Comment

1 June 2005

Driving data node to node

By Ellen Fussell Policastro

Wireless sensors gain ground in industrial settings.

A networking technology that ties sensor devices together, machine to machine, and aggregates data typically generated from sensors is finding a home in industrial automation space. Its role? Predictive maintenance, process monitoring, and process control; in fact, sensor networking enables the whole trend toward predictive maintenance rather than preventive maintenance, said Burlington, Mass.-based Millennial Net's vice president of marketing, Mark Pacelle.

"Predictive maintenance involves the monitoring of process conditions," and sensors that measure vibration, temperature, humidity, or acidity are attached to these machines to continuously monitor things like bearing temperature. "When the temperature goes out of an acceptable range, it triggers an alarm. When it goes through a wireless sensor network, it travels back to a control point. Likewise, if you don't want to send an alarm, it'll send a continuous hourly reading of that bearing and the application end of the network," Pacelle said. "And the network would keep a continuous log."

Mesh networking has been around for a long time, and in and of itself, it's not a new phenomenon. The vast field of applications for wireless technology range from dusting a process with tiny wireless sensors to more traditional deployment. At the low end, wireless sensors use mesh networking with auto-adaptive, self-healing capabilities. The news is the way it's being applied to sensor networks in wireless and to some extent to control networks. "It's a nice application for wireless because it allows for mobility," said Hesh Kagan, director of technology for a ventures group at Invensys in Foxboro, Mass.

Installing these wireless sensors on ships for the oil and gas industry to use for temperature and vibration measurement is one instance where mobility comes in handy, said Mike Horton president and CEO of Crossbow in San Jose, Calif. "The drive to go wireless is to eliminate the manual data collection," he said. Historically the process was completely manual. Now, data collectors can get readings automatically, and they automatically network back to the central depository. The advantage is the ship runs with a lean crew to maintain all the equipment. "Having someone walk around and collect data is cumbersome and expensive," Horton said. "With wireless, they can inexpensively install sensors and automatically collect data."

Data drives

But while mobility is nice in the industrial process world, it's not a driving force for wireless, Kagan said. Mobility addresses a niche market in industrial process, "but it's just a piece of the pie." The bigger question is how wireless is being adapted in industry. What are the constraints? What are the driving forces? "Mesh is just a subset of that," Kagan said. "It's a nice way of bringing multiple things together that are scattered around and of having devices joining the network, removing themselves, and rejoining the network."

You might think one driving force is of course the luxury of not having the cost of running the wire. "The short term view is wireless is slick and can be significant in terms of dollars," he said. (Wiring costs range from hundreds to thousands of dollars.) But of course, that's still a short-term, immediate benefit.

The real driving force is having access to all the incremental data and being able to co-incorporate the data that can save you money day after day, Kagan said. Optimizing and preventive maintenance are big deals in the wireless game. "Most maintenance in a plant today is done in a break-fix modality," Kagan said. "People wait for something to break, and then they fix it." And in the short run, it's probably the cheapest way to do it. But it's "when you get into trouble, when that thing that breaks causes a shutdown, that you have bigger problems," he said. Preventive maintenance gets pretty expensive too because, "essentially, you're throwing away good equipment because a manual tells you after so many hours it needs to be replaced, and it's not necessarily broken," Kagan said.

Reliable routes

But the wireless concept is a challenge to get people to adopt because of reliability and cost, said Horton. "The way a mesh network addresses reliability, every node acts as a router, and that's true mesh. That means when you deploy a set of sensors with the true mesh technology built into it, every node can relay messages to every other node," he said. "If a customer deploys 50 sensors in an industrial factory environment, any given sensor will be able to talk to four or five other sensors reliably with one radio jump."

Horton likened the mesh network to the all-too-familiar spider's web—"strong and reliable because it's not dependent," he said. Losing a connection, like you would using a cellular phone, is "totally unacceptable in the industrial environment," he said. The way mesh overcomes that is it doesn't talk to just one base station, it talks to four or five other nodes, and they relay the message back to a central controller. "So there's an increase in strength and reliability that allows technology to be deployed in an industrial environment," Horton said.

Measure for measure

So, reliable routes and mobility are great. But how are you going to measure it? The ultimate way of managing assets is through a predictive type of maintenance as opposed to preventive, Kagan said. "You have to get new measurements on how well that equipment is performing, and you want to bring that new data into an application that can intelligently determine how it's degrading or wearing." Based on that, you can determine what the equipment's maximum use is and how to schedule. "You have to, so to speak, Velcro on these sensors, and low-cost wireless sensors make it very convenient to do that." And it's the only way you can do it in a meaningful way because it's too expensive to run wires for these incremental sensors for asset management, he said.

"Look at the way pumps and valves run until it becomes painfully obvious they need to be maintained. If you sensor that valve, you can determine how it's behaving, so you can determine overhaul. There are innumerable opportunities for process equipment to be maintained in a more rational fashion; you can save tons of dollars. To rebuild a big valve might be $20,000. If you can maintain it in such a way as you prolong its life, there are lots of savings involved," Kagan said.

Temperature measurement is one common application in industrial processes on the factory automation and process automation side, Horton said. "We've built these wireless temperature modules that plug in to an existing system and replace hardwired RTD connections, or thermocouple connections. So instead of a wired connection from the temperature sensor back to the measurement system, you have a battery-powered node, completely wireless, and that takes the temperature measurement and reports it into the same type of control system people have used for years."

Pacelle said his company was working with Emerson Process Control for oil and gas exploration in petroleum plants and refineries. "Those types of customers are interested in implementing this notion of predictive maintenance to save money and to minimize downtime," he said. The main reason is "predictive maintenance is shown to be a more effective way to maintain your physical assets."

One field trial coming out now in a big petrochemical plant is equipment motor bearings (the motors driving big refinery fan assemblies). The motor bearings have a temperature sensor that senses the temperature of lubricant within the bearings. And as long as that bearing housing remains within an acceptable range, the motor is deemed healthy and running appropriately. If the temperature rises outside an acceptable range, "that's an indication something's wrong with that bearing," Pacelle said.

Because temperature is a leading indicator of failure, manufacturers could have perfect information on the bearing temperatures and motor assemblies, and thus, they could predict failure. Wireless allows economical installation of these sensors onto motor bearings. "So it's now economically feasible to send information back to a single control point with wireless networking," Pacelle said. "If they have to run wires to collect this data, then it becomes economically unfeasible because it's so expensive to run wires in an industrial plant. You have to run them through conduit or bury them in cement if it's an explosive environment, and that breaks the feasibility of implementing this type of predictive monitoring and maintenance."

Inferential measurements

One of the most immediate benefits of being able to use inexpensive wireless sensors is the manufacturer's ability to make inferential measurements—inferring the measurement you want by measuring as many things around the process as possible. Whenever you try to refine oil or make a better cookie, you need to control your process, that includes the temperature, the oven, the speed of the conveyor belt, Kagan said. "You're constantly measuring. If you measure everything around the oven, such as the cookie dough, the temperature, even the weather, and you understand what those other measurements are, you can determine what the moisture of the center of the cookie is."

Collecting data and building models is nothing new, Kagan said, "but you can always build a better model by having more data. If you ask a process operator what's the one measurement they wish they had to control their process better," Kagan said, "they'll say 'I wish I could measure the smell, the texture, the color of the flame'—measurements that are difficult to get."

In fact you could get previously impossible measures. "When you're running a plant, whether an oil refinery or cookie company, if you can improve your process by a tiny fraction, there are big dollars involved," he said. Real-time strategic measurements of asset availability, asset utilization, condition monitoring, smell, taste, turbidity, quality, value, and color would be all but impossible through a physical sensor. "But with enough data points in your model, you can infer real-time measurements," he said, and save a bundle through such inferential measurements.

When you can deploy sensors into a process and collect data inexpensively, it allows you to deploy more sensors in a process, Pacelle said. "You have better information about how a process is working. And because you have better and more information about the health of a process, the better you can tune the efficiency of that process. More theoretically, you have better visibility and better insight into the characteristics of a dynamic process. The dynamic process is what you're looking at in your refinery or manufacturing environment. The more visibility you have, the better decision you can make and the better adjustments you can make in the process to make it more efficient."