitoring onto fixed-network AMR a good fit for Connellsville. American Water contacted several technology providers, and Itron and Hexagram Inc. showed interest in coordinating their efforts to provide combined leak detection and AMR over a fixed network.
Separating Leak and AMR Data
Itron had already envisioned an AMR interfacing capability and created an AMR-compatible protocol for transmission of its leak-monitoring data to a Web server hosted by parent company Itron. The two providers also worked out the number of data transmissions that could occur in a day. In regard to AMR data, “Instead of doing reads probably twice a day, we’re probably doing four reads a day, but that’s sufficient to convey the information that goes from our 500 transmitters to nine collectors that are scattered throughout the Connellsville area,” Hughes says. “The AMR portion of the system has performed very well from the start.”
 |
Photo: Itron |
| Meter transmission connects data logger units to water meters. |
The Hexagram system uses Meter Transmission Units (MTUs) that connect to water meters and Data Collector Units (DCUs), network devices that receive, process, and store meter reading information. The DCUs transmit these data to a central computer for processing. “It provides the meter data and gives us a file that we can input into our billing system on the day we designate,” says Hughes. “The other thing it does is it pulls out the information that Flow Metrix has assembled and put through the transmitter and brings it out into a series of files that can be downloaded into a computer equipped with the Flow Metrix software.” The data collected by the MTUs are “packaged” in such a way as to facilitate the separation of leak-detection and AMR information into distinct files.
“We’re sending one condensed set of information through the AMR system in a small meter-reading-like message,” Lander notes. “Our Web server collects data from all the MLOGs in the world every day and determines whether or not the leak status of each MLOG should be red, yellow, or green” indicating high, moderate, or low probability of a leak.
“Once the Itron main server has made these determinations, then the individual utility users can get the information in one of two ways: They can either come to our Web site and log online and browse the Web site and see results for all of their utility, or they can receive an e-mail or a regular text message that alerts them to new red loggers.
 |
Photo: American Water |
| One benefit of fixed networks is the availability of data on a daily basis. |
“Many of the leaks that MLOG finds are at the beginning of their lives, which is a great time to find them because you don’t lose all of the water over the next few years and you don’t risk all of the liability or damage,” Lander says. “When they’re at the beginning of their life, they might be inaudible and invisible, with no water present during the day because at nighttime, when there’s no flow, the pressure is highest and it may cause the leak to appear. It may cause a crack or a small pinhole in the pipe to open up and admit water, and then when people start using water in the morning, that crack may close and then for all intents and purposes that leak is not findable.”
The method of finding leaks using this fixed network occurs in reverse order of “conventional” leak detection, in which individuals use listening devices to get a general idea of where leaks are located and then deploy correlators to pinpoint the leaks. This fixed-network method allows for the varying behavioral characteristics of small leaks between nighttime, when water pressure is highest and extraneous noise lowest, and daytime, when staff is more typically available to perform leak maintenance. The listening devices provided by Itron calculate numerical values where the higher the value, the closer the device is to the leak along a pipe’s horizontal axis. This technology provides an independent confirmation of the data provided by the fixed network.
 |
Photo: Flow Metrix |
| Depending on the type of leak present, a monitored pipe can yield different results. For example, leaks may only appear when there is no water flow, when pressure is highest. |
“What the MLOG will do is to record several hundred times a night, looking for the quietest period on the theory that you can’t turn the leak off,” says Lander. “The sound level at the quietest time of the night is the best representation of the sound of any leaks that are present in the area. We measure the frequencies and the quality of the sound and then we turn that information into a message that is sent out through the AMR system. The MLOG sensor only knows what kind of quality of sound is present at the quietest time of the night. It doesn’t know if that is a leak or if that is an ice machine in the basement or if that is a pump station down the road; it doesn’t know those things. This is where MLOG online comes into play; MLOG online will take raw information at the quietest time from every MLOG sensor, and it is the Web software on the PC server that makes a determination of whether the sensor is hearing a leak or something else, or nothing at all.”
A key to ensuring the validity of the fixed network’s leak data is the ability of the Web server to group historical data from different data loggers and determine if a leak is actually present. The data-analysis function of the system demonstrates that computers are superior to humans for the purposes of making a high volume of mathematical calculations very quickly.
“The server software knows which MLOG sensors are neighbors and which MLOG sensors are located in similar environments—for example, cul de sacs or on a street with a large main,” Lander says. “So it’s able to form large populations of sensors and look for recordings that are unusual. The other thing the server software is able to do is look back down through the history of recordings from each sensor and use each sensor as its own historical reference so it can ask the question of how has the sound at this location changed over the last couple of days, the last week, the last month, the last season, or even the last year. It uses all of that information to determine whether it should assign a status of red, yellow, or green.”
Lander reports that after the fixed network was installed in Connellsville in June 2005, American Water found about 40 leaks in the first full month of operation, compared with only four and 12 during each of the preceding two months. “We were finding that the local people using conventional leak noise correlators and listening devices weren’t confirming any leaks that the MLOG found and it was a little worrying at first,” Lander says. “We were thinking, My gosh—is the MLOG hearing things that go bump in the night, or is it finding leaks that aren’t detectable during the day?” In this instance, Itron used conventional correlators to verify the precise location of the roughly 40 leaks and confirm the accuracy of the fixed network.
 |
Photo: Flow Metrix |
| Automated leak detection systems can result in huge economic benefits. |
Hughes argues that the return on investment provided by fixed-network leak detection helps make the case for fixed-network AMR. “We felt that the investment was justified and, in fact, I think that what may happen is that because of this leak detection ability being fairly economical, it will probably help push the business case for AMR.” American Water determines the payback period for major capital investments such as a fixed network, Hughes adds. In determining the payback for AMR, the company took into consideration the savings that would be derived from more accurate readings and fewer customer complaints and discovered that the fixed-network AMR payback period would still be about 12.5 years. However, in determining the payback period for a fixed network for both leak detection and AMR, “If we factored in savings for Connellsville—and Connellsville may be unique because of the high cost of water—the payback is two and a half years,” Hughes points out. “All of a sudden, something that might not have even gone onto your economic horizon becomes a can’t-miss situation.”
A fixed network can allow a utility to schedule leak repairs in advance when problems are minor and get a handle on maintenance costs. “We talked about saving water, but that’s not where it stops,” says Hughes. “If we find the leaks early enough, we can schedule a leak repair. Instead of paying our guys overtime to go out on a winter night, messing around with ice, they’re out there on a regular fall workday under more or less ideal conditions, making the repair. That’s a huge cost savings.
“Another thing we’re investigating—we haven’t really quantified it at this point, but the suspicion is that when we find them early enough, we’re not talking about putting in huge amounts of modified material or backfill material into the trench because it really hasn’t run that long and washed the soils away; it hasn’t destabilized the pavement above. The suspicion is that we’re also going to save in the restoration cost of the repairs.”