Erosion during on-line compressor washing and subsequent blade failures on some F-class gas turbines has prompted a major study of the effects and influence of an air-assisted washer nozzle that can achieve a greater spread of the washing fluid over the first rotating blade row.

By Eric Jeffs

On-line compressor washing is an established practice among gas turbine operators, but as gas turbines have become bigger some problems have come to light. Specifically, the larger air flows into the compressors, particularly of the F-class machines, tended to deflect the wash water flow, and so reduce the area of the blade that can be cleaned. Excessive impact of water at the blade roots can lead to erosion, and subsequently stress cracking that could bring about failure.

The advantage of on-line compressor washing is that it results in some power recovery from cleaning the first stage compressor blades, which are the most vulnerable to the deposition on the leading edge of fine particles that come through the filters. It therefore slows the rate of loss of output and extends the time between outages required to perform an off-line wash. The rate of power loss and the frequency of washing to recover it depend on the ambient conditions at the plant.


The Mk 3 washer nozzle, with a flat water spray bracketed by parallel air sprays
Click here to enlarge image

The Swiss-based company Turbotect was the first to introduce on-line compressor washing in 1991, with the first installations on ABB Type GT 9 gas turbines at a district heating plant in Utrecht, the Netherlands. Extensive studies of the effects of washing these gas turbines, which were ISO-rated at 36 MW, effectively proved the value of on-line compressor washing.

Prior to this compressor washing was performed on the stationary gas turbine. The first on-line washing system used a spherical nozzle set in the surface of the bell mouth duct of the compressor intake plenum. A number of nozzles were placed in the intake plenum so as to cover the maximum blade surface, i.e. the first 20 cm above the root where the majority of dirt was likely to be deposited.

As the concept took hold, and with the development of the merchant plant market, with its emphasis on high availability and reliability, on-line compressor provided a way to control power loss and avoid the need to shut down frequently for an off-line wash. With the introduction of the F-class gas turbines some manufacturers developed their own on-line wash systems.

In 1993 Turbotect approached GEC-Alstom, who were licensees of GE, and sold five washer sets for the 9FA gas turbines. These turbines were for some of the first single shaft combined-cycle blocks, which were being installed in the Netherlands, at Eemscentrale, near Groningen.

Blade failure

A crisis arose in 2001 following the catastrophic failure of a front row (R0) compressor blade on an early model GE Frame 9FA at Black Point, Hong Kong, which was put down to erosion resulting from a frequent and aggressive compressor wash routine. GE immediately sent out an advisory notice to all its 9FA and 7FA operators recommending a reduced level of water use in on-line washing to retard the rate of erosion.

Lower erosion rate

Among these operators were the Netherlands division of Electrabel which had some units that, like Black Point, had the GE OLWW system. However, at its Eemscentrale station, which had been operating since 1995, all five units had the Turbotect Mk 1 on-line compressor wash-ing system. Using this system Electrabel performed a daily on-line wash for 30 minutes and a single off-line wash every year when they shut down for the annual inspection or maintenance outage.

The off-line wash recovered between 6-7 MW (about 2.8 per cent) of output lost by a dirty compressor. Eemscentrale is in a relatively benign environment on a wide river estuary: a flat coastal site backing on to farmland. It is susceptible to salt spray, and in certain wind conditions sand and dust from the beach and the farmland bordering the site, while in spring and early summer there is pollen in the air.


The Mk 3 nozzle air jets push the water spray into the intake air flows
Click here to enlarge image

During 2001 the gas turbines at Eemscentrale had undergone its first major overhaul, which had allowed the operators to examine the compressor first row blading for signs of erosion. What they found was that the R0 leading edge roughness had built up to an average of 226 microns over six years of operation, during which time they had been performing a daily on-line wash lasting 30 minutes. The gas turbines had run a total of 51,936 hours and during that time had experienced a total of 1500 on-line washes of 31 minutes each for a total of 790 wash hours. Although the measured erosion rate was only half of that which had caused the failure in Hong Kong, it pointed to a way of further reducing the erosion rate while maintaining on-line washing over the lifetime of the gas turbine.

At this time the blades were blended to eliminate surface roughness. This means polishing followed by shot blasting to improve surface stress. All five units went back into service with the Mk 1 on-line nozzle, but in accordance with GE recommendations, in light of the Hong Kong and other blade failures, reduced to a 20-minute daily wash. GE had also requested that blade roughness be measured after 100 hours of on-line washing.

Of the five units at Eemscentrale, two units are connected to the 330 kV network, while the other three are outputting to the 220 kV system. One unit on each system is designated stop/start, with on average 80 starts per year. The other three units are base loaded running up to 8000 hours/year.

On-line washing entails spraying a water-based detergent solution into the gas turbine from nozzles strategically placed in the walls of the compressor intake. As gas turbines got bigger, with higher rates of air flow, it was observed that the washer spray was not spreading out into the main air flow, but running parallel to the wall. This meant that from the inner surface of the bell mouth the spray was concentrating in the lower 8 cm of the first rotating blade, exacerbated by the impact on the upstream stationary vanes. With a high rate of flow at high pressure this could cause erosion of the leading edge at the blade root and eventually failure.

Nozzle evolution

Turbotect had been looking at the erosion problem for some time, and deliberately designed its Mk 1 system, with a low pressure and flow rate to minimize the possibility of erosion. From this it had conducted experiments on units of existing customers with the smaller Frame 9E gas turbine in the UK and Thailand, using a new nozzle system from which the Mk 3 unit evolved.

The new unit has since been installed in one of the Frame 9Es at Peterborough, UK, in a GT13E2 at Kuala Langat, Malaysia, and on one of the base-loaded Frame 9FAs at Eemscentrale.

Operational test

The principle of the Mk 3 nozzle is a low-pressure, low-flow system that is designed to fit in the same mounting as the Mk 1 nozzle. The water flow is 17.6 litres/min at 4 bars, which is projected in a flat-profile spray, sandwiched between two flat-profile, high-velocity air sprays, also at 4 bars. At each plant, eyeglass observations have shown that the spray is pushed out into the main air stream and impinges further up the surface of the first blade row, and therefore covers a wider area. There is less concentration of the spray at the leading edge root, which was the main site of erosion.

In June 2004 R0 roughness measurements were take on four of the engines. The longest running unit was the EC6 with 23,352 hours of operation since the blending at the first major overhaul outage. It received 107 hours total wash time at the reduced rate, and its mean roughness was measured at 199 microns on the blades.


Graph A shows blade erosion against number of operating hours, while Graph B shows erosion against number of washing hours
Click here to enlarge image

At Eemscentrale, the Mk 3 nozzle was installed on one of the base-loaded units at the hot gas path inspection in 2004. The original installation of 38 Mk 1 nozzles was reduced to 30 Mk 3 nozzles, with the unused positions blanked off. No refurbishment of the compressor first row blades was undertaken, but a planned combustion inspection in early 2006 provided an opportunity to examine for blade erosion over the two years since the new nozzles were installed. What Electrobel observed was that the same wash results were obtained with fewer Mk 3 nozzles than in the Mk 1 case, and therefore with less water consumption spread over a larger area of the blade.

In early 2006 Electrobel shut down the modified Unit EC-6 with the Mk 3 nozzle and two of the unmodified Units EC-5 and EC-7 with the Mk 1 nozzle. An off-line wash was performed on each unit before they were returned to service. Under effectively similar operating conditions Units EC-5 and EC-6 were washed on-line three times a week for 17 minutes, while EC-7 was washed daily for 20 minutes.

Results showed that because the Mk 3 nozzle wets a bigger area of the blade with the same volume of water, the rate of power degradation is some 25 per cent lower than for the Unit EC-5 with the MK 1 nozzle, while the higher washing frequency of EC-7 achieved the same degree of power recovery as EC-6, with the Mk 3 nozzle over a period of 1000 operating hours.

Solving the wash problem

There has probably been no better investigation into the effects of compressor washing over an extended operational period than with the units at Eemscentrale.

It has shown for the first time the rate of erosion of the R0 blading, and the way that it can be controlled, including the influence of blending on the rate of power loss and the incorporation of other stress relief measures by the OEM in the blade root.

Although the Mk 3 nozzle was initially developed in response to a specific problem with GE’s F-class machines, in reality it provides a solution to the problem of wash water distribution in larger gas turbines, i.e. over 120 MW. The Mk 3 nozzle not only improves the efficiency of washing, but also reduces the rate of power degradation.