Innovations in communications and computing hardware and software have made it easier than ever to collect minute details regarding just about any topic of interest. For technology and manufacturing interests, small, low-powered sensors can be embedded in almost any machine for data collection. Thanks to wireless technology, these embedded devices can continuously and unobtrusively provide measurements of performance and environmental data. Analysis of this data offers vast opportunities for fine-tuning performance and process.
Given the benefits that the technology provides, it is not surprising that many organizations have deployed sensors to the point that they are spending more time collecting, storing, and sorting data than improving operations; the sensors become a task unto themselves and not an enabler.
Within this article, we’ll show how many of these common problems can be avoided by making some operational considerations early on. With the addition of microprocessing and management at the sensor level, deployed sensors can be united into a cohesive, low-maintenance communications infrastructure that exists separately from the system it monitors.
A Sensor’s Return on Investment
Sensors become more powerful, and useful, when they are partnered with microprocessors and networked. A single sensor is limited in the information it can provide, but a network of connected sensors paints a clearer, more comprehensive picture. The network becomes even more powerful when a historical record of sensed data is recorded and analyzed. The network then evolves from a means of alerting users to an event that has already occurred, to an active means of analyzing and optimizing a system or environment.
Moving from event monitoring to business intelligence collection can be a complicated process. The first challenge is determining what sensors to deploy and what information to collect. Just because sensors can collect and store data, that does not mean the data will lead to insight. For a sensor network to be truly valuable, it is important to first determine the business value of the information that can be gleaned from the network.
A sensor network must provide a return on investment. The return could be a reduction in unplanned downtime, enhanced workplace safety, or even insight into a topic that could not have been explored without multiple points of data collection. You have to know what data you want to collect, and how you are going to use that data once it is collected.
The temptation to collect multiple data points is rooted in legacy wired systems. Previously, when a wired product or environment was being designed, data collection and monitoring had to be built in at the beginning. Wireless technology has changed the process by making it easy to add sensors after the fact.
Modern sensors, and the microcontrollers that accompany them, are small enough to be deployed in hard-toaccess, hard-to-maintain places, and a well-designed system can be powered by batteries alone for well over a year without maintenance. Sensors can be deployed to monitor one aspect of a project, and later expanded to add additional data points without disrupting the existing setup.
Consider a restaurant owner, for example, who wants to monitor a freezer and alert the staff when food spoilage may occur. Freezers are well insulated so temperatures do not shift radically, and a single sensor is sufficient for an entire freezer. The freezer has a small power demand, so a single battery-operated sensor that activates once an hour, takes a measurement, and transmits the result to a microprocessor is sufficient.
The microprocessor stores collected information and provides an alert if temperatures exceed a predetermined range. More importantly, the restaurant owner now has a running record of freezer temperatures, which eliminates the manual process of recording freezer information for health code compliance. If this is all that is ever done with the sensor, it is already saving the owner time and money.
The Importance of Context
Although a single sensor provides a lot of information, the device does not necessarily provide context. Context requires more sensors and a means for analyzing collected data. With proper tools, a network of sensors is much more valuable than the sum of the parts.
Using the freezer example above, imagine a restaurant owner who has temperature sensors in several stores. Just reviewing the records of temperature readings from a handful of stores would likely expose patterns, such as which freezer brands are more efficient, or which freezer is consuming the most power. If any location is out of profile with the others, it would be an immediate prompt to look for inefficiencies.
Even within a single location, there will be fluctuations based on routine traffic, and incidents like stock days where the door of the freezer is propped open. But what if, over the course of several weeks, the temperature of a single freezer slowly rises? How does the owner determine if the compressor is going to fail?
One solution might be to attach a vibration sensor or power meter to the compressor and record the times that the compressor is running. The chart of data, when compared to the temperature readings, would show how long the compressor runs in order to maintain a constant temperature. As a compressor moves toward failure, it would likely run longer, and consume more power, than a compressor at optimal performance. By providing insight that would not have been available before, the sensors ultimately prevent unnecessary power usage and help eliminate unplanned downtime.
In this way, modern sensor networks are evolving beyond simple monitoring to become the central mechanism for a continuous feedback and response ecosystem. As more sensors and microprocessors are added to these ecosystems, the best ones expand beyond merely providing insight and context. The devices leverage existing hardware and infrastructure and have mechanisms for updates to allow the network to adapt to the conditions it is sensing.
Making Sensors Mesh
The ability to correlate sensed information across multiple sensors paints a more comprehensive picture, but what happens when the sensors in a network are spread across a large area that does not already have a wireless infrastructure in place?
One method for overcoming the challenge is to unite the sensors in a mesh network. In a mesh network, sensors are paired with microcontroller nodes at the point of deployment. The nodes act as small computing platforms that establish communications routing among the various sensor packages in the installation. When a remotely located sensor needs to transmit a reading, the data is relayed through intermediate nodes until the information reaches a central collection point. In this way, each node within the mesh acts as a signal repeater for the other nodes. A remote sensor only needs to be in communications range of one other node to be a part of the network. If a single sensor should fail, the sensors surrounding it can automatically route information around the failed node without compromising the rest of the network.
This kind of intelligent routing is essential for a robust wireless sensor network. Monitored sites are seldom static, and the everyday operations of an enterprise can alter the wireless landscape and leave remote sensors stranded. For optimal performance, the system must be able to evolve and respond to challenges. To achieve an effective level of adaptability, sensor placement and connectivity must be a concern during the installation of the network.
Site survey and specification tools provide a stable foundation for the network. Just understanding how the site environment affects wireless signals will help users avoid many common problems. Once a system is deployed, however, how do you guard against interference?
One of the best ways is to ensure that the nodes at the edge of the network have intelligent, embedded application software built in. Individual devices must know if communications fail so their collected data can be maintained until a successful connection is made.
The addition of “intelligence” at the node level provides much flexibility, so it is possible for the same node that senses a leaky water pipe to turn off the water supply until the leak is repaired. Even if the node is currently out of communications range, the device will continue to do the job. The node is given a measure of autonomy, performing both individually and as a part of the network.
A Secure Network
Of course, a more connected sensor network makes security a greater concern. Security should be considered between wireless nodes, and also between the nodes and the Internet if the network is connected. Communications among nodes must have best-in-class security provisions and encryption at all layers of the platform. Optimally, the security should be built into the node to ease implementation burdens on developers and speed development times. To reduce chances of signal interception, wireless sensor networks should also be isolated to localized intranets or connected to the public Internet.
A final component in a strong and secure wireless sensor network is the ability to manage and update connected nodes. If the software that accompanies a sensor is updated, you can make repairs, or security improvements to the software, or even add new functionality. The updates grant the sensor network a level of “future proofing” and provide a significant level of flexibility beyond traditional “hard-wired” systems that cannot evolve.
The unification of low-cost, low-power sensors has made it possible to monitor just about anything, from anywhere. Sensor density creates a challenging environment where the amount of data being collected can hinder analysis efforts, and evolving needs can quickly render the sensor network obsolete.
Careful planning and the combination of sensors and wirelessly enabled microcontrollers help to prevent these conditions by providing a platform for the new sensors, updates, and error corrections needed to keep a sensor network relevant.
This article was written by Jonathan Heath, Systems Architect at Synapse Wireless (Huntsville, AL). For more information, Click Here .