article_097_tnail
article_097_tnail
article_097_tnail
article_097_tnail
article_097_tnail

Machines staying healthy by staying smart

May 9, 2006
Machine health monitoring is going beyond mainstream vibration and oil analysis to embrace increasingly high-resolution sensing and prioritized data processing technologies.
By Jim Montague, Executive Editor

IF YOU'VE got your health, then you’ve got everything. Though slightly condescending, this old saw is as true for machines in process applications as it is for the people that operate them. And, in yet another similarity to human healthcare, recent advances in sensing technologies are allowing users to look ever deeper and with greater resolution into their machines and process applications. This and focused handling of the resulting data is allowing many users to find and diagnose potential problems even sooner, and make more intelligent decisions on scheduling maintenance, allocating resources, managing inventory, and scheduling to minimize the cost of maintenance and repairs.

Organizationally, it’s true that machine health monitoring and/or management fits within larger, less tangible concepts, such as condition monitoring, preventive and proactive maintenance, asset and lifecycle management, and even enterprise resource planning. However, machine health comes in far more direct contact with the processes it monitors than any ostensibly higher-level methods, and often helps supply them with data they need to function.

Healthy Evolution 
“As users move from straight, time-based or run-to-failure maintenance, they find they need information on the condition of their assets. Next, it’s important to integrate that physical, process data into a condition monitoring platform, which can correlate it, and help measure performance,” says Scott Breeding, product line leader for Bentley Nevada, a division of GE Energy. “This is what gives users an indication that a bearing may be running hot before they can see any signs of wear or damage. For example, an engineer may use process data indicating a vibration to deduce that a centrifugal pump is cavitating because an operator is applying inadequate suction-head pressure, and this can support a decision to improve that process.

“Process data also can be used to write new rules around assets, derive subsequent indicators, and screen data to produce alerts when a problem may be likely to occur. This allows users to focus more closely on the ‘bad actors’ in their applications.”

     FIGURE 1: BAD VIBES

An amplitude/phase/time (APHT) plot of vibration data from the drive rollers on the Wisaforest mill’s lime kiln shows their motor speeds varying between 991 and 1,083 rpm, and demonstrating a structural resonance at 36 Hz as the kiln moved from low- to higher-speed operation, which required stiffening and strengthening the drives’ supports.

Similarly, using screened process data, and performing failure mode of effects analyses (FMEA) can help users pick the most appropriate machine health technology for each asset based on its relative criticality. These typically include trebology (lube oil analysis), vibration analysis, thermography, online and offline motor current monitoring, and several other less well known technologies. “This is more than protecting a machine when it trips. This is showing operators where they’re causing wear or breaks, and preventing trips entirely. And the best users put this information right in front of their operators, and even those that want to keep on operating until failure can do it as an intelligent decision.”

For example, Bently Nevada recently implemented its System 1 software, Trendmaster Pro data acquisition hardware and Dynamic Scanning Modules, and 3500 monitoring system and proximity probes on almost 50 process machines at UPM-Kymmene’s Wisaforest pulp and kraft/sack paper mill in Pietarsaari, Finland, and reportedly has solved 10 separate machinery problems. This project also was part of the mill’s overall installation of a new recovery line to maintain its production of 800,000 air-dried tons per year.

Though most of its machines use rolling-element bearings monitored by accelerometers, the mill’s huge lime kiln uses fluid-film bearings monitored by X-Y proximity probes and accelerometer transducers. To let operators see basic condition information data and let rotating machinery engineers see diagnostic data, the mill’s process control system set up a bi-directional OPC interface, which imports amplitude and alarms from System 1 into the DCS, and exports process variables from the DCS into the System 1 database. The main problems that Bently Nevada helped solve at the mill included:

  • Identifying intermittent, high-frequency vibration amplitudes in the inboard bearing in the recovery boiler’s 600-kW air fan, which helped the mill’s engineers better schedule lubrication system repairs and bearing replacements.
  • Correcting elevated vibration levels in the 15.4 x 443-ft lime kiln’s drive rollers by evaluating their phase, rpm, amplitude and frequency data, which found a structural resonance at 36 Hz as the kiln moved from low- to higher-speed operation, and required stiffening and strengthening the drives’ supports (See Figure 1).
  • Locating a cracked inner ring in a bearing in the recovery boiler’s exhaust gas fan, shown by a characteristic “ringing” phenomenon in the amplitude/phase/time (APHT) plot for the fan, which was repaired.
  • Identifying excess vibration and a faulty rubber element in a coupling on the recausticizing unit’s white liquor pump, which also was repaired.
  • Finding inboard bearing deterioration in the recausticizing unit’s totally unmonitored mixer, which was adjacent to its monitored rotary filter, and required repairing the mixer’s a broken lubrication pipe.

Implementing these solutions reportedly helped the Wisaforest mill achieve payback on its upgrade investment in just eight weeks.

In addition, though basic vibration has been used to check machine health for decades, new algorithms and software are helping users monitor more operating parameters and applications. For example, proximity sensors and software included in Bently Nevada’s Asset Condition Monitoring Solution can take signals that indicate the trueness of shafts in motors. “We have new algorithms and software that can set up alerts and alarms based on asset conditions,” says Jeff Schnitzer, Bently Nevada’s general manager. “We also have Rule Packs that allow users to write rules based on vibration, temperature, or other indicators, and create subsequent functions and responses.

“These and other machine health monitoring capabilities can then be combined with maintenance records, assets classifications, and other data to create baselines for entire plants, and help target future investment where it will return the most improvement. This means no longer touching or replacing components based on past outages and history, but touching or replacing them based on real-time reliability data.”

Sharpening Senses
Robert Skeirik, machinery health product manager for Emerson Process Management, says it links the five main senses of machine health—vibration, oil condition, motor current, infrared thermography, and ultrasonics—into its AMS Machinery Health Manager, AMS Suite software, CSI 2130 portable analyzer, CSI 4500 analyzer that connects to a multiplexer, and CSI 9210 machinery health transmitter.   

Skeirik adds that Emerson and its CSI division recently helped upgrade the 20% worst-performing of 27 coal pulverizers at American Electric Power’s (AEP) John E. Amos plant in St. Albans, W.V. The units were previously serviced on a 12-month schedule, and collectively experienced 3-4 unexpected failures per year.  Besides preventing $240,000 worth of breakdowns in the project’s first year, CSI and Emerson helped AEP and Amos’ engineers achieve savings and return on investment (ROI) in several machinery health monitoring areas:

  • Lube oil analysis saved more than $74,039 for a 4-to-1 ROI
  • Infrared thermography saved more than $343,000 for a 5-to-1 ROI
  • Motor current analysis saved more than $344,000 for a 20-to-1 ROI
  • Motor testing saved more than $381,000 for a 20-to-1 ROI
  • Leak detection saved more than $1.2 million for a 60-to-1 ROI
  • Vibration analysis saved more than $4.2 million for a 42-to-1 ROI
  • Total savings of $6.5 million for a 27-to-1 ROI

“Machinery health is now just one component of a plant’s overall status, which is continually seeking to answer the plant manager’s two basic questions about reliability and availability,” says Skeirik.

An Ultrasonic Heartbeat 
Though vibration detection and analysis is a bread-and-butter machine health monitoring method, and one that’s also increasing in sensitivity and sophistication, some developers believe ultrasonic monitoring may be even more useful. Based on military and aerospace monitoring methods, Stress Wave Analysis (SWAN) technology from Swantech reportedly can detect even earlier when a machine is starting to demonstrate symptoms that will lead to failure (See Figure 2 below).

FIGURE 2: SOUNDING STRESSED
Ultrasonic stress wave energy (SWE) plots for two machines show the one on the right was damaged in a “seeded fault” by an insufficient quantity and inappropriate quality of lubricating oil. The plots indicate the SWE levels on the “faulty” machine were immediately higher than its twin, trended upward, and accelerated, as the machine accumulated actual wear and damage.

“Ultrasonic technology provides the earliest detection of machine problems by detecting the unique sounds made by friction, impact events, and minor surface damage,” says Ralph Genesi, Swantech’s president and CEO. “Traditional machine health has been based on vibration, but this is too late because some damage may already have occurred and a safety shutdown may be needed. Ultrasonic detection gives users more time to gain the knowledge to plan maintenance and a response, and lets users decided if they need to shut down now or if they can safely run their planned quota.

“In addition, electrical signal histories collected over time, or histograms, can show if new signals are skewed, and even help indicate if a machine has lube problems, cracks, or seal damage. Ultrasonics can even show when an external load is placed on a machine, which can help operators change bad habits, and extend the lives of their devices.”  

For example, to help it process more than 2 million tons of reclaimed steel per year, North Star BlueScope Steel in Delta, Ohio, recently asked Swantech to help maintain its flat roll machines, which usually run 24/7 and can cost $1,000 per minute of unplanned downtime. The steelmaker’s SWAN system gathers information from Stress Wave data collectors associated with each bearing, analyzes it on a SWANserver in the plant’s offices, reports on any early signs of wear in the mill’s equipment, and allows managers to plan maintenance or repairs up to six months in advance.

“We’re concerned mainly about failed bearings and failed gears,” says Matt Morris, North Star BlueScope’s reliability-team leader. “All of the gearing is unique to the machine, and we have no spares. Because getting a new one usually takes at least six months, limping along or shutting down completely can add up to large losses. However, the field of available methods is very narrow because many of our large bearings turn very slowly. Also, the sensors mounted on each bearing have to withstand the high temperatures and heavy vibration in our processes.

“Initially, the SWAN system was standalone, but it was eventually added to our Level 1 network for backup purposes. Now, it’s also accessible through routers from our main Level 3 business systems.”

Not only has access to this information reduced production losses, but it also has allowed NorthStar to shrink its $3 million inventory of spare parts, and avoid paying thousands of dollars for rush orders. 

Networking, Education Needed
While machine health monitoring technologies have made technical advances on their own, they’ve also been aided by increasingly reliable and widespread networking solutions and protocols. Breeding says developers had been trying to integrate process data for years, but that previous networks weren’t durable and secure enough. Now, not only are hardwiring and connectors more robust and less expensive, but fieldbus protocols, such as OPC, Profibus, Foundation fieldbus, and Modbus are becoming more standardized and accepted by end users. Likewise, many former worries about network security have been lessened by the fact that many process industry companies have established virtual private networks (VPNs), and are growing more comfortable about allowing authorized access to them. Also, once any initial network connection is made, it’s possible to distribute machine health monitoring data up to the enterprise level, over wireless links, or via the Internet.

Though many increasingly sophisticated technologies are coming into play in machine health monitoring, none can be effective if they simply aren’t used. “Many customers think their job is done when they buy a machine health product. Our company found that many users had very nice monitoring technologies suffering from ‘dusty keyboard syndrome” because they were still using schedules and weren’t relying on condition-based monitoring in their daily routine,” says Steve Sabin, editor of GE Energy’s Orbit magazine. “Though it started out making instruments, this was why Bentley Nevada started its machine diagnostic services business. Having monitoring instruments was fine, but companies that got the best results were those that trusted their instruments to drive their behavior, and change they way they did their work. In the best plants, 80% of the maintenance work is planned, while 20% is reactive, but this still isn’t they way most plant are run.”

So, while machine health monitoring is still solidly based on empirical observation, the power of that observation, the focused analysis of the resulting data, and the faster and wider distribution of its results can give user unprecedented benefits, but only if they use them.