METHOD OF DETECTING A PREDETERMINED CONDITION IN A GAS
TURBINE AND FAILURE DETECTION SYSTEM FOR A GAS TURBINE
Background of the invention
This invention relates to a method of detecting a predetermined condition in a gas turbine. The invention further re- lates to a failure detection system for a gas turbine.
In such a gas turbine known in the state of the art, for ex¬ ample as it is disclosed in EP 1 953 454 Al, a gas duct or gas flow path is routed through a combustion section/system located between a compressor and a turbine section. The combustion section may include an annular array of combustors. High pressure air from the compressor flows through the combustion section where it is mixed with fuel and burned. As mentioned above, the combustors each comprise a burner for igniting the air/fuel mixture especially during start up of the gas turbine.
Combustion gases exit the combustion section to power the turbine section which drives the compressor. In single-shaft arrangements a high-pressure and a low-pressure turbine of the turbine section are mechanically connected and together drive an output power shaft. In twin-shaft arrangements a low-pressure turbine (power turbine) is mechanically inde¬ pendent, i.e. only drives the output power shaft, and a high- pressure turbine, or so called compressor turbine, drives the compressor. This combination acts as a gas generator for the low-pressure turbine. The combustion gases exit the turbine section through an exhaust duct. An unwanted introduction of liquid fuel in the combustor while operating on gas fuel can cause an increase of tempera¬ ture in the combustion section. The increase of temperature can damage turbine components, i.e. nozzles. Hardware fail- ures in components as a burner braze failure can be a cause for such unwanted conditions.
A solution for hardware failures is to prevent failures by designing hardware solutions that reduce a possibility for it to happen. But hardware failures can not be totally avoided, therefore a reliable detection system for hardware failures should be available. Various systems to detect gas turbine faults are known in the state of the art. Known systems are based on fixed tempera¬ ture thresholds or temperature measurements spread increase above certain limits. EP 1 953 454 Al discloses a method for detecting a partial flame failure in a gas turbine by measur- ing a temperature over time at two different locations using a rate of change in temperature to determine the flame fail¬ ure .
JP 56133636 A discloses another gas turbine fault detection system comparing maximum and minimum values of temperature measured by temperature detectors. A signal - indicating an abnormality - is generated when the difference between the maximum and minimum values of temperature detection signals of a plural of temperature detectors arranged in a circular flow passage of exhaust gas in a circumferential direction exceeds a set value.
Summary of the invention It is a first objective of the present invention to provide a method for detecting a predetermined condition in a gas turbine, particularly to detect an abnormal condition and/or a hardware failure, for example a burner braze failure, whereby the above-mentioned shortcomings can be mitigated, and espe- cially a more reliable detection of a predetermined condi¬ tion, particularly an abnormal and/or a failure condition, in a gas turbine is facilitated. It is a second objective of the invention to provide an advantageous failure detection system for a gas turbine.
The first objective is achieved, according to the present in- vention, by providing a method of detecting a predetermined condition in a gas turbine. Said method comprises the steps of:
- measuring a temperature over time at each of at least three probing points located in said gas turbine,
- determining a variance for each of said measured tempera¬ ture over time of said at least three probing points,
- selecting one of said determined variances and comparing said one selected variance with a first threshold which is based on at least two of the other determined variances, - detecting said predetermined condition in said gas turbine based on said comparing of said one selected variance with said first threshold.
The second objective is achieved, according to the present invention, by providing a failure detection system for a gas turbine, in particular configured for conducting the method of detecting a predetermined condition in a gas turbine. Said failure detection system having evaluation means for detecting a predetermined condition, particularly an abnormal con- dition of said gas turbine and/or a hardware failure in said gas turbine, for example for detecting a burner braze failure in said gas turbine, from at least three temperature over time of temperature measurements at probing points in said gas turbine.
In other words, the invention analyses instantaneous tempera¬ ture readings from different locations in the gas turbine, i.e. in or outside/downstream a combustion system of the turbine, by calculating the variance, particularly a standard deviation, for each measurement location during a predetermined time period. The variance (or standard deviation) of one - preferable of each - measurement location is than com¬ pared - respectively - with a threshold that is based on at least two - preferable each of - the other variances (or standard deviations) of the other measurement locations. If the said one variance (or standard deviation) goes above the threshold the occurrence of the predetermined condition is detected.
The invention is based on the insight, that analysing meas¬ urement over time on basis of variances/standard deviations - in addition with a comparing with a dynamic threshold - can lead to more significant, robust results and reliable conclu¬ sions as an analysis of measurement over time on basis of single values of the measurement over time (exceeding fixed limits) . Particularly for complex dynamic processes with dy¬ namic altering process parameter, such as a process tempera- ture, the analysis and observation of "higher level" statis¬ tic characteristics could be much more significant.
A change in a process parameter caused by a "normal" process operation can lead to exceed a fixed, default limit and to a spurious emergency action by using such first level statis¬ tics. Higher level statistics in addition with dynamic thresholds allows preventing spurious detection of predetermined/abnormal conditions, i.e. failures, leading to a more stable operation of the turbine. The more stable operation leads to an increased life expectancy of the gas turbine and improves its performance in general. Albeit the invention - using higher level statistics, i.e. using the analysis of variances, - can be used in combination with such a first level analysis.
The invention is further based on the insight, that furthermore unusual temperature variations are characteristic for hardware failures such as leakages or burners braze failures. Leakages or burners braze failures could cause an abnormal condition, i.e. an unwanted introduction of liquid fuel in the combustor. Especially while operating on gas fuel - the introduction of liquid fuel is followed by an increase in temperature at a certain location in the gas turbine. There- fore, measuring temperature and/or providing temperature over time at different locations in and/or outside the combustion system, analysing the measurements by calculating the variances and comparing the variances - as dynamic thresholds - provide an indication of an abnormal condition and an indica¬ tion for a hardware failure in the turbine.
According to the inventive solution gas turbines conditions can be controlled and/or observed much more reliable - and hardware failures of the turbine such as leakages and/or burner braze failures are detected with higher reliability and earlier in time. This allows preventing further damages of turbine components. This can lead to an increased life ex¬ pectancy of the gas turbine and improves its performance in general.
Further the detection method/system according to the invention can be retrofitted in existing gas turbine installations - for example by implementing (software/computer) routines in gas turbine control systems - or it can be used as a fault detection (software/computer) tool by support or test engineers - testing gas turbines and/or detecting turbine opera¬ tional problems. This includes the use of such a fault detec¬ tion tool in a mobile way - using it for turbines being lo- cated at different locations while connecting such a tool with different turbines.
It is also included in the invention to use the detection method/system on real time basis in gas turbine while con- trolling and/or observing turbine operations. The detection method/system can also be used for simulations, forecasts and/or past performance analysis, particularly analysing his¬ torical or simulated measurements. It is also subject to the invention that the gas turbine could be stationary as well as mobile, for example for air¬ craft . In an advantageously embodiment the method comprises further the steps of
- calculating a difference value based on said comparing of said one selected variance with said first threshold,
- determining a control function value of a control function based on said difference value, particularly if said one se¬ lected variance exceed said first threshold,
- comparing said control function value with a second threshold, particularly with a second default threshold,
- detecting the predetermined condition in said gas turbine if said control function value exceed said second threshold.
It is further advantageous if the method steps are repeated at various successive points of time wherein said control function value is accumulated in said control function.
In other words, if the selected variance goes above the first threshold, the difference is then accumulated in the control function that is then compared with the second, fixed thresh- old. When the control function value goes above the second threshold a warning for a fault could be released and further actions might be implemented to preserve integrity of the combustion hardware. In another advantageous embodiment said temperature over time will be measured at several or various probing points located in said gas turbine, particularly located within said combus¬ tion system of said gas turbine and/or located downstream from said combustion system of said gas turbine. It is also advantageous if said first threshold is based on all other variances. Said first threshold could be calculated as an av¬ erage value using at least two of the other variances. It can be preferable to calculate the average value using several or all of the other variances.
It is further advantageous if said first threshold is calcu¬ lated by using at least one default parameter, particularly if said default parameter is selected according to a gas tur- bine model, a temperature measurement tool, a temperature measurement location, a temperature measurement time fre¬ quency and/or a range of the temperature over time. If is also advantageous if said control function value is re¬ set at regular time intervals in order to avoid spurious warnings dictated by small deviations accumulated during a long time period. Such a warning and/or safety action can be a warning message and/or a change of fuel and/or a shut down and/or a trip to idle of said gas turbine wherein the warning and/or safety action can be generated, particularly by a control system of said gas turbine, if said predetermined condition, i.e. the abnormal condition, is detected.
Said predetermined condition to be detected can be an abnor¬ mal condition, particularly an unusual temperature variation or behaviour, particularly caused by an unwanted liquid fuel introduction within a combustor of said combustion system.
Unwanted liquid fuel introduction could be caused by a hard¬ ware failure, particularly a leakage and/or a burner braze failure . Advantageously, the temperature sensors for conducting the temperature measurements comprise thermocouples. Thermocou¬ ples used in the turbine should have a low thermal capacity and therefore a high response rate, so they are a good repre¬ sentation of the thermal field in the turbine at any operat- ing condition.
Brief description of the drawings
A detailed description of the present invention is provided below with reference to the following diagrammatic drawings, in which: Figure 1 is a sectional view of a embodiment of a gas tur¬ bine according to the invention;
Figure 2 is a cross-sectional view of a combustion system with a pilot burner contained in the gas turbine according to Figure 1 ;
Figure 3 is a flow chart of failure detection routine work¬ ing principles according to the invention;
Figure 4 is a flow chart of a (software) program structure of the failure detection routine working principles of Figure 3 according to the invention.
Detailed Description of the Illustrated Embodiment
Figure 1 is an embodiment of a gas turbine 10 in the form of a single-shaft gas turbine. The gas turbine 10 comprises a single rotor shaft 12 carrying both a compressor 14 and a turbine 16. A gas duct 34 guides a propulsion gas 18 through the gas turbine 10 starting from an inflow section 20 via the compressor 14, a combustion section/system 22, the turbine 16 and an exhaust duct 26.
At the left end of the gas turbine 10 according to Figure 1 the propulsion gas 18 in the form of air flows via an inflow section 20 into the compressor 14. The compressor 14 thereupon compresses the propulsion gas 18. The propulsion gas 18 then enters the combustion system/section 22 of the gas turbine 10, in which it is mixed with fuel and ignited in com- bustors 24. The combustion section 22 contains an annular array of (six) combustors 24, of which only one of six is shown in Figure 1 and which lead into the gas duct 34. The com¬ busted propulsion gas 18 flows through the turbine 16 expand¬ ing thereby and driving the rotor shaft 12. The expanded pro¬ pulsion gas 18 then enters an exhaust duct 26. The (six) combustors 24 each comprise a burner 36 for intro¬ ducing fuel into the inside of the corresponding combustor 24 and igniting the fuel/air mixture. A burner 36 comprises a pilot burner 37. The pilot burner 37 contains a fuel inlet 38 for introducing the fuel into the pilot burner 37. The fuel is subsequently guided to a burner face 40 or burner tip of the pilot burner 37. Furthermore, each pilot burner 37 con¬ tains a temperature sensor 42 in the form of a so-called pi¬ lot tip thermocouple arranged for measuring the temperature at the burner face 40.
At an exit 28 of the turbine 16 into the exhaust duct 26 sev¬ eral temperature sensors 30 in the form of so called turbine exit thermocouples are positioned at different probing points 32. By placing the temperature sensors 30 at the power tur¬ bine exit 28 the probing points 32 are located downstream from the combustors 24.
The temperatures measured over time by the (six) thermocou- pies 42 are evaluated by evaluation means 44 while control¬ ling and observing conditions of the gas turbine 10 by that means 44. As a result of this evaluation hardware failures could be detected, i.e. a burner braze failure in a subset of the combustors 24 could be assessed.
The evaluation routine (implemented in the evaluation means 44), i.e. the detection routine 100, will be described later in the text. However, it should be noted, that the thermocou¬ ples 42 does not necessarily needed to be located in the pi- lot burner 37. Other locations in the gas turbine 10, for ex¬ ample downstream from the combustors 24, allowing for measuring the temperature in the turbine are also possible.
The evaluation means 44 according to Figure 1 are adapted to conduct the detection routine 100 - shown in Figures 3 and 4 - described in the following. The detection routine 100 is implemented - as software or programmed computer components - on the turbine control system in order to detect a burner braze failure and take prompt actions to avoid the damage of nozzles. However, it should be noted, that the detection rou¬ tine 100 could be implemented on a turbine test beds as a support for test engineers while performing a turbine test.
The detection routine 100, 200 (Figures 3 and 4) monitors the pilot tip temperature instantaneous measurements and analyzes their variations over a certain interval of time, for example 10 seconds target to be tested. The unusual temperature variations that are characteristic of leakage or burner braze failures are captured and this allows the system to respond promptly with the required actions and warning messages.
Detection routine 100 working principle (Figure 3) :
The routine 100 applies on combustor hardware temperature re¬ cordings that are intended to follow similar trends. In the following description the temperature instantaneous readings from N different measurement locations are identified as (110) : ... TN(t)
Where t indicates the time of recording (120) . For each Tt (t) the standard deviation is calculated using the last x read¬ ings recorded during the time interval AtA (130)
StdT^T^ At^ for i = \ ...N Each standard deviation value for the single Tt is then compared to a function of the average value using the standard deviations calculated for the other temperature measurements locations (140) : AvgExTt StdTK ) with k = 1 to N excluding k = i in the averaging calculation . The control function is calculated for each time instant us¬ ing the last x readings available according to the following equation ( 150 ) : FT, (t) = StdT, - M * AvgExT, - Bp
M and Bp are two parameters which are selected according to the temperature measurements time frequency and Tt range of values. The suggested values as default are: M = 3; Bp = 1.
If FT^O , its value is added on FTTt (160) .
When FTTt reaches a value that is above a threshold limit FTTL (170), the control system recognizes the existence of a fault and warning signal (message) can be issued and further actions could be prompted (180) .
The function value FTTt can be brought to zero at regular time intervals ( Atc ) in order to avoid spurious warning mes- sages dictated by small deviations accumulated during a long time period. A default value suggested for Atc is 2 hours.
The following parameters can be tuned and selected according to the gas turbine model considered, the temperature measure- ment tool and the measurement location:
AtA ; M ; Bp ; ABFL ; Atc . These parameters guarantee flexibil¬ ity for the application of the detection routine to various types of combustor hardware. Burner braze failure mode
A burner braze failure is associated to a component identi¬ fied by the name "Liquid Core" 49 as it is shown in Figure 2 as part of the combustion system 22.
Brazed joints 50, 51 are also shown in Figure 2 and their failures would determine a leakage of liquid fuel into the area between the liquid core 49 and the pilot burner 37. The liquid fuel would therefore reach the combustion chamber not only from the standard path through the nozzles, but also through the gap between pilot burner 37 and swirler. This second path would not generate a spray but it would determine extra fuel to be injected in proximity of the pilot burner surface .
The consequent combustion of this fuel would determine a drastic increase of the pilot tip temperature. This type of failure is more likely to happen when a fuel change over from gas to liquid take place. The liquid fuel, cold in comparison to the liquid core metal body, is suddenly injected in the channel between the braze joints inducing thermal stresses. From the analysis of the pilot tip thermocouples readings ac¬ cording to the detection routine 100 several characteristics can be observed: the pilot tip temperatures maintain a con¬ stant average value during time intervals away from fuel change over or power output request change; during these in- tervals the temperature variations of subsequent readings ap¬ pear to be uniform between the six pilot tip thermocouples; the pilot tip temperatures are not the same for all six ther¬ mocouples and the differences rearrange after a fuel change over or a power output change; the only feature that is in- troduced by the burner braze failure is the unusual variation of only one thermocouple reading during a period that could last few minutes.
Detection routine 100, 200 for the burner braze failure mode (Figure 4) :
Based on these characteristics the detection mode of a burner braze failure can be based upon the observation of pilot tip temperature variations as shown in Figure 4 (according to the principles as shown in Figure 3) .
All six thermocouples time line readings from the six pilot burners of the turbine are associated with variations when there is a change in power output, a gas to liquid changeover or liquid to gas changeover. But the variation in pilot tip temperature associated to burner braze failure is character¬ istic of only one burner out of six. Instead of using the simple difference between two subsequent recordings, it is more robust to use the variance, i.e. standard deviation, of at least several pilot tip temperature recordings.
For each pilot tip temperature the standard deviation is cal- culated and it can be compared with the standard deviations average obtained from the readings of the other 5 thermocou¬ ples 230, 240, 250.
The difference between these two values can be used as an in- dication of the extra fuel release over the pilot burner sur¬ face. The use of 10 readings will guarantee a robust detec¬ tion system that will not incur in spurious warnings.
The threshold level used to detect the burner braze failure against a thermocouple is calculated using the average of the other five thermocouples standard deviations multiplied by three and increased by one 240, 250. Adding one and applying a factor equivalent to three on the average allows avoiding spurious warnings associated to small differences that are typical during normal operation.
The pilot tip temperatures are sensibly different from burner to burner but they are steady at constant power operation and no fuel changeover. When the power or the fuel is changed, they usually rearrange on different value levels that are not linearly correlated to the values before the change. This does not interfere with the standard deviation calculation because it will be captured by the variations of all six readings .
As it could be seen in Figure 4 the routine further provides discrimination between a failure condition and a carbon build up 290. Both cases can only be based upon the temperature level. The carbon build up happens at low load and therefore at low pilot tip temperatures, the temperature value 750°C can be used as a cut off level for the burner braze failure routine to avoid spurious warnings 290.
When the Energy Indicator level (threshold level) will reach the value 50 an action is prompt from the control system that could lead to the following thread (270, 280) : - trigger a change to gas fuel, if gas fuel is available
- warning message on the screen to contact service centre (if internet is available on site a warning email could be auto¬ matically be sent to service provider)
- trip the turbine to Idle if gas fuel is not available or shutdown the turbine.
The Energy Indicator based on 1 minute rate readings will re¬ cord the sum of the heat release indicator during the last 60 recordings in order to cancel the memory of Energy Indicator values that might have already produced a warning 160. The Energy Indicator values will have to be set to zero every time there is a fuel change over too.
It has to be underlined that some features and checks could be added to the routine in order to improve its robustness.
For instance protections against malfunctioning thermocouples or wrong thermocouples recordings are required. If one ther¬ mocouple is not working it could be simply taken out of the routine checks and average values calculations.