阅读说明：本技术 用于确定测量变量的方法和传感器系统 (Method and sensor system for determining a measurement variable ) 是由 H.米卡特 N.R.波杜图里 J.C.卡布雷雅斯 于 2019-06-11 设计创作，主要内容包括：一种用于确定测量变量的方法包括以下的步骤a)提供数据组(S1),所述数据组包括多个输入变量的值和所述测量变量的值；b)求得在所述输入变量的值与所述测量变量的值之间的相关性(S2)；c)根据所求得的相关性创建传感器模型(S3)；d)检测所述输入变量的至少一个的至少一个另外的值(S5)；以及e)基于所述输入变量的至少一个的至少一个另外的值和所述传感器模型确定所述测量变量的值(S6)；其中,所述输入变量的至少一个的值在所述步骤a)和/或e)中自身相应于所述步骤a)至e)地确定。此外,提供一种传感器系统和燃气涡轮驱动装置。(A method for determining a measured variable comprises the steps of a) providing a data set (S1) comprising values of a plurality of input variables and values of the measured variable; b) finding a correlation between the value of the input variable and the value of the measured variable (S2); c) creating a sensor model from the determined correlation (S3); d) detecting at least one further value of at least one of the input variables (S5); and e) determining a value of the measured variable based on at least one further value of at least one of the input variables and the sensor model (S6); wherein the value of at least one of the input variables is determined in steps a) and/or e) itself in accordance with steps a) to e). Furthermore, a sensor system and a gas turbine drive are provided.)

1. Method for determining a measured variable, the method comprising the steps of:

a) Providing a data set (S1) comprising values of a plurality of input variables and values of the measured variables;

b) Finding a correlation between the value of the input variable and the value of the measured variable (S2);

c) Creating a sensor model from the determined correlation (S3);

d) Detecting at least one further value of at least one of the input variables (S5); and

e) Determining a value of the measured variable based on at least one further value of at least one of the input variables and the sensor model (S6);

Wherein the value of at least one of the input variables is determined in steps a) and/or e) itself in accordance with steps a) to e).

2. The method according to claim 1, wherein the value of the measured variable provided in step a) is provided or is already provided by a sensor measuring the measured variable.

3. Method according to claim 1 or 2, wherein in step c) a subgroup of input variables of the plurality of input variables is selected and used for creating the sensor model.

4. Method according to one of the preceding claims, wherein the sensor model is determined by means of a self-learning algorithm.

5. the method according to one of the preceding claims, wherein the sensor model comprises a correlation matrix and/or is ascertained by means of a correlation matrix.

6. The method according to any of the preceding claims, wherein the input variable and the measured variable are variables of a gas turbine drive (10).

7. The method of any one of the preceding claims, wherein at least one of the input variables is indicative of pressure, temperature and/or volumetric flow.

8. The method according to one of the preceding claims, wherein the measured variable represents a pressure, a temperature, a volume flow or a manipulated variable of an adjustable component, in particular a valve (162).

9. method according to any one of the preceding claims, wherein the values of the input variables provided in step a) comprise values measured by means of at least one sensor (42A-42J) and/or simulated values.

10. Sensor system (4, 4 ') for determining a measurement variable, in particular configured for carrying out the method according to any one of the preceding claims, the sensor system (4, 4') comprising:

A correlation module (40) which is designed to acquire a data set comprising values of a plurality of input variables and values of the measured variables, to determine a correlation between the values of the input variables and the values of the measured variables, and to determine at least one sensor model from the determined correlation; and

-an analysis module (41) which is set up for acquiring the at least one sensor model from the correlation module (40), wherein the analysis module (41) has at least one input (411) and an output (412) and is designed to determine a value of the measured variable on the basis of the at least one further value of the at least one of the input variables and the at least one sensor model and to output it at the output (412) when the at least one further value of the at least one of the input variables is provided, in particular to the at least one input (411), wherein the at least one value of the measured variable determined by the analysis module (41) is used by the analysis module (41) as the value of the at least one input variable and/or is provided to the correlation module (40).

11. The sensor system (4, 4') according to claim 10, wherein the correlation module (40) comprises a storage mechanism (400) on which the data sets are stored.

12. sensor system (4, 4') according to claim 10 or 11, further comprising at least one sensor (42A-42J) which is constructed and set up for measuring the value of an input variable and provided at least one input (411) of the analysis module (41).

13. The sensor system (4, 4') according to claim 12, wherein the sensor (42A-42J) is a pressure sensor, a temperature sensor or a volumetric flow meter.

14. Gas turbine drive (10), in particular for an aircraft, comprising virtual sensors, wherein the output of one virtual sensor is coupled to the input of another virtual sensor.

Technical Field

The present disclosure relates to a method for determining a measurement variable according to claim 1, to a sensor system according to claim 10, and to a gas turbine drive.

Background

In many cases, in particular when the machine is running, it is necessary to recognize a measured variable or measured variables, but the sensors used to detect these measured variables are defective or completely unusable. In other cases, the measured variable needs to be determined redundantly, however only a single sensor of the measured variable is available. In this case, the failed sensor can be repaired, replaced or added. For this reason, it may be necessary to interrupt operation while the machine is running.

Disclosure of Invention

The object of the invention is to improve the determination of the measured variable.

According to one aspect, a method for determining a measured variable, in particular in a gas turbine drive, is provided, comprising the following steps: providing a data set in step (a), the data set comprising values of a plurality of input variables and values of a measured variable to be determined. In a step (b) following step a, a correlation (in particular the overall correlation) between the supplied values of the input variables and the measured variables is determined. In a step (c) following step b, a sensor model is created from the determined correlations. In a further step (d), optionally performed after step c, at least one further value of at least one of the input variables is detected. In step (e) following step c and following step d, at least one value of the measured variable is determined based on at least one further value of at least one of the input variables and a sensor model, for example a plurality of input variables are provided to the sensor model, which then outputs the measured variable. The value of at least one input variable (or a plurality of, optionally all input variables) in steps a and/or e is determined and/or detected (and the measured variable is shown) in its own correspondence with steps a to e. These values can be determined and/or detected in this way in the case of the method before they are used in steps a and/or e.

The method enables the use of virtual sensors as input variables for further virtual sensors. The method described makes it possible to detect not only these further measurement points but also the input variables of the virtual sensor with respect to reliability by comparing the virtual redundancy with the existing measurement points (i.e. by comparing the output of the virtual sensor with the output of an equivalent measurement point, for example a hardware sensor).

Furthermore, a method for determining a measured variable is provided, which is improved in particular in that, starting from step d, a sensor for directly detecting the measured variable (for example a temperature sensor arranged at a point for detecting the temperature at the point) is no longer necessary, wherein a particularly large number of measured variables can be determined with a limited number of sensors. The value of the measured variable can be determined from the sensor model and the further value(s) of the input variable(s) or further values. Thereby enabling to compensate for a failure of the sensor. Alternatively, it is possible to dispense with one or more sensors and to determine the measured variable further. In this way, the configuration of the sensor arrangement used in steps d and e, which provides at least one further value of the at least one input variable, can be simplified. Furthermore, sensors that detect the measured variable can be monitored.

The determination of the correlation in step b makes it possible to create an accurate sensor model in a simple manner without any knowledge of the underlying physical relationships. For example, in step a, variables measured at all available measurement points, for example the measured values of all sensors of the gas turbine drive, are included as input variables. By means of the correlation analysis, input variables can be selected which have a significant correlation with the measured variable. In this way, it is possible to react to a failure of any of the sensors in a particularly short time. The method can be used in particular in machines in which all variables to be measured and determined are dependent on one another. For example, in a gas turbine drive, all variables measured by means of sensors, in particular all thermodynamic variables within the gas turbine drive, can be correlated with one another.

The method provides a virtual sensor in which the measured variable is determined via an input variable (which does not directly characterize the measured variable).

Optionally, the value of the measured variable provided in step a is provided by a (physical) sensor (directly) measuring said measured variable or already provided by such a sensor. The sensor is constructed and arranged to act (directly) on the measured variable.

In step c, a subgroup of input variables of the plurality of input variables can be selected. The subgroup includes a portion of the input variables. Furthermore, it can be provided that only a subgroup of the input variables is used in the creation of the sensor model, without the remaining input variables being used. The subgroup is selected, for example, on the basis of the determined correlation. It can thereby be provided that all input variables of a subgroup are correlated with the measured variable, in particular to at least one predetermined degree.

The sensor model is determined, for example, by means of a self-learning algorithm. It is thus possible to determine the exact value of the measured variable without knowledge of the underlying physical relationships. The sensor model is determined, in particular, without applying a model of physics that combines the values of the input variable(s) with the values of the measured variables (optionally even without recognizing the model of physics). Determining a physical model of the sensor model is typically particularly time-consuming. In contrast to this, the application of the self-learning algorithm can be carried out particularly quickly. The self-learning algorithm comprises, for example, an artificial neural network. The artificial neural network can be trained, for example, using supervised learning methods. Alternatively or additionally, an evolutionary algorithm, in particular genetic programming, is used for determining the sensor model. It is shown here that particularly accurate results can be obtained when more operators than input variables are used, for example two or three more operators than input variables. Optionally, the number of iterations is set as an abort criterion, for example for rapid determination of the sensor model. In particular for measured variables to be determined particularly precisely, the quality of the values of the measured variables determined by means of the sensor model can be taken into account as a pause criterion. The criterion can be, for example, that the deviation between the determined value of the measured variable and a reference value (for example a value measured by means of a sensor of the entity) is below a predetermined threshold value. Alternatively or additionally, the sensor model can be determined using a genetic algorithm. Furthermore, it is possible to perform at least one regression analysis for determining the sensor model. Optionally, a sliding average (moving average) of one or more input variables and/or of measured variables is formed (in particular in connection with one of the above-mentioned methods) when the sensor model is determined.

The sensor model can include a correlation matrix. Alternatively or additionally, the sensor model is determined by means of a correlation matrix. The strongest correlations are determined, for example, by means of a correlation matrix and the sensor model is based on these correlations.

Optionally, the at least one input variable and the measured variable are variables of a gas turbine drive and/or thermodynamic variables. For example, sensors of the gas turbine drive which detect the measured variable can be controlled. It is also possible that the measured variable continues to be determined as long as the sensor measuring the variable fails. If a sensor measuring a variable fails, for example during the flight of an aircraft with such a gas turbine drive, the value of the measured variable can be determined continuously. Furthermore, measured variables can be determined in which no sensor can be used during operation of the gas turbine drive (for example during flight), for example for structural reasons.

The at least one input variable characterizes, for example, pressure, temperature and/or volume flow.

The measured variable can be a characteristic of the pressure, temperature, volume flow or a manipulated variable of a component that can be adjusted, in particular a valve.

The values of the input variables provided in step a can comprise values measured by means of at least one sensor and/or simulated values. For example, these values of the input variables are determined in one or more reference measurements, in particular together with the values of the measured variables.

According to one aspect, a sensor system for determining a measured variable is provided, which is designed and set up to carry out a method according to any of the embodiments described herein.

According to one aspect, a sensor system for determining a measured variable is provided, which comprises a plurality of virtual sensors, for example connected to one another in cascade. In this case, the output of one virtual sensor can be coupled to the input of another virtual sensor.

According to one aspect, a sensor system for determining a measured variable is provided. The sensor system comprises a correlation module, which is designed to acquire a data set comprising values of a plurality of input variables and values of measured variables, to determine a correlation between the values of the input variables and the values of the measured variables, and to determine a sensor model or a plurality of sensor models as a function of the determined correlation; and the sensor system comprises an evaluation module which is designed to obtain a sensor model or a plurality of sensor models from the correlation module, wherein the evaluation module has at least one input and one output and is designed to determine the value of the measured variable on the basis of the at least one further value of the at least one of the input variables and the sensor model or the plurality of sensor models and to output it at the output when the at least one further value of the at least one of the input variables is provided to the evaluation module, in particular to the at least one input. The sensor system is configured such that the value of one or more measured variables determined by the analysis module is used by the analysis module and/or provided by the analysis module (for example as part of a data set) to the correlation module as a value/values of the input variable. The sensor system can be designed and set up to carry out a method according to any of the embodiments described herein.

In this way, a sensor system for determining a measured variable is provided, which is improved in particular in such a way that a sensor is not absolutely necessary for detecting the measured variable. The sensor system, in particular the analysis module, provides a virtual sensor.

Optionally, the correlation module comprises a storage mechanism on which the data sets are stored.

furthermore, the sensor system can comprise at least one sensor which is designed and set up to measure the value of the input variable and to provide it to at least one input of the evaluation module.

The sensor is, for example, a pressure sensor, a temperature sensor or a volume flow meter.

According to one aspect, a gas turbine drive, in particular for an aircraft (luffahrzeug), is provided. The gas turbine drive includes a plurality of virtual sensors, wherein an output of one virtual sensor is coupled to an input of another virtual sensor. Optionally, the gas turbine drive comprises a sensor system according to any of the embodiments described herein, which sensor system is in particular able to provide one or more virtual sensors. Each virtual sensor is capable of processing a plurality of different input variables and determining a measured variable.

It will be appreciated by the skilled person that a feature or parameter (which is described in relation to one of the above aspects) can be applied in any other aspect as long as it is not mutually exclusive. In addition, any of the features or any parameters described herein can be used in any aspect and/or in combination with any other feature or parameter described herein, provided that they are not mutually exclusive.

Drawings

Example embodiments are now described with reference to the drawings; in the figure:

FIG. 1 shows a side cross-sectional view of a gas turbine drive;

FIG. 2 shows an enlarged side sectional view of a section of a gas turbine drive with a sensor system;

FIG. 3 shows a schematic view of a sensor system of a gas turbine drive;

FIG. 4 shows a schematic view of a sensor system for a gas turbine drive; and

Fig. 5 shows a method for determining a measured variable.

List of reference numerals

4. 4' sensor system

40 correlation module

400 storage mechanism

41 analysis module

410 storage mechanism

411 input

412 output

413 input

42A-42J sensor

43 data processing mechanism

5 display mechanism

9 main axis of rotation

10 gas turbine drive

11 core driving device

12 air inlet

14 low pressure compressor

15 high-pressure compressor

16 combustion mechanism

160 fuel injection part

161 fuel line

162 valve

163 combustion chamber

17 high-pressure turbine

18 bypass propulsion nozzle

19 low pressure turbine

20 core propelling nozzle

21 drive device cabin

22 bypass channel

23 Fan

26 shaft

27 connecting shaft

30 drive mechanism

A core air flow

B bypass the air flow.

Detailed Description

Fig. 1 shows a gas turbine drive 10 with a main axis of rotation 9. The gas turbine drive 10 comprises an air inlet 12, which generates two air flows, and a fan 23: a core air stream a and a bypass air stream B. The gas turbine drive 10 comprises a core portion 11 containing a core air flow a. The core drive 11 includes, in axial flow order, a low pressure compressor 14, a high pressure compressor 15, a combustion mechanism 16, a high pressure turbine 17, a low pressure turbine 19, and a core motive nozzle 20. A drive bay 21 surrounds the gas turbine drive 10 and defines a bypass passage 22 and a bypass motive nozzle 18. The bypass air flow B flows through the bypass passage 22. The fan 23 is arranged at the low-pressure turbine 19 via a shaft 26 and an epicyclic (epicyklisches) planetary gear train 30 and is driven by it.

in operation, the core air stream a is accelerated and compressed by the low pressure compressor 14 and directed into the high pressure compressor 15 where further compression is achieved. The compressed air discharged from the high-pressure compressor 15 is introduced into the combustion mechanism 16, where it is mixed with fuel and the mixture is combusted. The hot combustion products produced are then diffused through the high-pressure turbine 17 and the low-pressure turbine 19 and thereby drive the high-pressure turbine 17 and the low-pressure turbine 19 before they are discharged through the nozzle 20 for providing a certain propulsive force. The high-pressure turbine 17 drives the high-pressure compressor 15 via a suitable connecting shaft 27. The fan 23 typically provides the major portion of the propulsive force. The epicyclic (epicyklisches) planetary gear train 30 is a reduction gear train.

Optionally, the transmission can drive auxiliary and/or alternative components (e.g., a medium pressure compressor and/or a recompressor (nachverdcter)).

As a further example, the gas turbine drive shown in FIG. 1 has a split nozzle (Teilungsstromed ü se) 20, 22, meaning that flow through the bypass passage 22 has its own nozzle separate from the drive core nozzle 20 and radially outward of the drive core nozzle 20. however, this is not limiting and any aspect of the present disclosure can also apply to drives in which flow through the bypass passage 22 and flow through the drive core 11 are mixed or combined before (or upstream of) the only nozzle, which can be referred to as a mixed flow nozzle.

The geometry of the gas turbine drive 10 and its components are defined by a conventional axis system comprising an axial direction (which is aligned to the axis of rotation 9), a radial direction (in the direction from bottom to top in fig. 1) and a circumferential direction (perpendicular to the view in fig. 1). The axial, radial and circumferential directions extend perpendicular to each other.

Fig. 2 shows additional details of the gas turbine drive 10. The gas turbine drive 10 includes a sensor system 4. The sensor system 4 includes a plurality of physical (hardware) sensors 42A-42J disposed at various locations of the gas turbine drive 10. In the illustrated example, sensors 42A-42J are mounted at high pressure compressor 15, at fuel line 161, at combustion chamber 163 of combustion mechanism 16, at high pressure turbine 17, and at low pressure turbine 19, respectively. These sensors 42A-42J are connected to the analysis module 41 via signal lines. Additional sensors, which are not shown in fig. 2, can be coupled to the evaluation module 41.

A sensor 42A for measuring the volumetric flow of the air flowing through, a sensor 42B for measuring the temperature of the air flowing through and a sensor 42C for measuring the pressure of the air flowing through are arranged at the high-pressure compressor 15.

The combustion mechanism 16 includes a fuel injection portion 160 by which fuel supplied via a fuel line 161 is injected into a combustion chamber 163. The quantity of fuel injected can be regulated by means of an adjustable valve. A sensor 42D for measuring the volumetric flow of fuel flowing through is arranged on fuel line 161. A sensor 42E for measuring a temperature and a sensor 42F for measuring a pressure in the combustion chamber 163 are arranged at the combustion chamber 163.

Sensors 42G, 42I for measuring the temperature of the air flowing through and sensors 42H, 42J for measuring the pressure of the air flowing through are arranged on the high-pressure turbine 17 and the low-pressure turbine 19, respectively.

It should be noted that the gas turbine drive 10 need not necessarily include all of the mentioned sensors 42A-42J.

Analysis module 41 is configured to acquire signals from sensors 42A-42J.

The sensor system 4 provides one or more virtual sensors. Optionally, sensor system 4 provides a respective virtual sensor for each sensor 42A-42J. If one of the sensors 42A to 42J fails during operation of the gas turbine drive 10, for example the sensor 42G for measuring the temperature in the high-pressure turbine 17, the measured value of this sensor can be replaced by a value for the same measured variable, which value is determined by means of a virtual sensor. Alternatively or additionally, the value of the measured variable (e.g. the temperature in the high-pressure compressor) can always be determined, which can then be compared with the measured value of the sensor 42G in order to verify the measured value of the same measured variable. If the value of the measured variable determined (virtually) by the sensor system 4 deviates from the value of the measured variable determined by means of the (physical) sensor 42G, a defect of the sensor 42G can then be inferred.

Optionally, in response to identifying a defect of a sensor, a virtual sensor for its measured variable is automatically created.

FIG. 3 illustrates the sensor system 4 of the gas turbine drive 10, wherein not all of the sensors 42A-42J are shown for simplicity.

The evaluation module 41 is designed, for example, as a data processing means or comprises such a data processing means. The analysis module 41 includes a storage mechanism 410 for storing computer readable data. The sensor model is stored on storage mechanism 410 or multiple sensor models are stored on the storage mechanism. The analysis module 41 includes a plurality of inputs 411, wherein a sensor 42A-42J is coupled at each input 411. The sensors 42A-42J provide values of input variables at the respective coupled inputs 411 that are indicative of the variables to be measured by the respective sensors 42A-42J. The analysis module 41 calculates the value of the measured variable by means of the sensor model and from the value of the input variable. The analysis module 41 outputs this value of the measured variable at the output 412 and/or uses it as a further input variable for the sensor model of the analysis module 41 or for other sensor models, wherein its output variable can then again be output at the output 412. Optionally, one or more of the inputs 42A-42D are coupled at the output 412 or at the output 412 of the further sensor system 4. The sensor system 4 can comprise or provide a plurality of virtual sensors coupled (e.g. in communication) with each other.

A display means 5, for example, displaying the value of the measured variable, is coupled to the output 412.

The sensor model is based on values of the input variables and of the measured variables, wherein these values of the measured variables are detected by means of the sensors of the entity. The sensor system 4 includes a correlation module 40 for creating a sensor model.

The correlation module 40 comprises a memory unit 400, on which data sets are stored, which comprise values of input variables (for example values of some or all of the sensors 42A-42J of the gas turbine drive 10), furthermore, which comprise values of measured variables, which are detected by means of physical sensors, for example by means of the sensors 42g for measuring the temperature in the high-pressure turbine 17, the correlation module 40 is designed to determine a correlation between the values of the input variables and the values of the measured variables from the stored data sets, a subgroup (Untergruppe) of the input variables, which is particularly strongly correlated with the measured variables, is selected as a function of the determined correlation, as a result of which weakly correlated input variables are rejected, which can improve the quality of the sensor model to be created, the sensor model is created as a function of the determined correlation, the correlation module 40 uses machine learning, in the example shown, the correlation module 40 comprises a genetic programming (programemorganing), in which the iterative model is optimized as a function of the evolution rule ä.

The correlation module 40 provides the sensor model to the analysis module 41. To this end, the analysis module 41 comprises an input 413 via which the correlation module 40 is or can be connected with the analysis module 41.

The correlation module 40 is configured as a data processing means or comprises such a data processing means, for example. In the example according to fig. 3, the correlation module 40 and the analysis module 41 are spatially separated from each other. The analysis module 41 is arranged at the gas turbine drive 10 (or alternatively at an aircraft with the gas turbine drive 10). Optionally, but not mandatorily, the correlation module 40 is arranged at the gas turbine drive 10 (or aircraft).

When the correlation module 40 is arranged at the gas turbine drive 10, it can continuously take measurements of one, more or all of the sensors 42A-42J (e.g. via connections not shown to the sensors 42A-42J or via the analysis module 41). Using these measured values, which can characterize the input variables and/or measured variables, the correlation module 40 can newly create and/or optimize the sensor model continuously or in discrete time intervals (zeitlichen Schritten) and provide it to the analysis module 41. Thereby, for example, sensor drift can be compensated.

Alternatively, it is possible for the correlation module 40 to be spaced apart from (and/or communicatively separated from) the analysis module 41 after the acquisition of the data set, the creation of the sensor model and the provision of the sensor model (via the connection shown in fig. 3, alternatively wire-free via a data network or data carrier) to the analysis module 41.

It is then possible, for example, for test measurements to be carried out at the gas turbine drive 10 (or a gas turbine drive of identical or similar construction) in which physical sensors for measuring variables are provided, these test measurements providing a data set for creating (and optionally for verifying) a sensor model, in the case of operation of the gas turbine drive 10, the measured variables can thus be determined by means of the analysis module 41 (which provides virtual sensors) while physical sensors are not necessary, as a result of which, in the operation of the gas turbine drive 10, one or more sensors can be dispensed with, which can reduce the weight and improve the reliability, furthermore, the test measurements of the gas turbine drive 10 can be continued even if one or more sensors fail, the time available for the gas turbine drive at the test station (test ä denn), for example, being limited regularly and the effective load of the test station can be improved by means of the sensor system 4, since, for example, replacement of failed sensors is not mandatory.

In the example, the measured variable shows the temperature in the high-pressure turbine 17. In this case, in particular the volume flow (for example of the air flow or of the fuel flow), the air pressure upstream of the combustion chamber 163, the air pressure in the combustion chamber 163 and the air temperature downstream of the combustion chamber 163 are taken into account as input variables.

In other examples, the measured variable shows the actuating position of the actuating element, for example of the valve 162 (alternatively a movement derived therefrom). In this case, in particular the volumetric flow of the fuel in the fuel line 161, the pressure (for example in the combustion chamber 163) and the temperature (for example in the combustion chamber 163) are taken into account as input variables.

Fig. 4 shows an alternative embodiment of a sensor system 4' for the gas turbine drive 10 according to fig. 1. In contrast to the sensor system 4 according to fig. 3, a single data processing unit 43 is provided, which includes both the correlation module 40 and the evaluation module 41. The correlation module 40 and the analysis module 41 are configured, for example, in the form of software modules, which are connected or can be connected to one another via a software interface. The operating principle of the sensor system 4' corresponds, in addition, to the operating principle of the sensor system 4 according to fig. 3.

Optionally, the correlation module 40 and/or the analysis module 41 are integrated into a drive unit controller (ECU).

Fig. 5 shows a method for determining a measured variable, in particular of the gas turbine drive 10. In the case of the method, in particular, one of the sensor systems 4, 4' described above can be used.

In a first step S1, a data set is provided. The data set comprises values of a plurality of input variables and (then) values of measured variables to be determined. The values of the measured variables in the data record are, in particular, the values measured by the hardware sensors, i.e. it can be provided that the measured variables are measured by means of the hardware sensors and the values thus obtained are registered in the data record. In the case of a gas turbine drive, the data set can be determined, for example, within the framework of a drive test, alternatively or additionally during ongoing operation.

In a second step S2, a correlation between the value of the input variable and the value of the measured variable is calculated. An input variable subgroup of the plurality of input variables is selected based on the calculated correlations. For this purpose, for example, a correlation matrix can be created. All input variables that are significantly correlated with the measured variable can be selected into the subgroup. Alternatively, a predetermined number (for example between 3 and 10) or a predetermined share of the input variables with the strongest correlation can be selected as subgroups.

In a third step S3, a sensor model is created. The sensor model combines the input variables of the subgroup of input variables with the measured variables. Self-learning algorithms can be used in particular. The self-learning algorithm is trained, for example, using the data set.

In an optional fourth step S4, the quality of the sensor model is detected. For this purpose, a further data set (optionally a total data set, which is divided into data sets and further data sets) can be provided, which comprises the values of the plurality of input variables and (then) the values of the measured variables to be determined. The value of the measured variable can then be determined by means of the sensor model and on the basis of the values of the input variables of the further data set. These values can then be compared with the measured values of the measured variables present in the further data set. When the quality is insufficient (deviation exceeds a previously determined highest value, for example), it is possible to return to any of steps S1 to S3. For example, further, for example, larger-scale data records can be provided. For this purpose, it is possible, for example, to supplement the data set with simulated values of the input variables, in particular with simulated values representing extreme values. If the method is used, for example, in a gas turbine drive 10, the extreme values can indicate an overload situation, for example overheating of the gas turbine drive 10. Furthermore, the arrangement of the input variables in the data set can vary. For example, in order to suppress noise signals and thus to calculate the correlation more accurately, an average value, for example a sliding average value, can be formed. The creation of the sensor model can also vary, for example, when using genetic programming, for example, changing the number of iterations and/or the number of operators (opertoren).

Alternatively or additionally, in step S4, the quality parameter to be detected is the response time of the sensor model. If this time is too long, the sensor model can be simplified. Further alternatively or additionally, the stability of the sensor model, for example whether strong deviation values occur, can be detected.

If sufficient quality has been determined (generally or for the determined application) in step S4, the fifth step S5 is continued.

In a fifth step S5, at least one further value of at least one of the input variables, in particular of a plurality of input variables, in particular originating from each of the subgroups of input variables, is (respectively) detected. For this purpose, the respectively associated sensor is read.

In a sixth step S6, the value of the measured variable is calculated on the basis of at least one further value of at least one of the input variables (in particular the values of a plurality of input variables) and the sensor model. In this way, for example, a failed sensor and/or the value of a verification sensor can be replaced.

It is further possible that the one or more input variables are not measured by means of sensors, but are likewise determined according to the method described above (vorbechriebeenen), that is to say by means of corresponding sensor models. This can provide a cascade-type virtual sensor. At least one output variable of at least one virtual sensor can be used as an input variable of a further virtual sensor. The method can use a plurality of virtual sensors coupled (e.g., communicatively) to each other.

Steps S1 to S4 can be performed once in advance, for example, to acquire a sensor model. For example, a first gas turbine drive is used for this purpose, which is equipped with a sensor or a plurality of sensors for providing measured values of the measured variable, which sensor or plurality of sensors are not provided in mass-produced gas turbine drives. Steps S5 and S6 can thus be carried out at a later point in time, for example, in accordance with the mass production, for example, in one or more further (in particular structurally identical) gas turbine drives.

With the described method, the input variable can optionally also be monitored by comparing the determined measured variable with the measured value of the sensor for the measured variable. In this way, failures or faults can also be determined in the sensors of the input variables. The method can generally be used in order to achieve a redundant determination of the measured variable in order to correct faulty measurement locations (Messstelle) and/or in order to replace failed sensors. This makes it possible to avoid interruptions in the test (for example of the gas turbine drive 10) due to repair of defective sensor systems or repetitions due to defective sensor systems. Furthermore, sensors can be saved, which enables a simple construction.

optionally, one or more of the input variables are not measured by means of sensors, but are calculated either analogically or on the basis of models.

Furthermore, the sensor system 4, 4' and the method described above are particularly suitable for the following application in the case of a gas turbine drive.

Any input variable can be used in development testing that does not require real-time monitoring. Here, steps S5 and S6 can be performed online or offline. The quality and the validity of the test measurement can be improved by determining the measured variable by means of a sensor model. For the sensor model, for example, genetic programming and/or artificial neural networks can be used. Furthermore, such a test can be carried out in particular with fewer sensors. This improves the operating characteristics (Laufverhalten) of the gas turbine drive.

in the case of development test requiring real-time monitoring, steps S5 and S6 are performed online. In this case, the determination of the measured variable by means of the sensor model can in particular replace a failed sensor in order to prevent interruptions or delays in the test.

In the case of an application requiring real-time monitoring and authentication, steps S5 and S6 are performed online. In this case, the input variables are selected as reliability-relevant input variables. The operational reliability of the gas turbine drive can be improved by determining the measured variable by means of a sensor model. Here, a strict modeling is used for the sensor model, wherein the mathematical relationship between the input variables and the measured variables is known.

It will be understood that the invention is not limited to the embodiments described above and that various modifications and improvements can be made without departing from the concepts described herein. Any of the features described can be used separately or in combination with any other features as long as they are not mutually exclusive, and the present disclosure extends to all combinations and subcombinations of one or more of the features described herein and includes these.