DE102015225654A1 - Method and device for determining the state of a measurement data channel in a duplex measurement data channel - Google Patents

Abstract

The invention relates to a method for determining the state of a measurement data channel in a duplex measurement data channel, characterized in that
a) an approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) between measurement data (M ₁ , M ₂ ) of the measurement data channels (1, 2) of the duplex measurement data channel and at least one additional measurement date (Z ₁ , Z ₂ ), where
b) the at least one additional measurement date (Z ₁ , Z ₂ ) is causally, in particular physically, related to the measurement data (M ₁ , M ₂ ) of the duplex measurement data channel (1, 2) and
c) depending on the determined approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) automatically at least one state signal (10) is generated, which indicates the state of the at least one measurement data channel (1, 2).
The invention also relates to a device.

Description

The invention relates to a method for determining the state of a measurement data channel in a duplex measurement data channel having the features of claim 1 and a device for determining the state of a measurement data channel in a duplex measurement data channel having the features of claim 13.

In many technical systems, a large number of measurement data is transmitted via duplex measurement data channels, eg. For control or monitoring. It may happen that one of the two measurement data channels in the duplex measurement data channel is working incorrectly. However, it is usually not possible to determine which of the measurement data channels is faulty.

Therefore, it makes sense to develop methods and devices with which the state of a measurement data channel can be determined.

The object is achieved by the method having the features of claim 1.

The method is used to determine the state of a measurement data channel in a duplex measurement data channel. A state is understood here in particular as whether one of the channels (that is to say comprising the sensor, line, etc.) operates incorrectly or not.

For this purpose, an approximation of the correlation coefficients between measurement data of the measurement data channels of the duplex measurement data channel and at least one additional measurement data is determined, wherein the at least one additional measurement date causally, in particular physically, related to the measurement data of the duplex measurement data channel. The at least one additional measurement date thus forms a kind of reference, at which the actual measurement data are assessed by means of the calculation of the approximation of the correlation coefficients.

Then, depending on the determined approximation of the correlation coefficients, at least one state signal is automatically generated which indicates the state of the at least one measurement data channel. The status signal can z. B. may be an error message in connection with a channel indication.

In one embodiment, the status signal is generated as a function of the value of at least one of the approximated correlation coefficients and / or of a temporal change of at least one of the approximated correlation coefficients. Depending on the operating situation, it may be useful to use absolute values for comparison or temporal changes of the values. The methods can also be combined with each other. In this case, in particular, the at least one status signal can be output as a function of exceeding or falling below at least one predefined limit value. It can also be decided case by case which measured values are used as measured data and which measured values are used as the at least one additional measuring date.

In a further embodiment, the at least one additional measurement data is also transmitted in a duplex measurement data channel, so that analogous conditions to the actual measurement data are available.

For an efficient calculation, in one embodiment, the approximation of the correlation coefficients on the basis of the acquisition of measurement data using temporal averaging, a simple difference between a maximum and a minimum and / or the calculation of a scattering measure of the measured data, in particular over a predefined time window made , The use of a time window allows u. a. also a use in systems where strong dynamic effects play a role, as a sliding averaging takes place. It can be z. B. the time averaging over a sliding time window with a length between 0.1 and 6 seconds, in particular between 1 and 2 seconds, in particular between 4 and 240, in particular 100 measured data per time window are detected.

For an efficient calculation, in one embodiment, the measurement data for the approximation of the correlation coefficients are scaled and / or normalized. This ensures in particular that the numerical values are in a numerically meaningful range, so that z. B. different measures (temperature, pressure, etc.) do not distort the result due to their numerical values. As a rule, different physical signals can not be directly compared since their numerical values can vary considerably depending on the choice of unit of measurement or measured variable. The normalization to a certain number range prevents occurring problems. For example, number series can be transformed to the range [0, 1], by subtracting the minimum value of the measurement series from each measured value and dividing by the difference maxValue - minValue.

In case of disturbed measurement series - a frequently occurring case - on the other hand, the subtraction of the row mean value with respect to the fluctuation width of each signal, i. H. uses the empirical variance. In the definition equation of the empirical correlation coefficient (see below) the normalization is implicitly included. Since an MAD approximation is used, reference is made to normalization.

In one embodiment, the measurement data and / or the at least one additional measurement datum are electrical measurement data, in particular current data or voltage data. This can z. B. the electrical measurement data and / or the additional measurement data from the operation of an electric motor, in particular a torque motor, and / or a tachometer be.

In particular, the following pairings can be made: measurement data Additional measurement date Differential pressure sensors in the oil or fuel system Oil or fuel flow Differential pressure measurement on the oil filter Differential pressure measurement of the oil flow Low pressure shaft speed High pressure shaft speed Thrust lever adjustment fuel flow Position of the hydraulically actuated flaps of the thrust reverser Limit switch thrust reverser

Additionally or alternatively, the measurement data and / or the at least one additional measurement datum may be temperature data and / or pressure data.

In one possible application of the method, the measurement data channels are used to acquire measurement data and / or the at least one additional measurement data of a turbomachine, in particular an aircraft engine. In this case, the measurement data and / or the at least one additional measurement datum can be determined as a function of measurements of the rotational speed of the turbomachine. In this case, the measurement data and / or the at least one additional measurement date are not directly measured by a sensor, but the measured variable "rotational speed" can be converted by predetermined calculation models into another measured variable, which is then used instead of the measurement data.

The object is achieved by a device having the features of claim 13.

Exemplary embodiments of the method and the device are described in conjunction with the figures. It shows

1 a schematic sectional view through a two-shaft aircraft engine with measurements;

2 a schematic representation of two duplex measurement data channels, wherein a duplex measurement data channel detects an additional measurement date;

3 a schematic representation of an embodiment of an apparatus for determining a faulty measurement data channel in a duplex measurement data channel.

In 1 is an aircraft engine in a sectional view 100 in particular, measured variables (for example temperatures, pressures) are given which play a role in the described embodiments. Basically, but also others, in the 1 not shown measured variables are used. So z. As measured data M ₁ , M ₂ also by means of computer models from the measured variable "speed" of the aircraft engine 100 be derived. These derived measurement data are also measurement data M ₁ , M ₂ in the sense of the following description.

Also is the aircraft engine 100 just an example of a turbomachine. The embodiments are also applicable to other machines or turbomachinery, such. B. stationary gas turbine, if the measuring channels are designed in duplex-type.

The aircraft engine shown here as an exemplary embodiment 100 is designed as a double-wave. A low pressure compressor 101 (here only a single Fanstufe) is by a two-stage low-pressure turbine 104 driven. A ten-stage high pressure compressor 102 is powered by a two-stage high-pressure turbine 103 driven. Between the high pressure compressor 102 and the high-pressure turbine 103 is an annular combustion chamber 105 arranged.

The following example introduces some terms for measures, namely pressures and temperatures, which are useful for measurement and control purposes.

The in the aircraft engine 100 entering air A has a pressure P0. Before the fan stage 101 the air has a pressure P20 and a temperature T20.

Behind the high pressure compressor 102 the compressed air has a pressure P30 and a temperature T30. Behind the combustion chamber 105 there is a temperature T46.

At the outlet of the low-pressure turbine 104 there is a pressure P50.

The temperature and pressure measurement data M ₁ , M ₂ are over in 1 measurement data channels not shown 1 . 2 to an electronic evaluation unit 106 (EEC: Engine Electronic Control).

In the 1 illustrated embodiment is only to be understood as an example. The method and apparatus illustrated herein may also be applied to turbomachinery of a different geometry, purpose, or design, such as a three-shaft aircraft engine 100 , be applied. Also, the temperatures and pressures selected here are to be understood as measures only as an example.

In 2 are exemplary two measurement data channels 1 . 2 represented, each having a measured variable (for example, a pressure or a temperature) via sensors 11 . 12 to capture. In 2 As an example, the temperature T30 is used as the measured variable.

This is a so-called duplex channel (duplex channel), in which two signals (ie measured data) of the same measured variable are measured. If the duplex measurement data channel 1 . 2 works error-free, the two measured data M ₁ , M ₂ must be the same.

In this example it is assumed that the temperature T30 after the high pressure compressor 102 in both measurement data channels 1 . 2 is measured. The two sensors 11 . 12 in each case take on a temperature signal M ₁ , M ₂ .

If in one of the measurement data channels 1 . 2 an error occurs, there is a divergence of the measured data measured M ₁ , M ₂ .

This is by default in the evaluation unit 106 indicated by a so-called "cross check failure" (Xch failure). However, due to this information, it is not possible to identify which of the two measurement data channels 1 . 2 is defective.

In the method described here and the device described here, the two measured data M ₁ , M _{2 are combined} with additional measured data Z ₁ , Z ₂ (ie measured data of an additional measured variable Z) which are physically related to the measured data M ₁ , M ₂ via physical laws. In principle, the relationship can also exist via other causal laws, such as chemical or data-related relationships. In the example shown here, the two additional measurement data Z _1, Z ₂ are also in a duplex auxiliary measuring data channel 3 . 4 transmit, with sensors 13 . 14 take the additional measurand Z

The duplex measurement data channels 1 . 2 can get an x-check error. If then the additional measurement data channels 3 . 4 duplex and error-free, you still have to decide on a, but error-free, additional measurement data channel 3 . 4 (ie a reference channel) fall.

The additional measurement data channels 3 . 4 can have stochastic or transient differences among each other, but they do not trigger an X-check error. If the first duplex measurement data channel 1 on the first additional measurement data channel 3 and the second duplex measurement data channel 2 to the second additional Measurement data channel 4 the difference is reflected as an additional risk in the calculation of the correlation coefficients. This should be avoided by opting for one of the (error-free) channels.

The duplex additional measurement data channels 3 . 4 but could also report an X-check error, then the process could easily be reversed if at the same time the duplex measurement data channels 1 . 2 are error-free. The accuracy can z. B. be recognized that no range error or EEC error are reported and the signals are within the expected range.

In principle, it is also possible that the additional measured variable Z only via a Zusatzmessdatenkanal 3 is transferred alone. If the additional measurement data channel 3 only simplex exists, this decision is omitted.

It then simply refers to this channel, which is present. The disadvantage of this constellation is that the method can not be reversed, but this does not seem to be necessary because the simplex channel can not have an X-Check error. Errors in a simplex channel must evolve into a range error before they become identifiable or there are other specific assessment criteria.

In the present example, changes in the measured signals M ₁ , M _{2 considered are} also found in the additional measured data Z ₁ , Z ₂ . An increase in the measured temperature T30 is z. B. due to physical conditions with an increase or decrease of the additional measure Z causally connected (in particular by a physical law or an approximately linear relationship).

In the following it shall be assumed that there is a linear relationship between the measured values M ₁ , M ₂ and the additional measured data Z ₁ , Z ₂ .

The linear relationship between two signals, which may be stochastically disturbed, is determined by the correlation coefficient KT. Its value is a number in the range [-1, 1]. +1 is synonymous with a directly proportional, linear relationship and -1 is equivalent to inverse proportionality. A correlation coefficient of 0 means that the two signals have no relationship.

In the following example, the measurement signals P30 and T30 in the pressure chamber of the high pressure compressor are considered.

As the volume of the pressure chamber of the high pressure compressor 102 is almost constant, P30 and T30 must have the ideal gas equation pV = T · constant directly proportional and linearly related. That is, as P30 increases, T30 must also increase because at approximately constant volume, this is the only way to maintain the constancy of the ratio pV / T can be maintained. A correlation coefficient of the measurement signals P30 and T30 will have a value close to or even equal to +1.

There are in the duplex measurement data channel 1 . 2 (Channels 1 and 2 ) two measurements of the temperature T30 with T30 ₁ and T30 ₂ as measured data M ₁ , M ₂ .

There is in the duplex additional measurement data channel 3 . 4 (Channels 1 and 2 ) two measurements of the pressure P30 with P30 ₁ and P30 ₂ as additional measurement data Z ₁ , Z ₂ .

Furthermore, it is assumed that the pressures P30 ₁ , P30 ₂ (ie the additional measurement data Z ₁ , Z ₂ ) are measured without errors, but not necessarily free of disturbances (including stochastic disturbances). Furthermore, it is assumed that between the two temperature measurement data T30 ₁ and T30 ₂ a "Cross Check Failure" (Xck-Failure) is detected.

Between the measured data T30 ₁ and P30 ₁ and the measured data T30 ₂ and P30 ₁ , the correlation coefficients (KT ₁ , KT ₂ ) are continuously calculated (eg in a 25 ms cycle) according to one embodiment of the method. They must both be (nearly) the same as well as close to +1. Where the equality represents the superficially tested property.

Now, if the Xck failure occurs in the T30 ₁ , T30 ₂ measurement data, the correlation coefficient KT of the defective measurement data channel will decrease and tend to 0. In case of error, the correlation goes between temperature and pressure, linked via the ideal gas equation, lost. Thus, the proportionality and the linear dependence of P30 no longer exists.

In the case of an Xck failure in the T30 measurements with a defective first measured value T30 ₁ , the correlation coefficient KT ₁ approaches 0 while the correlation coefficient KT ₂ ≈ 1 remains.

Thus, in this case, the measured value M ₁ (ie T30 ₁ ) as the defective signal, or the temperature channel 1 as a defective channel. The error can z. B. in the first sensor 11 for T30 ₁ and / or in the line of the measurement data channel 1 lie yourself.

Similarly, a faulty T30 measurement in the second measurement data channel 2 KT ₁ ≈ 1, while KT ₂ approaches 0.

By linking the measurement data M ₁ , M ₂ with the physically related additional measurement data Z ₁ , Z ₂ , a statement can be reached as to which of the measurement data channels 1 . 2 is flawed.

The embodiment of the method also works in the reverse case of error, if in an Xck failure in the pressure measured variables P30 (T30 error-free), the faulty pressure signal to be determined. The following applies: KT ₁ = f (T30 ₁ , P30 ₁ ) KT ₂ = f (T30 ₁ , P30 ₂ ) where the function f represents the physical relationship.

When KT ₁ goes to 0 while KT ₂ remains substantially equal to 1, the first pressure measurement channel is 3 malfunction. If KT ₁ remains approximately equal to 1 while KT ₂ approaches 0, then the second pressure gauge will be 4 malfunction.

This makes it clear that the measurement data M ₁ , M ₂ and the additional measurement data Z ₁ , Z _{2 can} also be reversed, as long as the causal relationship remains.

Starting from this method, the following describes how the correlation coefficients KT ₁ , KT ₂ can be calculated efficiently. The point here is that the correlation coefficients KT ₁ , KT _{2 are determined} as a function of the measured values M ₁ , M ₂ and the additional measured values Z ₁ or Z ₂ .

A complete calculation of the empirical correlation coefficient K _xy takes place in two steps, where x, y stand for the measured quantities.

With i = {1, ..., n} the arithmetic mean values x _m , y _{m are} determined.

Then the correlation coefficient KT is _xy

The calculation of the correlation coefficients KT _xy during operation, z. B. an aircraft engine 100 , requires a considerable computing power, so that in the method proposed here an approximation in the calculation of the correlation coefficients KT is used, which is based on the mean absolute differences (MAD: Mean Absolute Differences). The calculation is faster and can be made recursive. The recursive calculation can be used to calculate mean and scatter. In this case, the correlation coefficients KT are calculated over a time window with the length w.

If, in general, running index i = {1,..., N}, the calculation is now carried out over a window of width w. Thus, for the run index i = {n + 1 - w, ..., n}. The arithmetic mean values are then

The following standardization is performed on the empirical scattering s _m :

So that applies to MAD:

The MAD coefficients thus calculated provide a measure of how close the measurement signals (here denoted by x, y) lie on average.

This approximation MAD of the correlation coefficient KT is now applied to the above example of the temperature and pressure signals T30, P30.

For the respective first measuring channel 1 . 3 of the duplex measuring data channels for the temperature T30 ₁ and the pressure P30 ₁ , the following arithmetic mean values apply to a window with the length w:

The following standardization is then carried out:

Thus, the approximation of the first correlation coefficient MAD _{1 is} :

For the second measuring channel 2 . 4 of the duplex measuring data channels for the temperature T30 ₂ and the pressure P30 ₁ , the following mean values apply to a window with the length w:

The following standardization is then carried out:

For the value P30 _1norm , the calculation does not have to be done twice. Thus, the approximation of the second correlation coefficient MAD ₂ results in:

The approximated correlation coefficients MAD ₁ and MAD ₂ are u. U. not close to +1, but are in the error-free case almost the same.

Thus, it does not depend on the absolute values of the approximated correlation coefficients MAD ₁ , MAD ₂ , but on the difference. Over time, the changes of the approximated correlation coefficients MAD ₁ , MAD _{2 are} of particular importance.

In the Xch-Failure case (here it is assumed that T30 ₁ fails), the approximated correlation coefficient MAD ₁ , which belongs to the defective measurement data channel, greatly decreases compared to the approximated correlation coefficient MAD ₂ , which belongs to the intact measurement data channel.

If z. B. MAD ₁ greatly reduced compared to MAD ₂ and MAD ₂ remains near 1, it can be assumed in the Xck error case of the T30 sensors that the first measurement data channel 1 is faulty. It then becomes a status signal 10 generates that indicates the faulty state here.

It is assumed that the pressure measurements P30 did not generate a simultaneous error message.

The calculations of the MAD coefficients are cyclic. However, the comparison of the MADs should only be activated if an Xck failure in temperature or pressure signals is detected. A permanent comparison is also possible and could possibly detect an "imminent" Xck failure.

In 3 schematically an embodiment of a device is shown, which serves to a faulty measurement data channel 1 . 2 (please refer 2 ) of a duplex measurement data channel. In the embodiment, it is assumed that this device is connected to the evaluation unit 106 (eg EEC) is coupled. In principle, this device can also be designed in another way, in particular also form an autonomous computer unit.

The measurement data M ₁ , M ₂ and the additional measurement data Z ₁ , Z ₂ described above constitute input data for an approximation unit 20 , which calculates the approximated correlation coefficients MAD ₁ , MAD ₂ .

Depending on the determined approximated correlation coefficient MAD ₁ , MAD ₂ then the state signal 10 formed and submitted, by which it can be determined which of the measurement data channels 1 . 2 is faulty.

LIST OF REFERENCE NUMBERS

1

first measurement data channel of a duplex measurement data channel

2

second measurement data channel of a duplex measurement data channel

3

first measurement data channel of a duplex additional measurement data channel

4

second measurement data channel of a duplex additional measurement data channel

10

state signal

11

first sensor for a measured variable

12

second sensor for a measured variable

13

first sensor for an additional measurand

14

second sensor for an additional measurand

20

Approximationsvorrichtung

21

signaling device

100

Turbomachine, aircraft engine

101

Low pressure compressor (power amplifier)

102

High pressure compressor

103

High-pressure turbine

104

Low-pressure turbine

105

combustion chamber

106

evaluation unit

A

incoming airflow

KT

correlation coefficient

M

measurement data

M ₁

first measured data

M ₂

second measured data

MAD

Approximation of the correlation coefficient

w

Time window

Z

Additional measured variable

Z ₁

first additional measured value

Z ₂

second additional reading

Claims (13)

Method for determining the state of a measurement data channel in a duplex measurement data channel, characterized in that a) an approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) between measurement data (M ₁ , M ₂ ) of the measurement data channels ( 1 . 2 ) of the duplex measurement data channel and at least one additional measurement date (Z ₁ , Z ₂ ) is determined, b) the at least one additional measurement date (Z ₁ , Z ₂ ) causally, in particular physically, with the measurement data (M ₁ , M ₂ ) of the duplex Measurement data channel ( 1 . 2 ) and c) depending on the determined approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) automatically at least one status signal ( 10 ) which generates the state of the at least one measurement data channel ( 1 . 2 ) indicates. Method according to Claim 1, characterized in that the at least one status signal ( 10 ) is generated as a function of the value of at least one of the approximated correlation coefficients (MAD ₁ , MAD ₂ ) and / or of a temporal change of at least the approximated correlation coefficients (MAD ₁ , MAD ₂ ). Method according to claim 1 or 2, characterized in that the at least one status signal ( 10 ) is issued in response to exceeding or falling below at least one predefined limit value. Method according to at least one of the preceding claims, characterized in that the at least one additional measurement date (Z ₁ , Z ₂ ) in a duplex measurement data channel ( 3 . 4 ) be transmitted. Method according to at least one of the preceding claims, characterized in that the approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) on the basis of the acquisition of measurement data (M ₁ , M ₂ ) using a temporal averaging, a simple Difference formation between a maximum and a minimum value and / or the calculation of a scattering measure of the measured data (M ₁ , M ₂ ), in particular over a predefined time window (w), takes place. A method according to claim 5, characterized in that the time averaging over a sliding time window (w) takes place with a length between 0.1 and 6 seconds, in particular between 1 and 2 seconds, in particular between 4 and 240, in particular 100 measured data per time window (w) are recorded. Method according to at least one of the preceding claims, characterized in that the measurement data for the approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) are subjected to scaling and / or normalization. Method according to at least one of the preceding claims, characterized in that the measurement data (M ₁ , M ₂ ) and / or the at least one additional measurement datum (Z ₁ , Z ₂ ) are electrical measurement data, in particular current data or voltage data. A method according to claim 8, characterized in that the electrical measurement data (M ₁ , M ₂ ) and / or the at least one additional measurement date (Z ₁ , Z ₂ ) from the operation of an electric motor, in particular a torque motor and / or a tachometer , come. Method according to at least one of the preceding claims, characterized in that the measurement data (M ₁ , M ₂ ) and / or the at least one additional measurement date (Z ₁ , Z ₂ ) are temperature data and / or pressure data. Method according to at least one of the preceding claims, characterized in that the measurement data channels ( 1 . 2 ) for acquiring measured data (M ₁ , M ₂ ) and / or for detecting at least one additional measuring datum (Z ₁ , Z ₂ ) of a turbomachine ( 100 ), in particular an aircraft engine, serve. Method according to at least one of the preceding claims, characterized in that the measurement data (M ₁ , M ₂ ) and / or the at least one additional measurement date (Z ₁ , Z ₂ ) in dependence on measurements of the rotational speed of the turbomachine ( 100 ) be determined. Device for determining the state of a measurement data channel in a duplex measurement data channel, characterized by a) an approximation device ( 20 ) for correlation coefficients (KT ₁ , KT ₂ ) of the measurement data (M ₁ , M ₂ ) of the measurement data channels ( 1 . 2 ) of the duplex measurement data channel, wherein at least one additional measurement date (Z ₁ , Z ₂ ) for the Determination of the approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ) which are causally, in particular physically, related to the measurement data (M ₁ , M ₂ ) of the duplex measurement data channel and b) a signaling device ( 21 ) for automatically generating at least one status signal ( 10 ) for characterizing the state of the at least one measurement data channel ( 1 . 2 ) as a function of the determined approximation (MAD ₁ , MAD ₂ ) of the correlation coefficients (KT ₁ , KT ₂ ).

Images