During the 1980s, online monitoring meant using expensive, permanently installed protection systems to detect the onset of catastrophic failures, sound alarms and shut down equipment to protect it (and the surrounding area) from consequential damage. The systems were typically cost-effective only for large, costly, complex turbines and generators.
Since then, lower-cost systems capable of detecting impending failures have slowly gained acceptance, driven by advances in networks, integration with control and information systems, stronger diagnostic and decision capabilities, innovative sensor technologies, lower costs and above all, powerful pressures for higher plant performance and maintenance productivity.
Modern systems are being applied to equipment ranging from those complex turbines to production machinery, compressed air systems, electrical cabinets, boilers, fans, pumps and valves. They’re solving chronic reliability problems, streamlining maintenance labor and inventories, updating ancient equipment to compete in the 21st century and helping even the most advanced facilities reach higher levels of reliability.
“We’ve had a long history of continuous improvement — 2005 set several records,” says Scott Brown, reliability engineer at Vectren Power Supply’s Evansville, Ind., plant. “We’ve got a 92% operating requirement, and when you’re that high, you’ve got to look at doing new and different things. This is a tool to get us to the next level. And the cost isn’t prohibitive — payback is less than one year based on PMs alone.”
As opposed to portable
Most plants perform condition monitoring by sampling equipment vibration, temperatures, oil, ultrasonic emissions or electric power parameters at intervals using portable equipment. This approach can be very effective, but online monitoring may be better in certain applications (see sidebar, “Online advantages”).
“Permanently installed, continuous monitoring systems that are intelligent, scalable, support a distributed architecture, and are based on open industry standard networks represent the most significant development in condition monitoring over the last several years,” says Tom Alford, product marketing manager, condition monitoring hardware, Rockwell Automation (www.ra.rockwell.com). Plants can deploy intelligent monitoring systems that meet their specific needs and are easily installed, integrated and maintained because they leverage existing plant networks.
Still, many plants have had a tough time making a business case for online systems.
Although they’re becoming less expensive due to advances in sensor and system technology, increasing sales volumes and competition among suppliers, they typically still bear a higher initial cost per point than mobile equipment, which many facilities already own. How do you explain that your investment in mobile technology is inadequate and gathering dust for lack of manpower?
Talk about manpower. It’s obvious that done right, many hours are invested in doing rounds.
So many plants don’t get around to doing it right, even for well-established PdM programs like infrared and vibration. But even partial discharge monitoring is increasingly going online, says Greg Stone, dielectrics engineer, Iris Power Engineering (www.irispowr.com).
On motors at 4 kV and higher, “80% to 90% of stator problems that lead to rewinds start as sparking in the insulation,” Stone says. “Detecting those usually gives about a two-year heads-up.” Technically, online monitoring isn’t needed because you could test every six months or so with portable equipment. “Continuous monitoring is done as a matter of convenience,” he says. “Other plant priorities too often displace walkarounds. It’s not unusual to see measurements missed for two or three years.”
But plans to switch to online may run into cultural resistance. For example, many plants have adopted an approach that uses operators to collect data, and machinery/maintenance specialists to interpret it. “These companies want their operators to get out in the plant, take ownership and actively participate by noticing asset conditions as they make their daily rounds,” says Steve Sabin, editor of General Electric’s ORBIT magazine (www.bently.com/orbit.htm). “Some do not want to fully automate data collection because it would tend to disengage these operators.”
However you slice it, today’s pressures for higher productivity mean predictive maintenance must become more prevalent. “More and more companies will be adopting predictive maintenance and using it as the key element to drive their maintenance programs,” says Jack Nicholas, senior technical advisor, Allied Reliability (www.alliedreliability.com). “Instead of the individual going out to get the data, the data will come to the individual. That will be a much more efficient, effective way to do it.”
To protect and serve
Developments in online technologies are exciting, but some of the most compelling reasons to implement them focus on people. The first is safety. “It can be dangerous or impossible to collect data from some assets,” says Sabin. Many assets are difficult and unsafe to access, particularly when operating.
“The paper and steel industries are looking at $20,000 to $25,000 per hour of downtime, and there are fire hazard and safety issues,” says Ed Bondoc, product manager, surveillance systems for SKF Reliability Systems (www.skf.com). “In some countries, you are not allowed to walk under a machine. Essentially, you are not allowed to do rounds.”
Remote, inaccessible and mobile assets also beg for online monitoring. Examples include outdoor pumps and compressors, mining conveyors and screens, and mobile equipment such as mining shovels and trucks. “Nuclear power plant operators can’t get close to their equipment except every two years during a shutdown,” says Ernesto Wiedenbrug, Ph.D, project manager, online monitoring, Baker Instrument (www.bakerinst.com). Closer to home for most plants, “Arc flash dangers mean you want to leave the cabinet closed,” he says. “Cooling towers are high up and hard to get close to.”
Unmanned or operator-only facilities, such as hydroelectric generators, also are good candidates. On such equipment, “Service providers are installing systems to reduce their costs,” Bondoc says. “Typically, 30 to 40 points can be accessed via the Internet.”
Steam and compressed air systems are often decentralized to bring generation closer to the point of use for higher efficiencies. The remote boilers or compressors can be monitored by centralized control systems, which also can report on ancillary equipment.
“You can link boilers in individual buildings,” says Dan Willems, vice president, product engineering, Cleaver Brooks (www.cleaverbrooks.com). “It’s safer, and it takes less manpower.
The controller resides in one location and monitors the condition of multiple boilers connected through a system. One Cleaver-Brooks customer with 42 boilers can monitor them via the Internet from anywhere, in or out of the office. The controller gives real-time access to conditions in the boiler room including alarms and shut downs, sends an e-mail or page on alarm, and provides trending data such as efficiencies over time and how many hours the pumps have been running since last serviced.
Other equipment such as water softeners, chemical feed systems and deaerators can be integrated over the Web. “You can have a flow switch on the outlet of a pump — if the flow should be on but isn’t high enough, it will alarm,” says Willem. “Load cells under the chemical feed tanks can tell when chemicals run low.”
Built-in diagnostics and decision-making capabilities can help turn raw readings into condition status indicators, time-to-failure (TTF) predictions and alarms, allowing operators and managers to react appropriately.
“There is now a broader array of approaches from online to portable, but they’re just the vehicle to get data to make the decision,” says Jeff Schnitzer, general manager, Bently Nevada asset condition monitoring, Bently Nevada (www.bently.com). “You can’t underestimate the value of the infrastructure — having information in a central structure lets you pull it all together and analyze it. You can write the rules to tell you what is happening and when a situation is abnormal.”
Online data collection allows you to correlate machine condition with process conditions. Without that information, you may be able tell that a machine is failing, but may not be able to ascertain why. “People end up becoming very good at predicting the problem and performing just-in-time maintenance, but never actually solving the root problem that is causing the asset to fail in the first place,” Sabin says.
Built-in diagnostics can save a lot of time and money. For example, if a protection system shuts down a machine because it detected excessive vibration and the cause isn’t obvious, it can take an expert to troubleshoot the machine. If a diagnostic system were being used, the vibration signature may provide enough information to indicate the exact nature of the fault or problem. “In many cases, the cost of the diagnostic system could be less than the cost of opening up the machine for inspection,” says Alan Friedman, senior engineer, DLI Engineering (www.dliengineering.com).
People think in terms of monitoring vibration and bearings, but what they really want to know is when the equipment is going bad. “You want an alert rather than an emergency shut-off,” says Alex Nino, applications and sales engineer, Ludeca (www.ludeca.com). “More diagnostics are being embedded into the hardware and software. You see a change, but is it an imbalance? A technician might not know what that looks like — we can program it in.”
The savings aren’t in expert labor alone. Continuous monitoring also can help determine root causes and prevent recurrences. “On a paper machine with a hundred bearings, paybacks are less than a year, sometimes six months,” says Andre´ [accent on Andre] Smulders, director of business development, surveillance systems, SKF Reliability Systems (www.skf.com). “But it’s not always apparent to plants that are comparing online systems to walkarounds.
“You do walkarounds with maybe $35,000 or so for equipment. Online systems might cost $2,500 per channel installed, maybe $1.5 million to instrument a mill,” Smulders says, but you learn a lot more. You can extract the root cause and do a modification to prevent recurrence. “Walkarounds don’t collect enough information, and if there’s a breakdown, there’s no data,” he says.
An approach that distills condition data, feeds it into a decision-support system and serves the results in concise reports to the right people can have a powerful effect on how well a plant follows up and takes advantage of its predictive prowess.
“A power company came to us with two problems: equipment they couldn’t monitor, and monitoring-based information they couldn’t manage,” says Neil Cooper, general manager, Avantis for Invensys (www.avantis.net). In one case, a main cooling pump had failed and the back-up pump didn’t start. The turbine alarmed, overheated and seized at a cost of $6 million.
“This shows three problems: not enough condition monitoring in the first pump, insufficient information when the back-up pump did not kick in, and no alarm until after the turbine had thrown blades,” Cooper says. “Plants need intelligent condition monitoring, and they need to actually use the information. You want to manage problems, not gather data, so don’t put in a tool that doesn’t tell you what to do about it.”
Where does the intelligence come from? “Some is knowledge, some is experience,” Cooper says. Large equipment manufacturers give condition-monitoring specifications, and some facilities have been collecting data for a long time that can be used to program the system (Figure 1).
“Time to failure (TTF) is the holy grail,” says Kevin Fitzgerald, senior program director, new ventures group, Invensys (www.invensys.com). How long do you have? Once it’s fixed, where is it on the probability of failure (PF) curve? All the way back? “Equipment manufacturers offer some data, but say, ‘Your installation may be different,’” Fitzgerald says. “It’s a liability issue — they can’t say with confidence.”
Vendors will facilitate collecting, transmitting and analyzing, but interpretation stays with the users.
Falling costs, rising capabilities
Multiple-channel, networked systems have been the key to affordability. “Traditional online systems consisted of a central processor to which all of the sensors were connected,” says Friedman. This meant installing a great deal of cabling, often in conduit, to get the signals to the central processor. “The cable installation costs would often dwarf the actual hardware and software costs required for the online system,” he says.
Hardware costs are falling. “A four-channel system runs $2,500; a 32-channel system is $5,500. That’s $172 per channel,” says Jack Dischner, president, Commtest (www.commtest.com). “With an accelerometer at about $85, it’s less than $300 per channel for hardware.”
The price of sensors has been dropping. “Accelerometer costs are coming down — some are less than $100,” says Brian Graney, product manager, Vibra-Metrics (www.vibrametrics.com). “Sensitivity and frequency response are good enough to get the job done.” You can use temperature, vibration and acoustic emissions sensors that send 4-20 mA signal to multiplexors, or directly to PLCs.
Online sensors don’t have to be as rugged or expensive as mobile sensors — they don’t get handled or dropped, and online, they’re measuring the same point, same pressure or same position, so repeatability is more important than absolute accuracy. “You can’t get away with that with walkaround sensors,” Dischner says.
On the other hand, online applications put higher emphasis on reliability. “As monitoring technology moves from mobile to continuous, it must be made more reliable so it will operate 24/7,” Stone says. “It has to handle lots of data, and get rid of noise so it gives actionable information.”
As instrumentation becomes less expensive, more of it is being installed in more places.
“Oxygen sensors used to cost $3,000, now they’re a few hundred dollars,” says boilermaker Willems. “We can trim oxygen to the most efficient level instead of setting up with excess oxygen, which is less efficient. And we have oxygen readings we can analyze, trend and alarm on.”
Hardware is now easier to install and maintain. “If a $50 sensor costs $2,000 to install and then $200 per year to maintain, it hardly qualifies as low-cost,” says Sabin. Hardware can be DIN-rail mounted or installed in off-the-shelf conduit fittings. Innovative sensor mounting methods don’t require in-shop drilling and tapping of the machine case — they can be mounted in the field using portable drills, adhesives, magnets and the like. “This has a dramatic impact on total installed costs of a system,” he says.
Hazardous area approvals are important for applications in hazardous areas. A sensor-bus architecture typically only turns on one transducer at a time, allowing a single intrinsic safety (IS) barrier to be shared among several dozen or hundred transducers.
Monitoring systems also can share wiring with control and, in some cases, power systems.
“Power cables that take electricity to a motor may also used to transmit data back to a point where they can be diverted for analysis,” according to a presentation by Elsa Anzalone, account manager, Azima (www.azimainc.com) and Jack Nicholas, Jr., P.E., CMRP, senior technical advisor, Allied Reliability (www.alliedreliability.com) at RCM-2006. This requires some sophisticated electronics to protect the sensors, transmitters and receivers from the electrical energy on the power cables, and to ensure it’s unaffected by power line frequency and energy used to drive the machine being monitored.
Vibration, temperature and oil are the traditional mainstays of condition monitoring parameters. But power parameters and ultrasonics are on the rise, and the results of using these and other less-well-known technologies show great promise.
“Load monitoring is a huge application,” says Wiedenbrug. “The load, the motor and the power quality are three links in a chain. You can use the motor as a sensor and find cavitation — it pulls and pushes the rotor. You can use bearing numbers and gear tooth ratios to see gear and bearing problems.”
Analyzing the frequency spectrum and amplitude of motor current spectra can detect degradation and built-in defects not only in some types of AC motors, but also in the machine the motor is driving, say Anzalone and Nicholas. They described emerging technology that can use this information to predict equipment life. When applied to a set of cooling tower fans driven through gearboxes, for example, the system will recognize equipment anomalies that arise from fan degradation, electric line problems, motor degradation, excessive bearing and gear box wear, imbalance of rotational equipment, and misalignment between coupled equipment.
Equipment health is represented as a calculation of the cumulative damage, or ageing, of the equipment “The proposed predictive machinery management systems will use the results of the anomaly detection function and the probabilistic risk engine to estimate an instantaneous value for overall equipment health by combining the condition of all of the equipment’s failure modes into a single representation of health,” Nicholas says. Given the current health and the projected utilization, the system estimates the probability of equipment failure relative to a time scale.
Though not yet credited with the ability to predict TTF, ultrasonic systems are showing high versatility. Contact sensors can detect leaks and mechanical problems, such as whether a valve is bypassing or leaking. “They detect friction and turbulent flow,” says Mark Goodman, vice president, engineering, UE Systems (www.uesystems.com). Non-contact acoustic sensors can hear electrical discharge. “If you have a critical cabinet but don’t want to open it, you can hear arcing and tracking,” Goodman says (Figure 2). “In high-voltage [more than 1,000 V] applications you can hear corona discharges, which you can’t see with infrared.
A chemical company uses sniffers to detect valve leaks, but on outdoor valve sets, if the wind is blowing the wrong way, the sniffer won’t see the leak. Ultrasonic sensors on each side of the valve alarm and trigger an ignition system to flare the leak.
An instant coffee plant uses a baghouse to collect the fines. If a bag develops a hole, the machine sucks up whole coffee beans, which destroy the blowers. An ultrasonic sensor detects a coffee bean banging on the ductwork and shuts down the machine before the bean can hit the fan.
“We have sensors on submerged pump bearings, on ball mills to tell when a load is finished, and for quality control to detect leaks in propane tanks and torque converters,” Goodman says. “Most people think only of vibration for rotating equipment. Ultrasound is more specific — you can hear a single bearing. You need more sensors, but you can tell earlier when a bearing is going bad, or even when it needs lubrication.”
Pressure your OEMs
Plants that want to do more online monitoring should be asking their equipment suppliers to embed appropriate sensors, include condition data in the control systems where applicable, and supply the intelligence to turn data into information, and information into decisions. “OEMs need to embed modules,” says Dischner. “Reliability becomes dependability, and it should be built into the capital investment. Operations should be pushing for this.”
This approach is nothing new to users of large rotating machines, where OEMs began including standard protection systems years ago. “This was largely a direct response of the machinery OEMs to give their customers what they were demanding,” Sabin says. Systems are included in specifications developed by the American Petroleum Institute (API) for pumps, compressors, turbines and gearboxes, and API began specifying acceptance criteria for many of their machines based on vibration readings from proximity probes.
The machinery manufacturer had to make provisions for these probes to do factory acceptance testing of the machine, so it became that much easier for the user to simply specify that these probes remain in the machine permanently, and be connected to some kind of permanent monitoring system.
“Refineries sometimes are having equipment vendors install CM, but not always, so it doesn’t become a standard product,” Fitzgerald says. “Vendors could instrument economically if users would ask for it.”
In the meantime, plants are increasingly able to justify online systems for a larger percentage of their assets, and these systems are delivering substantial value. “ROI is based on avoidable costs,” says Cooper. “Typically, the most recent failure was six or seven figures, and a system is five figures. But it’s presumptive — that a failure will be avoided — and not everyone buys that.”
You also have to figure in the value of efficiency — of being able to plan the work, have parts on-hand, and schedule and deploy crews. “It’s how you do your maintenance: maximize wrench time, use lean crews, empower planners and manage stores and purchasing,” says Fitzgerald.
“You can invest in being proactive or spend it on reactive,” says Brown at Vectren Power. “It’s your choice.”
- Safety: Where equipment type or location can jeopardize personnel.
- Hazardous failures: Where detection prevents consequences such as explosion, fire or hazardous substance releases.
- Remote locations: Where it’s impractical to visit the equipment.
- Sampling interval: When it’s impractical to sample frequently enough to prevent failures.
- Manpower: When manpower is unavailable or too expensive to consistently perform rounds.
- Context: When readings depend on varying equipment loads, process parameters, etc.
- Communications and workflow: Automatic sampling via networks facilitates information flow, analysis, decision-making and follow-up.
- Repeatability: Studies show readings can vary 20% depending on how the technician holds a portable transducer against a machine.
“Online monitoring and diagnostic systems can be piggy-backed on top of existing protection systems,” says Alan Friedman, senior engineer, DLI Engineering Corporation (www.dliengineering.com). “This means that they can use shared sensors or be integrated into existing hardware.”
This approach was implemented at a Consumers Energy plant in Weadock, Mich. An antiquated turbine supervisory instrumentation (TSI) systems with paper recorders was upgraded by adding Rockwell Automation systems to monitor and display condition parameters. “It monitors all the same things — bearing vibration, casing expansion, eccentricity, valve positions, phase relationship — using the original sensors, but it can display more information and make it available to other systems, says Ed Denzer, project manager, equipment services department, Consumers Energy.
Each system is covered by its own screen with built-in predetermined trends. “We can bring it into an office computer and monitor it in real time, and correlate with exactly what’s causing a problem,” says Tim Histed, instrument & control supervisor, Consumers Energy.
It does a better job of integrating supervisory information like valve position and temperature with the vibration information so, for example, we can correlate steam temperature with vibration,” Histed adds. “We did exactly that and solved a problem.”
It’s easier to see the conditions at a specific time, such as when operators say there was a problem or when something happened while no one was there, and the results are easier to present to plant staff. Other equipment can be added to the system as needed. “We’ve already added the boiler, and we’re looking at adding feedwater pumps and fans,” Denzer says.
“We have no DCS and this approach gives us many of the advantages at lower cost than a DCS,” Denzer adds. “In older facilities with equipment scattered all over the plant, there is a great benefit to this approach.”
Vectren Power Supply’s Evansville, Ind., plant is well instrumented, and a great deal of process data flows into the historian. “We could analyze upsets and events after the fact, but we could not anticipate them,” says Scott Brown, Vectren reliability engineer. “The control system alarms are too simple.”
The plant added an Avantis work order system to access the historical data and use it to build condition-based alerts. “For example, a cooling tower has a make-up water pump with a valve at the tower controlling water level in the sump. Low pump amperage might indicate a worn pump, but the valve modulates, so you can’t tell,” Brown says. “But if the valve is 100% open and amperage is low, the pump is worn out. Now we can see multiple variables at the same time, and build rules to give us condition-based proactive alerts.
“It lets us optimize our preventive maintenance. We were calendar-based, but now can create virtual hour-meters and do things based on measurements, like pressure drop across a filter, temperature difference across a heat exchanger, etc. PM savings come from not doing it too often, and not risking failure.
“We have a lot more variables than we have eyeballs to look at them. Now we can have the computer watch things 24/7 and issue a predesigned work order or send an e-mail.”
Many plants have incomplete monitoring on compressed air systems, and are spending $20,000 each for semi-annual audits. The system leak rates increase over time, and to compensate, the system operator may increase the line pressure. High line pressures and leaks waste energy.
The Hollingsworth and Vose paper mill in Ayre, Mass., wanted to detect inefficiency with more frequent inspections but avoid costly system inspections. It sought an approach that would qualify for reimbursement from its local utility, National Grid.
“Proposals for traditional wired networking were too expensive and also fell outside the eligibility range for reimbursement from National Grid,” says Gary Ambrosino, CEO, Sensicast (www.sensicast.com). A SensiNet system was deployed to monitor the compressed air system in the plant.
Wireless capability made the system cost-effective and easy to install in the crowded paper mill. It was configured to monitor key operating variables including temperature, line pressure, airflow and energy use of the compressor. SensiNet software displays the data, and the graphical user interface and data display make it easy to spot trends.
A complete system profile is measured every minute and small changes in system performance became immediately visible. This enables facilities personnel to take corrective action immediately to restore the system to full efficiency and avoid increased costs do to unnecessary air compressor activity.
Results include a 50% efficiency gain on the system, which reduced annual operating costs by $26,000 for a five year projected savings of $130,000. With a system cost of $10,500, payback is five months, and five-year ROI is 1,300%.
“Wireless capability made the system very cost effective,” Ambrosino says, “$10,500 to deploy SensiNet compared to a projected cost of $75,000 for a wired system.” This cost-effectiveness qualified the system for the subsidy, and National Grid reimbursed Hollingsworth and Vose for the full price.