Page 2By CIOinsight | Posted 12-01-2004
Sensor Networks Make Early Inroads
Last year, BP plc, the $233 billion global oil and gas giant, had a big problem on its hands: It was literally losing track of 900 railcars—5 percent of its U.S. oil and gas inventory—every year in the convoluted rail and shipyard systems of the U.S. Each sidelined car cost the company anywhere from $50,000 to $125,000. For years, this was considered a cost of doing business. But P.P. "Daru" Darukhanavala, BP's vice president and CTO of digital and communications technology, had other ideas. Daru's team is charged with overseeing a number of pilot programs that test how sensor networks can help BP get smarter about its operations. His solution: to fit each railcar with a network of sensors that not only track where the railcar is at all times, but can also monitor characteristics such as weight, internal temperature and physical impact. The goal is to eliminate losses altogether and create greater efficiencies in transportation logistics.
Sensors are nothing new, of course. You can find them all over—in smoke detectors, home security systems, auto theft devices, even simple thermometers. But unlike RFID chips, which simply store data, sensors actively monitor and collect data on virtually everything from temperature, light and movement, to chemical compounds, wind velocity and pressure. And sensor networks can constantly send information back to a central system for interpretation.
The concept of a sensor network is fairly simple. Take a couple of wireless nodes (a small battery with a radio transmitter, roughly the size of a plum), add a microprocessor, some sensors and a reader (such as a desktop computer, a laptop or even a PDA) to collect the data, and you've got a wireless sensor network. But unlike traditional wired networks, which send data from each point back to a central computer, a sensor network works more like a mesh, passing data from unit to unit, creating a kind of sensor force field. Because each sensor requires just a tiny bit of power to transmit its information to the next closest sensor, the mesh network can cover a far greater area without the complexity and cost of stringing wires. And, since the nodes depend on one another to deliver the data back to the hub, the more nodes you add, the stronger the network becomes.
According to Business Communications Co., a Norwalk, Conn., market research firm, the U.S. market for industrial sensors will reach $7.6 billion by 2009. The ZigBee Alliance, a nonprofit consortium of more than 100 companies working to create open global standards for sensor solutions, estimates that there are more than eight billion embedded microprocessors shipped worldwide annually—more than the global population—any one of which could act as part of a network. But at the moment, the vast majority of those microprocessors are not part of any network. Robert Metcalfe, who invented Ethernet and is now chairman of Ember Corp., a developer of sensor networks, estimates that only 2 percent of those eight billion microprocessors are currently networked.
"If you boil it all down, there are two things that every object on the planet shares," says Accenture Technology Labs Chief Scientist Glover Ferguson, who is based in Chicago. "One is identity; the other is location. Sensors are all about using that information better, but there is more you can learn from them." Knowing the state of an environment, or even particular piece of equipment the sensor is monitoring, can create huge business opportunities, he says. "Add that to the explosion of wireless communication opportunities, and now we can inexpensively acquire information—temperature, weight, pressure, any attribute you can dream of." For a hotel owner, this could mean cutting costs on heat and electricity by wiring up hotel rooms that sense when a guest has left his or her room. For manufacturing, it could mean replacing parts in a machine before it breaks down. For firefighters, it could mean locating one another in hazardous situations when radios fail.
Want the latest news, reviews and analysis of sensor networks?
But as with most new technologies, the learning curve is steep. Wireless nodes and sensors are still pricey, leaving the early implementation to those who can afford the initial outlay. Limited battery life means that someone will eventually have to hunt them all down and replace their power supplies; right now the batteries last no longer than a couple of years. And because each node is crucial to the operation of the network, if one goes down, the network can fail. Setting up a stand-alone sensor network isn't that difficult, but integrating it into your key business systems (such as your supply chain) is no small accomplishment. Additionally, the lack of standards and security measures, as well as some lingering wireless interference issues, can keep data from properly transmitting. All of which means that sensor networks are still in their early days. "Getting the technology right is going to take a long time"—up to ten years, says Martin Reynolds, a vice president at Gartner Inc.
Some executives may balk at the thought of spending millions on a technology that has yet to prove itself in the enterprise, but Darukhanavala's position on the future of sensor networks is clear: "This technology will transform everything we touch, not just in business but our personal lives as well."
At Pickberry Vineyard, in Sonoma, Calif., owner Lorna Strotz thought a sensor network might help her better understand why grapes weren't growing on a particular plot of land. "We were having trouble growing in the soil, and we thought moisture could be a part of the problem." So when Accenture approached Strotz in fall 2003 and asked if she'd let them use her vineyard as a test bed for a mesh network that measures soil moisture, rainfall, wind velocity and direction, and air and soil temperature, among other things, she jumped at the offer. "It's a great way to gather detailed information," she says. The project started in January and while it's too soon to tell about the soil issue, Strotz expects the mesh network to be very useful in the coming months as the weather gets cold. "For example, if you have a network that tells you your grapes are about to freeze, you can do something about it promptly. Some vineyards use wind blowers to keep the air moving to make sure the grapes don't freeze, for example." And that can be significant, since each ton of grapes can bring in anywhere from $1,000 to $4,000.
Pickberry Vineyard, which grows grapes for Ravenswood Winery, is one of the few small companies that has the luxury of a sensor network (Accenture installed it for free as part of the research deal). Most small and medium-sized businesses aren't that lucky. Each sensor node can cost upwards of $200. "At this point it's really only the larger companies that can afford sensor networks," says Dennis Gaughan, a research director at AMR Research.
But there are some inexpensive ways you can prepare for what's coming, and the IT department is a good place to start. Vendors such as Crossbow Technology Inc. and Ember Corp. sell starter sensor network kits for as little as $2,000, which means that, for a small investment, CIOs can begin to automate asset tracking. "Right now, that's done with bar-code scanners and sticky tape," says Gartner's Reynolds. "The time spent doing all that tracking manually is expensive, but the real problem is when the IT department realizes they're missing a $50,000 server. So there's an immediate benefit in terms of keeping track of the asset." Whereas RFID has physical limitations in how far the signal can transmit, sensor networks can blanket entire campuses, constantly relaying information on the location and status of assets.
Another good starting point is preventive maintenance. At Intel Corp., researchers aren't just conducting extensive R&D into wireless sensor networks that they could potentially sell to their customers, they're also deploying them in their own fabs—the factories that make the Pentium 4 microprocessors. "You have all these pumps and motors that process all the needs of the fab; it's extremely important for those machines to be running at all times," says Lama Nachman, an Intel researcher who develops sensor networks. "If anything goes down, it's a huge loss, basically in the millions of dollars." At the moment, most of Intel's manufacturing equipment is inspected manually, which takes months. "It's very error prone, and we're happy if we can hit each piece of equipment once a quarter, which doesn't give you the detection you're looking for." Nachman estimates that the manual process uncovers about 80 percent of all potential breakdowns annually. "If you can do it once a day, you can catch more like 95 percent of the cases." So Intel is fitting all its equipment with sensors that can measure vibration and predict when a piece of machinery needs to be replaced.
Talk to your business unit heads and do some brainstorming on how sensor networks could create greater efficiencies or deliver deeper analytics. Analysts say most companies can achieve ROI in the supply chain, asset tracking and building HVAC systems, even in these early stages. There are a slew of vendors out there that specialize in different areas of the market. Some focus, for example, on creating efficiencies in heat and electrical systems. Others develop only the nodes, not the sensors. "Look at the integrators and service providers who know the limitations and challenges of this technology; those are the folks you want involved in the process," says AMR's Gaughan.
Ultimately, you'll want to fold your sensory network into a larger system, such as order management or your supply chain. It's those integration issues that will be your largest headache, say analysts. "It's not a trivial task," says Gaughan. "A lot of factors will have to come together in terms of standardization." The ZigBee Alliance claims it will have standards ready by the end of the year, easing integration woes and helping to secure the networks. "If you are controlling a building's lighting and air conditioning," says Bob Heile, chairman of the ZigBee Alliance, "you need to make sure that's as secure as if you were doing a $1 billion financial transaction. Can you imagine someone turning off all the lights in your high-rise building?"
BP's investment in sensor technology began a mere 18 months ago, during one of the company's Blue Chalk events—two-day seminars that Daru's team holds twice a year to demonstrate the value of emerging digital technologies to company executives. "It's a carefully crafted event in terms of what technology means to BP," says Daru. "We bring in various case studies or visionaries who can paint the landscape about what the possibilities are, and then the internal executives can make that translation into how it applies to their business." After hosting a Blue Chalk event on sensors in May 2003, Daru was bombarded with ideas from business managers, and his team had to pick which projects would be of greatest value to the company. "We look at sensory networks as complete solutions to business problems," says Daru. "This is not just about putting sensors on pieces and parts. This is total process change."
In the U.K., BP is using sensors to monitor its customers' liquefied petroleum gas fuel tanks. The company has seen a 25 percent improvement in efficiency. "Before, we had no way of knowing how much inventory was in those tanks," Daru says. The tanks had gauges but customers would forget to look at them until it was too late. As a result, BP would either rush to fill empty tanks, or send a tanker truck far too early to fill a minimally depleted supply. Thanks to the sensors, Daru says, "we have substantially improved the use of our fleet and increased customer satisfaction."
A number of other projects are also under way at BP, including fitting refinery workers with special sensors that track them for safety. "If you have lone workers in a plant, it would be important to track where they are to avoid hazardous areas where they may be at risk," Daru says. His team is also working to ensure pipeline integrity and worker safety at sites along its more than 11,000 miles of pipeline throughout the U.S. When a construction company informs BP of work being done near a pipeline, BP will install a portable geo-sensing network that will alert workers when they get too close to the pipeline. "We want to make sure that the whole operation is very safe," he says.
But Daru stresses the need for a business case in all his company's endeavors. "All of this has to be cost effective. We can't just try out something that works in the lab but doesn't translate to a real business solution."
For most companies, mesh networks are still a long way off. But for some, analysts say the sensor revolution is closer than we might think. In the automotive industry, for example, car makers are very interested in figuring out how mesh networks can create greater efficiencies and service for drivers. "Cars are getting much more intelligent already," says AMR's Gaughan. "The question is whether the car manufacturers can monetize it. They already have OnStar and in-vehicle services. There is a potential to pull in some additional sensor data and leverage it."
At Norwich Union, one of the largest car insurance companies in the U.K., customers have the option of enrolling in the "Pay As You Drive" program, which uses GPS to track drivers and charge variable insurance rates based on where they go, rather than a flat fee. But that's just the beginning, says program director Robert Ledger. In future stages of the plan, he envisions an entire sensory network that can inform the insurance company the moment an incident occurs on the road. "For example, we could tell if the air bag has been deployed. If the car has been in an accident we could sense the severity of that impact," he says. That could allow Norwich to process claims faster and with greater accuracy.
Gartner's Reynolds envisions an office where sensor nodes are installed on office walls roughly 30 feet apart from each other, creating an invisible sensor field that can track anything that has a sensor on it. "Pretty soon, the CIO can actively track all the IT assets in the building," he says. "You could also see, for example, how many wheelchairs are on the third floor of a hospital, and how often they've been used." The added intelligence of sensor networks can augment existing RFID solutions. A passive RFID asset tracking system could be incorporated into a broader and more active sensor network to provide real-time, varied information on not just location, but status of valuable inventory. A purveyor of frozen foods, for example, would be able to not only locate his merchandise, but know when the temperature has warmed to unacceptable levels.
Gaughan expects to see a lot of consolidation in the nascent sensor market, "because no one wants to buy pieces and fit them together by themselves." But Intel's Nachman disagrees. "I think it's too early to see where the market is going to go," she says. "If you look at the sensors themselves, the majority of them are analog, some are digital, and they all use different interfaces. If all these interfaces become defined, then you can easily have different providers offering different pieces of the puzzle."
Ultimately, the thing to remember is that sensor networks are about data collection and using that information for deeper business analytics. Accenture's Ferguson believes this is a natural evolution of what he calls an "insight" economy. "The whole practice of lifting data has to be raised to another level," he says. "The winner will be the one who asks the data the most interesting questions."