Portland, Ore. — It is the holy grail of the factory floor: hundreds of sensors wirelessly connected, monitoring motors for problems and drastically reducing energy consumption — all with the precision and rhythm of a philharmonic orchestra.

The need is there, the software is there, the topology is fairly well understood and the silicon costs are falling. One market forecaster sees 169 million nodes and a $5.9 billion end-user market by 2010. Still, it's not as easy as it looks. Wireless mesh is a new paradigm with lingering unknowns, and some wireless silicon is still more expensive than wired solutions. The goal, in the eyes of many, remains a ways off.

GE Global Research, Sensicast Systems Inc. and Rensselaer Polytechnic Institute have teamed up to push the agenda forward. They are engaged in a three-year, $6 million proof of concept funded by the U.S. Department of Energy that is scheduled to yield a working prototype within a year and a complete wireless-factory installation by 2006.

"Today a motor has to be 1,000 horsepower or more to warrant the expense of a wired sensor network to monitor it," said Peter Stein, vice president of Sensicast (Needham, Mass.). "Our goal is to enable cost-effective wireless networks that are cheap enough for almost any size motor."

Industrial motors (not including facility heating and ventilating) consumed 679 billion kilowatt hours in the United States in 2003. That's 63 percent of all electricity used in industry and 23 percent of all electricity sold in the United States last year, according to the Energy Department.

DOE figures a wireless sensor network could increase a motor's efficiency by 10 to 20 percent, which corresponds to a total U.S. energy savings of 120 trillion BTUs (35.1 billion kW-hr) per year, according to Sensicast's Stein.

By monitoring their vibrations, temperature and power quality, the DOE reasons that 10 million BTUs a month could be saved. Its award to GE, with Sensicast and Rensselaer as subcontractors, has three targets: first, an engineering blueprint that catalogs the technical challenges; then a prototype development stage including a complete proof-of-concept motor; and finally, a fully implemented control system at a Texaco petrochemical facility from which the operational data of all motors can be collected and analyzed.

Machine, heal thyself

"We can monitor motors with a low-power, wireless-mesh network and send all the data back to operators in real-time," said Stein. "The network is self-healing, too, so that in case a forklift driver blocks your transmission, the network will automatically search for another route that maps around the forklift."

Each sensor node will use an inexpensive microcontroller and wireless-transceiver chip operating with the ISO-802. 16.4 wireless protocol, which involves secure handshaking between nodes. Nodes can pass along messages from other nodes, eventually reaching a "parent" gateway node that concentrates the data for transmission back to the operators. Even dead gateway nodes on the mesh can be automatically routed around if they fail.

Sensicast has devised a suite of control programs that have frequency "agility" — that is, they seek out the clear frequencies and attempt to evenly spread out the simultaneous communications going on among the nodes on the mesh. In an industrial setting, multipath — or reflected transmissions that arrive slightly later to confuse a receiver — is also a problem that must be automatically overcome.

The biggest advantage of smart mesh topologies is that the overall system is completely autoconfiguring. Because each node has two-way communications, new nodes can be added by merely turning them on. Their autorouting algorithms transparently weave them into the mesh.

Each wireless sensor will operate for years on a single battery charge, but for the future, GE is working on new technology that could harvest the vibrational energy of the motor itself to power the transceivers. Rensselaer, for its part, will be developing a physics-based model that will enable turnkey analysis and lifetime predictions from the sensor outputs of a motor.

By reducing the cost of owning a motor-monitoring system, the DOE hopes to accelerate adoption of the technology and enable nationwide energy savings. The effort is part of its $61 million Industries of the Future initiative to improve energy efficiency throughout strategic U.S. industries.

But some think the problem may prove a tougher nut than DOE anticipates. The wireless paradigm in the factory environment is far from understood, said Francis daCosta, founder and chief technologist of Mesh Dynamics (San Jose, Calif.). In the wired world, "The paradigm exists. There's a company called Cisco and they make routers," daCosta said. "The problem [with a wireless-mesh system] is, now you're dealing with a pipe that's leaky. You're planning on using the pipe to get the info and for potentially actuating sensors. I don't think anyone understands how leaky that pipe is." And when the forklift runs in front of a node, many systems simply won't have enough logic to figure out redundant paths fast enough, daCosta suggested.

Power and cost are also factors, said Carl Brasek, wireless-USB product manager at Cypress Semiconductor Corp. (San Jose). "If your nodes are battery-operated and they're doing mesh networking, there's a requirement for them to be on all the time," he said. "If you're on the air all the time, you're burning power." In addition, the cost of silicon radios remains high compared with wired ones. For the short term, Brasek foresees hybrid networks where certain nodes go wireless and others plug into power outlets, even in some cases using power line communications methods to shunt data. The piezoelectric techniques for harvesting power from vibrating engines, he added, aren't ready for prime time.

— Additional reporting by Brian Fuller.