After the safe shutdown, Barnes adds that Entergy's engineers diagnosed and disassembled the unit, and found a 2.5-inch deep crack running 180° around the end of the rotator shaft (Figure 3). Analysis revealed that the new crack was due to stress fatigue from a previous arc gouge that had been repaired several years earlier.
"The consensus of our rotating equipment and generator experts was that, if we hadn't taken this unit offline, then the shaft would have cracked through, and we would have had a catastrophic failure. The 3,600 rpm section would have hit the stationary area, which would have completely wreaked the generator, done serious damage to the turbine, and probably caused hydrogen coolant and oil fires. In other over-speed events, large pieces of metal also have penetrated their facility's shell, and flown up to a half mile away. An incident like this would have caused $40-50 million or more in damage, and this unit would have been down for months, if not years. Because of our safe shutdown, the repair cost only $5 million, and the unit was only down six weeks."
Seeking, Showing Small, Significant Signs
Tim Holtan, manager of SmartSignal's Availability and Performance Center in Lisle, Ill., reports that, "The initial value proposition of early warning in process applications is revenue improvement, but we're seeing signs that SmartSignal can mitigate a lot of accidents, too."
For example, Holtan reports that SmartSignal is used at another client's major refinery to detect very early signs of possible leaks into the seal oil system on a centrifugal compressor that circulates hydrogen through the desulfurization unit. These leaks occur over time because the seals are in contact with and eventually degraded by H2S-rich gas in the unit. The seal oil is circulated to a reservoir that is open to atmosphere, and in the past that reservoir could be vented to flare when an H2S alarm sounded. However, EPI*Center can detect subtle but significant shifts in temperature, indicating that a leak may be starting, and then quickly call for the oil to be changed, so the process doesn't have to go to flare so often.
To process applications' data and detect changes, EPI*Center begins with a personalized, empirical model of each piece of equipment in the process where it's being implemented. These data may include pressure ratios, temperature ratios, power inputs and other typical operating norms. Next, these models are compared to actual operating data, which reveal the small but significant shifts that don't turn up with traditional performance methods.
"Another refiner using EPI*Center found a decrease in efficiency in their diethanolamine system, which picks up H2S in the process unit and delivers H2S to the filter unit," says Holtan. "Only four measures are collected—suction flow, suction temperature, motor current and discharge pressure—but our model was able to pick up 25 gpm less suction flow in one pump, even though it was using the same power.
"So, the refiner's staff shut the pump down, investigated, and found the cause was significant corrosion. If that line had broken, there could have been an H2S leak and exposure to staff and the atmosphere. This pump's loss of efficiency wouldn't have been noticed otherwise because it was no one's job to check what amps it was pulling. This is how we get to the needle in the haystack. There are different levels of attention paid to equipment in facilities, and so EPI*Center can monitor 400 pieces of equipment, but maybe only send alerts on the two or three that need it.”
Taking Control of Downtime
Because most U.S. power generating utilities are past their designed lives, Barnes reports this makes it even more important for these older and often smaller units to have the right sensors and instruments in place. "There's always some risk in running and maintaining any power generation system, and so it's also part of Entergy's policy to have pattern-recognition capabilities on all our units. EPI*Center mitigates our failure rate, and improves our cost of doing business. In fact, we had zero recordable accidents in 2008 for our fossil operations.”
Likewise, Barnes adds that having PI data and past failure data makes it easier to do failure modes and effect analyses, and then build in improved safety features to make future machines, applications or systems less prone to failure. "For us, the real, day-to-day benefit of detecting problems early is simply that we can have time and ability to make informed decisions," explains Barnes. "Consequently, instead of simply running a process until we have to shut down a unit right away, now we can know enough about a problem to know how long we can continue to safely run its process—perhaps for additional days or weeks until we can fix it as part of a planned outage. This lets us decide when to shutdown and make repairs, and makes it easier to keep other processes running. In general, unplanned outages are more costly than planned ones, and so it's especially useful that we can now turn unplanned outages into planned ones. For example, we buy a lot of our fuel on the spot market, and so if we know that we're going to have a shutdown soon, then we can buy fuel ahead of time and not have to buy it when it's more costly.”
These and other benefits are how PMDC helps save Entergy millions of dollars per year in avoided repair costs, according to Barnes, who adds these savings quickly grow to tens of millions of dollars when the costs of saved replacement power and avoided repair costs are added. "We used to rely on traditional plant alarms and staffers making the rounds, and they're very competent, but they're also busy with day-to-day operations. They don't have the time to pull up weeks of data, and look for potentially beneficial trends," explains Barnes. "This is why it's good that our analysis technology and procedures have evolved from having a failure, looking at strip charts, and trying to figure out what happened with simple root-cause analysis to, now, gathering more subtle indicators, proactively figuring when something will need to be fixed in the near future, doing more minor repairs, and avoiding catastrophic failures. Not only is there less risk to staff and facilities by doing proactive repairs, but our whole process is now inherently safer. Everyone should implement this kind of monitoring to identify the anomalies in their mountains of data, train their engineers and operators what to look for, and use this new data to make better judgment calls."