EP3774487A1

EP3774487A1 - Method and apparatus for diagnosis and monitoring of vehicles, vehicle components and routes

Info

Publication number: EP3774487A1
Application number: EP19727863.3A
Authority: EP
Inventors: Daniela DEJACO; Bernhard Lukas GIRSTMAIR; Gerald Grabner; Andreas Haigermoser; Justinian Rosca; Johannes Simon
Original assignee: Siemens Mobility Austria GmbH
Current assignee: Siemens Mobility Austria GmbH
Priority date: 2018-05-16
Filing date: 2019-05-15
Publication date: 2021-02-17
Also published as: RU2763414C1; US20210217256A1; CN112469613A; WO2019219756A1

Abstract

The invention relates to a method and an apparatus for diagnosis and monitoring of vehicles, vehicle components, routes and route components, wherein at least one first sensor (1) is used to perform measurements and at least one computing unit (10) is used to effect a signal processing (11). It is proposed that the at least one computing unit (10) be supplied with at least measured first signals (4), that at least one first characteristic value be formed from the at least first signals (4), that the at least one first characteristic value or at least one first characteristic value combination be classified by means of at least one first statistical model (15), or a prediction (14) be performed, and that at least one technical first condition indicator for at least one first vehicle component or at least one route component be determined. This achieves safe detection of faults, damage, excess wear, etc., and effective, condition-oriented maintenance of vehicles and infrastructures.

Description

Method and device for diagnosing and monitoring vehicles, vehicle components and routes

The invention relates to a method for diagnosing and monitoring vehicles, vehicle components, routes and track components, in particular for rail vehicles and infrastructures of rail vehicles, wherein by means of at least one first sensor measurements are performed and by means of at least one arithmetic unit

Signal processing takes place.

Vehicles, especially rail vehicles must have a high level of driving safety. An accurate assessment and

Prediction of technical conditions of vehicles,

Landing gears, driveways and other vehicle as well as

Infrastructure components is therefore important.

Another important factor is the effective and efficient maintenance and servicing of vehicles and infrastructures.

DE 101 63 148 Al is known from the prior art, which is a monitoring and diagnostic method for a

Rail vehicle shows, with operating parameters of the

Rail vehicle detected and supplied to characterize a driving behavior of the rail vehicle, the operating parameters of a monitoring hierarchy.

The mentioned approach has in its known form the

Disadvantage is that although measurement signals are evaluated by means of statistical methods, but no evaluations or forecasts of evaluation results under use

statistical methods are evident.

The invention is therefore based on the object, a developed over the prior art, especially

specify reliable method.

According to the invention this object is achieved with a method of the type mentioned, in which at least one arithmetic unit is supplied with at least measured first signals,

in which at least one first parameter is formed from the at least first signals,

wherein the at least one first characteristic value or at least one first characteristic value combination is classified by means of at least one first statistical model or a prediction is carried out on the basis of the at least first characteristic value or the at least first characteristic value combination, and wherein at least one technical first state indicator with respect to at least one first vehicle component or at least one infrastructure component of at least one

Classification result or from at least one

Prediction result is determined.

This ensures a particularly reliable detection of damage as well as effective, condition-based maintenance and repair

Maintenance of vehicles and infrastructures, assuming a tolerable or negligible number of false alarms.

This is brought about by the fact that in the method according to the invention a classification or a prediction of

Characteristics or combinations of different

Characteristics is carried out by means of statistical models. This means that e.g. erroneous or error-free states or a faulty or error-free behavior of a vehicle component (for example a damper) or a faulty or error-free behavior

Track component (e.g., a track) no longer needs to be detected on the basis of rigid limits, but that, for example, limit values or threshold curves etc. depending on operating conditions of a vehicle or infrastructure, a failure behavior of a vehicle or infrastructure component o.ä. are customizable. Condition indicators may be e.g. out

Frequency distributions of classification results or prediction results. An advantageous embodiment is obtained when the

at least first statistical model is formed by means of a machine learning method.

As a result of this measure, as the quantity of acquired operating data increases, an ongoing reliability improvement or enhancement of the method according to the invention with respect to detection and evaluation of technical conditions is achieved since the first statistical model is trained on the basis of the operating data and thus fault states of faultless states increasingly differentiated as time progresses.

It is advantageous if the at least first statistical model is formed on the basis of an adjustment calculation.

For the classification and / or prediction of characteristic values, various statistical models are thus available in connection with the method according to the invention, which can also be combined with one another. So it is

For example, not only is it possible to detect instantaneous error conditions, but also to estimate error conditions by means of interpolation or extrapolation

Predict operating limits, etc.

A favorable solution is achieved if the at least first status indicator has a first probability value for entry by the at least one

Classification result or the at least one

Prediction result is assigned to be indexed technical condition.

As a result of this measure, risks with respect to false alarms are further reduced since detection of error conditions is based not only on limit values or threshold value courses but also on probabilities for false alarms

Violations of these limits or these threshold curves etc. can be made. These probabilities may be formed, for example, based on frequencies of said violations.

It is furthermore advantageous if signals from one to nl signal categories and one to n2 determined from them

Characteristics one to n3 state indicators are formed with one to n4 probability values, each of the one to n3 state indicators one of the one to n4

Probability values are assigned and from the one to n4 probability values a combination status indicator is formed with an associated combination probability value.

The combination status indicator can be formed, for example, by means of a probabilistic graphical model.

By using more than one status indicator, technical conditions can be evaluated with better quality. For example, indicates the first state indicator or a first state indication with a high first one

Probability of an error of the first

Vehicle component and pushes a second

Condition indicator or a second state indication with a low second probability value from a fault of a second vehicle component, so the first

Condition indication by means of the second state indication (and vice versa) plausibility and hedged.

An advantageous embodiment is obtained when data from at least one method step in an on-board

Diagnosis and / or monitoring device of a vehicle can be used. For example, characteristic values or

Condition indicators are used.

In the on-board diagnostic and / or monitoring device, a further evaluation and / or a formation of warning events, status or error outputs, etc. can take place. These can be used, for example, as visual and / or audible warnings and / or status information or monitoring and / or diagnostic information is output in a cab of the vehicle.

However, it is also advantageous if data from at least one method step is transmitted to an infrastructure-side device.

By this measure, a land-side evaluation of characteristic values, status indicators, etc. can take place (for example, in a maintenance status), where data of entire vehicle fleets

can be brought together and compared with each other.

The invention is based on

Embodiments explained in more detail.

They show by way of example:

Fig. 1: A flowchart for an exemplary

Variant of a

inventive method, Fig. 2: A result diagram with by means of a

Support Vector Machine - Method

certain classification results for an exemplary embodiment of the method according to the invention, and

Fig. 3: A rail vehicle with sensors, a

Arithmetic unit and a

Data transmission device for

Implementation of the invention

Procedure.

An example shown schematically in Fig. 1, exemplary

Embodiment variant of a method according to the invention, partially computer-implemented method is in a computing unit 10, which in a car body 21 one in Fig. 3rd

arranged rail vehicle is arranged,

implemented.

On a chassis frame 22 of a chassis 23 of the

Rail vehicle, a first sensor 1 is provided which measures accelerations of the chassis frame 22, i. is designed as an acceleration sensor. Furthermore, a sensor 2 designed as a temperature sensor is connected to a damper 24 of the chassis 23 and is as

Strain gauge formed third sensor 3 on a spring 28 of the chassis 23 is arranged.

The second sensor 2 measures damper temperatures, the third sensor 3 deformations of the spring 28th

The first sensor 1, the second sensor 2 and the third sensor 3 continuously perform measurements.

The first sensor 1, the second sensor 2, the third sensor 3 and the arithmetic unit 10 are supplied with electricity via a power supply device, not shown.

In the arithmetic unit 10, which is connected to the first sensor 1, the second sensor 2 and the third sensor 3, there is a current signal processing 11 and evaluation of continuously detected first signals 4 of the first sensor 1, second signals 5 of the second sensor 2 and third signals 6 of the third sensor 3.

Furthermore, in evaluation processes of the arithmetic unit 10 already preprocessed reference signals are included, i.

recorded in time before the first signals 4 and

processed first reference signals, temporally recorded before the second signals 5 and processed second

Reference signals and temporally recorded before the third signals 6 and processed third reference signals. The signal processing 11 comprises a storage of the first signals 4, the second signals 5 and the third signals 6 and their preparation for the evaluation processes.

The first signals 4 and the first reference signals are acceleration signals of the

Chassis frame 22, the second signals 5 and the second reference signals to temperature signals of the damper 24, the third signals 6 and the third

Reference signals to deformation signals of the spring 28th

From the first signals 4 are first characteristic values, from the second signals 5 second characteristic values and from the third

Signals 6 third characteristic values formed (characteristic value determination 12). Temporally before the first characteristic values, the second characteristic values and the third characteristic values become from the first

Reference signals 7, from the second reference signals second reference characteristics 8 and from the third reference signals third reference characteristics 9 formed. In the first characteristic values, the second characteristic values, the third characteristic values, the first reference characteristic values 7, the second reference characteristic values 8 and the third

Reference parameters 9 are statistical

Characteristics, in this exemplary embodiment of a method according to the invention by means of averages

Acceleration, damper temperature and

Spring deformation curves. According to the invention, however, it is also possible that minima, maxima, etc. are formed as statistical characteristic values.

The statistical characteristic values relate to defined path sections or the path sections via a relationship with a travel speed of the rail vehicle

assignable periods of time. About this way

Time periods are the characteristics associated with the reference characteristics, i. the first characteristic values are on the same path and time segments as the first reference parameters 7

related, the second characteristics on the same way and Time segments such as the second reference characteristic values 8 and the third characteristic values correspond to the same path and time segments as the third reference characteristic values 9.

The first characteristic values, the second characteristic values and the third characteristic values are inserted into a first statistical model 15, a second statistical model 16 and into further statistical models which are implemented in the arithmetic unit 10 and are continuously classified and predicted by means of these statistical models (Classification 13 and

Prediction 14).

The first statistical model 15 and the second statistical model 16 are formed by means of a method of machine learning, the so-called Support Vector Machine.

The method of the Support Vector Machine is well known in the art and will be exemplary of this

Embodiment of a method according to the invention used in a linear binary variant to classify characteristic values or characteristic value combinations.

The first statistical model 15 will be out of the first

Reference characteristics 7 and the second reference characteristics 8 and the second statistical model 16 from the first

Referenzkennwerten 7 and the third reference characteristics 9 formed or trained. The first reference characteristics 7, the second reference characteristics 8 and the third reference characteristics 9 thus function as learning data for the first statistical model 15 and the second statistical model 16.

It is known, for example, from a past observation of the damper 24 that when the undercarriage frame 22 is over a long time, i. with an observed frequency an acceleration in a defined first

Acceleration interval and the damper 24 for a long time, ie at an observed frequency a temperature Within a defined first temperature interval, an error-free state of the damper 24 is present. Furthermore, it is known from the past observation of the damper 24 that when the undercarriage 22 over a long time, ie with an observed frequency acceleration within a defined second acceleration interval and the damper 24 for a long time, ie with an observed frequency, a temperature within a Have defined second temperature interval, a damage or a defective or defective state of the damper 24 is present.

The acceleration intervals and the temperature intervals which indicate a faulty or faultless state of the

Dampers 24 may vary depending on the

Change the operating behavior of the rail vehicle. It is essential that the damper 24 is observed while doing so

Acceleration intervals and temperature intervals with erroneous or error-free states of the damper 24 are brought into connection.

The first statistical model 15 is based on

Identification combinations or reference characteristic combinations formed, i. spatially and temporally correlated pairings of first reference characteristics 7 and second

Reference characteristics 8, depending on whether they in the first acceleration interval or in the second

Acceleration interval or in the first

Temperature interval or in the second temperature interval, favorable or unfavorable states of the damper 24 at.

Reference characteristic combinations, which are a favorable

Indicate damper state (i.e., pairings of first

Reference characteristics 7 and second reference characteristics 8, which in the first acceleration interval and the first

Temperature interval lie), define a first

State area 25, which is below a separation line 27, indicating an unfavorable damper state

Reference characteristic combinations (ie pairings from first Reference characteristics 7 and second reference characteristics 8, which in the second acceleration interval and the second

Temperature interval lie) a second state region 26, which is above the separation line 27. The first

State area 25, the second state area 26 and the separation line 27, which are shown diagrammatically in FIG. 2, can be localized via vectors.

The separation line 27 is formed from the reference value combinations in a manner such that normal distances from the

Separation line 27 to the reference characteristic combinations of the first state area 25 are mathematically negative and normal distances from the separation line 27 to the

Reference characteristic combinations of the second state region 26 are mathematically positive. A corresponding

Formulation rule for the separation line 27 is known from the prior art and is described for example in Ethem A.,

Introduction to Machine Learning, Third Edition, The MIT Press Cambridge, Massachusetts, London, England, 2014

described.

In this exemplary embodiment of a

inventive method is additionally provided, statistical models based only on a

Reference characteristic value category (e.g., based on the second reference characteristics 8 and their temporal behavior, etc.), i. do not use reference characteristic combinations. This can also damper states

are estimated when e.g. only one reference characteristic category is available on sensor defects.

The second statistical model 16 is formed according to the scheme described above, which is also used to form the first statistical model 15. However, the formation of the second statistical model 16 will be

Accelerations and acceleration intervals of the

Chassis frame 22 in combination with deformations and

Deformation intervals of the spring 28 used. By means of the first statistical model 15 and the second statistical model 16, the classification 13 of

Characteristics performed.

Here are from the first characteristics and the second

Characteristics formed first characteristic combinations in

Depending on their position with respect to the separation line 27 associated with the first state area 25 or the second state area 26 of the first statistical model 15.

From the first characteristic values and the third characteristic values

formed second characteristic combinations are according to the same scheme corresponding state areas of the second

assigned to statistical model 16.

In this exemplary embodiment of a

inventive method is additionally provided, the classification 13 not only on the basis of

Characteristic combinations, but also on the basis of individual characteristics (for example, the second characteristic values and their temporal behavior) to perform. For this purpose, for example, the second characteristic values are used in that statistical model which is formed by means of the second reference characteristic values 8 without the use of reference characteristic value combinations.

By means of further statistical models based on a

Adjustment calculation, the prediction 14 is based on characteristic values. For this purpose, known from the prior art, linear regression functions are provided or implemented in the arithmetic unit 10. This

Regression functions are formed from the first characteristics, the second characteristics, and the third characteristics by means of educational rules known in the art and used to perform interpolations or extrapolations based on the characteristics.

For example, regressions of the second characteristic values and of the first characteristic value combinations are dependent on the

Driving speed performed to a functional relationship between the second characteristics and the first characteristic combinations on the one hand and the driving speed on the other hand to be able to form. As a result, for example, maximum travel speeds can be determined, up to which the damper 24 can be operated without premature excessive wear, etc. occurs.

Frequency distributions of classification results and prediction results thus become continuous

Condition indicators determined (indicator determination 17).

In the case of classification results, this exemplary embodiment of a method according to the invention involves allocations of characteristic values or

Characteristic combinations for status areas of statistical models.

Prediction results are e.g. Forecasts of medium

Damper temperatures as a function of the driving speed.

A first state indicator is formed by a number of first characteristic combinations associated with the second state region 26 being set to a total number of first

Characteristic combinations, i. is set in proportion. From this, a first probability value is formed, which gives a statement about how

probably a damper defect. The first

Thus, the status indicator contains information "damper defective" with an associated probability for this damper defect.

On the same scheme becomes concerning the second

Kennwertkombination a second state indicator relating to a spring defect with an associated second

Probability value formed.

Other status indicators are added to the same scheme from the second characteristic values or their assignment

State areas formed in by means of the second reference parameters 8 without the use of Referenzkennwertkombinationen statistical models. Status indicators from prediction results

formed by an extrapolated middle

Damper temperature a confidence level

is assigned, which depends on a

Occurrence frequency of lying in a defined tolerance range, extrapolated mean damper temperatures is determined. State indicators from prediction results thus give u.a. Forecasts of a condition of the damper 24 (e.g., a prediction that high wear of the damper 24 when operating above a certain vehicle speed and a life reduction of the damper 24 are anticipated) or another component associated therewith

Confidence level.

Furthermore, an ongoing indicator combination 18 is performed such that signals from one to nl = 3

Signal categories (first signals 4, second signals 5, third signals 6) and therefrom determined one to n2 = 3 characteristic values (first characteristic values, second characteristic values, third characteristic values, n2 is greater than or equal to nl) one to n3 = 2 status indicators (first status indicator, second state indicator) with one to n4 = 2 probability values (first

Probability value, second probability value) are formed (see above), where each of the one to n3 = 2 state indicators one of the one to n4 = 2

Probability values are assigned (see above) and from the one to n4 = 2 probability values

Combination status indicator with an associated one

Combination probability value P _{K is} formed.

For this, each of the one to n4 = 2 probability values is made from a frequency of a specific one

Classification result based on a total number of classification results formed (see above) and the combination probability value P _K as conditional

Probability from the one to n4 = 2

Probability values formed. The combination probability value P _K is formed by means of a probabilistic graphical model based on a method of machine learning.

In principle, the following educational regulation is used:

The combination probability value P _K is determined from a first parameter P _DF , which indicates a probability for an indication of an error of the damper 24 by the first state indicator and the second state indicator in the actual presence of a faulty state of the damper 24, from a second parameter P _F which describes a probability for an actual occurrence of a faulty state of the damper 24 and a third parameter P _D , which is a probability for an indication of an error by the first

State indicator and the second state indicator indicates determined.

The combination probability value P _K describes a probability of an actual occurrence of a failed condition of the damper 24 under a condition where an error is indicated by the first condition indicator and the second condition indicator.

Shows the first directly related to the damper 24

For example, if the condition indicator is likely to fail, and the second low probability indicator related to the spring 28 indicates an error, then the combination condition indicator will likely indicate a failure of the damper 24.

The first parameter P _DF , the second parameter P _F and the third parameter P _D are continuously adapted via operating observations. For example, the second parameter P _{F is} increased when increased over a defined period of time Frequency of a damper defect is observed. These operational observations become learning data for one

functional relationship of the second parameter P _F over a time formed by means of which an increase of a

Failure probability of the damper 24 parameterized with increasing duration of use and the probabilistic graphical model is trained.

In this exemplary embodiment of a

According to the method of the invention, the first state indicator is related to a first vehicle component (the damper 24) and the second state indicator relates to a second vehicle component (the spring 28). However, according to the invention it is also conceivable that e.g. both the first state indicator and the second state indicator point to the first one

Obtain vehicle component. This is realized, for example, in that two different sensors are provided on or in the damper 24, which are different

Generate signals (e.g., temperature signals and oil pressure signals, etc.).

In this exemplary embodiment of a

The method according to the invention also becomes a third one

Condition indicator formed on the basis of the first characteristic values and the driving speed, which erroneous and

faultless conditions of a track, i. a guideway or a track component indicated.

From observations of a track damage behavior is explained above scheme with reference characteristic combinations of accelerations of the chassis frame 22 and

Driving speeds a statistical model formed by means of a Support Vector Machine - method, which

Status areas for faulty and error-free

Has track conditions.

Kennwertkombinationen from continuously measured accelerations of the chassis frame 22 and measured speeds are assigned to the state areas, creating a Classification 13 of characteristic combinations takes place. From a frequency distribution corresponding

Classification results, the third condition indicator is determined.

Data formed in the method step of the indicator combination 18 (data from the first status indicator, the second status indicator and the third status indicator) are used in a diagnostic and monitoring device implemented in the arithmetic unit 10 and continuously evaluated there. Furthermore, these data are shown for further evaluation via a shown in Fig. 3

Data transmission device 19 of the rail vehicle also continuously to a maintenance level, i. to one

infrastructure-side device transmitted (data transmission 20).

FIG. 2 shows a diagram with a first state area 25, a second state area 26 and a separation line 27 of a first statistical model 15, which are also described in connection with FIG. 1, are formed on the basis of a support vector machine method and via which technical conditions of a damper 24 shown in Fig. 3 are evaluable.

Accelerations are plotted on an x-axis of the diagram, and damper temperatures on a y-axis.

The first state region 25 is a first function value 29 and a second function value 30 first

Identification combinations, which are described in connection with FIG. 1 assigned.

The second state area 26 is a third function value 31 and a fourth function value 32 first

Assigned characteristic value combinations.

The first function value 29, the second function value 30, the third function value 31 and the fourth function value 32 are formed from first characteristic combinations. The first Characteristic combinations are in turn made up of first characteristic values relating to accelerations of one shown in FIG

Chassis frame 22 and second characteristics with respect to temperatures of the damper 24 formed.

With respect to a Cartesian coordinate system 33, the first function value 29 and the second function value 30 have negative y 'coordinates, and the third function value 31 and the fourth function value 32 have positive y' coordinates.

The first function value 29 and the second function value 30 are assigned to a favorable damper state, the third function value 31 and the fourth function value 32 to an unfavorable state of the damper 24.

As described in connection with Figure 1, from a cluster of adverse damper conditions (e.g., more than three times a day, such conditions occur), a first condition indicator is formed, optionally indicating a damper defect.

Also, according to the scheme shown in FIG. 2

Function values, which are formed from second characteristic combinations, associated with corresponding state areas of a second statistical model 16 described in connection with FIG. 1. The second characteristic combinations are, as mentioned in connection with FIG. 1, formed from the first characteristic values as well as from third characteristic values with respect to a deformation of a spring 28 shown in FIG. 3. As described in connection with FIG. 1, a second state indicator is formed from an accumulation of unfavorable spring states based on an assignment of acceleration-spring deformation function values to state regions, which optionally indicates a spring defect (eg more than three times a day when such spring states occur). ,

According to the invention, an assignment of the second characteristic values to state regions of a further statistical model is also present Base of a support vector machine method provided.

This further statistical model is used with temporal

Steps of second reference parameters 8 described in connection with FIG. 1 are formed as a temperature-time relationship.

Second characteristic values, which branch with increasing time increasing temperatures, which lie above a defined period of time above a temperature-time separating straight line formed especially for this further statistical model, indicate a damper defect or excessive damper wear.

In Fig. 3, a rail vehicle with a car body 21 and a chassis 23 is shown. The chassis 23 has a damper 24, which is a primary damper, and a spring 28, which serves as a primary spring

is executed, whose technical conditions are determined and evaluated by means of a method according to the invention described in connection with FIG.

On a chassis frame 22 of the chassis 23 is a first sensor 1 for measuring accelerations of

Chassis frame 22, i. an acceleration sensor

intended. Furthermore, a second sensor 2, which is designed as a temperature sensor, with the damper 24th

connected .

In addition, the spring 28 has a third sensor 3, which is designed as a strain gauge.

The first sensor 1, the second sensor 2 and the third sensor 3 are provided with a computing unit 10 in the car body 21

Signal and current conducting connected.

In a roof area of the rail vehicle is a

Data transmission device 19, which is designed as a radio device and signal and current conducting with the

Arithmetic unit 10 is connected, provided. The data transmission device 19 also has a Global Positioning System unit

Locating device on.

The arithmetic unit 10 is supplied with power via a vehicle electrical system (not shown) of the rail vehicle and in turn supplies the first sensor 1, the second sensor 2, the third sensor 3 and the data transmission device 19 with electricity.

In the computing unit 10, a computer program product is installed, by means of which process steps of the method according to the invention according to FIG. 1, i. a

Signal processing 11, a characteristic determination 12, a

Classification 13, a prediction 14, a

Indicator 17 and a combination of indicators 18.

Off in the arithmetic unit 10, which as a diagnostic and

Monitoring device of the rail vehicle acts, certain status indicators are warning events

formed, which via a train bus, not shown

Rail vehicle to a display unit in a likewise not shown cab of the rail vehicle and transmitted there as warnings or status information (for example, to display a damper defect or a residual life of the damper 24).

Furthermore, the state indicators on the

Data transmission device 19 by radio to a

Infrastructural facility, which is located in the

Connection with this exemplary embodiment of a device according to the invention is a maintenance status, transmitted for further evaluation via a vehicle fleet. List of terms

1 first sensor

2 second sensor

3 third sensor

4 First signal

5 second signal

6 Third signal

7 First reference parameter

8 Second reference characteristic

9 Third reference parameter

10 arithmetic unit

11 Signal _P rocessing

12 characteristic value determination

13 classification

14 prediction

15 First statistical model

1 6 Second statistical model

17 Indicator

1 8 indicator combination

1 9 Data transmission device

20 data transmission

21 car body

22 chassis frame

23 suspension

24 dampers

25 First state area

2 6 Second state area

27 separation line

2 8 spring

2 9 First function value

30 Second function value

31 Third functional value

32 Fourth function value

33 coordinate system

P _K combination probability value PDF First parameter

PF Second parameter

PD Third parameter

Claims

claims

1. A method for diagnosing and monitoring vehicles, vehicle components, routes and infrastructure components, in particular for rail vehicles and infrastructures of rail vehicles, wherein by means of at least one first sensor measurements are carried out and by means of at least one processing unit signal processing, characterized

in that the at least one arithmetic unit (10) is supplied with at least measured first signals (4),

in that at least one first parameter is formed from the at least first signals (4),

in that the at least first characteristic value or at least a first characteristic value combination by means of at least one first characteristic value

statistical model (15) is classified or one

Prediction (14) is performed on the basis of at least the first characteristic value or the at least first characteristic combination, and

in that at least one technical first status indicator with respect to at least one first vehicle component or at least one infrastructure component comprises at least one

Classification result or from at least one

Prediction result is determined.

2. The method according to claim 1, characterized in that the at least first state indicator of a

Frequency distribution of classification results or prediction results is determined.

3. The method according to claim 1 or 2, characterized in that the at least first characteristic value is formed as related to a path portion or a period of time statistical characteristic value.

4. The method according to any one of claims 1 to 3, characterized in that the at least first statistical model (15) is formed by a method of machine learning.

5. The method according to claim 4, characterized in that the at least first characteristic value or the at least first

Characteristic combination is classified by means of a Support Vector Machine method.

6. The method of claim 4 or 5, characterized in that from before the at least first signals (4) processed reference signals reference characteristics as

Learning data are formed.

7. The method according to any one of claims 1 to 3, characterized in that the at least first statistical model (15) is formed on the basis of a balance calculation.

8. The method according to claim 7, characterized in that the at least first characteristic value is used in a regression function.

9. The method according to any one of claims 1 to 8, characterized in that the at least first state indicator, a first probability value for an occurrence of an at least one classification result or the at least one prediction result to be indexed

technical condition is assigned.

10. The method according to claim 9, characterized in that the first probability value is formed from a frequency of a specific classification result based on a total number of classification results.

11. The method according to any one of claims 1 to 8, characterized in that from signals from one to nl

Signal categories and determined from one to n2

Characteristic values one to n3 status indicators with one to n4 Probability values are formed, where each of the one to n3 state indicators one of one to n4

Probability values are assigned and from the one to n4 probability values, a combination status indicator with an associated combination probability value (P _{K) is} formed.

12. The method according to claim 11, characterized in that each of the one to n4 probability values is formed from a frequency of a specific classification result related to a total number of classification results.

13. The method according to claim 11 or 12, characterized

in that the combination probability value (P _{K) is formed} as a conditional probability from the one to n4 probability values.

14. The method according to any one of claims 11 to 13, characterized in that the combination status indicator is formed by means of a probabilistic graphical model.

15. The method according to claim 14, characterized in that the probabilistic graphical model is formed on the basis of a method of machine learning.

16. The method according to any one of claims 1 to 15, characterized in that from second signals (5) at least a second characteristic value and a technical second

State indicator with respect to the at least first

Vehicle component are formed.

17. The method according to any one of claims 1 to 15, characterized in that from second signals (5) at least a second characteristic value and a technical second Condition indicator are formed with respect to a second vehicle component.

18. The method according to any one of claims 1 to 17, characterized in that data from at least one

Step in an on-board diagnostic and / or monitoring device of a vehicle can be used.

19. The method according to any one of claims 1 to 18, characterized in that data from at least one

Step to be transferred to an infrastructure-side device.

20. Device for carrying out a method according to one of claims 1 to 19, characterized in that the at least first sensor (1) on a chassis (23) of the

Vehicle is arranged and the at least one computing unit (10) is provided in the vehicle and connected to the at least first sensor (1).

21. The device according to claim 20, characterized in that the at least one arithmetic unit (10) in one

Car body (21) of the vehicle is provided.

22. Device according to claim 20 or 21, characterized

characterized in that for data transmission (20) at least one data transmission device (19) is provided in or on the vehicle, which is connected to the at least one computing unit (10).