WO2018178196A1

WO2018178196A1 - Method for determining a degree-of-damage uncertainty of a motor vehicle

Info

Publication number: WO2018178196A1
Application number: PCT/EP2018/058009
Authority: WO
Inventors: Rafael Fietzek; Stéphane Foulard; Stephan Rinderknecht
Original assignee: Technische Universität Darmstadt
Priority date: 2017-03-30
Filing date: 2018-03-28
Publication date: 2018-10-04
Also published as: DE102017106919A1; DE112018001704A5

Abstract

The invention relates to a method (1) for determining a degree-of-damage uncertainty (18) of a component of a motor vehicle (3). In a determining step (2), during the operation of the motor vehicle (3), motor vehicle parameters (6) are ascertained continuously at predefined time intervals on the basis of items of sensor and driving state information (4, 5) available in the respective motor vehicle (3). The motor vehicle parameters (6) are subsequently processed in one or more parallel and/or successive evaluation steps (7, 9), and a degree of damage (11) of the component is ascertained. A degree-of-damage uncertainty (18) is ascertained on the basis of predefined parameter uncertainty variables and/or functions (12), which for each determined motor vehicle parameter (6) represent a degree of uncertainty for whether the determined motor vehicle parameter (6) corresponds to a real motor vehicle parameter (6), and on the basis of predefined evaluation step uncertainty variables and/or functions (14, 15), which for each evaluation step represent a degree of uncertainty for whether evaluation step influencing variables, independent of the motor vehicle parameters (6) and taken into consideration in the respective evaluation step (7, 9), correspond to the real evaluation step influencing variables. The degree-of-damage uncertainty (18) represents a measure for whether the component has actually sustained damage corresponding to the ascertained degree of damage (11).

Description

Darmstadt University of Technology

Method for determining a damage measurement uncertainty of a motor vehicle

The invention relates to a method for determining a damage degree uncertainty of a motor vehicle, wherein in a determining step during operation of the

Motor vehicle at predetermined time intervals

continuous vehicle parameters based on available in the respective motor vehicle sensor and

Fahrzustandsinformationen be determined and wherein the vehicle parameters are subsequently in one or more parallel and / or successive

Evaluation steps are processed and a

Damage measure of the component is determined.

Various methods for the determination of damage measurements are known from the prior art. For example, it is known from the prior art, based on determined in the determination step vehicle parameters, first a load spectrum for the component such

For example, to determine a gearbox. By doing

Load collective while the course of the load acting on the component, such as the forces acting or voltages occurring over time using known methods mapped. Starting from the load collective thus determined in a first evaluation step, in a subsequent second evaluation step based on a previously determined for the component Wöhler-Kurve, which shows a result of a Wöhler experiment carried out in advance for a structurally identical component, using the linear damage accumulation hypothesis according to Miner- Haibach to determine the degree of damage.

In addition, from the prior art and methods

known in which, starting from the in the

In a first evaluation step, one or more features that have a comparatively high correlation with a damage of the component (for example, so-called condition monitoring and / or condition prediction methods) are first determined in a first evaluation step. Subsequently, these features are processed in a further evaluation step with a previously determined mathematical model and determined in this way the damage measure. The mathematical model is usually using appropriate

Parameter identification method based on

existing vehicle parameters and associated damage measures determined.

The vehicle parameters can either be determined directly in the determination step by suitable measuring devices or can be determined on the basis of mathematical models, for example between measurement information determined by different measuring devices.

For example, it is known in the motor vehicle field, that of an internal combustion engine to a crankshaft of the

Engine provided torque by a mathematical model in dependence on a metrologically detected engine speed and a predetermined by a control unit of the motor vehicle target injection quantity determine. For the purposes of the invention, both the torque determined on the basis of the mathematical model and the engine rotational speed used for determining the torque and detected by the sensor as well as the engine torque

Solleinspritzmenge characteristic control size than

Understood motor vehicle parameters.

The determination of the degree of damage with the aid of the methods known from the prior art is fundamentally subject to uncertainties. This uncertainty in the

Determining the degree of damage results, for example, from unavoidable manufacturing tolerances, inter alia due to fluctuating material quality and differences between the respective production equipment used in the production of the component.

An object of the invention is to provide a method for determining a damage measure uncertainty, which defines the damage measure uncertainty first and then continuously by the inclusion of further

Information and data minimized.

This object is achieved in that based on predetermined parameter uncertainties

and / or functions that are specific to each one

Vehicle parameters an uncertainty measure for it

represent that the particular vehicle parameter corresponds to a real motor vehicle parameter, and based on predetermined evaluation step uncertainty quantities and / or functions, for each evaluation step

Uncertainty measure that of the

Automotive parameters independent and in the respective Considered evaluation step

Evaluation step influencing variables the real one

Evaluation step influencing variables correspond to one

Damage uncertainty is determined, the

Damage measurement uncertainty represents a measure of the fact that the component has actually suffered a damage corresponding to the determined damage level. When determining the damage measurement uncertainty, both

Parameter uncertainty quantities or functions, such as

For example, a known probability distribution of measurement noise one for determining a

Motor vehicle parameters used sensor or a model uncertainty of a for determining a

Automotive parameters used mathematical model considered as well as various

Evaluation step uncertainty quantities or functions that, for example, a scatter or a

Probability distribution of a Wöhler curve used to determine a partial damage describes. Based on the previously determined

Different uncertainty quantities or functions can be calculated using known methods from the

Probability theory and statistics the

Damage measurement uncertainty are calculated.

In the automotive parameters, it may be next to

According to the invention, the parameters directly influencing the operation of the motor vehicle are also variables which influence the operation of the motor vehicle, for example the respective ambient temperature and current weather data. The different uncertainty measures, in particular the

Parameter uncertainties and functions that

Evaluation step uncertainty variables and functions and the measure of damage uncertainty advantageously each indicate a probability that the

determined value corresponds to the actual value or that the actual value lies in a certain value range around the determined value. Advantageously, the uncertainty measures each bound a range of values within which the true value of the observed

Size, such as a metric with a

certain probability lies. This means that the measure of uncertainty is a measure of whether the

observed value corresponds to the actual value.

The damage measurement uncertainty is determined by the

Consideration of all uncertainty measures

Advantageously, a measure or a measure of probability, with which a damage measurement range is described, in which the damage measure with a certain

Probability is. In other words, the damage uncertainty describes the probability with which the determined damage corresponds to the actual one. The measure of injury inaccuracy represents a measure, namely a measure of the probability of determining whether the damage identified is actually the case

corresponds.

In a particularly advantageous embodiment of the

The method according to the invention provides for the parameter uncertainty quantities and / or functions and / or the evaluation step uncertainty quantities and / or functions each by one

Probability density function is. With the aid of a probability density function, a probability can be determined by integration that the respectively determined motor vehicle parameter or the variable determined in the respective evaluation step actually corresponds to a corresponding real variable. On the basis of the known probability density functions can also be a common

Likelihood density function can be specified, in which the various uncertainties in the determination of the vehicle parameters and in the

Further processing in the different

Evaluation steps can be considered. The

Probability density functions can, for example, under the assumption of a normal distribution in the determination of the vehicle parameters and the variables determined in the respective evaluation steps by a

Expected value and a variance or a

Standard deviation of the vehicle parameters or sizes are described.

By taking into account the various uncertainties, a scattering of the motor vehicle parameters and scattering of intermediate variables determined in the respective evaluation steps therefore essentially results according to the invention

which, among other things, are due to manufacturing tolerances as well as factors not taken into account in the mathematical models used.

Advantageously, the invention provides that, starting from the damage measure and the Damage probability uncertainty a default probability for predefined points in time in the future of the component is determined, whereby the probability of failure for each time is a measure of the fact that the component will fail at this time. From the state of

Technique is known, an expected

Failure time of the component in the future due to a linear updating of the damage determined up to a current point in time or of the determined one

Determining the extent of damage in the future. On the basis of the damage uncertainty determined with the aid of the method according to the invention, it is advantageously possible, with additional consideration of an inaccuracy of use, which represents an uncertainty measure for the fact that the damage occurred in the past also in the future

Expect a probability of default for it

be specified that the respective component will actually fail at the determined failure time. According to the invention, it is advantageously provided that a load collective is determined in a load collective evaluation step with the determined motor vehicle parameter. According to the invention, for example, a time profile of a torque load of a motor vehicle operation can be mapped in the load collective.

Advantageously, the invention provides that the damage measure in an accumulation evaluation step is determined by a linear damage accumulation using a Wöhler curve determined in advance for the component. Advantageously, it is provided according to the invention that a dispersion of the Wöhler curve can be achieved by accumulation evaluation results and / or functions

is taken into account. Through the Wöhler curve is

usually an average of a Wöhler experiment carried out in one or more experiments

Stress amplitude plotted over the sustainable number of cycles. A scatter of measurement results in the Wöhler experiment is usually comparatively large, so that the consideration of this scattering for the

Determination of damage measurement uncertainty in particular

is relevant.

In the evaluation step uncertainty variables or functions used for the accumulation evaluation step in the determination of the damage uncertainty, according to the invention, inter alia, it can be taken into account that a temporal course of load changes acting on the component is not taken into account when determining partial damages based on the Wöhler curve time course basically has an impact on the actual injury, but off

Simplification in the determination of the

Damage measure is disregarded. At a

Particularly advantageous embodiment of the method according to the invention is therefore intended that by the

Accumulation evaluation step uncertainty quantities and / or functions in a caused by an unconsidered course of load changes uncertainty in a

Determination of partial damage in the context of linear damage accumulation. Advantageously, the invention provides that the determined damage measure and the

Damage measurement uncertainty in a diagnostic system (for example, using condition monitoring and / or

Condition prediction methods) of the motor vehicle

Determining a cause of error is used. The damage measurements and damage uncertainties determined from a damage calculation can be used particularly advantageously to determine a faulty or damaged component which has led to a detected fault condition in the context of an error detection on the basis of fault detection functions. Appropriate

Error detection functions that are part of

Diagnostic systems are used, are known from the prior art.

Advantageously, the invention provides that, when determining the probability of failure, results of a diagnostic system of the motor vehicle are taken into account. By considering information about

Errors and causes of errors from a diagnostic system in the determination of the probability of failure can be taken into account certain events that are not taken into account when determining the damage uncertainty.

Advantageously, the invention provides that the determined damage dimensions and the associated

Damage uncertainties and / or the

Failure probabilities along with the

Automotive parameters and / or the load collectives are transmitted to a central database. According to the invention, it is provided that the parameter uncertainty variables and / or functions and / or the

Evaluation step uncertainty quantities and / or

functions based on the central database of numerous comparable components of a

Car fleet transmitted data centrally determined and subsequently transferred as required to individual vehicles. Advantageously, the

Parameter uncertainty quantities and / or functions and / or the evaluation step uncertainty variables and / or

functions to all motor vehicles of the

Car fleet transmitted. In this way, the determination of the damage dimensions and

Damage uncertainties and the

Default probabilities are continuously adjusted and improved.

According to the invention, it is advantageously provided that the parameter uncertainty variables and / or functions and / or the evaluation step uncertainty variables and / or functions are determined and continuously minimized with the aid of machine learning algorithms. By using machine-learn-ring algorithms, the parameter uncertainty quantities and / or functions and / or the evaluation step uncertainty variables and / or functions can be processed and minimized particularly efficiently on the basis of the large data volumes transmitted to the database.

Advantageously, the determined damage dimensions and the associated damage uncertainties and / or the failure probabilities together with the

Motor vehicle parameters and / or load collectives regularly transferred to the central database. But it is also possible and provided according to the invention that the determined damage dimensions and the associated

Damage uncertainties and / or the

Failure probabilities along with the

Automotive parameters and / or the load collectives are then transmitted to the central database if the associated component has failed unexpectedly or has caused an error or if the component is not

failed, although the failure was expected.

According to the invention, the data is then transmitted to the database if the probability of default for

Time of failure or failure of the component smaller than a first predetermined

Default probability limit is or if the

Default probability at the current time greater than a second predetermined

Failure probability limit is and the component has either failed or caused an error.

Further advantageous embodiments of the method according to the invention are based on a in the drawing

illustrated embodiment illustrated. It shows:

Figure 1 is a schematically illustrated flow of the

method according to the invention in a determination of a damage measure for a transmission according to the method of Miner-Haibach and Figure 2 shows another schematically illustrated sequence of the method according to the invention.

FIG. 1 schematically shows the sequence of a method 1 for determining a damage measure uncertainty

shown. In a determination step 2, during the operation of a motor vehicle 3 with the aid of an observer-based method known from the prior art based on metrologically detected physical quantities 4 of the motor vehicle 3 and possibly further variables 5 characterizing the operation of the motor vehicle 3, a transmission 5 of the motor vehicle 3 is determined acting torque continuously calculated. This continuously calculated

Torque represents a motor vehicle parameter 6.

On the basis of the continuously determined torque, a load collective 8 acting on the transmission is first of all calculated in a load collective evaluation step 7.

Subsequently, in an accumulation evaluation step 9, a damage measure 11 is determined by linear damage accumulation using a Wöhler curve 10 determined beforehand for the transmission.

At the same time, for each determined and acting on the transmission torque based on predetermined

Parameter Uncertainty Quantities and / or Functions 12 a torque parameter uncertainty 13 determines that the determined torque actually a real on the

Gear acting torque corresponds. The

Torque parameter uncertainty 13 is particularly dependent on an absolute value of the torque, since the

Observer-based determination of torque at higher Torque values a less accurate calculation of the

Torque allowed.

In addition, based on predetermined

Accumulation evaluation step uncertainty quantities and / or functions 14, 15 caused by an unconsidered course of load changes

Load change uncertainty 16 in a determination of

Partial damages in the context of the linear

Damage accumulation and a Wöhler curve uncertainty

17, by which a dispersion of the Wöhler curve is taken into account.

On the basis of the torque parameter uncertainty 13, the load change uncertainty 16 and the Wöhler curve uncertainty 17, a damage uncertainty 18 is determined. In a not shown advantageous

Extension of the procedure 1 is based on the

Damage measure 11 a residual life or a

Time of failure determined and for this loss time based on the damage measure uncertainty 18 and with additional consideration of a failure uncertainty, which represents an uncertainty measure that the

calculated failure time corresponds to a real failure time, a default probability determined.

In Figure 2 is schematically another

Embodiment of the method according to the invention explained in more detail. In order to determine a damage measure (D) 11 using methods known from the prior art, at predetermined times (t), t ₂ , t ₃ , t _n ) 19

Partial damage measurements AD (t) 20 determined. Amongst other things Due to measurement uncertainties relevant vehicle parameters (F) 21 for determining the Teilschädigungsmaße not accurate, but for example with

Uncertainties of measurement are detected. The respective measurement uncertainty is determined by a

Parameter uncertainty quantity or function (P (F (t)) 12 at each instant (ti, t ₂ , t ₃ , t _n ) 19. This parameter uncertainty quantity or function (P (F (t)) 12 can narrow a range of values within which the true or actual value of the observed

Motor vehicle parameters (F) 21 with a specific

Probability is.

In addition, for example, too

Manufacturing tolerances by a corresponding

Probability distribution are taken into account, which at the different times (ti, t ₂ , t ₃ , t _n ) 19 are subject to no change. From the various Teilschädigungsmaßen (AD (t)) 20 can at any time (ti, t ₂ , t ₃ ) 19 also a

Partial damage uncertainty (P (AD (t))) 22 can be determined. On the basis of the partial damage dimensions (AD (t)) 20 and the partial damage uncertainties (P (AD (t))) 22, the damage measure (D) 11 and a

Damage uncertainty (P (D)) 18 at each time point (ti, t ₂ , t ₃ , t _n ) 19 are determined. With increasing

Measuring duration, probability distributions, which represent the damage measure uncertainty 18, flatter, since at each time point (ti, t ₂ , t ₃ , t _n ) 19 more

Uncertainties must be considered. This has the consequence that a probability of failure calculation of a Component based on the damage measure 11 also

becomes increasingly inaccurate.

To counteract this, data 23 of each vehicle under consideration is transmitted to a central database 24 where it is augmented with metadata 25 such as weather conditions at the different times (ti, t ₂ , t) 19 at the locations where the respective vehicle is different Time points (ti, t ₂ , t ₃ ) 19 has found.

The method can advantageously be expanded in such a way that recorded, actual failures of a component based on the determined damage measure 11 and the damage uncertainty 18 are compared with the predicted failure of the component. Self-learning algorithms 26 can use this information

if appropriate taking into account the metadata 25, the parameter uncertainty variables or functions (P (F (t)) 12 and also evaluation step uncertainty variables and / or functions and / or the damage uncertainty 18 and / or the partial damage uncertainties 22 are improved in such a way that a more accurate prognosis is made possible ,

Claims

PATENT APPLICATIONS

1. Method (1) for determining a

Injury uncertainty (18) of a component of a

Motor vehicle (3), wherein in a determination step (2) during operation of the motor vehicle (3) at predetermined time intervals continuously motor vehicle parameters (6, 21) based on in the respective motor vehicle (3) available sensor and driving condition information (4, 5 ), and wherein the motor vehicle parameters (6, 21) are subsequently determined in one or more parallel and / or successive evaluation steps (7, 9).

and a damage measure (11) of the component is determined, characterized in that on the basis of predetermined parameter uncertainty variables and / or functions (12), which are determined for each

Motor vehicle parameters (6, 21) represent an uncertainty measure that the particular vehicle parameters (6, 21) corresponds to a real motor vehicle parameters, and based on predetermined

Evaluation step uncertainty quantities and / or functions (14, 15) that enter for each evaluation step

Uncertainty measure that of the

Vehicle parameters (6, 21) independent and in the respective evaluation step (7, 9) considered

Evaluation step influencing variables the real one

Evaluation step influencing variables correspond to one

Injury uncertainty (18), where the injury measure uncertainty (18) represents an uncertainty measure for the fact that the determined damage level (11) has suffered corresponding injury.

2. Method (1) according to claim 1, characterized in that it concerns the parameter uncertainty variables and / or

functions (12) and / or at the

Evaluation step uncertainty quantities and / or functions (14, 15) are each a probability density function.

3. Method (1) according to claim 1 or claim 2, characterized in that starting from the damage measure (11) and the damage measure uncertainty (18) a

Failure probability for predetermined times in the future of the component is determined, the

Probability of default for each time is a measure of the fact that the component will fail at this time.

4. Method (1) according to one of the preceding claims, characterized in that in one

Lastkollektivauswertungsschritt (7) with the determined motor vehicle parameters (6, 21) a load collective (8) is determined.

5. The method (1) according to claim 4, characterized in that the damage measure (11) in a

Accumulation evaluation step (9) by a linear damage accumulation using a previously determined for the component Wöhler curve (10) is determined.

6. Method (1) according to claim 5, characterized in that

Accumulation evaluation step uncertainty quantities and / or functions (14, 15) a scatter of the Wöhler curve (10) is taken into account.

7. Method (1) according to claim 5 or claim 6, characterized in that by the

Accumulation evaluation step uncertainty quantities and / or functions (14, 15) an uncertainty caused by an unconsidered course of load changes in a determination of partial damage in the context of

linear damage accumulation is taken into account.

8. Method (1) according to one of the preceding claims, characterized in that the determined damage measure (11) and the damage measure uncertainty (18) in one

Diagnosis system of the motor vehicle (3) is used to determine a cause of the error.

9. Method (1) according to claim 3, characterized in that in the determination of the probability of failure

Results of a diagnostic system of the motor vehicle (3) are taken into account.

10. The method (1) according to one of the preceding claims, characterized in that the determined

Damage measurements (11) and the associated

Damage uncertainties (18) and / or the

Failure probabilities along with the

Vehicle parameters (6, 21) and / or the Load collectives (8) are transmitted to a central database (24).

11. The method (1) according to claim 10, characterized

characterized in that the parameter uncertainty quantities and / or functions (12) and / or the

Evaluation step uncertainty quantities and / or functions (14, 15) on the basis of the number of comparable components of a central database (24)

Data transmitted to the vehicle fleet (23) are determined centrally and subsequently transferred to individual vehicles (3) as needed.

12. The method (1) according to claim 11, characterized

Evaluation step uncertainty quantities and / or functions (14, 15) are determined using a machine learning algorithm and continuously minimized.