Methods for identifying and correcting faulty load collectives in motor vehicles

Abstract

The invention relates to a method for verifying a load collective determined in a vehicle (0), comprising the steps of: providing (S1) a value (1) of a load collective determined in the actual operation of the vehicle (0); providing (S2) one or more determined influencing variables (3) on the load collective; executing (S3) a machine learning (ML) model with the at least one influencing variable (3) as input, wherein an estimate (5) of the load collective is obtained as output after the execution of the machine learning (ML) model; and determining (S4) a deviation between the estimate (5) of the load collective from the value (1) of the load collective determined in the actual operation of the vehicle (0) and checking the deviation for exceeding a predetermined limit condition and / or checking for the absence of a value (1) of the load collective from the actual operation of the vehicle (0);

Description

[0001] The invention relates to a method for checking a load spectrum determined in a vehicle, as well as a computing system for stationary arrangement outside a vehicle for carrying out the method.

[0002] Load and stress spectra are often measured in road vehicles. These are parameters that determine the service life of vehicle components.

[0003] In this context, DE 10 2012 005 679 A1 relates to a method for load spectrum-based operational optimization of a motor vehicle, comprising the following steps: a) recording a frequency distribution of at least one operating parameter of a motor vehicle during practical operation, b) defining a functional dependency of a characteristic variable on the at least one operating parameter, in particular in the form of a characteristic map, c) defining a plurality of predetermined values ​​of an optimization parameter, d) calculating a resulting characteristic variable for each of the values ​​of the optimization parameter predetermined in step b) by summing the characteristic variable over the at least one operating parameter, taking into account the frequency distribution of the at least one operating parameter recorded in step a), e) determining an optimal value of the optimization parameter from the plurality of resulting characteristics calculated in step d).

[0004] Furthermore, DE 10 2010 047 820 A1 describes the creation of load spectra for tire pressure and axle load. DE 10 2007 018 176 A1 discloses a method for determining an aging index based on a catalyst load spectra.

[0005] DE 10 2008 019 463 A1 also predicts failure events of vehicle components using load spectra and a damage function. DE 10 336 597 A1 also discloses a method for compressing load spectra data to reduce storage requirements in the vehicle.

[0006] Load spectra can be used on-board in the vehicle for additional functions or collected centrally in a database and later evaluated for product improvement. In the former case, errors in the load spectra can impair the functionality of the associated functions. In the latter case, errors in the load spectra collection do not cause any directly noticeable negative consequences for the customer. Nevertheless, load spectra errors also bring with them various disadvantages: Incorrect conclusions can be drawn from faulty load spectra, leading to incorrect decisions in product development; missing load spectra reduce the sample of usable data for product improvement and thus the statistical reliability; and very expensive and scarce storage space on vehicle control units (permanent read / write memory, especially EEPROM) is not used effectively in the event of a fault.

[0007] Load spectrum errors can have various causes: - On-board errors in the vehicle, e.g. due to a sensor failure or as a result of a control unit flash; - Off-board errors, e.g. when entering data into the database due to an incorrect hexdump translation.

[0008] If an incorrect load spectrum forms the basis for calculating other parameters, the error propagates.

[0009] The object of the invention is therefore to be able to identify occurring load spectrum errors quickly and reliably in order to be able to correct them and thus avoid possible subsequent errors.

[0010] The invention is based on the features of the independent claims. Advantageous developments and refinements are the subject of the dependent claims.

[0011] A first aspect of the invention relates to a particularly computer-implemented method for checking a load spectrum determined in a vehicle, comprising the steps: - Providing a value of a load spectrum determined during real operation of the vehicle with respect to a component of a vehicle; - Providing one or more determined influencing factors that occur during and / or before the determination of the load spectrum; - Executing a machine learning model with the at least one influencing variable as an input variable of the machine learning model, which, through appropriate pre-training, indicates a relationship between the at least one influencing variable and a respective associated value of the load spectrum, wherein an estimate of the load spectrum is obtained as the output value after the execution of the machine learning model; and - Determining a deviation between the estimate of the load spectrum and the value of the load spectrum determined during actual operation of the vehicle and checking the deviation for exceeding a predetermined limit condition and / or checking for the absence of a value of the load spectrum from actual operation of the vehicle;

[0012] In particular, the method is carried out entirely on a computer. Particularly preferably, a computer program is provided and designed to carry out the method.

[0013] At this stage of the application, the machine learning model is already pre-trained and validated. The machine learning model is expediently trained using existing, older relationships between influencing variables and load spectra. The relationship between one or more influencing variables and a respective associated load spectra is automatically improved in the machine learning model through training as the amount of training data increases. This advantageously ensures that unbiased relationships between influencing variables and load spectra are used for training. This data validation can be automated, but may need to be performed manually.However, once the machine learning model has been trained, it can be used again in the future to provide an estimate of the associated load spectrum based on current influencing factors. This estimate can then be compared with the actual load spectrum determined during vehicle operation and checked for deviations. A variety of state-of-the-art architectures and structures for machine learning models can be applied, for example, an artificial neural network.

[0014] Advantageously, completely missing load spectra, also called load spectra, can be reconstructed to a good approximation during the determination of real-world vehicle operation, and / or faulty load spectra can be identified and corrected if necessary. Errors in the load spectra can be found and corrected more easily and quickly with the help of the machine learning model. The overall quality of the load spectra data can thus be improved, and a plausibility check is performed. Implausible load spectra can be identified automatically. Missing or faulty load spectra are corrected or replaced, particularly based on forecast values ​​from the machine learning model. Therefore, analyses and decisions in the product development process based on the determined load spectra are more reliable.

[0015] According to an advantageous embodiment, the provision of the determined load spectrum, the execution of the machine learning model, and the determination and verification of the deviation take place in a central computing unit separate from the respective vehicle.

[0016] According to a further advantageous embodiment, if a value of the load spectrum from the actual operation of the vehicle is missing, the associated value of the estimate of the load spectrum is used.

[0017] According to a further advantageous embodiment, the predetermined limit condition comprises a predetermined tolerance range which is determined as a function of the absolute value of the load spectrum determined during real operation of the vehicle and / or the associated value of the estimate.

[0018] According to a further advantageous embodiment, a value of the load spectrum determined during real operation of the vehicle, which leads to an exceedance of the predetermined limit condition of the deviation, is replaced by the associated value of the estimate in a database of a central processing unit.

[0019] In the central processing unit's database, the faulty value from the vehicle's actual operation can not only be identified but also replaced with the estimate if necessary. This allows for further development of the vehicle component with the relevant load spectrum, for example, at the manufacturer.

[0020] According to a further advantageous embodiment, a value of the load spectrum determined during real operation of the vehicle, which leads to an exceedance of the predetermined limit condition of the deviation, is replaced by the associated value of the estimate in a local memory of the vehicle.

[0021] This can be particularly useful if the respective value of the load spectrum under consideration is required for another function in the vehicle itself. The function thus receives a corrected value for the load spectrum, thus avoiding any subsequent errors, especially those that could propagate to other dependent load spectrums.

[0022] According to a further advantageous embodiment, the at least one influencing variable comprises at least one measurement of a first physical variable, while the respective value for describing the load collective comprises at least one measurement of a second physical variable different from the first.

[0023] According to a further advantageous embodiment, the at least one influencing variable comprises the respective value of another load collective with respect to another component of the vehicle.

[0024] A further aspect of the invention relates to a computing system for stationary arrangement outside a vehicle, which is designed to repeatedly carry out the method as described above and below for a plurality of vehicles.

[0025] Advantages and preferred developments of the proposed computing system result from an analogous and analogous transfer of the statements made above in connection with the proposed method.

[0026] Further advantages, features, and details will become apparent from the following description, which – where appropriate with reference to the drawings – describes at least one embodiment in detail. Identical, similar, and / or functionally equivalent parts are provided with the same reference numerals.

[0027] They show: Fig. 1: A method for checking a load spectrum determined in a vehicle according to an embodiment of the invention. Fig. 2: Application of the procedure according to Fig. 1 according to an embodiment of the invention.

[0028] Fig. 1 shows a method for checking a load spectrum determined in a vehicle 0, comprising the steps: - Providing S1 of a value 1 of a load spectrum determined during real operation of the vehicle 0 with respect to a component of a vehicle 0; - Providing S2 one or more determined influencing factors 3 which occur during and / or before the determination of the load spectrum; - Executing S3 a machine learning model with the at least one influencing variable 3 as an input variable of the machine learning model, which, through appropriate pre-training, indicates a relationship between the at least one influencing variable 3 and a respective associated value of the load spectrum, wherein an estimate 5 of the load spectrum is obtained as the output value after the execution of the machine learning model; and - Determining S4 a deviation between the estimate 5 of the load spectrum and the value 1 of the load spectrum determined during the actual operation of the vehicle 0 and checking the deviation for exceeding a predetermined limit condition and / or checking for the absence of a value 1 of the load spectrum from the actual operation of the vehicle 0;

[0029] Fig.Figure 2 shows an example application of the method. Various vehicles 0 with instances A to C transmit their diagnostic values ​​1 determined in real operation at times D1 to D4 over time t of load spectra X1, X2, X3, X Tto a database of a central processing unit 7, for example, of the vehicle manufacturer. There, the load spectra can be evaluated, for example, for the purpose of product improvement. Unexpected high loads can thus be taken into account in future products, and unnecessary design changes in components of the vehicle type for which lower than expected load components exist can be avoided. The data transmission from the vehicles 0 to the database of the central processing unit 7 can take place during a workshop diagnosis of the diagnostic interface and / or continuously wirelessly during operation of the respective vehicle. While workshop visits only occur at longer intervals, the wireless transmission from the respective vehicle 0 can be significantly more frequent. Each vehicle 0 (for example, A to C) transmits the current values ​​1 of all its load spectra X1, X2, X3, X for each diagnosis. T. In order to estimate errors of an exemplary specific load spectrum X T To be able to meet this challenge, a machine learning model ML was developed for this specific load collective X T trained. The load collective X T is the target variable for which the machine learning model (ML) is to provide an estimate 5 in the sense of a forecast using the influencing variables 3, also called predictors, X1 to X3. If the predictors 3 are appropriately selected and the training data set is sufficiently large, the forecast values ​​5 of the machine learning model (ML) will agree well with the values ​​1 of the load spectrum X. Tand both are within a specified tolerance range. In this case, the machine learning model (ML) has good predictive power. In the present example, for vehicle C, out of vehicles A, B, and C, an error occurs between diagnosis D3 and diagnosis D4 with the values ​​1 determined in real operation of vehicles 0 for the respective load spectrum. If a value 1 is missing for load spectrum X T If the value 0 is completely missing for this vehicle, it can be reconstructed using the ML machine learning model, and the estimate 5 of the ML machine learning model can be used instead of the missing value 1. Alternatively, if the value 1 is identified as erroneous because it exceeds a specified tolerance range compared to the estimate 5, at least an anomaly can be detected, and the value 1 determined for the load spectrum in real operation of the C vehicle 0 can be discarded or replaced by the estimate 5.

[0030] Although the invention has been illustrated and explained in detail by preferred embodiments, the invention is not limited by the disclosed examples, and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention. It is therefore clear that a multitude of variations exist. It is also clear that exemplary embodiments are truly only examples and should not be construed as limiting the scope, possible applications, or configuration of the invention in any way.Rather, the preceding description and the description of the figures enable the person skilled in the art to implement the exemplary embodiments in concrete terms, whereby the person skilled in the art, with knowledge of the disclosed inventive concept, can make various changes, for example with regard to the function or the arrangement of individual elements mentioned in an exemplary embodiment, without departing from the scope of protection defined by the claims and their legal equivalents, such as further explanations in the description. List of reference symbols 0 vehicle 1 value of a load spectrum determined in real operation 3 influencing factor(s) 5 Estimation of the load spectrum 7 central processing unit ML model of machine learning S1 Provision S2 Provision S3 Execute S4 Determine QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2012 005 679 A1

[0003] DE 10 2010 047 820 A1

[0004] DE 10 2007 018 176 A1

[0004] DE 10 2008 019 463 A1

[0005] DE 10 336 597 A1

[0005]

Claims

[1] Method for checking a load collective determined in a vehicle (0), comprising the steps: - Providing (S1) a value (1) of a load spectrum determined during real operation of the vehicle (0) with respect to a component of a vehicle (0); - Providing (S2) one or more determined influencing variables (3) which occur during and / or before the determination of the load spectrum; - Executing (S3) a machine learning model (ML) with the at least one influencing variable (3) as an input variable of the machine learning model (ML), which, through appropriate pre-training, indicates a relationship between the at least one influencing variable (3) and a respective associated value of the load spectrum, wherein an estimate (5) of the load spectrum is obtained as an output value after the execution of the machine learning model (ML); and - determining (S4) a deviation between the estimate (5) of the load spectrum and the value (1) of the load spectrum determined during actual operation of the vehicle (0) and checking the deviation for exceeding a predetermined limit condition and / or checking for the absence of a value (1) of the load spectrum from actual operation of the vehicle (0); [2] Method according to claim 1, wherein the provision of the determined load spectrum, the execution of the machine learning (ML) model, and the determination and checking of the deviation are carried out in a central computing unit (7) separate from the respective vehicle (0). [3] Method according to one of the preceding claims, wherein in the absence of a value (1) of the load spectrum from the actual operation of the vehicle (0), the associated value of the estimate (5) of the load spectrum is used. [4] Method according to one of the preceding claims, wherein the predetermined limit condition comprises a predetermined tolerance range which is determined as a function of the absolute value (1) of the load spectrum determined during real operation of the vehicle (0) and / or the associated value of the estimate (5). [5] Method according to one of claims 1 to 4, wherein a value (1) of the load spectrum determined during real operation of the vehicle (0), which leads to an exceeding of the predetermined limit condition of the deviation, is replaced by the associated value of the estimate (5) in a database of a central processing unit (7). [6] Method according to one of claims 1 to 4, wherein a value (1) of the load spectrum determined during real operation of the vehicle (0), which leads to an exceeding of the predetermined limit condition of the deviation, is replaced by the associated value of the estimate (5) in a local memory of the vehicle (0). [7] Method according to one of the preceding claims, wherein the at least one influencing variable (3) comprises at least one measurement of a first physical variable, while the respective value for describing the load collective comprises at least one measurement of a second physical variable different from the first. [8] Method according to one of the preceding claims, wherein the at least one influencing variable (3) comprises the respective value of another load collective with respect to another component of the vehicle (0). [9] Computing system for stationary arrangement outside a vehicle (0), which is designed to repeatedly carry out the method according to one of the preceding claims for a plurality of vehicles (0).

Images