In Condition Monitoring, Earlier is Better

Jan. 1, 2006
Innovations help manufacturersdetect problems sooner to avoid costly breakdowns.

A 20-degree Fahrenheit rise in a bearing’s temperature in just three hours did not escape the attention of the reliability engineering group at the Sundance power plant near Edmonton, Alberta, Canada. The company’s condition-monitoring system had flagged the slight rise and posted the 6,000-horsepower boiler feed pump to the watch list. The posting occurred automatically as the system continued to track more than 3,500 parameters there and in the two other power plants in the Alberta Thermal group of Calgary-based TransAlta Corp.

And it was a good thing that it did. The system had given the reliability group the early warning that it needed to prepare for the marked change that would occur seven days later, when the temperature would surge by nearly 70 degrees F.

Closer tabs

The incident demonstrates the tremendous insights that Alberta Thermal has gained into its operations ever since TransAlta joined the ranks of progressive corporations investing in condition-monitoring equipment. In TransAlta’s case, the upgrade was part of a wider initiative to infuse its fleet of facilities with wireless automation to keep closer tabs on the aging equipment in some of its 49 plants throughout North America and Australia. Consequently, the monitoring system was built on a wireless communications network of sensors and other devices that report to a software application known as EPI*Center, from SmartSignal Corp., of Lisle, Ill.

Corporate management selected the six-unit Sundance facility for the first infusion because the coal-fired facility, the largest in western Canada, is more than 30 years old. Management then connected the other two plants in the Alberta Thermal group to the EPI*Center at Sundance. The Operations Technology Team responsible for the installation could then spread the success to other plants and groups within the company.

The key to the monitoring system’s success is its ability to detect small deviations and isolate the ones that are unusual, such as the rise in bearing temperature on the boiler feed pump. “By itself, a 20-degree rise in a bearing temperature isn’t necessarily a significant indicator,” notes Keith Christianson, a project manager for TransAlta’s Generation Operations. “Throughout the day, load varies, ambient temperature changes and so on. When it doesn’t come back down, though, you have to start asking what’s going on.”

Since the bearing temperature failed to fall as usual, the reliability-engineering group began asking that question, and dispatched technicians to provide the answer. After conducting a vibration analysis, the technicians reported that the bearing was beginning to fail. Just when it would fail was anybody’s guess, but the engineers doubted that it would last until the next scheduled downtime for routine maintenance. They, therefore, went about formulating a repair plan, which included gathering the necessary parts and instructing the operator to take the pump out of service immediately when the deterioration progressed enough to trigger the alarm.

Cost avoidance

“Normal practice is to run the pump until an operator has an opportunity to investigate the problem,” says Christianson. “If we had done that, it would have scored the shaft minutes later, and we would have had to replace that 6,000-horsepower motor to have the shaft rebuilt.” Because the parts were on hand and shutting off the pump immediately limited the harm to minor surface damage, maintenance had it back on line in a record nine hours, which is about 25 percent less time than would otherwise have been required. Moreover, the repair was much less costly.

The reliability-engineering staff credits results like this largely to a predictive model that vendor SmartSignal calls similarity-based modeling. The models in most other condition-monitoring software that the staff had evaluated focus on engineering design specifications, such as the maximum operating temperature for a particular bearing. SmartSignal’s model, on the other hand, computes where the bearing temperature, pressure, flow and other readings should be now in real-time, based on trends in a year’s worth of readings stored in the historian. “If it’s outside the expected value by a statistically significant amount, then you post an alarm to the watch list, and somebody checks on it,” says Christianson.

The software scans the data in Alberta Thermal’s historian every five minutes. The engineering staff took SmartSignal’s advice to scan in ten-minute intervals at the start, but decided later to cut the time in half to catch events—such as vibration during start up—that happen too fast to notice in the larger interval. Consequently, little gets past the software. “It’s like having your best operator on 24-seven, looking at every piece of equipment in the fleet every five minutes,” says Christianson.

He also finds the software’s method of calculation to be better suited for reliability and maintenance. Software based on design limits assumes that each unit in a particular make of boiler feed pumps, for example, are essentially identical, because they are built in the same way to deliver the same flows and operate at the same pressures and temperatures. “From a plant engineer’s perspective, however, the pumps are very different,” notes Christianson. “One might have been rebuilt. Another might have a new motor. Yet another might have coupler problems and so have high vibration.” The EPI*Center accounts for these differences.

Results like these are the product of continuing technical advances in condition-monitoring technology, the biggest of which have been computerized data acquisition, analysis and reporting. “In years past, it was time-prohibitive to collect large amounts of condition monitoring data because it was largely a manual process,” says Jim Frider, product manager for plant intelligence products at Wonderware, a Lake Forest, Calif.-based unit of Invensys Systems Inc. “Modern plant automation and computerized monitoring can collect data cost effectively across an entire process plant and provide a wealth of accurate, timely and detailed information.”

Frider adds that manufacturers of sensors have made significant contributions, too. “Modern solid-state sensors have lowered the cost to monitor plant equipment, and have improved the reliability and quality of the data,” he says. “Now, thousands of low-cost sensors can be deployed in a process plant not only to help control plant processes, but also to provide real-time data to condition monitoring systems.”


Moreover, vendors have been busy developing some unusual applications. An example is the ultrasonic sensors that North Star BlueScope Steel (NSBS), of Delta, Ohio, uses for detecting the first signs of stress and wear on the massive rollers working 24 hours a day to produce more than 2 million tons of steel a year. The technology is helping the flat-roll steel producer to slash unplanned downtime, which can cost more than $1,000 per minute.

“We are concerned mainly about failed bearings and failed gears,” says Matt Morris, 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. In fact, the cost pressures are so intense that the company once replaced a cracked gear temporarily with one made of softer material because it was faster to make than the final hardened one and could serve as a crutch in the meantime.

Given these pressures, the reliability team decided to replace manual vibration analysis and infrared thermography with a condition-monitoring system capable of detecting problems much earlier and automating data collection and documentation. “The field of methods was very narrow, since many of our large bearings turn very slowly,” recalls Morris. Moreover, the

sensors had to withstand the high temperatures and heavy vibration in the process.

The technology that fit the bill was ultrasonics, from Swantech, of Fort Lauderdale, Fla. The vendor’s stress wave analysis measures slight shocks and even friction as much as six months earlier than the condition-monitoring technologies that NSBS had been using. “Conventional accelerometers (vibration sensors) have been used, and still are the sensors used in most condition-monitoring applications,” says Ralph Genesi, chief executive officer of Swantech. “However, the information that they can provide is limited by the physics of their design.”

“Hearing” friction

He notes that ultransonics and other sensing technologies have been evolving to overcome these physical limitations and to provide more levels of information. An analogy that he draws is the various levels of diagnostic equipment used by the medical profession. “Ultrasonic technology provides a new dimension to measurement and deeper insight into the health of a machine in real time,” he says.

The ability to “listen” for friction allows users to detect and track the stress and the progression of wear and other damage from the beginning.

The sensors are mounted at each bearing and send a continuous stream of data directly to the Swantech card in the server. “Initially, the Swantech system was stand-alone,” says Morris. “Over time, though, it was added to our level 1 network for backup purposes. Now, it is also accessible through routers from our main business systems (level 3).” Not only has access to this information reduced production losses, but it also has allowed NSBS to shrink its $3 million inventory of spare parts and avoid paying thousands of dollars for rush orders. So the consensus there is that learning of problems earlier is indeed much better.

For more information, search keywords “condition monitoring” at

See sidebar to this article: Avoid These Mistakes