METHOD FOR DETECTING DAMAGE DURING THE OPERATION OF A GAS TURBINE

Abstract

A method for detecting damage during the operation of a gas turbine, including the following steps: calculating an average value of individual temperature measurement values over a defined sampling time period from an ensemble of temperature sensors in or on the gas turbine; calculating the individual temperature differences between the average value and the individual temperature measurement values over the defined sampling time period; calculating the individual temperature differences for successive sampling time periods over a defined time interval; creating a first distribution by dividing the temperature differences associated with a temperature sensor for the defined time interval into temperature difference intervals; comparing the first distribution with a second distribution of temperature differences likewise divided into temperature difference intervals; and producing an operation signal on the basis of a negative result of the comparison.

Claims

1. A method for detecting damage during operation of a gas turbine, comprising: calculation of an average value (T.sub.avg,k) of individual temperature measurement values (T.sub.i,k) over a predetermined sampling period (t.sub.k) of an ensemble of temperature sensors (S.sub.i) in or on the gas turbine, calculation of the individual temperature differences (T.sub.i,k) between the average value (T.sub.avg,k) and the individual temperature measurement values (T.sub.i,k) over the predetermined sampling period (t.sub.k), calculation of the individual temperature differences (T.sub.i,k) for chronologically successive sampling periods (t.sub.k) over a predetermined first time interval (dt.sub.1), compilation of a first distribution (D.sub.1) by dividing the temperature differences (T.sub.i,k) assigned to one temperature sensor (S.sub.i) for the predetermined first time interval (dt.sub.1) into temperature difference intervals (dT.sub.i,j), comparison of the first distribution (D.sub.1) with a second distribution (D.sub.2) of temperature differences (T.sub.i,k) likewise divided into temperature difference intervals (dT.sub.i,j), and generation of an operating signal on the basis of a negative outcome of the comparison.

2. The method as claimed in claim 1, wherein the second distribution (D.sub.2) relates to temperature differences (T.sub.i,k) of the same temperature sensor (S.sub.i), which are divided into the same temperature difference intervals (dT.sub.i,j) but which were calculated for a second time interval (dt.sub.2).

3. The method as claimed in claim 1, wherein the comparison of the first distribution (D.sub.1) with the second distribution (D.sub.2) is carried out by a comparison of the maximum (Max.sub.1) of the first distribution (D.sub.1) with the maximum (Max.sub.1) of the second distribution (D.sub.2).

4. The method as claimed in claim 2, wherein the comparison of the first distribution (D.sub.1) with the second distribution (D.sub.2) is carried out by plotting the two distributions (D.sub.1, D.sub.2) in a common diagram with an axis representing a time profile over the first and second time intervals (dt.sub.1, dt.sub.2).

5. The method as claimed in claim 1, wherein the comparison of the first distribution (D.sub.1) with the second distribution (D.sub.2) is carried out by a comparison of the position of the maximum (Max.sub.1) of the first distribution (D.sub.1) with the boundaries of a state space (Z) which has been determined from a distribution width of the second distribution (D.sub.2).

6. The method as claimed in claim 2, wherein the first distribution (D.sub.1) and/or the second distribution (D.sub.2) are calculated cumulatively for all sampling periods (t.sub.k) of the first and second time intervals (dt.sub.1, dt.sub.2).

7. The method as claimed in claim 1, wherein the second distribution (D.sub.2) is derived from the first distribution (D.sub.1) by a sliding calculation over the first time interval (dt.sub.1).

8. The method as claimed in claim 1, wherein the second distribution (D.sub.2) relates to temperature differences (T.sub.m,k) of a different temperature sensor (S.sub.m), which are divided into temperature difference intervals (dT.sub.m,j) but which were calculated for the first time interval (dt.sub.1).

9. The method as claimed in claim 8, wherein the comparison of the first distribution (D.sub.1) with the second distribution (D.sub.2) is carried out by a comparison of the maximum (Max.sub.1) of the first distribution (D.sub.1) with the maximum (Max.sub.1) of the second distribution (D.sub.2).

10. The method as claimed in claim 9, wherein the comparison of the maxima (Max.sub.1) applies for all maxima of the temperature sensors (S.sub.i) of the ensemble, so that the ensemble difference (E) between the greatest and smallest maximum (Max.sub.1) from the ensemble is determined.

11. A control device which is configured in order to carry out the method as claimed in claim 1, comprising: a calculation unit being contained which carries out the calculation of the average value (T.sub.avg,k) of the individual temperature measurement values (T.sub.i,k), the calculation of the individual temperature differences (T.sub.i,k), the compilation of a first distribution (D.sub.1) and of a second distribution (D.sub.2), and the comparison of the first distribution (D.sub.1) with a second distribution (D.sub.2), as well as a signal generation unit which generates an operating signal in the event of a negative outcome of the comparison.

12. A gas turbine, comprising: a control device as claimed in claim 11.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In the figures:

[0037] FIG. 1 shows a diagrammatic representation of a first frequency distribution D.sub.1 of individual temperature difference values for a predetermined time interval in temperature difference intervals dT.sub.i,j;

[0038] FIG. 2 shows a diagrammatic representation of the first frequency distribution D.sub.1 shown in FIG. 1 with indication of the frequency maximum Max.sub.1 at a temperature of 4.0 K (=dT.sub.Max1);

[0039] FIG. 3 shows a color value histogram of a chronological representation of the variation of individual distributions as a function of time t;

[0040] FIG. 4 shows a diagrammatic representation of the ratios of the position of the first frequency maximum Max.sub.1 (or Max.sub.1) of the first distribution with the state space Z derived from the second distribution for different powers P of the gas turbine;

[0041] FIG. 5 shows a diagrammatic representation of the variance of the maxima Max.sub.i for an ensemble of temperature sensors, respectively for a predetermined time interval;

[0042] FIG. 6 shows a diagrammatic representation of the variance of the maxima Max.sub.i of temperature sensors of an ensemble of temperature sensors as shown in FIG. 5, but with a much smaller spread; and

[0043] FIG. 7 shows a flowchart representation of one embodiment of the method according to the invention for detecting damage during the operation of a gas turbine.

DETAILED DESCRIPTION OF INVENTION

[0044] FIG. 1 shows a diagrammatic representation of a first frequency distribution D.sub.1, which has been normalized. The ordinate therefore corresponds to a unitless relative frequency H. As can be seen, the frequency distribution D.sub.1 has a distribution width of about 10 K, a clear maximum being formed at a different temperature dT of about 4 K. The first distribution D.sub.1 was compiled by calculating a predetermined number of temperature differences T.sub.i,k for chronologically successive sampling periods t.sub.k over a predetermined time interval dt.sub.1 and correspondingly dividing a temperature difference interval dT into individual numerical intervals dT.sub.i,j. After corresponding normalization and matching of the temperature to a predetermined zero value, the represented first distribution D.sub.1 is obtained.

[0045] At this point, it should be mentioned that the preceding and following indices i and m range from 1 to the number of temperature sensors of the ensemble in question. If individual sensors are picked out by way of example, the indices are indicated explicitly, for example by 1 or 2. The index k furthermore ranges from 1 to the number of sampling periods which lie within a time interval dt. This may be just one sampling period, or significantly more, for instance even of the order of 10.sup.6 or more. The index j relates to the number of individual temperature difference intervals for a distribution. In this case, the index j typically lies between 1 and less than 1000.

[0046] According to one embodiment of the method according to the invention, the represented first distribution D.sub.1 may then furthermore be compared with a second distribution D.sub.2, the second distribution D.sub.2 likewise having temperature differences T.sub.i,k divided into temperature difference intervals dT. If there is a deviation during the comparison of the first distribution D.sub.1 with the second distribution D.sub.2, a negative outcome of the comparison may possibly be deduced, an operating signal being generated on the basis thereof and the operator of the gas turbine for instance being informed that a thermal damage event will occur in the future, or has already occurred.

[0047] FIG. 2 shows the first distribution D.sub.1, as already represented in FIG. 1, but now with an indicator of the maximum of the first distribution D.sub.1. Since the temperature values have been matched to a zero value, the present distribution shows a maximum value dT.sub.Max1=Max.sub.1 of 4 K. According to the embodiment, this maximum value Max.sub.1 may be used for a further comparison of the first distribution D.sub.1 with the second distribution D.sub.2. In this case, for example, as explained above, the maximum values Max.sub.1 and Max.sub.2 (=dT.sub.Max2) of the two distributions D.sub.1 and D.sub.2 may be compared with one another, or alternatively a correspondence of the maximum value Max.sub.1 with a state space Z which has been derived from the distribution width of the second distribution D.sub.2.

[0048] FIG. 3 shows a color value histogram of individual distributions in a time profile. In this case, the distributions D divided into temperature difference intervals dT are plotted against the time t. The time t is itself given by continuous representation of individual time intervals dt which follow one another. Typically, the time axis has the unit of weeks. The present case thus shows the time profile of the temperature differences T.sub.i,k divided into temperature difference intervals dT.sub.i,j over a total time of 22 weeks. The time intervals dt in this case have a length which is not further specified. If, however, the first distribution D.sub.1 is taken at the time of three weeks, it is found that the distribution has a maximum in the range of about +5 K. Changes of this maximum value of the distribution D.sub.1, for instance at a time of five and a half weeks, may be explained by different operating conditions of the gas turbine in partial load. Thermal damage, however, may be ruled out in this case.

[0049] If, however, a distribution is considered in the sense of a second distribution D.sub.2 at the time of 17 weeks, it is found that the distribution has been shifted toward larger temperature difference interval values dT.sub.i,j. The maximum of the distribution is now only at about 7.5 K. This condition results because of damage to the combustion turbine components in the gas turbine. In other words, a change which can visually be identified easily can be seen clearly from the difference of the first distribution D.sub.1 and the second distribution D.sub.2, both of which have been generated from the division of different temperature differences T.sub.i,k of a temperature sensor S.sub.i for equal temperature difference intervals dT.sub.i,j. This deviation of the two distributions D.sub.1 and D.sub.2 continues to increase from the seventh week to the twenty-second week. At the twenty-second week, the operator of the gas turbine in question could then infer failure of the gas turbine because of damage to the hot-gas components in the combustion chamber. This damage event at the twenty-second week, however, could already have been foreseen from the sixteenth week. By continuous monitoring of the temperature measurement values, or by suitable evaluation as proposed in the scope of the present invention, a future damage event can easily be predicted with the aid of the visual as well as numerical deviations of the calculated individual distributions.

[0050] FIG. 4 shows a diagrammatic representation of a visual comparison between the maximum value Max.sub.1 (or Max.sub.1) of the first distribution D.sub.1 in comparison with a distribution width, the state space Z, derived from the second distribution D.sub.2. In this case, the distribution width is plotted as a function of the power P of the gas turbine. As can be understood clearly, although this distribution width does not increase with increasing power, it is however shifted linearly toward higher temperature values. As already noted above, it is known that, in the event of different loads of the gas turbine, different temperatures also occur during the combustion, and particularly for gas turbines it is for instance known that the combustion chamber temperature also changes with an increasing load according to certain operating methods.

[0051] The comparison of the first maximum Max.sub.1 of the first distribution D.sub.1 with the state space Z of the second distribution D.sub.2 is carried out at a power of about 215 MW. As can be clearly seen, the maximum value Max.sub.1 lies inside the shaded region of the marked state space Z. In other words, the deviation of the first distribution D.sub.1 from the second distribution D.sub.2, which in the present case is not explicitly shown but would be made pictorially representable with the aid of the state space, is sufficiently corresponding. Damage to the hot-gas parts of the gas turbine can therefore substantially be ruled out.

[0052] The situation, however, is different with the maximum value Max.sub.1 likewise represented in the diagram, which is intended to represent approximately the maximum value of another first distribution D.sub.1. As can be seen easily, this alternative maximum value Max.sub.1 does not fall within the gray-shaded region and therefore lies outside the state space Z of the second distribution D.sub.2. This circumstance may now be rated as a negative outcome of the comparison of the two distributions D.sub.1 and D.sub.2, and may lead to the generation of an operating signal. On the basis of the deviation of the values, the operator may then infer that a possible thermal damage event has occurred.

[0053] FIG. 5 shows a representation of all maxima Max.sub.i (=dT.sub.Maxi) of an ensemble of temperature sensors S.sub.i. Overall, the maxima Max.sub.i of 24 temperature sensors S.sub.i are represented, the spread comprising about 75 K. The spread corresponds to the ensemble difference E, which may be used as a reference quantity for future comparisons.

[0054] It should be mentioned that the maxima shown have again been subjected to a zero-value match, that is to say the maximum values have been reduced by a constant factor to the extent that they spread around 0 K. These technical diagrammatic simplifications are merely used for improved representation.

[0055] FIG. 6 shows in comparison to FIG. 5 a diagram of a number of maxima Max.sub.i of the same ensemble of temperature sensors S.sub.i, the spread now being much smaller and lying at 21 K. By a comparison of the ensemble difference E of the two representations, it can now be seen that there are less great thermal deviations according to the operating state according to FIG. 6 within the gas turbine in question. Particularly in gas turbines with moving or rotating components, for instance the rotor, it can therefore be assumed that the moving components are exposed to increased alternating thermal stress.

[0056] If the maxima Max.sub.i shown are derived for instance from temperature sensors which are arranged radially in the flue gas exit channel of the gas turbine, it may be inferred from the greater spread according to FIG. 5 that, in particular, the rotating turbine blades are subject to a stronger alternating thermal stress during their movement. This in turn makes possible to deduce that the thermally loaded components are subjected to a more rapid aging process and therefore need to be serviced or replaced earlier. If it is then found in the course of operation that the spread of some maxima Max.sub.i increases very greatly, or the ensemble difference E changes significantly, this may be evaluated as a negative outcome and the operator of the gas turbine may be informed that a servicing measure should be carried out in some way.

[0057] FIG. 7 shows a flowchart representation of one embodiment of the method according to the invention for detecting damage during the operation of a gas turbine, which comprises the following steps: calculation of the average value T.sub.avg,k of individual temperature measurement values T.sub.i,k over a predetermined sampling period t.sub.k of an ensemble of temperature sensors S.sub.i in or on the gas turbine 1 (first method step 101); calculation of the individual temperature differences T.sub.i,k between the average value T.sub.avg,k and the individual temperature measurement values T.sub.i,k over the predetermined sampling period t.sub.k (second method step 102); calculation of the individual temperature differences T.sub.i,k for chronologically successive sampling periods t.sub.k over a predetermined time interval dt.sub.1 (third method step 103); compilation of a first distribution D.sub.1 by dividing the temperature differences T.sub.i,k assigned to one temperature sensor S.sub.i for the predetermined time interval dt.sub.1 into temperature difference intervals dT.sub.i,j (fourth method step 104); comparison of the first distribution D.sub.1 with a second distribution D.sub.2 of temperature differences T.sub.i,k likewise divided into temperature difference intervals dT.sub.i,j (fifth method step 105); generation of an operating signal on the basis of a negative outcome of the comparison (sixth method step 106).

[0058] Further embodiments may be found in the dependent claims.