Performance Monitoring of a Three Shaft Aero-Derived Gas Turbine
Table of Contents
8.0 Power Loss
Gas turbine performance deterioration has a big impact on engine performance. Loss of engine performance results in loss in power output and increased heat rate. Loss in power can result in lost revenue and the increase in heat rate results in increased operating costs. Both these factors reduce life cycle costs and profit.
This report discusses the results obtained from the application of the advanced performance monitoring system (GPAL system), developed by GPA Ltd, to an Aero derived three shaft industrial gas turbine. The gas turbine is employed in gas transmission and has an ISO rating of about 20 MW.
The GPAL system is capable of detecting performance related faults at engine component level. The detection of faults at engine component level is achieved by the computation of Fault Indices. Fault indices, developed by Gas Path Analysis Ltd, represent the percentage change in component characteristics and further details may be found in this web site under Gas Turbine Performance Monitoring and Diagnostics (XPGT3). Fault indices are applicable to any component or equipment, for which a unique steady state characteristic can be defined. The GPAL system can be integrated with any SCADA system, DCS or engine control system. On this occasion it was integrated with the Honeywell SCAN 3000 system.
Generally two fault indices can be computed for an engine component. The first, the fouling fault index, which reflects the change in flow or capacity of the component due to faults and the second, the efficiency fault index, which reflect the efficiency changes in the component due to faults. These fault indices should remain to within ± 1% when no faults are present in the engine components. We have set this bandwidth to allow for any errors in computing steady state conditions.
Instrument faults can give misleading results and a successful performance monitoring system should be able to highlight instrument faults and be able to give some indication which instrument is faulty. The report also discusses how the GPAL system can be used to detect instrument readings that are suspect and describe a methodology based on Aero-Thermal analysis to detect the faulty instrument.
Changes in fault indices for compressors imply one or more of the following:Fouling Fault Index
Efficiency Fault Index
- Incorrect variable inlet guide vanes (VIGV) / variable stator vane operation (if present)
- Increase in clearance between the rotor tip and the casing. This can occur due to excessive rubbing.
- Incorrect variable inlet guide vanes/ variable stator vane operation (if present)
- Increase in clearance between the rotor tip and the casing. This can occur due to excessive rubbing.
- Labyrinth seal ware/damage
For turbines changes in the fault index implies one or more of the following.Fouling Fault Index
Efficiency Fault Index
- Nozzle guide vane blockage
- Nozzle guide vane erosion
- Nozzle guide vane bowing
- Nozzle guide vane blockage
- Nozzle guide vane erosion
- Nozzle guide vane bowing
- Rotor tip seal ware/damage
- Labyrinth seal ware/damage
Figure 1.1 Schematic representation of the gas turbine being monitored is given below.
The following measurements were taken when the engine is operating under steady state conditions for the computation of fault indices.
- IP Compressor Inlet Pressure
- IP Compressor Inlet Temperature
- HP Compressor Inlet Pressure
- HP Compressor Inlet Temperature
- HP Compressor Exit Pressure
- HP Compressor Exit Temperature
- Power Turbine Inlet Pressure
- Power Turbine Inlet Temperature (EGT)
- Power Turbine Exit Pressure
- Power Turbine Exit Temperature
- IP Compressor Speed
- HP Compressor Speed
- Power Turbine Speed
- Fuel Flow
- Lower Heating value of Fuel (off-line)
The airflow or power measurements are not required for the computation of the fault indices. All measurements were taken on-line except the heating value of the fuel which is entered off-line via the keyboard.
Airflow and power measurements are often unavailable on mechanical drive application and the installation of such instrumentation can be very expensive, especially on an existing site. Airflow measurements are also notoriously unreliable especially after a compressor wash. Pressure measurements can also give spurious reading after a compressor wash and care is required to ensure that the impulse lines are clear of any debris and water after a compressor wash.
2.1 Fault Indices
The following fault indices were calculated and trended.
- IP Compressor fouling fault index
- IP Compressor efficiency fault index
- HP Compressor fouling fault index
- HP Compressor efficiency fault index
- HP Turbine fouling fault index
- HP Turbine efficiency fault index
- IP Turbine efficiency fault index
- Power turbine fouling fault index
- Power turbine efficiency fault index
Only the IP turbine efficiency fault index could be calculated because we cannot measure the inter-stage pressure between the HP and IP turbine.
By examining the trends of these fault indices specific faults can be highlighted at component level.
Compressor fault indices indicate the deviations of the compressor characteristics from the base condition. Factors that can change the compressor characteristics are fouling, VIGV schedule problems, tip rubs and labyrinth seal damage.
Figure 3.1 shows the trend in the IP compressor fouling (A-IPCFFI) and efficiency (A-IPCEFI) fault index during engine operation. There are three periods of engine operation and correspond to the following dates.
First period of operation from about the 20 Feb-99 to 12-Mar-99
Second period of operation about the 5-Mar-99 to 25-Apr-99
Third period of operation from about the 22-May-99 to 16-May-99
Compressor fouling is the most common performance related faults that occur in gas turbines. Fouling reduces compressor flow capacity and efficiency. However efficiency losses are relatively small compared with the reduction is flow capacity and this can be seen in figure 3.1.
Figure 3.1 shows the trend of the IP compressor fault indices on a 24 hours average
The data is averaged every 24 hours and trended. This is necessary to cover a reasonable period in time with clarity. Figure 3.1 shows the fouling fault index progressively reducing from about zero to about -2.5% when the compressor is fouled. The recovery is the IP compressor fouling after a wash is also clear when the fouling fault index returns to about zero.
The compress efficiency fault index is offset by about -1% and is probably due to an instrument fault where the IP compressor discharge pressure is reading lower than expected. This will be discussed latter in Appendix I where we consider Aero-Thermal techniques to detect instrument faults. The compressor efficiency fault index only falls by about 0.5% to 1% from this offset during fouling and we can notice a recovery after a compressor wash.
As stated above the compressor efficiency fall marginally compared with compressor flow changes. Systems that rely on compressor efficiency changes to detect fouling are likely to miss the effects of fouling and the fouling fault index is a significantly better indicator of compressor fouling.
Compressor washing is too frequent and can be delayed until the fouling fault index is reading about -3% to 3.5% when the compressor will be moderately fouled.
Figure 3.1 also shows the fouling index recovering at in-between periods. The engine is fitted with a variable inlet guide vane (VIGV). This is necessary to achieve a satisfactory surge margin for the IP compressor during transients operation.
The control of the VIGV is based on an open loop system. There is no feed back loop. As a result on occasions the VIGV position is either over compensated (VIGV is opened too much) or under compensated (VIGV is closed too much). When the VIGV is over compensated the compressor fouling decreases because the flow capacity of a given speed increases and conversely when the VIGV is under compensated the compressor fouling increases. Although the surge margin is improved when the VIGV is under compensated the surge margin is reduced when the VIGV is over compensated. Therefore there is an increased risk of compressor surge when the VIGV is over compensated.
This is shown more clearly in figure 3.2 where the fault indices for the IP compressor is trended on an 8 hour average for the second period of engine operation. In spite of the VIGV scheduling problem fouling is still clearly seen. Effect on the compressor efficiency fault index is also seen. We observe that the compressor efficiency slightly improves when the VIGV is over compensated and conversely a slight fall in compressor efficiency when the VIGV is under compensated. It might appear desirable when the VIGV is over compensated but the risk of compressor surge is increased.
Figure 3.2 shows the trend of the IP compressor fault indices on an 8 hours average
The Aero gas turbine from which this engine was derived was a high by-pass turbo fan engine. The engine was known to suffer from this VIGV schedule problem where the VIGV will close excessively on occasions during flight and result in a sudden increase in IP compressor speed. A loss in thrust will result and often referred to as "Pod Nodding".
Should the VIGV be significantly over compensated (i.e. opened much more than required) then compressor surge can occur during transients, resulting in major engine failure incurring increased maintenance costs and unscheduled down time. Again systems that use compressor efficiency changes to detect fouling or compressor faults are unable to detect VIGV scheduling problems.
The efficiency fault indices should remain with ± 1% when no faults are present. From figure 3.1 and 3.2 we notice that the IP compressor efficiency fault index has an offset of about -1%. As stated above, this is probably due to the IP compressor discharge pressure reading lower than expected by about 1 to 2% and this is discussed in more detail in appendix I.
If we accept this offset as the zero for the efficiency fault index, then we see after a wash the fouling indices are within ± 1%. Therefore we conclude that there are no other faults other than fouling and VIGV scheduling issues.
Figure 4.1 shows the trend in the HP compressor fouling (A-HPCFFI) and efficiency (A-HPCEFI) fault index. We observe that there is an offset in the fouling and efficiency fault index. The fouling fault index is offset by about 2% and the efficiency fault index is offset by about 1%. We will show later (Appendix I) that this could be due to a lower IP discharge pressure. We shall assume the zero or base line is now offset by 2% for the fouling fault index and 1% for the efficiency fault index for the HP compressor.
We observe possible light fouling of the HP compressor during the first period of operation. However if the fault index remains with ± 1% from the base we say no faults exists. We observe that all the points for the fouling fault index are within 1% and 3% expect for one point during the third period of engine operation (i.e. within ± 1% of the base, which is now 2%).
We do not observe the fluctuation in the fouling fault index as seen with IP compressor shown in figure 3.1. We should not see such activity on the HP compressor since it has no VIGV's. Furthermore we do not observe fouling except possible light fouling in the first period of operation because most of the fouling occurs in the IP compressor. Therefore the system correctly isolates the faults to their respective engine components, in this case the IP compressor highlighting fouling and VIGV activity. The efficiency fault index also remains between 0% and 2% during the three periods of engine operation (i.e. ± 1% of the base line).
Since the fouling and efficiency fault index for the HP compressor remains with ± 1% from their respective offsets we conclude that there are not faults present in the HP compressor for the period of engine operation shown.
Figure 4.1 shows the trend of the HP compressor fault indices on a 24 hours average
Figure 4.2 shows the trend of the fault indices for the HP compressor on an 8 hour average for the second period of engine operation. Again we observe the fouling and efficiency fault index remains within ± 1% of the base line. We do not observe any of the VIGV activity or fouling on the IP compressor reflected on the HP compressor. The system correctly isolates the faults to their respective engine components, in this case the IP compressor.
Figure 4.2 shows the trend of the HP compressor fault indices on an 8 hours average
Turbine fault indices indicate the deviation of the turbine characteristic from the base condition. Factors that can effect the turbine fault indices are blockages, nozzle guide vane and rotor damage, tip rubs damaging tips seal if present and increasing rotor casing clearances and labyrinth seal damage.
Figure 5.1 shows the trend in the HP turbine fouling (A-HTFFI) and efficiency (A-HPTEFI) fault index. We do not observe any of the fouling and VIGV activity we found in the IP compressor (figures 4.1). We observe that these fault indices remain within ±1% except the third period of operation where we observe that these indices fluctuate. We also observe that both indices fluctuate to gather. This could be due to the HP compressor discharge pressure measurement fluctuating. Figure 4.1 also shows that the HP compressor efficiency fault index increases when both HP turbine fault indices decrease during the third period of operation.
A similar analysis discussed in appendix I can be carried out to demonstrate the effect of changing HP compressor discharge pressure only on the fault indices. A model of the engine can be used to generate measurements. We can increase the value of the HP compressor discharge pressure to produce the observed pattern of the fault indices indicating that the HP compressor pressure is reading higher than expected. HP compressor fouling fault index is not significantly effected due to the fluctuation in HP discharge pressure because the compressor characteristics are very steep as explained in appendix I.
Figure 5.2 shows the trend in the HP turbine fault indices on an 8 hour average. We do not observe any effects of the IP compressor fouling or VIGV activity (figures 4.2) on any of the HP turbine fault index trends. The system is isolating the faults to the component where the fault exist. We conclude that there are no faults in the HP turbine.
Figure 6.1 shows the trend of the IP turbine fault index on a 24 hours average
Figure 6.1 shows the trend in the IP turbine efficiency (A-IPTEFI) fault index. We can only calculate the efficiency fault index because we cannot measure the inter-stage pressure between the HP turbine and IP turbine. We observe that the IP turbine fault index remains within ±1% except the third period of operation where we observe that the index becomes negative as shown in figure 6.1.
We can produce this pattern by using the engine model to generate measured parameters. We can simulate the instrument fault by reducing the IP turbine exit pressure. These measurements can be used to calculate the fault indices and compare the pattern of the fault indices with the observed pattern. Appendix I shows an example where the IP turbine exit pressure is reduced. By reducing the IP turbine pressure by 2% we can generate the pattern we observe. It should be noted that the IP turbine exit pressure also corresponds to the power turbine inlet pressure.
We need to consider the effect of this instrument fault on the power turbine fault indices. This will be discussed further in the next section when we discuss the power turbine fault indices.
Figure 6.2 shows the trend of the IP turbine fault index on an 8 hours average
Figure 6.2 shows the IP turbine efficiency fault index on an 8 hour average. Again we do not see any effect of IP compressor fouling or VIGV activity on any of these trends. The system isolates the fault at component level.
7.0 Power turbine fault indices
Figure 7.1 shows the trend in the power turbine fouling (A-PTFFI) and efficiency (A-PTEFI) fault index. We observe that these fouling fault index remain within ±1% except the third period of operation where we observe this index becomes positive. This also corresponds to the points where the IP turbine efficiency index becomes negative.
Figure 7.1 shows the trend of the power turbine fault indices on a 24 hours average
A similar analysis described in appendix I above can be carried out to demonstrate that this can be caused by the low power turbine inlet pressure measurement (which also corresponds to the IP turbine exit pressure). Brifly if the power turbine inlet pressure only was reduced then the power turbine swallowing capacity and the IP turbine pressure reatio will increase. The increase in IP turbine pressure ratio will reduce the IP turbine efficiency. We clearly see this in the trend where the power turbine fouling fault index increases and the corresponding IP turbine efficiency fault index decreases.
Examining the trend for the power turbine efficiency fault index we observe that the fault index is within ± 1% for the first two periods of operation. There is some fluctuation in this fault index in the third period of operation. We note the power turbine efficiency fault index increasing indication that the power turbine efficiency is improving. We do not belive that the power turbine can performance better than designed. This can be due to a higher power turbine exit pressure or a lower power turbine exit temperature.
We also observe that the power turbine efficiency becomes negative during the third period of operation. This could be a real fault, however this can also be due to a lower power turbine exit pressure or a higher power turbine exit temperature. If the pressure and temperature measurements are indeed correct then this could be due to turbine tip seals or labyrinth seals being damaged. Note this fault is not shown on any of the other fault index trends again the system isolating the fault to the component concerned.
If this represents a real fault then we would expect to see a persistently lower power turbine efficiency fault index and need to see the trend of this fault index over a longer period of time. The power turbine exit pressure signal is very noisy and it is likely an instrument fault. Damping the instrumentation should improve the quality of the measurements taken.
Figure 7.2 shows the trend of the power turbine fault indices based on an 8 hour average. Again we do not see any of the IP compressor fouling and VIGV activity of these trends. Again we see the system isolating faults on at component level.
8.0 Power Loss
The component fault indices represent the change in engine component characteristics. The revised or current characteristics determined from the fault indices can be used in an engine model to calculate the change in power and heat rate relative to the base line. We can trend these changes and figure 8.1 shows the changes in power during engine operation.
We observe a power loss during fouling. Figure 8.1 shows a power loss of about 2% due to fouling and the power loss recovering after a compressor wash. We also note that the power recovers during the fouling cycle and this largely corresponds to the points when the VIGV is over compensated (see figure 3.1 and 3.2). As stated earlier it might appear that this is very beneficial but the risk of compressor surge is increased during transient operation.
We observe that there is a significant power loss during the third period of operation. This is primarily due to the power turbine efficiency fault index decreasing as shown in figure 7.1. Some systems claim that compressor fouling can be determined from the power loss. If the power turbine fault discussed above is real, then it is almost impossible to determine compressor fouling during the third phase of operation by computing the power loss. However trending the compressor fouling fault index does not confuse the operator about how much the compressor has fouled (figure 3.1 and 3.2).
The change in heat rate can also be trended. This information with the power loss trend is invaluable to the operator to determine the impact of engine performance deterioration on operating costs, profit and on life cycle costs. Means to determine the effects of performance deterioration on plant operation is discussed in this web site under 'Benefits of Model based analysis'.
Gas turbine performance deterioration has a big impact of life cycle costs. Performance deterioration not only reduces power output and increase heat rate but it also reduces engine life and increases emissions.
The GPAL system uses component matching techniques and gas path analysis techniques to determine the change in component characteristics. These techniques calculate Fault Indices for each engine component, which reflect the changes in the flow and efficiency characteristic of engine components. Trending these fault indices enable the detection of fault developing.
The system successfully detected compressor fouling which is the most common form of performance deterioration. The effect of fouling on power loss has also being determined. Such information is invaluable to engine operators to determine the intervals of engine washing and the effectiveness of the wash.
This engine is fitted with and variable inlet guide vane (VIGV) to achieve safe operation through the operating range of the engine. The effect of VIGV activity has also being determined and VIGV faults can also be detected by GPAL system.
Instrument problems have also been detected and means to highlight the instruments that are faulty are also discussed. This is a very important feature of the GPAL system because if a system cannot detect instrument faults then the diagnostics can be very misleading.
The use of fault indices with 'Model Based Analysis' enables the operators to determine the impact of performance related faults on production, operating costs and engine life usage. All these factors effect life cycle cost and this technology will help reduce life cycle cost hence improving profit.
This section discusses the detection of instrument faults using an Aero-Thermal method. We consider the case where we suspect that the IP compressor discharge pressure is reading lower than expected. We should have suspected an instrument error when we observe that the HP compressor has a higher capacity and efficiency than design (see figure 4.1 and 4.2).
We have shown the offsets discussed above in figure 4.2 and suggested that these offsets were due to a lower IP discharge pressure. We observe the off set for the HP compressor fouling fault index is about 2% and the offset for the HP compressor efficiency fault index is about 1%. We also observe that the IP compressor efficiency fault index is about -1%.
Measured performance parameters such as pressures, temperatures, speeds and flows are primarily determined by component matching. When faults appear then the changes in component characteristics result in changes in the measured parameters from which we calculate the fault indices.
With a model of the engine we can use the model to generate the necessary measurements to compute the fault indices. We can also alter these derived measurements there by introducing instrument faults. We can change a particular derived measurement or many of these derived measurements. These modified parameters can be used to compute the fault indices and compare the fault index pattern with that observed in practice. Should these patterns compare closely we can determine what instrument that is suspect and by how much the measurements are deviating.
We can reproduce the offsets we see in figures 3.1, 3.2, 4.1 and 4.1 by reducing the IP discharge pressure by about 1 to 1.5%. Figure A1 shows the fault index pattern when only the IP compressor discharge pressure is reduced by 1%. We see that the fault index pattern for the IP and HP compressor matches closely to that observed where the IP compressor efficiency fault index is offset by about -1% and the HP compressor fouling and efficiency fault indices ate offset by about 2% and 1% respectively.
Figure A1 shows the fault indices when an instrument fault is simulated (IP Discharge Pressure)
We do not observe any change in the IP compressor fouling due to the IP compressor discharge pressure reading lower than expected. Axial compressor characteristics are very steep especially at high operating speeds due to near choking conditions of the front stages. Due to the steep compressor speed lines there is very little effect on fouling index due to a lower IP compressor discharge pressure than expected.
Note the power loss and thermal efficiency is very small. This is because the loss in component performance of the IP compressor is more or less compensated by the component gain of the HP compressor. This response is typical when the pressure measurement at some inter-stage measuring point is faulty. If the instrument reading at the boundary of the engine is faulty (i.e. inlet or exhaust) then a change in power loss and thermal efficiency is observed as we suspect with the power turbine efficiency fault index discussed in section 7.0 above.
We can also investigate the effect of reducing the power turbine inlet pressure in addition to the IP compressor discharge pressure. Figure A2 shows the fault index pattern when the power turbine inlet pressure is reduced by 2%. We observe that the power turbine fouling index and the IP turbine efficiency index has also changed. This pattern is very similar to that shown in figure 6.1, 6.2, 7.1 and 7.2 above where we suspect the power turbine inlet pressure is reading lower than expected during the third period of operation. We still observe the offsets present in the IP and HP compressors, due to the lower IP compressor discharge pressure (figures 3.1, 3.2, 4.1 and 4.2).
A similar analysis can be carried out check the validity of the HP compressor discharge measurement. Other instrument faults may, such as temperature, fuel flow and speed measurements can also be investigated.