Sensors Support Integrated Information Systems

April 1, 2005
Advances in sensors, transmitters and networking make real-time information available throughout an enterprise

Sensors are just as important to Bayer AG’s customer service as its fancy inventory and supply chain management software. From its largest plant in Leverkusen, Germany, the giant chemical and pharmaceutical company relies on radar, ultrasound and pressure sensors to monitor the levels of the remote storage tanks that it maintains at its customers’ plants. Based on reports from these sensors, the plant’s management software not only orders supplies and schedules production to satisfy demand automatically, but it also dispatches trucks to refill its customers’ tanks with the appropriate chemicals as needed.

The ability to provide this service was the result of a project that Bayer had initiated to streamline its supply chain, consolidate deliveries and get closer to its customers. Its backbone is an integrated manufacturing and delivery system designed and built by ATplan GmbH, a German systems integrator based locally in Leverkusen. The success of the project adds Bayer to the growing list of manufacturers learning how to use today’s sensor, transmitter and networking technologies and communications standards to integrate and automate the businesses with real-time information.

To streamline the initial implementation and later upgrades and expansions, ATplan chose interfaces and technologies that would reduce the plant’s dependence on proprietary technology. The sensor interface, for example, was the modular Snap Ethernet I/O from Opto 22, a remote monitoring and networking specialist based in Temecula, Calif. The Ethernet interface connects the sensors to conventional and wireless modems and to the Internet, which in turn link the sensors to Bayer’s enterprise information network. An Oracle database collects the data, puts a time stamp on it, and packages it for the various applications that need it, including Bayer’s SAP supply chain management software.

Integrating sensors requires three components: a sensor interface on the plant floor, some means for transmitting the information, and networking technologies to carry it to the appropriate database or application. “Sensors have been in place for quite a while, but you need some sort of interface [to make it work],” explains David Crump, a spokesman for Opto 22. “So we’re seeing more intermediary hardware [become available] for connecting enterprise databases and SQL (structured query language) servers with the process equipment.” Sometimes one device suffices, and at other times, several are necessary.

“The trend is toward aggregating information and putting intelligence behind it before feeding it to a higher-end application,” adds Mike Prowten, at Lantronix Inc., a vendor of networking devices and services from Irvine, Calif. “Automation companies are connecting equipment that was not normally network-enabled before and are providing easier access to the data and control of that equipment.” The two-step process of making that happen, he says, is putting the sensor information onto the network and pushing the intelligence as far down the hierarchy of devices as possible.

Smart links

Because of the relatively cheap computing power available today, an intermediate device gaining a measure of popularity in the process industry is the programmable automation controller (PAC), which brings together features of a programmable logic controller (PLC) and a personal computer (PC). The combination preserves the reliability and ruggedness of a PLC and adds the open systems computing and communications technology found on a PC. PACs can connect to a variety of sensors, analyze their signals in real time on the factory floor for immediate use there, and upload information to a central database. “One of the primary benefits of using PACs is the communications protocols that enable easy connectivity,” says Rahul Kulkarni, product manager, industrial data acquisition and control, at National Instruments Corp., a manufacturer of PACs based in Austin, Texas.

For example, PACs can recognize smart sensors that have their “personalities” defined according to the relatively new IEEE 1451.4 standard, promulgated by the Institute of Electrical and Electronics Engineers. Manufacturers of this class of sensors encode a set of coefficients describing a sensor in an electrically erasable programmable read-only memory (EEPROM) chip. The PAC can search for these coefficients and read them automatically, much like PCs do with plug-and-play hardware. Consequently, not only is changing sensors simpler than in the past, but keeping them calibrated also is easier.

“Our PACs even can have an embedded Web server in them, so you can transmit data directly in XML (eXtensible Markup Language) and talk directly to an ASP.Net page (provided by a third-party application service provider) via TCP/IP (transmission control protocol/Internet protocol),” says Kulkarni. “It opens up a whole variety of communication protocols that you cannot even imagine with a PLC.”

Such is the case on a glass press that makes Pyrex products. Displacement sensors, pyrometers, and other sensors report their various measurements to the National Instruments PXI PAC installed by Data Science Automation, a systems integrator based in Canonsburg, Pa. To ensure that the press produces quality products, the operators then monitor the data and review the analyses performed by the LabView software running on the PAC. “If the glass is not exhibiting the right properties, an operator then could take corrective action,” explains Kulkarni.

The operators monitor the press in two places. During setup and startup, they monitor temperatures in the molds, velocity of the hydraulic ram, and the timing of the events at the machine, adjusting the process as necessary. Once the machine begins producing good parts, they then go to the control room, where they monitor production from panels. They view screens that contain more details than are available on the floor, as well as show results of various analyses performed by LabView.

Not only does the software allow supervisors and engineers to access the data from their desks in other parts of the facility, but its built-in Web servers also permit supervisors to control the machine whenever their intervention is necessary. Periodically, a connectivity tool in the software sends the data to an Oracle database for the enterprise management system so that others can use the information for various analyses and reports.

Besides more intelligence throughout networks, another important enabler for integrating sensors tightly into business management systems is the standard networking technology driving the Internet. “IP [Internet Protocol] is probably the most ubiquitous networking technology,” notes Crump, at Opto 22. “It’s the one that the Internet is based on and the one that a lot of applications and other protocols rest on.”

A ramification of the cheap computing power and the communications standards available today is the ability to build Web servers into sensors. This development has the potential to spark a second wave of Internet fervor. “The first wave of the Internet was all about PCs talking to PCs and connecting people to people through PCs,” explains Ed Yenni, president, Beckett LogiSync LLC, an integrator in Elyria, Ohio, that upgrades embedded computers with current networking technology. “The second wave, which we are just scratching the surface of right now, is about devices talking with devices.”

For him, the key technologies are integration standards such as Ethernet and a 32-bit processor capable of speaking TCP/IP available for less than $10 from Waltham, Mass.-based NetSilicon Inc., a Digi International company. Sensors fitted with one of these processors, therefore, can host Web pages on a corporate intranet, log data on them, and send e-mail messages containing that data to the addresses and at the frequency programmed on the pages. “The system can send you an e-mail with, say, logged pressure data once an hour or once a day,” he says. “Or it can transmit an alarm immediately if a reading varies beyond certain control limits.”

Eight tons or 400 eggs?

The ability to speak the language of the Internet makes integrating sensors into a local area network much easier today than in the past. An integrator spends much less time interpreting the 4- to 20-mA signals coming from sensors. “A reading of 6.3 mA could mean eight tons, 400 eggs, or 16 degrees,” notes Yenni. Now the local processor attaches the appropriate units, as well as other interpretive information, and sends them on the network as information objects over a common communications path, such as e-mail or the Web, using something like XML.

An application could be a service agreement that an original equipment manufacturer of, say, pumps, offers for the pumps it has in the field. The processor on each pump would collect data about the bearings from accelerometers, interpret the data, and send status reports to the central office. Based on the trends in the reports, a service representative could call the end user and say, “We noticed a problem developing in one of the pumps that we sold you and would like to send you a replacement part and dispatch a technician to install it.”

Another application of this technology is in the gas detection system that a beverage producer bought from MSA Co., a Pittsburgh-based manufacturer of safety equipment. The company decided to integrate the new gas sensors into its information network when it upgraded the gas detectors that it uses to ensure that no ammonia leaks from its refrigeration units into the product or the plant. It wanted to be able to give its engineers access to the data from anywhere so they could troubleshoot problems quickly, as well as conduct process improvement projects and make the data available to auditors from the Food and Drug Administration (FDA).

In the past, integrating the gas detectors was an expensive proposition because the number of sensors was small. The software alone can cost $25,000, and a programmer must map the sensors’ output signals to the devices on the network, according to Kevin McKeigue, product group manager for MSA’s Instrument Division. Although this approach works well and makes economic sense when the number of sensors being connected is large, it tends to be too expensive for integrating the six to eight sensors typically required in many gas detection systems.

The IP connection established by Beckett LogiSync technology, however, changed the economics of integration significantly. The cost of integration on the beverage line was about a tenth that of conventional methods. Now, the Web page on the digital bus network installed by MSA queries each sensor address and provides instantaneous updates on the gas detection system. It then sends the data periodically as a File Transfer Protocol (FTP) file to a hard drive on the plant network, rather than simply dumping the data into chart recorders for FDA audits, as the old system did. “The Web server shows a two-hour or 20-minute graph to give you an idea of what occurred over that time within the process,” says McKeigue.

The system also can send reports and alarms by e-mail. “Someone within the facility can get an e-mail every day containing what the system read every minute for the day,” says McKeigue. If a sensor actuates an alarm, the system can transmit a text message explaining the problem to the appropriate technician’s cell phone to expedite corrective action.

Smile, you’re on camera

Many manufacturing facilities are incorporating video-based sensors into their enterprise networks. MSA, for example, uses video technology to complement its gas sensors. By clicking on the icon, an operator or manager in the control room can look at the area of the facility reporting a leak, determine the nature of the problem and whether an employee is in trouble, and organize an appropriate response. Access is through a camera linked through an enabled IP address associated with the camera.

More than likely, however, cameras connected to computer networks are checking products and processing the data locally to provide quality control data. These applications need quite a bit of processing power locally and tend to be more like the one at a pharmaceuticals plant in Puerto Rico. There, the company uses both a photoelectric sensor and a vision sensor to inspect blister packets of pills before the packets are sealed, cut and packed into boxes, according to Dan Holste, director of vision products for Banner Engineering Corp., in Minneapolis.

When the photoelectric sensor detects a packet, it actuates a Banner PresencePlus Pro vision sensor, which checks whether each blister pocket contains a pill, whether the pill is the right shape and color, whether it is intact and unbroken, and whether the pocket contains foreign matter. As the packets heads down the line, the sensor tells an automatic picker which ones contain flaws and must be taken off the line to prevent them from being packed and shipped.

Holste notes that vision sensors are a good choice for detecting multiple features on a product. “A vision sensor uses digital imaging rather than a phototransistor element, so instead of sending one piece of information, it sends 1.3 million,” he says. “Vision sensors been around for about 15 years, but now they are smaller, better, faster and cheaper. So companies can afford more of them and are finding more ways to use them.”

Besides using vision systems to sort good product from bad, manufacturers also use them to prevent costly waste along the production line. “Companies use vision sensors to detect potential problems and relay them upstream before scrap is generated,” he says. “They can monitor the process so when it deviates at the control points, they can correct it so the product stays within standards.”

Whether the sensor is a camera, radar unit, thermocouple, or pressure gage, systems integrators and automation vendors recommend installing it in a way that complies with standard protocols. “Otherwise, if you rely on proprietary protocols, you could be painting yourself into a corner,” notes Crump, at Opto 22. “Although the system in place might suit your needs now, your needs will change. Proprietary technologies might not be compatible with the technology you will need in the future.” So he advises using the open forms of sensors, transmitters and networking devices. They will make integrating equipment into the information systems much easier.

For more information, search keyword “sensors” at