SIEMENS

The ancient Greeks, Romans, and Assyrians supplied their major cities with water piped in from distant springs. Canals made of precisely shaped stones enabled these urban populations to survive. Canals had covers to minimize losses due to evaporation, and the Romans used concrete to prevent leakage. But the engineers of antiquity were unable to detect hidden leaks — and that’s a problem we still have today. Even in a modern city like London, millions of liters of water are lost to leakage every day. “That’s due to the dilapidated water mains from the Victorian era,” says Martin Geiger, Freshwater Director at the German section of the World Wide Fund for Nature.

According to official sources, London loses 217 liters of drinking water daily at each joint of its 30,000-kilometer-long water supply network. If the losses could be reduced by only one percent, 224,000 more people could be supplied with drinking water. Each leak is small and inconspicuous, but collectively the losses are enormous. A teaching manual reports that at a pressure of five bar a tiny leak only one millimeter wide lets 58 liters of water per hour seep into the ground. And it may be a long time before such a leak is discovered. Depending on how a water mains operator is organized, it may take three to six months.

“Leakage is an unsolved problem, especially in southern countries and when pipes are old,” says Geiger. In countries that have plenty of water, the losses cost “only” money and energy, because pressure-generating pumps require large amounts of energy. But in dry regions the leaks are a threat to life and health. According to the UN, over a billion people in the Third World suffer from inadequate supplies of drinking water. The World Health Organization reports that half of the world’s people live in regions at risk of drought. But water losses are only part of the problem. Leaks also decrease water quality, because impurities such as microorganisms can penetrate pipes through the holes.

“In some water supply networks in India, up to 60 percent of the water leaks out on the way to users,” says Geiger. Unfortunately, finding leaks is a very expensive and time-consuming process. One reason for this is the fact that flow measuring devices are installed only in major pipes, if at all. What’s happening in the smaller pipes is discovered only by accident, when inspectors use a listening device to detect the difference in sound between a leak and a normal diversion of the flow. Another reason is that there is seldom a network-wide assessment of measured values.

All of this could soon change thanks to a new system that listens to water flows and automatically detects anomalies, thus helping to minimize expensive losses. Working closely with their counterparts at Siemens Corporate Technology (CT) in Munich, researchers at Siemens’ Nuremberg-based Industrial Automation Division have developed a solution for this problem. “We offer our customers an electrotechnical package for managing their water supply, ranging from the tapping of a freshwater source to water distribution, management of the water supply network, and disposal of the wastewater,“ explains Dr. Andreas Pirsing, who is responsible for portfolio management at the Water & Wastewater business unit. “But thanks to our new SIWA LeakControl system, we can now find the leaks as well.”

To detect leaks, engineers divide a network section into zones whose rate of water use can be analyzed economically using only a few sonar-based ultrasound flow meters to measure incoming and outgoing flows. This flow measurement solution has already proved its value many times for customers ranging from oil refineries to wastewater authorities. It is easy to install, and it can be used on pipes of any diameter, made of any material. Another advantage is the fact that it can be attached to a pipe from outside. A service technician simply attaches the flow meters to pipes and connects them with measuring devices that transmit the resulting data by radio to a process control center. These data are then used to produce an overview of flow rates over time.

In the search for leaks, the measured values for night hours are important, because the amount of water that is used between 2 a.m. and 4 a.m. is typically minimal. Only the leaks in pipes continue to spew out water during this time. A sudden increase from one measured nighttime value to the next is a clear indicator of a new leak.

Siemens’ measurement algorithms require between one and two weeks to record normal use in a water pipeline network. After that, they can detect new leaks automatically. The most difficult step in this process was to develop sufficiently intelligent algorithms.

“A drinking water network is not a pipeline with clearly defined inflows and outflows,” says Roland Rosen, head of Siemens’ Modeling, Simulation, and Optimization (MSO) technology field. “We know exactly what’s flowing in, but our knowledge of where the water flows out is very vague. So we’re looking for deviations from a standard value that we don’t yet know.”

The laws of statistics provide the solution to this problem. If the network is sound, there is a characteristic distribution of measured values. This is similar to games of dice. “A leak has the same effect as loaded dice,” explains Dr. Jan Christoph Wehrstedt, a mathematician who works at MSO. The computer programs watch for unusual measured values every night. If they detect a distorted value distribution, an alarm goes off. In order to minimize false alarms, the software compares the results of several zones with one another. “If an unusual amount of water runs through all the pipes at night all of a sudden, that’s not a leak — it means there’s a soccer game on,” says Wehrstedt, who knows this from experience. That’s because at halftime everyone’s running to the toilet. The system does not take such peak use into account.

No Learning Phase Needed. “In addition to the distribution of water flow and pressure, we’ve also been using Monte Carlo simulations recently,” says Wehrstedt. These simulations are particularly complex statistical processes that can detect leaks very quickly. “To do this, we create a simulation of the water piping network. Our computers then calculate the pressure conditions in the water flow within the network using measured inflows and outflows. This enables us to simulate the water flow along the branches and at the points of confluence,” he explains. The result is a computer model that engineers can use to run experiment with.

Measured values are compared with sensor data, substituting coincidental profiles for the unknown behavior of consumers. The Monte Carlo simulation then generates fluctuations in the simulated values. These comparisons of measured and simulated values make it possible to skip the two-week learning phase.

In order to take a closer look at suspect sections of the network, the teams install additional sensors. This eliminates the need for expensive searches of suspect pipe sections using pickaxes and shovels. “The Monte Carlo algorithms also indicate old leaks — in other words, leaks that already existed during the learning phase, which we would not have found by means of simple statistical processes,” adds Wehrstedt. Engineers recommend that network operators have their entire pipeline network checked once every quarter. A conventional PC needs about ten minutes to analyze a network that is approximately 500 kilometers long and has a few hundred nodes. “Mathematically speaking, the problem of ensuring a metropolitan area’s water supply is a lot like analyzing the outcome of an election,” says Wehrstedt. “In both cases researchers conduct a statistical analysis of the measured values and then estimate what kind of behavior should be expected in the future. However, we’re not predicting how a group of citizens will vote — we’re predicting how much water they will use in a certain branch of a network.”

Siemens engineers were able to test their system under real-life conditions in a major city soon after they had developed it. The leak seekers were soon rewarded. “Even though we were investigating a well-kept European water piping network, the very first run-through of our software revealed one or more leaks in every part of the city — which the network managers had not been aware of,” recalls Pirsing. Compared with the conventional periods between inspections, the test run of the SIWA LeakControl system immediately saved the provider several hectoliters of drinking water per day at each (repaired) leak. All the same, one of the city’s inhabitants does not have pleasant memories of this test run. He had illegally tapped the water mains for many years in order to fill his swimming pool. After the water network analysis was carried out, a team of emergency technicians was sent out to investigate. The thrifty pool owner received a huge water bill. Without the analysis he would probably have been able to swim in stolen water for decades.