JUNE 2002
  Intelligent Systems
Table of Contents

MACHINE MONITORING/NETWORKING

Intelligent Wireless
Condition-Based Maintenance

More than $1 trillion is spent each year to replace perfectly good equipment because no reliable and cost-effective method is available to predict the equipment’s remaining life. Now through careful design and a balance of technologies, a wireless condition-monitoring system provides a dependable and economically viable solution.

Christopher McLean and Dave Wolfe, Techkor Instrumentation

photo
The future of condition-based maintenance will incorporate up-to-date machine condition information on all equipment in a manufacturing facility. Intelligent, low-cost wireless sensors provide detailed data, several times a day. Plant managers can make well-informed maintenance and production decisions to schedule downtime optimally. Facilities will be able to run more efficiently because they'll have more confidence in their ability to eliminate unscheduled downtime.
Top-level manufacturers have instituted comprehensive condition-based maintenance (CBM) programs. But in most cases, this involves a technician running routes with a data collector on a monthly or quarterly basis. Data are collected on machine vibration, temperature, oil levels, and running speed. With trending analysis software, you can assess and project machine health, and with the ability to predict failures and prevent downtime, you can achieve optimum production output.

The next step in the evolution of CBM is online surveillance systems, which provide more frequent machine-condition information than data collectors. The online systems include hard-wired sensors connected to multiplexers, which are networked to a main database computer. Most manufacturing facilities can’t fully implement a surveillance system because of the high capital costs, installation difficulties, and overall complexity of the system. Online systems are most often used on the most critical pieces of equipment.

Cost Justification
After up-front data collection hardware, software, and training costs, labor remains a continuing expense of running routes. The critical question is how often to monitor a piece of equipment. Monitoring critical equipment daily, you can detect most problems, but labor costs are prohibitive. The cost consequence of machine failure and how rapidly the machine will degrade after a fault is detected usually determines frequency. The average collection interval is 30 days, and this is sufficient to detect most faults [1]. With well-designed and -managed programs, data collectors are cost-effective CBM tools; however, the frequency of the readings is subject to the time constraints of technicians, budget cuts, and human error.

Online surveillance systems are more cost-effective than data collectors when the machinery they’re monitoring:

  • Is critical to the process
  • Has a rapid failure history
  • Operates with a known fault
  • Operates beyond its original design

If an inexpensive process pump can halt a $140,000/hr. production line, an online system is easy to justify. Some machines will indicate problems and propagate failure hourly, so even daily route running won’t be sufficient [2].

Frequent condition monitoring can change the way maintenance and production decisions are made. After a fault is discovered, the next question is, can the machine continue to run or can it make it to the next outage? A system of thousands of sensors monitoring a facility’s equipment (including some with developing faults) provides management with a clear, up-to-date picture of its infrastructure’s condition. With analysis software that can predict the life expectancy of a failing component, plant maintenance and production can schedule downtime when it is the least disruptive, which maximizes process throughput and minimizes costs [3].

With the new online technology, facilities will be able to run leaner because they’ll have more confidence in their ability to eliminate unscheduled downtime. Up-to-date information can also ensure that equipment is not replaced too early. Most manufactuêers have excess production capacity in their facilities to handle unexpected breakdowns. In petrochemical and petroleum refining plants, as much as 10% of production capacity can be required to make up for unplanned maintenance [4].

Wireless Systems
Eliminating the cables from online systems is a natural progression for CBM. If the cost of the online systems can be justified, a wireless implementation—with much lower installation costs—should apply to even more applications.

Several companies have developed telemetry systems that are a step in the right direction, such as Wilcoxon’s BlueLynx and Oceana Sensor’s ICHM 20/20. In addition, Rockwell has the HiDRA, and SKF promotes its wireless system. All of these systems connect a couple hard-wired sensors to a radio hub, so significant cable installation is still required.

Computational Systems has introduced a true wireless sensor (i.e., devices that use no cables at all). The 4100RF’s installed cost is about $1000 per sensor, which is comparable to an online surveillance system. All of these wireless systems work well in difficult or dangerous-to-reach locations or in moving applications; however, they’re still expensive and offer an alternative only to a small portion of a facility’s locations.

Desired Wireless Features
Sometimes it’s difficult to determine what customers want in new technology. SKF market research [3] lists some of the features plant operators are seeking in new products:

  • Low installed costs
  • Easy-to-use products that simplify maintenance and eliminate process problems
  • Fault-tolerant technology
  • Zero defects and 100% reliability from products out of the box
  • Compatibility with existing products

So how does this apply to wireless CBM? For such a system to support thousands of low-cost sensors and meet customer expectations, it must incorporate intelligent sensors that are the same size as traditional sensors and offer plug-and-play connectivity, high reliability, and long battery life.

Low-Cost Sensors
Although data collectors have lower installed costs than online systems, their operating costs are proportional to the number of points covered and the number of times per day the points are measured. Surveillance systems will have higher installed costs, but their operating costs are fairly flat, regardless of the number of points and the frequency of monitoring (see Figure 1).

figure
Figure 1. A wireless system provides the lowest overall cost for large-scale condition-based maintenance monitoring. While the up-front investment of handheld data collectors is less than other systems, the continual cost of labor, especially when frequent readings are required, makes this approach expensive.

ühe installed cost of any CBM system is critical to its financial success. A $100 accelerometer becomes a $600–$1000 sensor when you consider installed costs. At these levels, it’s common to use hard-wired systems only on critical assets. Wireless systems limit the installed cost to a fraction of that incurred with cabled solutions. A $400 wireless sensor becomes a $450 sensor when installed.

The operating costs of the data collector program accumulate every year, including labor and training. This method can have significant operating costs when faced with a high number of points and high frequency of measurement. Surveillance systems, especially wireless systems, have minimal operating costs.

Ease of Use
For wireless CBM systems, sensor nodes must be plug-and-play and have ad-hoc networking capability. This will allow additional sensor nodes to be added easily and let you quickly reallocate existing nodes. For a system incorporating thousands of sensors, anything less requires significant installation and upkeep effort.

The transceiver that receives the information from the wireless sensor network and connects to the main computer must also be easy to use and install. Of the existing enterprise networks available on the factory floor, Ethernet provides the most flexibility when compared with Modbus, DeviceNet, and Fieldbus. Many manufacturing facilities have Ethernet systems in place, so the transfer of data after the wireless portion is painless.

Once on the enterprise network, information on factory conditions can easily be sent anywhere in the world. Within this context, application software must be able to be used by multiple users. Typically, the sensor node data are stored in an Open Database Connectivity—compliant open architecture database using SQL language.

Wireless Communications
Areas of concern with digital radios are cost, size, battery drain, reliability, distance of transmissions, carrier frequency, wireless protocol for multiple sensor nodes, and compliance with FCC regulations.

figure
Figure 2. This illustration shows a simplified view of digital radio layering in wireless sensor nodes. Many possible combinations of the physical layer, communication protocol, and application exist for wireless communications. But a design must carefully balance performance, cost, size, and power consumption.
The radio begins with the physical layer, which is the actual radio circuitry, and it is the transmitter and receiver of information. The communication protocol is built on top of the physical layer, and it defines how information is packetized, routed, and encrypted, and how error checking and correction are handled. The final layer is the application, which defines what information must be transmitted through the communication protocol and what to do with received information once it is stripped out of the protocol (see Figure 2).

Physical Layer. The two potential choices for the physical layer are narrow band and spread spectrum. Both yield good results when properly implemented. Narrow-band radios can be broken down into two categories: those with only one frequency and those with multiple frequencies. A well-designed, multiple-frequency, narrow-band radio with spatial diversity can perform well in industrial environments. Spread-spectrum radios come in two main styles: frequency-hopping spread spectrum (FHSS) and direct-sequencing spread spectrum (DSSS). Spread-spectrum radios improve transmission reliability by distributing (hopping) the transmitted power over a large bandwidth. FHSS’s hopping rate is slower than the bit rate, but with DSSS, the hopping rate is faster. Spread-spectrum radios offer better physical-layer performance than narrow-band radios, but at the price of increased cost, size, and battery drain (see Table 1).

TABLE 1
Relative Performance for Various Physical Layer Solutions
Physical Layer Solution
Cost
Size
Battery Drain
Range
Interference Rejection
Frequency-hopping
spread spectrum
(Bluetooth)

Medium
Medium
Medium
Low
Medium
Direct-sequencing
spread spectrum
(IEEE 802.11b WAN)

High
Large
High
High
High
Narrow band
Low
Small
Low
Medium
Low
Multiple-frequency
narrow band

Low
Small
Low
Medium
Medium

One of the potential benefits of spread-spectrum transmissions is process gain, which can improve receiver sensitivity at the expense of additional battery drain. Process gain is proportional to 20 db 3 log (chirp rate). However, almost all spread-spectrum systems used in wireless sensor networks are FHSS (e.g., Bluetooth), which hop too slowly to take advantage of the process gain. You have to use a DSSS radio (e.g., IEEE 802.11) to benefit from process gain.

You can use spatial diversity to improve signal reception and transmission. Spatial diversity refers to using multiple antennas, and it’s useful in both narrow-band and spread-spectrum systems. Simple multiple-antenna designs in narrow-band receivers can improve spatial resolution by removing the probability of multipath interference. Sophisticated multiple antenna systems use the antennas to provide receiver directional gain through beam forming.

All transmitters in the U.S. must meet FCC guidelines for licensed or unlicensed operation. Licensed operation is impractical for most industrial users because of the regulatory hurdles, so unlicensed operation is preferred whenever possible. For industrial applications, the Industrial-Scientific-Medical (ISM) bands (902–928 MHz or 2.4000–2.485 GHz) meet FCC requirements for unlicensed operation. Within the ISM bands, narrow-band operation is limited to 1 mW of radiated power, while spread-spectrum transmitters are allowed up to 1 W of transmitted power. Higher-frequency bands provide more signal bandwidth, but lower frequencies offer better performance in multipath environments, where reflections are common.

Communication Protocol. A wireless data-gathering network must be reliable in industrial environments, where RF interference from motors, lighting, and other wireless systems is typical. According to Glenn Allgood of Oak Ridge National Laboratory, most industrial interference is caused by intermittent bursts of narrow-band signals, random EMI (background noise), and deterministic EMI (radio stations).

At any given time, the wireless link does not have to be reliable, but there must be a 100% probability the message will get through within a reasonable time. It’s also necessary to guarantee that any errors in the message will be detected and corrected. Lost or corrupt data will show up as distortions and could produce false equipment-condition reporting. Collisions between packets and interference from other radio sources must also be addressed.

The wireless CBM communications method must allow thousands of sensors to coexist in a single network. Various techniques are available to ensure dependable coexistence. Some of the more common are time division multiple access (TDMA), frequency division multiple access (FDMA), and code division multiple access (CDMA). The circuit cost and complexity of TDMA is the lowest, followed by FDMA and then CDMA.

Time domain and frequency domain multiple access are the easiest to understand. In both, timing or frequency bands separate communications of multiple sensors. In code division, several communications channels can be open simultaneously, as long as the spread-spectrum chirping rates are high.

For CBM monitoring, you don’t need to have multiple, simultaneous communications. For example, for a 256-point sensor cell, where each accelerometer makes a 800-line FFT reading eight times a day and each communication to the network takes 5 s, the total air time is 2.84 hr. per day, or 12% network loading. Either FDMA or TDMA will perform nicely in these environments.

Wireless standards offer additional interplay between sensor manufacturers. Any wireless sensor standard will have to specify the physical layer (e.g., FHSS, DSSS, and narrow band), the communication protocol, and significant aspects of the application. In effect, the standard must define a universal wireless sensor node. The IEEE 1451.X working group and the Bluetooth SIG are doing some work in these areas.

Many possible solutions exist for wireless communications; however, a design must carefully balance performance, cost, size, and power consumption. This balance is particularly important in wireless sensors. For example, CDMA spread spectrum is reliable, but at a significant cost, size, and power drain penalty. A better solution is to use a less reliable physical layer but improve the communication protocol to achieve the same reliability.

Using a multiple-frequency, narrow-band radio with spatial diversity and an aggressively designed communication protocol, you can achieve impressive results while still meeting cost, size, and power consumption requirements. It’s important to remember that the timeliness of CBM data is somewhat relaxed. It’s more important to have correct data that’s 10 s old than erroneous data that is prompt.

Power Sources
A poor power source can significantly limit the operational life or usefulness of a wireless sensor. Many power sources exist, including chemical batteries, RF, vibrational energy, solar power, heat differential, and AC/DC line. Such power sources as solar (photovoltaic), heat differential (thermoelectric), and vibrational energy (piezoelectric) are unreliable for most industrial applications. RF power works well for RFID products, but the range, bandwidth, and auxiliary power are limited for sensor proõucts. Using AC/DC line power significantly limits the freedom of the wireless solution and is to be avoided. Given that the sensor environment is not known in advance, man.ufacturers have tended to rely on chemical batteries for power sources.

Battery size is directly related to the power consumption and operational requirements of a device. Simply adding a wireless connection to existing sensor designs is not a good solution from the battery standpoint. Sensor electronic designs and operational methodology must be laid out with battery drain as a primary concern. Power-efficient circuit design and aggressive battery management (e.g., standby modes and duty cycling) can provide operational lives hundreds of times longer than those without.

Chemical batteries come in rechargeable and primary varieties. In either case, lithium chemistries exhibit the best performance, with 3.3 V output, high peak output, low self-discharge, low weight, high energy density, good cost-to-capacity ratio, and }ittle memory effect. For rechargeable and primary batteries, lithium ion and lithium thionyl chloride are superior choices.

Rechargeable batteries are used where recharging is desirable, such as in temporary testing. Only two viable rechargeable battery chemistries exist—lithium ion and lithium polymer. Both exhibit excellent energy density, good peak currents, low weight, hiýh voltage, and little memory effect. Lithium ion tends to be less expensive than lithium polymer, but lithium polymer is more rugged and environmentally robust. Lithium polymer can be formed into various shapes and lends itself to packaging. Both nickel cadmium and nickel metal hydride are poor solutions because of their self-discharge, higher weight, lower energy density, lower voltage, and memory effect. For permanent installations, recharging a battery is not desirable.

The life of the battery is critical, especially when considering the expense associated with either replacing or recharging it. A chemical battery should be designed to last more than three years under fairly heavy sensor use. For example, the Techkor wireless accelerometer is designed to operate on a half A-size lithium battery with a capacity of 1000 mAh. With a standby current draw of only 3 µA and an operational current of 16 mA—including the sensor, signal conditioner, microcontroller, memory, and transceiver—significant battery life results. For such a battery, the standby life would be 38 years, and that factor is neglected in this analysis because it is longer than the shelf life of the battery. The sensor should be able to perform 9000 operations (assuming 20 s duration) over its life. If the wireless sensor monitors eight times a day, this leads to a projected battery life of 3.08 years. The life is scalable, so for monitoring four times a day, the projected life would be six years. When performing battery projection, you must take into account network traffic and reliability of the physical layer. Additional transmissions caused by errors or collisions will reduce the battery life. In contrast, if the sensor is operating continuously, it would operate for only 50 hr. (2.08 days). For maximum confidence, wireless CBM sensors should be designed to perform periodic self-testing for power fluctuations and low battery power.

Miniaturization
Several miniaturization strategies are available for wireless sensors, including system-on-a-chip, MEMS, and ASIC. There are five main areas of volume consumption in a wireless sensor: the battery, radio, sensor, circuitry, and package. The radio, accelerometer, and circuitry can be reduced in size through hybrid, MEMS, or mixed-signal ASIC design. The battery and package must be dealt with separately.

Hybrids. Thin film hybrid technology or chip-on-board technology can be used to significantly reduce the size of wireless sensors. A standard 4 by 4 by 0.25 in. printed circuit board with surface-mount components can be reduced to approximately 1 by 1 by 0.25 in. by using chip dice, subminiature components, and thin film hybrid technology.

Combining the radio, accelerometer, and circuitry in a thin film hybrid is not a simple matter. As packages shrink, interference problems and power dissipation increase. Packaging the digital radio in such a small space with the microcontrollers and analog accelerometer involves significant analysis.

MEMS. MEMS system-on-a-chip capabilities are just beginning to become a reality. While the components of MEMS have been around for some time (e.g., microsensors, microactuators, microelectronics, and microstructures), the ability to combine them in a reliable, cost-effective manner has been lacking.

With today’s MEMS technology, combining the radio transceiver, accelerometer, and circuitry is still in the development stage. MEMS accelerometer elements are not new, but MEMS accelerometer technology tends to have low frequency response and is more sensitive to shock. Special attention needs to be given to account for these shortcomings.

Industry is just beginning to produce viable MEMS radio components. Work being done at Georgia Tech (Chahal and Bhattacharya), DARPA (Tang), Carnegie Mellon University (Fedder, Luo, and Xie), Berkeley (Bilic), and others is furthering the advancement of MEMS accelerometers and radios. The announcement by MEMSCAP in May 2000 of a component library of MEMS inductors and variable capacitors ushers in a new generation of radio design.

ASICs. Historically, developing mixed-signal ASICs is cost prohibitive, and the devices are sensitive to layout. Only when dealing with volumes in the millions of pieces are ACICs cost-effective. Mixed-signal ASICs would still require external components to complete the circuitry. This is not considered a viable alternative for wireless accelerometers.

Summary
Advances in wireless technology, battery chemistry, and miniaturization have made extensive, wireless CBM data gathering systems practical. Through careful design and balance of technology, inexpensive wireless data gathering systems are currently available.

The CBM market can justify the cost of widespread implementation of wireless condition monitoring systems. With up-to-date machine condition, maintenance, and production, decisions can be made to optimize production, and manufacturing facilities can operate more efficiently.

References
1. Bob Craft. “Appropriate Use of Wireless Technology in Condition Monitoring,” SKF Revolutions, Vol. 9, No. 2:4-5.

2. William Broussard. “How to Justify and Calculate the Cost of an On-Line Reliability System” [PDF], Computational Systems, Inc.

3. “The Future of Machinery Monitoring Technology,” SKF Condition Monitoring Technical Paper, CM1012.

4. Brad Law. “Asset Management and Condition Monitoring,” Bently-Nevada Orbit, Second Quarter 2000:18-21.

For Further Reading
Byington, Carl. 2000. “Systems and Automations Operation,” technology brief Web page, Applied Research Laboratory at Pennsylvania State University.

Federal Communications Commission, Office of Engineering Technology, Policy and Rules Division, “The FCC’s On-Line Table of Frequency Allocations,” 47 C.F.R. 2.106.

Harris, Cyril M. “Shock and Vibration Handbook,” 4th ed.

Mauk, Jeffrey M. 1999. “Collecting Data for Statistical Monitoring” Enteract ’99. Orlando FL.

Watkins, Jeffrey P. 1999. “RCM, Where Does it Fit?”, Enteract ’99, Orlando FL.

SIDEBAR:
A Watchdog for the Factory
photo
Photo 1. The Techkor Maintenance Watchdog system includes an REP-914 wireless repeater, a TACH-RF-L wireless laser tachometer, M9E-RF-1-50G wireless accelerometers, and a TELM-914 access point. The system operates in the license-free 902-928MHz ISM band.
The Techkor Maintenance Watchdog system (see Photo 1) can operate as many as 65,535 wireless sensor nodes from a single Ethernet-networked PC. The condition-based maintenance (CBM) system offers plug-and-play connectivity and sensors with battery life of more than three years, simplifying system maintenance. In large quantities, the sensor costs $400.

Data are sampled by individual sensor nodes, internally processed, and sent via wireless communications to the access points for upload and storage on a primary control PC. Operators can monitor sensor data, be informed of alarm conditions, and configure parameters from a standard PC.

The system operates in the 902–928 MHz license-free band. The access point uses a 16-bit Ethernet chip set that is Novell 2000 compliant. The Open Database Connectivity-compliant database engine provides a way of interfacing with various existing software packages.

The wireless CBM system addresses scalability, cost, miniaturization, reliability of RF communications, battery life, ease of configuration, and intelligence, while retaining the accuracy, bandwidth, and robustness of traditional sensors. The wireless CBM system includes accelerometers, thermistors, and laser tachometers. Repeaters are used when necessary (see Figure 3).

figure
Figure 3. In the system configuration shown here, the M9E-RF-1-50G accelerometers and TACH-RF-L laser tachometer record information and transmit the data wirelessly to the TELM-914 access point. The computer communicates with the access point via a standard Ethernet connection and stores data in Open Database Connectivity-compliant databases. The REP-914 wireless repeater is used when required.


Christopher McLean, Ph.D., is Division Manager, and Dave Wolfe is a Sales Engineer, Techkor Instrumentation, A Division of Advanced Conversion Technology Inc., 2001 Fulling Mill Rd., Middletown, PA; cmclean@techkor.com, dwolfe@techkor.com.

MORE!
For further reading on this and related topics, see these Sensors articles.

"Commercialization of RF MEMS," December 2000
"It's Time for Sensors to Go Wireless," April and May 1999
"Piezoelectric Sensors for OEM Applications," February 1999





 
E-NEWSLETTERS
SUBSCRIBE NOW!
Sensors Weekly
  What's New
  Product Picks





Questex Media
Home | Contact Us | Advertise
© 2009 Questex Media Group, Inc.. All rights reserved.
Reproduction in whole or in part is prohibited.
Please send any technical comments or questions to our webmaster.