Pictures of the Future      Spring 2008

Advent of an Invisible Army

Digital assistants are substantially increasing safety and efficiency in pipelines, wind parks, cement plants, andtunnels.

Customers expect products to function reliably, but ensuring that they do so under all conditions is anything but a trivial matter. Consider, for example, the gear train assemblies manufactured by the Siemens Drive Technologies Division for wind parks and for large-scale drives, such as those used in cement mills. In the case of wind parks, insurance companies insist on online monitoring, particularly if these facilities are in hard-to-reach areas such as the North Sea.

In such cases, Siemens uses an electronic condition monitoring system. In addition to continuously monitoring critical parameters and the functionality of the gear mechanisms, the system transmits results to the plant’s operator. Because of weather conditions, which may involve rough seas, offshore facilities are accessible only a few days per year. It is therefore all but impossible for an operator to get an up-to-date overview of the state of its equipment without an on-site condition-monitoring system.

Dr. Jörg Deckers is a diagnostics expert at the Siemens Condition Monitoring Center in the town of Voerde, Germany. "We do a detailed diagnosis of the vibrations and harmonics of the gear systems here," Deckers says. "This allows us to identify the smallest changes in operating characteristics, which signals the onset of damage at a very early stage."

The difficulty is that each individual gear has its own vibrational characteristics, and these vary continually during operation as a result of ambient conditions such as temperature and wind velocity, as well as oil level. Experts use warning and alarm limits to define the range in which the gear in question can still be rated as fault-free.

In the case of wind turbine gears, for instance, it often takes experts many hours      to tag between 400 and 700 characteristic frequency components with corresponding limit values. In the future, however, they will be supported by a new tool called a vibration diagnosis module (VDM)—learning software developed to the prototype stage by a research group at Siemens Corporate Technology (CT). The module combines several methods of machine learning for error analysis and prevention from the Siemens Machine Learning Library, a platform-independent software library (see Pictures of the Future, Spring 2006, Learning Software).

The VDM will be used in a system solution with Castomat, a general-purpose measurement acquisition and diagnostic system that can acquire data from a broad variety of sources and transmit it to VDM. In addition to frequency spectra, this data includes a large number of ambient influences in the interest of achieving the most precise possible classification of environmental conditions. Based on vibration and ambient data, VDM learns the state of gear mechanisms and then defines the limits which, when exceeded, indicate the first changes resulting from wear or defects. The warning and alarm limits that would otherwise be entered manually are thus set automatically with VDM.

On-Site Monitoring. Learning software systems based on the Siemens Machine Learning Library can also be applied to the field of oil and gas extraction. In Russia, for example, many of these resources are located in difficult-to-reach areas far from any infrastructure, where temperatures can reach -50 °C in the winter and 40 °C in the summer, when the permafrost at the surface thaws and turns the tundra into a wasteland of mud and swamp. With such adverse conditions, it is scarcely possible to keep enough experts on-site at all times.

In view of this, the pumps and generators used in oil extraction, and the compressors used in transporting oil and gas through pipelines, should be monitored remotely. Indeed, procedures used in wind parks can be applied here as well. "Because of the extreme weather conditions, system vibrations are constantly changing. That makes it all the more essential to have reliable analysis of the frequency spectra of the kind provided by our VDM," explains Bernhard Lang, head of the Fault Analysis and Prevention research group at CT in St. Petersburg, Russia, where these procedures were developed.

In addition, pipelines must be monitored for damage resulting from earthquakes, theft and sabotage. Sensors suitable for this would be able to react to decreases in pressure in pipelines, and to knocking and digging noises. "Instead of cable-based sensor systems of the kind used in the past for pipeline monitoring, we’re working on solutions for wireless, self-organizing sensor-actuator systems," says Project Leader Dr. Rudolf Sollacher from the Learning Systems center at CT in Munich. Such systems are expected to be used not only for oil pipelines and platforms but also in building services automation and process control. In the case of oil pipelines, distances of 25 to 40 km must be bridged between valve stations. To achieve this goal, an independent, energy-stingy power supply is needed.

"This conflict can be resolved by placing individual radio sensors between the valve stations at intervals of about 100 m and passing the messages on from one sensor to the next," says Sollacher. These small helpers must react reliably and quickly, especially in the event of alarm messages. And they must do so under environmental conditions that are sometimes extreme. "Our development work is moving in two directions in particular: energy efficiency coupled with zero maintenance, and self-organization of sensor nodes," says Sollacher.

Sollacher’s research team is also hoping to implement solutions that allow the sensors to put parts of their hardware, such as the radio portion, to sleep as often and as long as possible, which saves energy. The system’s microprocessor, which likewise uses little power, can automatically compress data delivered by sensors in the vicinity into diagnostic information. With this integrated intelligence, researchers want to "prevent the transmission of unnecessarily large amounts of raw data," Sollacher explains. "Then only critical events, such as unusual vibrations of the pipeline, would be reported to the valve stations, and from there onward to a monitoring center."

To ensure reliable communication, the sensor network must be well coordinated. For example, the radio sensors must know when and on which channel they will transmit data, and when they can put themselves to sleep without losing contact with their neighbors. Corporate Technology supplies self-organization solutions for this purpose (Pictures of the Future, Fall 2004, Sensor Networks), such as a completely decentralized allocation of radio channels for the sensors.

The sensors must also be able to independently determine their spatial position in order to localize events and data. "This is an important feature as it greatly reduces the demands on the operator team with regard to network management," says Sollacher.

Self-Organizing Networks. Reliable and energy-efficient wireless sensor-actuator networks are also the subject of a project called ZESAN, financed by the German Federal Ministry of Education and Research. "The topics dealt with include multi-antenna solutions for reliable radio transmission, extremely energy-efficient receivers for waking up the sensors, advanced self-management for updating software components on the sensors, automatic optimization of network operation, and data integrity," says Sollacher. Siemens is participating in this project through CT, the leader of the consortium, and through Siemens’ Industrial Solutions Division, which intends to integrate hardware components for radio sensor networks into its platform concept. In addition to the monitoring of pipelines, oil platforms, and containers, application scenarios being con-sidered by ZESAN’s partners include industrial process automation and energy use metering in buildings.

The market for these self-organizing sensor-actuator networks is still emerging, but Sollacher is not worried about the competition. "In building services automation, Siemens already markets wireless radio sensor networks such as the Apogee Wireless. What’s more, our first product solutions are being developed for manufacturing and process automation."

Digital Tunnel Keeper. Safety has a top priority in road tunnels as well, where accidents can have dire consequences. State-of-the-art safety systems are being developed for such uses—for example, in the "Citytunnel" in the Austrian town of Bregenz. In late 2007, Siemens equipped this 1,311-m-long, bidirectional tunnel with its Siveillance solution, a modern system of video sensors with intelligent camera surveillance.

Siveillance is a modular solution that can  detect fire and smoke. It also identifies and reports stalled vehicles, traffic congestion and traffic jams. "Siveillance even detects the passing maneuvers of individual cars, which is especially important for this single-tunnel environment," says Siemens Project Manager Christian Gobiet, enthusiastically. "Or if a car continues through on a red light, the traffic on the other side has to be given the red light until the driver who made the mistake has left the tunnel," he adds.

Citytunnel was outfitted with 17 fixed and five swivel-tilt cameras that deliver their analog data to a control center in Weidach, Austria, where it is converted to digital form and saved as compressed files in the MPEG-4 format (Pictures of the Future, Fall 2006, Video Surveillance). At the same time, the data is delivered to the Siveillance system for analysis. If a critical event is detected, an alarm is automatically triggered at the monitoring center in Hohenems, which is staffed around the clock, in contrast to the control center. Says Gobiet: "Potentially dangerous situations are automatically highlighted for the operators. That makes their job easier, and allows them to give their full attention to assessing the situation and making decisions."

Unlike human operators, Siveillance never tires or gets bored; the system learns what "normal" is, and triggers a warning message only in the event of an anomaly. That may sound simple enough, but this achievement is the result of more than 25 years of Siemens research in the field of video sensors.

In this context, Klaus Baumgartner of Siemens Building Technologies is directing an especially ambitious project in Karlsruhe, Germany. "We want to link the evaluation results of video cameras in intelligent ways," he says. "That means allowing cameras to zoom in on objects, and using several cameras to automatically monitor people." Known as NOOSE (Network of Optical Sensors), the project will be  based on the use of the Internet Protocol (IP) and on a large increase in computational capacity. These developments will make a range of new applications possible. For example, a traffic light equipped with intelligent IP cameras would be able to manage itself locally. It could, for instance, extend its green phase when a car approaches and no traffic is detected on the other street.

Digital assistants are being used in new fields all the time, whether it be condition monitoring, self-organized sensor monitoring, or intelligent camera technology. Increasingly powerful systems are automatically working together with one another and reaching independent decisions that they then present to human operators, who, of course, retain the ultimate decision-making authority.
                                                                                                                 Eduard Rüsing