By: Robert Swanekamp, P.E.,
Real-time monitoring of flame instability can save gas turbine owners from millions of dollars in damage due to “combustor humming.”
The development of dry low NOx (DLN) combustors for gas turbines has cut the industry’s consumption of water, while dramatically slashing air emissions. Unfortunately, this environmental enhancement has created an engineering challenge: The lean-burning DLN combustors are prone to flame instability, which can cause pressure pulsations large enough to destroy the combustor, launching debris downstream where it annihilates other hot-gas-path components. To detect and correct flame instability before it causes millions of dollars of turbine damage, savvy users have begun to install advanced sensors and robust software that monitor the combustor dynamics in real time.
For decades, gas-turbine emissions were controlled by the injection of steam or water into a type of combustor known as “diffusion flame.” The technology was reliable, but it consumed precious water, eroded combustor hardware, and achieved only moderate NOx reductions. Today’s industry standard is the DLN combustor. Rather than inject water or steam, DLN combustors lower the peak flame temperature (hence the formation of thermal NOx) by burning a leaner mixture of fuel and air. Early DLN designs typically reduced NOx emissions to the range of 25-42 ppmvd, while more advanced designs burn an even learner mixture to yield single-digit NOx emissions.
For example, the DLN-2, developed by GE Energy for its F-class gas turbines, limited adiabatic flame temperature to approximately 3000 F and guaranteed 25 ppmvd NOx. The manufacturer’s newer offering is the DLN-2.6, which limits adiabatic flame temperature to approximately 2800 F and achieves a remarkable 9 ppmvd NOx, without any post-combustion cleanup. The newer combustor physically differs from the DLN-2 by the addition of a sixth burner located in the center of the five existing burners, hence the “2.6” moniker. By fueling the center nozzle separately from the outer nozzles, the fuel-air ratio can be modulated relative to the outer nozzles for finer tuning of the combustor.
Today, all of the major gas turbine OEMs offer DLN combustion technology, and collectively they have accumulated millions of hours of low-emissions operation. Collectively, they also have accumulated scores of combustor failures due to dynamic pressure pulsations.
Regardless of manufacturer, DLN combustors operate at ultra-lean fuel/air ratios, on the ragged edge of flame instability. This inherently makes them more susceptible to pressure pulsations, which can be large enough to cause serious vibration in the combustor and an audible rumbling that is euphemistically referred to as “humming.” At the recent HRSG User’s Group Annual Conference (see sidebar), one plant reported that the humming from its DLN-equipped turbine actually shattered a window in the plant’s Administration Building. And that plant got off easy. When the vibrations are too large in amplitude, or when they occur at frequencies corresponding to natural resonances in the system, they have caused fatigue failure of combustor components, which when broken free are launched downstream to inflict secondary damage on other hot-gas-path components.
Gas-turbine designers are working to increase the flame-stability margin of their commercial offerings, by such steps as varying the combustor geometries and changing fuel-injector manufacturing tolerances. But to a large degree the stability margin is a function of site-specific, dynamic parameters—including fuel composition, the amount of wear on combustion-liner seals, and ambient conditions. As a result, each DLN combustor needs to be tuned for each site, on a recurring basis. “You have to tune these things individually,” explains one gas-turbine specialist, “because you just can’t simulate the interaction of all those changing parameters on all those combustor cans.” Many plants follow a regimen of semi-annual tuning as the seasons change, typically bringing in the OEM’s specially trained engineer to perform the sensitive adjustments using portable pressure-monitoring equipment.
A better approach—and one that is increasingly being accepted by OEMs, plant owners, and insurance carriers—is to permanently install an on-line monitoring system that continuously measures the dynamic pressure pulsations and provides early warning that the combustor is out of tune. According to Jeff Fassett, president of IEM Energy Consultants, Alexandria, Minn., the problems that can be detected through the proper application and use of these systems include:
- Combustion system hardware incipient failure
- Fuel control-valve calibration deficiencies
- Fuel nozzle pluggage/bypass
- Fuel-gas purity problems
There may be secondary benefits as well. For instance, a utility in the Southeast is currently planning to install a combustor dynamics monitoring system on two of its F-class gas turbines. The utility’s primary goal is to extend the service life of the combustors. It also hopes that the on-line system will help keep the combustors so well tuned that the gas turbines can safely be over-fired to produce more output during the hot summer months. The utility currently relies on inlet fogging to boost turbine output in the summer, but it wants to eliminate the use of that technology because of substantial erosion occurring in the first-stage blades.
To date, most combustor dynamics monitoring systems have been installed by the OEMs, primarily after modifying or changing the combustion hardware to ensure that the new hardware meets performance expectations. For example, Siemens-Westinghouse Power Corp. installed its first system nearly two years ago on its new W501G turbine. The W501G is Siemens-Westinghouse’s most advanced commercial offering—a partially steam-cooled gas turbine expected to deliver 58% combined-cycle efficiency (LHV) by incorporating closed-loop steam cooling of the first-stage vanes, and subsequently achieving higher turbine-inlet temperatures. Significant pressure pulsations at approximately 2200 Hz had been observed on the W501G during dynamics testing. The cause was identified as insufficient aerodynamic damping as a result of closed-loop steam cooling of the transition section. New resonators reduced the pressure pulsations, and as a further precaution Siemens-Westinghouse added a combustor dynamics monitor as standard supervisory instrumentation on the W501G.
A handful of non-OEM companies also offer combustor dynamics monitoring systems. These include Power Systems Manufacturing, Boca Raton, Fla., Control Center LLC, Orlando, Fla., and Alta Solutions, Escondido, Calif. The Control Center package has been installed on turbines from various OEMs, and has been endorsed and approved by one of the OEMs.
Most, if not all, of the systems from the turbine OEMs and from these non-OEMs use a pressure sensor that is located on the outside of the gas-turbine compartment, in a standoff (or “infinite”) tube that protects it from the high-temperature environment of the combustion chamber. This indirect measurement permits the use of widely available industrial pressure transducers such as strain gauges. The drawbacks, however, include a reduction in sensitivity due to pressure losses in the standoff tube, frequency-dependent attenuation of the signal as it travels down the tube length, and the need to routinely purge the standoff tube with nitrogen to eliminate condensation.
Figure 1. Piezoelectric sensors located on the combustor chamber enable direct measurement of pressure pulsations in sensitive DLN combustors. Photo Courtesy of KEMA.
Wanted: New Sensors
Because of these disadvantages, engineers have been searching for an alternative to the indirect measurement of dynamic pressure. That’s a significant challenge to sensor developers because of the high-temperature environment of the gas turbine combustion chamber. At a direct pick-up point, the sensor would be exposed to temperatures over 1000 F.
One alternative sensor currently in development is an electrostatic probe that would be integrated with the gas-turbine fuel nozzles. In a paper presented at POWER-GEN International 2003 in December, Tim Lieuwen of the Georgia Institute of Technology and George Richards of the U.S. Department of Energy (DOE) reported that the probe operates by recording the electrical conductivity of the flame as measured from two electrodes built into the fuel injector. Because the conductivity of the flame is related to the fuel-air ratio, the electrode close to the flame can record a signal that is related to the overall fuel-air ratio. The adjacent electrode can detect the transient passage of the flame into the premix passage during combustion dynamics, and provide a useful signal for detecting both pressure pulsations and “flashback”—a separate failure mechanism that can occur in dual-fueled DLN combustors. The electrostatic probe was developed at the DOE’s National Energy Technology Lab and currently is being commercialized by Woodward Industrial Controls.
Another advanced sensor, which allows direct measurement of dynamic pressure, is already proved in the field. Based on piezoelectric crystals, this sensor can be installed directly on the combustor chamber and reportedly will retain its pressure-measurement properties despite the high-temperature environment (Figure 1). One of the suppliers of piezoelectric sensors is Vibro-Meter SA of Switzerland. Its sensor is applied in combustor dynamics monitoring systems offered in North America by SKF USA, Kulpsville, Pa., and KEMA, Burlington, Mass. SKF and KEMA collaborated closely on the deployment of this system, which links the new piezoelectric sensor to KEMA’s analysis software, a trademarked product called “FlameBeat.”
Another advantage of the Vibro-Meter SA probe is that it automatically compensates for the system vibration inherent in any mechanical application. This yields a purer dynamic pressure signal being generated by the sensor, and consequently provides greater accuracy in the data.
The SKF/KEMA combustor dynamics monitoring system has been installed at five European sites, as well as two North American facilities on two different gas-turbine models: an Allegheny Energy plant powered by a W501F, and an AES Corp. plant powered by a 6FA. Since the installations, the 6FA turbine has accumulated far more service time because that plant is operated in baseload mode, while the W501F is in peaking duty.
The 6FA installation was completed at the AES Kingston plant in Bath, Ontario, in July 2003. As Plant Manager Steve Collings explains, the Kingston DLN-2 combustor had been operating without dynamics concerns since commissioning in 1997, with only minor tuning occasionally required. But the tuning performed in September 2002 following a hot-gas-path change-out introduced substantial changes to combustor settings. Significantly, the September tuning was conducted during warm weather. As the Ontario winter set in over the next few months, the combustor developed severe pressure pulsations and in December the combustor failed catastrophically, wrecking many of the new hot-gas-path components with domestic object debris.
Repairs were made, the combustors were retuned (using the customary portable monitoring equipment), and the plant was restarted. But after only 590 service hours, Collings and his team noted that exhaust-temperature spreads were increasing, and when they shut down to borescope they found substantial damage had already occurred to the new combustor hardware. Fortunately, no combustor components had yet broken free, but the root cause of AES Kingston’s combustor dynamics problem clearly had not been found.
So in January 2003, AES Kingston and GE rigged up the portable monitoring equipment and linked it to the OEM’s remote monitoring center in Atlanta. In parallel, AES Kingston awarded a bid to SKF/KEMA to install a permanent dynamics monitoring system by July 2003. Because AES Kingston is dispatched off-line most weekends, the plant also had the opportunity to perform frequent borescope inspections, to keep a close eye on the problem. Although the OEM’s dynamics monitoring equipment showed no serious pressure pulsations over the next several months of operation, a borescope inspection in June 2003 revealed cracking was occurring on the fusion plates around the burner tips. When the end caps were removed, Collings and his team found burner tubes “ready to liberate.” So for a second time, the root cause of excessive pressure pulsations had not been found, and a turbine wreck had been avoided only because of exceptional vigilance.
Fortunately, the permanent SKF/KEMA system, with its piezoelectric sensor and analysis software, was now ready and was installed at this time. The 6FA is a cannular design, and therefore it requires one dynamic pressure sensor per combustor. The characteristics of the dynamic pressure spectrum are significantly influenced by turbine operating conditions, so it is essential for effective diagnostics to merge the dynamic pressure data with Distributed Controls System (DCS) data. The SKF/KEMA system accomplishes this at a software database level by a dedicated server computer, which imports operating data from the DCS via standard network protocols and performs robust analysis in the “FlameBeat” software (Figure 2). Note that the AES Kingston system, like all combustor dynamics monitoring systems installed on GE turbines, is operated in a “passive mode,” with the on-line system providing only indication and alarms to plant personnel, and manual intervention required to actually vary the gas-turbine controls. Some OEMs will allow the combustor dynamics monitoring system to operate in an “active mode,” with the system able to make automatic changes in gas-turbine control.
Figure 2. Robust software and skilled analysts are needed to interpret the data collected by a combustor dynamics monitoring system. Photo Courtesy of KEMA.
After its on-line combustor dynamics monitoring system was installed in July 2003, the AES Kingston team—which now included analysts from the plant, the OEM, SKF, and KEMA working together—had greater amounts and more accurate data to study. As explained in the technical paper “Dynamic Pressure Monitoring in Gas Turbines,” authored by Chris G. James of SKF USA and Adriaan J. L. Verhage of KEMA, the interpretation of flame-instability data is a learning process, communicated to the analysts by dynamic pressure spectra. After months of observation with changing turbine speeds, starts, stops, and occasional trips, the correlation between observed pulsation signatures and other turbine parameters are revealed.
At AES Kingston, months of observation with the new combustor dynamics monitoring system, coupled with a weekend of testing with the OEM’s tuning engineers, revealed that the combustor damage was being caused by very low-amplitude pressure pulsations (as low as 0.4 psig) in the high-frequency range (approximately 2400-2500 Hz). Previously, pulsations below about 1.0 psig in this so-called “screech zone” were not thought to be destructive.
With its newfound understanding, AES Kingston could tune its combustor to stay below the new, lower alarm limits, and the unit has performed reliably since. Recent borescopes, after eight months of operation, show the combustor to be “in mint condition,” Plant Manager Collings reports. “We now know the danger areas, and can stay out of them.”
HRSG User’s Group Workshop Will Address Instrumentation
During the past decade, the Annual Conferences of the HRSG User’s Group have become the premier gathering place for owner/operators, manufacturers, and other combined-cycle/cogen professionals to discuss the steam-plant issues facing their power plants. But with the combined-cycle industry expanding and field experience growing, the HRSG User’s Group has begun to supplement its wide-ranging annual conference with a more in-depth, technical workshop focused on a specific set of maintenance issues. Last October, the HRSG User’s Group focused its Maintenance Workshop on (1) Managing P91/T91 Components, and (2) Optimizing HRSG Layup.
This fall, in conjunction with POWER-GEN International 2004, the HRSG User’s Group will host a Maintenance Workshop focused on instrumentation and diagnostic systems for combined-cycle/cogen plants. Day One of the Workshop will cover design issues, case studies, and O&M experience covering the optimum locations, brands, and types of field devices—such as thermocouples, pressure transducers, flow meters, on-line chemistry analyzers, and so on. The Workshop also will explore software and data acquisition issues, including DCS and duct-burner logic choices, types of data archives, and analysis software.
The second day of the Workshop will focus on operating procedures that boost HRSG durability and safety. Veteran users, manufacturing specialists, and consulting engineers will examine the wide variety of startup and shutdown procedures in use today at combined-cycle/cogen plants—including those affecting the critical control of water chemistry.
The HRSG User’s Group Maintenance Workshop will be held November 30-December 1, in conjunction with POWER-GEN International 2004 at the Orlando County Convention Center in Orlando, Fla.