DE102023125034A1 - Determining wear of a vehicle component using an AI module - Google Patents

Abstract

The invention relates to a vehicle (1) comprising at least one vehicle component (2) and an evaluation unit (3) with an artificial intelligence module (AI module) (4), wherein the evaluation unit (3) is set up to determine a value (11) of a wear index that is related to a degree of wear of the vehicle component (2) as a function of a time series (10) of operating data that is related to an operation of the vehicle component (2) at different points in time and with the aid of the AI module (4), wherein the AI module (4) is set up to determine a value (11) of a wear index that is related to a degree of wear of the vehicle component (2) as a function of further time series (20) of further operating data, wherein the further operating data of the respective further time series (201, 202,...20j,...20m) are related to a respective further previous operation of a respective further vehicle component that functions similarly to the vehicle component (2) at respective different further points in time, and as a function of training values (211, 212,...21j,...21m) of the wear index that is respectively assigned to the respective further Time series (201, 202, ... 20j, ... 20m) of the further operating data are assigned, is trained, and the evaluation unit (3) is set up to output the value (11) of the wear index.

Description

It is known that wear and tear on vehicle components depends on the type and manner of use of the respective vehicle component. This can particularly affect components of an internal combustion engine, such as a particle filter, and sensors for monitoring exhaust gas behavior of the internal combustion engine of a vehicle. Depending on how the internal combustion engine is used in everyday operation, wear and tear on the components of the internal combustion engine and the sensors can develop differently over a period of time during which the vehicle is used.

A vehicle is proposed, having at least one vehicle component and an evaluation unit with an artificial intelligence module (AI module). The evaluation unit is set up to determine a value of a wear index that is related to a degree of wear of the vehicle component as a function of a time series of operating data that is related to operation of the vehicle component at different times and with the aid of the AI module. The AI module is trained as a function of further time series of further operating data and as a function of training values of the wear index. The further operating data of the respective further time series are related to a respective further previous operation of a respective further vehicle component that functions similarly to the vehicle component at respective different further times. The training values of the wear index are each assigned to the respective further time series of the further operating data. Furthermore, the evaluation unit is set up to output the value of the wear index.

In the sense of the invention, the vehicle component can be any component of the vehicle. For example, the vehicle component can be a component of a drive train of the vehicle, such as an engine, a component of the engine, a transmission, a part of the transmission, a clutch, a battery, a component of a chassis of the vehicle, a wheel, a tire, a component of an interior of the vehicle, such as a seat, a window lifter or a control element for operating the vehicle, a component of a braking system of the vehicle or a component of the vehicle that enables autonomous driving of the vehicle.

The time series of operating data is a chronologically ordered summary of the operating data that is recorded at different points in time. The operating data of the time series can be understood as at least partial documentation of the operation of the vehicle component at different points in time. Depending on how the vehicle component is operated, different values of the operating data of the time series result.

Similarly, the additional time series of the additional operating data each represent a chronologically ordered summary of the respective additional operating data of the respective additional time series. The additional operating data of the respective additional time series can be understood as at least partial documentation of the operation of the respective additional vehicle component at the respective different additional points in time. The respective additional points in time comprise sets of additional points in time, whereby the respective set of additional points in time can each be assigned to one of the additional time series of the additional operating data. In particular, the additional operating data of the respective additional time series are each recorded at one of the additional points in time of the respective set of additional points in time.

The time series of the operating data includes at least one respective operating date for each of the different points in time. In principle, two or more respective operating dates of the time series can also be provided for the respective point in time of the points in time. The same applies to the other time series of the other operating data. This means that the other time series of the other operating data can include at least one respective additional operating date or two or more respective additional operating dates for the respective point in time of the respective set of the other points in time. Furthermore, the other time series of the other operating data at least partially describe the operation of the other vehicle component at the different additional points in time of the respective set of the other points in time.

The operating data are preferably recorded automatically at different times during operation of the vehicle component, hereinafter also referred to as component.

Analogously, the additional operating data are recorded automatically during the respective operation of the respective additional vehicle component, hereinafter also referred to as the respective additional component, at the additional points in time of the respective set of additional points in time.

The operating data can be obtained using a vehicle control unit or the evaluation control unit. Similarly, the other operating data can be recorded using a respective control unit or a respective evaluation unit of a respective additional vehicle in which the respective additional component is installed.

According to one embodiment, the time series of the operating data comprises a time series of user inputs for controlling the vehicle component. The user inputs are generated by a user of the vehicle at the different points in time. In this embodiment, the further time series of the further operating data comprise further time series of further user inputs for controlling the respective further vehicle component. The further user inputs are generated during the operation of the respective further vehicle component by a respective further user at the respective further points in time of the respective set of further points in time.

The user inputs can be described, for example, by a position of an operating means, such as an accelerator pedal, brake pedal, gear lever or control button. The respective user input is preferably related to the operation of the component at the respective point in time in such a way that the respective user input specifies, in particular controls, the operation of the component at the respective point in time. Similarly, the respective further user input is preferably related to the respective operation of the respective further component at the respective further point in time of the respective set of further points in time in such a way that the respective further user input specifies, in particular controls, the operation of the respective further component at the respective further point in time of the respective set of further points in time.

Alternatively or additionally, according to one variant, the vehicle can have a sensor for detecting a state of the vehicle component. In this variant, the time series of the operating data comprises a time series of sensor values generated using the sensor. The time series of the sensor values is generated depending on the operation of the vehicle component at the different points in time. In this variant, the further time series of the further operating data comprise further time series of further sensor values. The time series of the further sensor values are each generated depending on the respective further previous operation of the respective further vehicle component at the respective further points in time of the respective set of further points in time.

The additional sensor values of the respective additional time series of the additional sensor values are preferably generated using a respective additional sensor of the respective additional vehicle.

The sensor is preferably arranged on or in the component in order to detect a respective state of the component at the respective point in time. The respective sensor value is preferably related to the operation of the component at the respective point in time in such a way that the respective sensor value results as a result of the operation of the component at the respective point in time.

Analogously, the respective additional sensor is preferably arranged on or in the respective additional component in order to detect a respective state of the respective additional component at the respective additional point in time of the respective set of additional points in time. The respective additional sensor value of the respective additional time series of the additional sensor values is preferably related to the operation of the respective additional component at the respective additional point in time of the respective set of additional points in time in such a way that the respective additional sensor value of the respective additional time series of the additional sensor values results as a result of the operation of the respective additional component at the respective additional point in time of the respective set of additional points in time.

According to one variant, the value of the wear index can indicate the remaining time of the vehicle component's service life. The value of the wear index can also directly indicate the degree of wear of the component, for example in the form of a percentage between zero and one hundred percent.

It is also possible that the value of the wear index indicates a probability of failure of the component, preferably in percent. In any case, the wear index is related to the degree of wear of the component. If the degree of wear changes, the value of the wear index also changes.

According to one variant, the greater the wear of the component, the greater the value of the wear index. This is the case, for example, if the value of the wear index indicates the probability of failure or the degree of wear. Alternatively, the greater the wear of the component, the lower the value of the wear index. This is the case, for example, if the value of the wear index indicates the remaining time of the service life.

Because the evaluation unit is set up to determine the value of the wear index depending on the time series of the operating data, the evaluation unit can determine the value of the wear index depending on a past manner in which the component was operated in the past. This distinguishes the proposed vehicle from a vehicle in which the value of the wear index would, for example, only be determined depending on current measurement data. In other words, the proposed vehicle makes it possible to take a history of use of the component into account when determining the value of the wear index.

This makes it possible to carry out maintenance or repairs on the component in good time, i.e. before the component fails. This is primarily possible because the evaluation unit is set up to output the value of the wear index. The value of the wear index can be output visually so that the user of the component can see the value of the wear index. Alternatively or additionally, the value of the wear index can be output by sending the value of the wear index to an external computer. The external computer can be a workshop computer, a user's mobile device, or an external server. The external server is preferably set up to establish communication with the user's mobile device to exchange the value of the wear index.

It goes without saying that the time series of the operating data is in most cases related to an individual usage behavior of the vehicle, in particular the component, on the part of a user of the vehicle, in particular the component, and environmental influences. This follows simply from the fact that the time series of the operating data results as a consequence of the way in which the component was used by the user in the past. This applies both in the case where the operating data includes the user inputs and in the case where the operating data includes the sensor data.

Since the use of a vehicle or individual components of a vehicle generally varies greatly from person to person, the degree of wear and tear of the component after a specified period of use or after a specified number of uses of the component can also vary greatly depending on the person operating the component.

With the proposed vehicle, it is now possible to determine the value of the wear index depending on the time series of the operating data, which is directly related to the past usage behavior of the component by the user. The past usage behavior of the component by the user can also be understood as a driver profile.

The other vehicle component that functions in a similar way to the vehicle component can in particular be a component that is identical in construction to the component. The term "functioning in a similar way" means that the other component that functions in a similar way has a similar or, more preferably, the same physical function as the component. This makes it possible to compare information from the other operating data that can be used to assess wear on the other component with information from the operating data that can be used to assess the degree of wear on the component.

A kind of indirect comparison between the operating data and the other operating data in relation to the wear of the component or in relation to the wear of the respective other component can be carried out by using the AI module. This is because the AI module is trained depending on the other time series of the other operating data and the training values of the wear indicator.

The respective training value of the training values of the wear index can be and/or be determined for the respective additional component, for example by checking the respective additional component, for example in a workshop.

How the AI module can be used to calculate the value of the wear index depending on the time series of operating data and how the AI module can be trained is described below.

The term “module” as used herein describes any type of known or future developed hardware, software or combination of hardware and software, such as an executable program, artificial intelligence (AI), fuzzy logic, capable of performing any of the functions associated with the “module”.

The AI module can preferably have a machine learning module (ML module). The ML module is preferably designed and configured to perform a calculation of an output data set of the ML module depending on an input data set of the ML module. The input data set of the ML module can be understood as a request to calculate the output data set of the ML module depending on the input data set. The input data set and the output data set can each have values, preferably real numbers, as respective elements of the input data set or the output data set.

The input data set preferably contains the time series of the operating data or is preferably calculated depending on the time series of the operating data. The output data set can also contain only one value, for example the value of the wear index.

The ML module is preferably designed and configured to read the values of the input data set in the form of current or voltage values at several inputs of the ML module. Accordingly, the ML module is preferably designed and configured to output the values of the output data set in the form of current and voltage values at at least one output, preferably several outputs, of the ML module.

The ML module preferably has a neural network, a convolved neural network, a radial basis function network (RBF network) and/or similar mathematical functions that are suitable for approximating data, in particular functions in multidimensional spaces. The ML module is preferably designed and configured to calculate the values of the output data set depending on values of parameters of the ML module. In the event that the ML module has a neural network, the parameters can be weights that each indicate a strength of a connection between two neurons of neurons of the network.

According to one possible embodiment, the values of the output data set of the ML module can each represent a probability with which a respective class can be assigned to the input data set. For example, the values of the output data set of the ML module can each indicate a confidence value for the respective class. The respective confidence values of the output data set of the ML module can indicate a measure for the respective class of how likely it is to correctly assign the input data set to the respective class.

For example, the output data set can contain three real numbers, each of which indicates a probability between zero and one hundred percent that the input data set, in particular the time series of the operating data, will be correctly assigned to a first class, a second class or a third class. For example, the first class can represent low wear, the second class medium wear and the third class high wear of the component. The AI module can select the class that has the highest probability of being correctly assigned to the input data set. According to one possible variant, the value of the wear index can be a natural number between one and three and thus indicate the class that the AI module assigns to the input data set. In this variant, three stands for high wear, two for medium wear and one for low wear of the component.

According to the invention, the AI module, in particular the ML module of the AI module, is trained depending on the additional time series of the additional operating data and depending on the training values of the wear indicator. The following describes by way of example how the AI module, in particular the ML module, can be trained using the additional time series and the training values of the wear indicator.

In an untrained state of the ML module, the values of the parameters of the ML module are not yet adapted to the training values of the wear indicator. For example, the values of the parameters in the untrained state can be random numbers. The random numbers can be calculated, for example, by an initialization process of the ML module. Here, each value of the respective parameter of the ML module is set equal to a random number. Training of the ML module can preferably be carried out using training data sets in order to achieve a trained state of the ML module. The training data sets each have an input data set and an output data set corresponding to the respective input data set.

The input data set and the output data set of the respective training data set each have values, preferably real numbers. The input and output data sets of the training data sets preferably have the same dimension as the above-mentioned input data set or the above-mentioned output data set of the ML module.

The input data set of the respective training data set preferably has the further Operating data of the respective additional time series or is calculated depending on the additional operating data of the respective additional time series. In particular, one of the training data sets is generated for each of the additional time series of the additional operating data.

The output data set of the respective training data set can preferably have the respective training value of the wear index that is assigned to the respective further time series, depending on which the input data set of the respective training data set is calculated. In the event that the value of the wear index is determined by determining the above-mentioned probability of the respective class, the output data set of the respective training data set can have three numbers, each of which indicates a probability of correctly assigning the respective further component to the respective class of the above-mentioned three classes.

When training the ML module, the values of the ML module's parameters are adapted to the training data sets. Preferably, the values of the parameters are changed in such a way that an error in an approximation of the output data sets of the training data sets is reduced depending on the corresponding input data sets of the training data sets.

The error can preferably be formed in the form of a sum of squares of the errors. The error is advantageously calculated as a function of deviations. The respective deviation results in particular from a respective deviation between calculated values of a respective training output data set of the ML module and the respective output data set of the respective training data set.

The respective training output data set is calculated using the ML module depending on the values of the parameters of the ML module and depending on the input data set of the respective training data set, which also has the respective output data set. Accordingly, the respective training output data set corresponds to the input data set and to the output data set of the respective training data set. The training output data sets of the ML module have the same dimension as the output data sets of the training data sets. In particular, the respective elements of the respective training output data set correspond in type to the respective elements of the output data set of the respective training data set. A calculation of the respective training output data set using the ML module and depending on the input data set of the respective training data set can be understood as an approximation of the output data set of this respective training data set.

A change in the values of the parameters of the ML module depending on the deviations can reduce the error in the approximation of the initial data sets of the training data sets. If the error falls below a predetermined limit, the ML module can be considered trained, i.e. the ML module is in the trained state. The adaptation of the values of the parameters of the ML module can preferably be carried out using one or more learning algorithms, such as linear regression, backpropagation and/or a determination of K-means values, and is referred to below as "machine learning", hence the name ML module (machine learning module).

By adapting the values of the parameters to the training data sets, a functional relationship between the other time series and the training values of the wear indicator can be approximated. Information about this functional relationship is provided in the form of the trained AI module, for example in the form of the values of the weights of the network after training the network. The advantage of providing this functional relationship through the trained AI module is in particular that the functional relationship does not have to be described by concrete physical formulas. It is sufficient to record the training values of the wear indicators and the other time series of the other operating data and use them to train the AI module. This can also be done by a person who has little knowledge of how the component or the other components work.

Furthermore, the functional relationship can be approximated more accurately using the AI module than is possible using physical formulas. This may be because, although expert knowledge can be used to create basic physical formulas to calculate the degree of wear of the component depending on the time series of operating data, in most cases these physical formulas cannot fully reflect the relationship between the time series of operating data and the degree of wear of the component.

On the other hand, the accuracy of the AI module can be increased as desired, provided that the number of training data sets can be increased as desired. To increase the accuracy of the AI module, for example, the number of additional Vehicles, in each of which one of the additional components is installed, can be increased. This generally results in more workshop visits, during which the additional components can be examined with regard to the degree of wear of the respective additional component. The result of such examinations is logged in order to increase the number of additional time series of the additional operating data with their respective associated training values of the wear index. The additional time series of the additional operating data can advantageously be read out from the control unit of the respective additional vehicle. In principle, the training of the ML module described above can be repeated if new additional time series with new additional operating data with their respective associated new training values of the wear index are available. This corresponds to a continuous learning approach for neural networks. The new additional time series and the new training values are preferably obtained in the same way as the additional time series of the additional operating data with their respective associated training values of the wear index.

It goes without saying that the number of additional time series can be increased particularly easily if the additional components do not have to be constructed identically to the component, but merely function in a similar way to the component. However, if the additional components are constructed identically to the component, a higher level of accuracy of the AI module can be achieved when training the AI module. This applies if the number of additional time series is comparably the same as in a variant in which the additional components are not constructed identically but function in a similar way.

If the time series of operating data includes the time series of user inputs, this makes it possible to directly link the degree of wear of the component to the way in which the component is used by the user.

According to an advantageous variant, the evaluation unit can be set up to output the value of the wear index to the user at regular intervals. The time series of the operating data is preferably updated at regular intervals in order to calculate an updated value of the wear index depending on the updated time series of the operating data analogous to the value of the wear index using the AI module.

If the operating data includes user inputs, the user can see an influence of updated user inputs on the wear index value when the updated wear index value is output at regular intervals. This gives the user the opportunity to change a way of using the component in order to reduce wear of the component over time. This can be particularly useful if the component is a brake or a clutch of the vehicle.

If the additional operating data includes the additional sensor values, this can make it possible to more accurately approximate the functional relationship between the additional time series of the additional operating data and the training values of the wear indicator. This may be because the additional sensor values usually record the consequences of physical processes that take place in the additional components or are triggered by the additional components. On the other hand, it is in principle possible that physical effects of different user inputs of the additional user inputs compensate each other. This can make training the AI module more difficult. It follows that training the AI module can be carried out more quickly if the additional operating data includes the additional sensor values.

According to a further development, the vehicle is provided with a further sensor for detecting a state of an environment of the vehicle. In this development, the time series of the operating data comprises a time series of environmental data generated using the further sensor, which describes the state of the environment during operation of the vehicle component at different points in time. In this development, the further time series of the further operating data comprise further time series of further environmental data, each of which describes the state of the environment during the respective further previous operation of the respective further vehicle component at the respective different further points in time. This development makes it possible to take the time series of the environmental data into account when determining the value of the wear of the component using the AI module. This can increase the accuracy of the determined value of the wear of the component.

The environmental data can, for example, include weather data that describe a respective weather situation at the different points in time. In this case, the additional operating data includes additional weather data that each describe a weather at the respective different additional points in time of the respective set of additional points in time.

In an advantageous embodiment, the time series of the operating data comprises a time series of data for describing a respective state of a respective roadway at the different points in time. In this embodiment, the further operating data comprises further time series of further data that describe a respective state of a respective further roadway at the respective different further points in time of the respective set of further points in time. In this embodiment, the respective state of the respective roadway on which the vehicle is driving at the different points in time is taken into account for calculating the value of the wear of the component. This can further increase the accuracy of the determined value of the wear of the component.

The time series of data for describing the respective condition of the respective roadway can, for example, specify values of a waviness, a curvature, a gradient and/or a roughness of the respective roadway. Similarly, the further time series of the further data can describe a respective further waviness and/or roughness of the respective further roadway at the respective different further points in time of the respective set of further points in time.

According to a further embodiment, it is provided that the time series of the operating data comprises a time series of control values that describe a respective control of the vehicle component using a control unit of the vehicle at the different points in time. In this embodiment, the further time series of the further operating data comprise further control values, each of which describes a control of the respective further vehicle component using a respective control unit of the respective further vehicle in which the respective further vehicle component is installed, at the respective different points in time. This embodiment has the advantage that the control values and the further control values are easily available, since the component is controlled by the control unit using the control values independently of the invention. The control values and the further control values only have to be read out from the control unit of the vehicle or from the respective control unit of the respective further vehicle. The embodiment thus represents a variant of the proposed vehicle that can be implemented with little effort. It is assumed here that the control values have a direct influence on the wear of the component.

In a further embodiment, the AI module is trained depending on simulation data. The simulation data describe results of a simulation of operation of the vehicle component. To obtain the simulation data, several simulations of use of the component can advantageously be carried out on a computer, preferably a mainframe computer. The respective simulation can simulate a respective wear of the component depending on a respective time series of simulated operating data. The time series of simulated operating data can include time series of simulated user inputs and/or time series of simulated sensor values and/or time series of simulated control values and/or time series of simulated environmental data. Using the respective time series of the simulated operating data and a simulated value of the wear of the component depending on this, a respective further training data set can be generated for the training of the AI module described above. This allows the number of training data sets to be increased, whereby the accuracy of the calculation of the value of the wear can be increased using the trained AI module.

According to a further variant, the AI module is trained depending on test bench data that describe the results of tests with the vehicle component on a test bench. To obtain the test bench data, several tests can advantageously be carried out, in each of which the component is operated differently. In each test, the component is preferably operated depending on a time series of test operating data.

In each test, a respective value of the wear of the component can be measured depending on the respective time series of the test operating data. The time series of the test operating data of all tests and the associated measured values of the wear of the component can be considered together as the test bench data.

The time series of the test operating data can each include time series of user inputs made during the respective test and/or time series of sensor values measured during the respective test and/or time series of control values generated during the respective test. Using the respective time series of the test operating data and a value of the wear of the component measured as a function of this, a respective additional training data set can be generated for the training of the AI module described above. This allows the number of training data sets to be increased, which can increase the accuracy of the calculation of the value of the wear using the trained AI module.

Furthermore, a method for determining a value of a wear index of a vehicle component using an artificial intelligence module is proposed. The value of the wear index is related to a degree of wear of the vehicle component. The method has the following steps. In a first step, a time series of operating data is recorded that is related to operation of the vehicle component at different points in time. The time series of the operating data can be designed according to one of the variants described above. For example, the time series of the operating data can include the time series of the user inputs for controlling the vehicle component, the time series of the sensor values, the time series of the environmental data and/or the time series of the data for describing the respective state of the respective roadway at the different points in time.

In a second step, the value of the wear index is determined using the AI module and depending on the time series of the operating data. The AI module is, as described above, trained depending on the further time series of the further operating data, whereby the further operating data of the respective further time series are related to a respective further previous operation of a respective further vehicle component that functions similarly to the vehicle component at respective different further points in time, and depending on training values of the wear index, which are each assigned to the respective further time series of the further operating data. In a third step, the value of the wear index is output.

Furthermore, a further method for training an artificial intelligence module to determine a value of a wear index of a vehicle component is proposed. The value of the wear index is related to a degree of wear of the vehicle component. The further method has the following steps. In a first step of the further method, the further time series of the further operating data are recorded. The further operating data of the respective further time series are related to the respective further previous operation of the respective further vehicle component functioning similarly to the vehicle component at the respective different further points in time. In a second step of the further method, the training values of the wear index are recorded, which are each assigned to the respective further time series of the further operating data. In a third step of the further method, the AI module is trained using the further time series of the further operating data and the training values of the wear index.

• 1 eine Draufsicht eines Fahrzeugs mit einer Fahrzeugkomponente und einer Auswertungseinheit mit einem Künstliche-Intelligenz-Modul (KI-Modul) zum Berechnen eines Wertes eines Verschleißes der Fahrzeugkomponente;
• 2 eine Zeitreihe von Betriebsdaten und das in 1 gezeigte KI-Modul;
• 3 ein neuronales Netz des in 2 gezeigten KI-Moduls;
• 4 weitere Zeitreihen von Betriebsdaten zum Trainieren des in 2 gezeigten KI-Moduls;
• 5 Schritte eines Verfahren zum Trainieren in 2 gezeigten KI-Moduls;
• 6 Schritte eines weiteren Verfahrens zur Berechnung des Wertes des Verschleißes der in 1 gezeigten Fahrzeugkomponente.

Preferred embodiments are explained in more detail with reference to the following figures. Here, schematically

• 1 a plan view of a vehicle with a vehicle component and an evaluation unit with an artificial intelligence module (AI module) for calculating a value of wear of the vehicle component;
• 2 a time series of operating data and the 1 AI module shown;
• 3 a neural network of 2 shown AI module;
• 4 additional time series of operational data to train the 2 shown AI module;
• 5 Steps of a procedure for training in 2 shown AI module;
• 6 Steps of a further procedure for calculating the value of the wear of the 1 vehicle component shown.

1 shows a vehicle 1, having at least one vehicle component 2 and an evaluation unit 3 with an artificial intelligence module 4 (AI module 4). The evaluation unit 3 is set up to determine a value 11 of a wear number depending on a time series 10 of operating data and with the aid of the AI module 4. The value 11 of the wear number is related to a degree of wear of the vehicle component 2. The time series 10 of the operating data is related to operation of the vehicle component 2, hereinafter also referred to as component 2, at different points in time. In the following, it is assumed that a number of different points in time is equal to n. The number n of different points in time can, for example, be equal to 1000 or 100.

1 further shows a detection means 5 for recording the operating data of the time series 10 of the operating data. The detection means 5 can be designed, for example, in the form of a sensor for recording a state of the component 2. In this case, the detection means 5 can send a current sensor value, which is generated with the aid of the sensor as a function of a current state of the component 2, to the evaluation unit 3. The evaluation unit 3 is preferably set up to store a respective current sensor value determined with the aid of the detection means 5 at the respective time of the different times. and to generate a time series of the sensor values based on the stored sensor values. The time series 10 of the operating data can contain the time series of the sensor data. It is also possible for the detection means 5 to generate the time series of the sensor data. For example, the detection means 5 can determine and save the current sensor value at regular intervals in order to generate the time series of the sensor data. The detection means 5 can then send the time series of the sensor data to the evaluation unit 3.

According to a further embodiment, the detection means 5 can be designed in the form of an operating means for operating the vehicle component 2. The operating means can be, for example, a switch or a button with which the vehicle component 2 can be operated, in particular controlled, by a user of the vehicle 1, in particular the component 2. According to a variant, the operating means can send information about a current state of the operating means to the evaluation unit 3 at any time of the different times. The respective state of the operating means at the respective time of the different times can be regarded as a respective user input for controlling the vehicle component 2. The evaluation unit 3 is advantageously set up to store the respective user input that is implemented by the user at the respective time. The respective user input at the respective time forms part of a time series of user inputs for controlling the component 2. The time series 10 of the operating data can comprise the time series of the user inputs. For example, the respective user input can describe whether a switch for operating the component 2 is closed or open at the respective time. The switch can, for example, be a switch for operating a windshield wiper system.

2 shows a first operating date 10 ₁ , a second operating date 10 ₂ , an i-th operating date 10 _i and an n-th operating date 10 _n of the time series 10 of the operating data. The operating data 10 ₁ , 10 ₂ , ... 10 _i , ... 10 _n are recorded with the aid of the detection means 5 at a first point in time, a second point in time, an i-th point in time or an n-th point in time of the different points in time. It should be assumed that in the figures an ascending sequence of numbers represents a temporal order. Thus, in chronological terms, the first point in time is before the second point in time, the second point in time is before the i-th point in time and the i-th point in time is before the n-th point in time.

If the time series 10 of the operating data includes the time series of the user inputs, the first operating data 10 ₁ in the form of a first user input at the first point in time, the second operating data 10 ₂ in the form of a second user input at the second point in time, the i-th operating data 10 _i in the form of an i-th user input at the i-th point in time and the n-th operating data 10 _n in the form of an n-th user input at the n-th point in time can each be designed to control the component 2 at the corresponding point in time. If the time series 10 of the operating data has the time series of the sensor values, the first operating data 10 ₁ can be in the form of a first sensor value, the second operating data 10 ₂ in the form of a second sensor value, the i-th operating data 10 _i in the form of an i-th sensor value and the n-th operating data 10 _n in the form of an n-th sensor value, which each describe a state of the component 2 at the first time, second time, i-th time or n-th time, respectively.

According to a possible embodiment, the evaluation unit 3 can be set up to send the time series 10 of the operating data in the form of an input data set of the AI module 4 to the AI module 4. This can also be implemented within the evaluation unit 3. For example, the AI module 4 can be implemented on a processor of the evaluation unit 3 (not shown in the figures). According to a 2 In the variant shown, the AI module 4 can be set up to process the input data set in such a way that the operating data of the time series 10 of the operating data are each sent to a respective input of the AI module 4.

For example, the first operating data 10 ₁ can be sent to a first input 4 ₁ of the AI module, the second operating data 10 ₂ to a second input 4 ₂ of the AI module and the i-th operating data 10 _i to the i-th input 4 _i of the AI module 4 and the n-th operating data 10 _n to the n-th input 4 _n of the AI module 4. The AI module 4 is advantageously set up to process the values present at the inputs of the AI module 4 with the aid of mathematical functions in order to calculate the value 11 of the wear index.

3 shows a variant in which the AI module 4 is designed in the form of a neural network. The inputs 4 ₁ , 4 ₂ , ... 4 _i , ... 4 _n can be designed in the form of input neurons. 3 shows, by way of example, a first hidden layer 31 of the neural network with at least one first neuron 31 ₁ , a second neuron 31 ₂ and a third neuron 31 ₃ and a second hidden layer 32 of the neural network with at least a first neuron 32 ₁ and a second neuron 32 ₂ .

A number of hidden layers of the network 30 can be significantly higher in practice, and also a number of the respective neurons in the respective hidden layer is in most cases significantly higher than in 3 For example, a number of hidden layers can be between 10 and 100, and the respective number of neurons of the respective hidden layer can vary, for example, between 5 and 100. The 3 The network 30 shown is intended only to illustrate by way of example how the value 11 of the wear index can be calculated by forward propagation of calculated values within the network 30 depending on the time series 10 of the operating data.

For example, the processor can use first connection weights 41, which can each have different values, to calculate an input value for an activation function of the first neuron 31 ₁ , the second neuron 31 ₂ and the third neuron 31 ₃ of the first hidden layer 31 depending on the operating data of the time series 10 of the operating data. Depending on the respective input value of the activation function of the respective neuron 31 ₁ , 31 ₂ , 31 ₃ of the first hidden layer 31, the processor can calculate an output value of the respective neuron 31 ₁ , 31 ₂ , 31 ₃ of the first hidden layer 31.

With the aid of second connection weights 42, the processor can calculate an input value of a respective activation function of the first neuron 32 ₁ and the second neuron 32 ₂ of the second hidden layer 32 as a function of the respective output value of the first neuron 31 ₁ , the second neuron 31 ₂ and the third neuron 31 ₃ of the first hidden layer 31. Similarly, an input value of an activation function of an output neuron 33 of the network 30 can be calculated as a function of the output value of the first neuron 32 ₁ of the second hidden layer 32 and the output value of the second neuron 32 ₂ of the second hidden layer 32, taking third connection weights 43 into account, using the processor.

The output neuron 33 calculates the value 11 depending on the input value of the output neuron 33. The neurons of the network 30 generally calculate their respective output value depending on the respective input value of the activation function of the respective neuron. The respective activation function of the respective neuron of the network 30 can be, for example, a hyperbolic tangent or a logistic activation function.

The AI module 4 is trained depending on further time series 20 of further operating data. The further time series 20 of the further operating data include, as in 4 shown, a first further time series 20 ₁ of further operating data, a second further time series 20 ₂ of further operating data, ... a j-th further time series 20 _j of further operating data up to an m-th further time series 20 _m of further operating data.

4 shows a first operating date 20 ₁₁ , a second operating date 20 ₁₂ , an i-th operating date 20 _1i and an n-th operating date 20 _1n of the first further time series 20 ₁ of the further operating data. Furthermore, 4 a first operating date 20 ₂₁ , a second operating date 20 ₂₂ , an i-th operating date 20 _2i and an n-th operating date 20 _2n of the second additional time series 20 ₂ of the additional operating data. Accordingly, in 4 a first operating date 20 _j1 , a second operating date 20 _j2 , an i-th operating date 20 _ji and an n-th operating date 20 _jn of the j-th further time series 20 _j of the further operating data are shown. For the sake of clarity, 4 Furthermore, a first operating date 20 _m1 , a second operating date 20 _m2 , an i-th operating date 20 _mi and an n-th operating date 20 _mn of the m-th further time series 20 _m are shown.

The operating data of the respective additional time series 20 ₁ , 20 ₂ , ... 20 _j , ... 20 _m are advantageously recorded during a first, second, ... j-th, ... or m-th time interval. The time intervals can also have different lengths. However, for training the AI module 5, it has proven advantageous if at least a respective number of operating data that are recorded during the respective time interval is the same. For this reason, the respective number of operating data of the additional time series 20 in the 4 shown example is the same size, namely equal to the number n. Furthermore, it has proven advantageous if the number of operating data of the respective further time series is equal to the number of operating data of the above-mentioned time series 10 of the operating data.

The respective time interval therefore has n different points in time at which the respective operating date of the respective further time series 20 ₁ , 20 ₂ , 20 _j , 20 _m is recorded. For example, the first operating date 20 ₁₁ , the second operating date 20 ₁₂ , the i-th operating date 20 _1i and the n-th operating date 20 _1n of the first further time series 20 ₁ of the operating data are recorded at a first point in time, a second point in time, an i-th point in time and at an n-th point in time of the first time interval, respectively. Analogously, the first operating date 20 ₂₁ , the second operating date 20 ₂₂ , the i-th operating date 20 _2i and the n-th operating date 20 _2n of the second further time series 20 ₂ of the further operating data are recorded at a first point in time, a second point in time, an i-th point in time or at an n-th point in time of the second time interval.

Analogously, the first operating date 20 _j1 , the second operating date 20 _j2 , the i-th operating date 20 _ji and the n-th operating date 20 _jn of the j-th further time series 20 _j are recorded at a first point in time, a second point in time, an i-th point in time or at an n-th point in time of the j-th time interval. Analogously, the first operating date 20 _m1 , the second operating date 20 _m2 , the i-th operating date 20 _mi and the n-th operating date 20 _mn of the m-th further time series 20 _m are recorded at a first point in time, a second point in time, an i-th point in time or at an n-th point in time of the m-th time interval. Recording the respective operating date can, for example, mean measuring or determining the respective operating date using a respective measured value.

The time series 10 of the operating data can, for example, represent a period of use of the vehicle 1, in particular the component 2, by the user. The period of use can, for example, begin with a first day on which the user gets into the vehicle 1 for the first time and end with a day on which the component 2 is serviced or replaced in a workshop. The time period has the above-mentioned different points in time at which the operating data of the time series 10 are recorded.

The further operating data of the respective further time series 20 ₁ , 20 ₂ , ... 20 _j , ... 20 _m are related to a respective further previous operation of a respective further vehicle component functioning similarly to the vehicle component 2 at respective different further points in time. The respective different further points in time comprise the above-mentioned points in time of the respective time intervals.

The further operating data of the respective further time series 20 ₁ , 20 ₂ , 20 _j , 20 _m can, for example, be related to the respective further previous operation of the respective further component by recording the respective operating date of the respective further time series 20 ₁ , 20 ₂ , 20 _j , 20 _m at the respective point in time of the first time interval, the second time interval, the j-th time interval or the m-th time interval.

The operating data of the respective additional time series 20 ₁ , 20 ₂ , 20 _j , 20 _m can, as described above, comprise a respective additional time series of respective additional user inputs of a respective user of the respective additional vehicle component. The respective additional user inputs are generated during operation of the respective additional vehicle component by the respective additional user at the respective additional point in time of the respective first time interval, second time interval, j-th time interval or m-th time interval.

According to a further variant, the further time series 20 of the further operating data can comprise a respective further time series of respective further sensor values. In this variant, the respective further sensor values are each generated as a function of the respective further previous operation of the respective further vehicle component at the respective further point in time of the respective first time interval, second time interval, j-th time interval or m-th time interval using a respective sensor of the respective further vehicle.

After the respective additional vehicle component has been operated for a duration of the respective time interval and the additional operating data of the respective additional time series 20 ₁ , 20 ₂ , 20 _j , 20 _m of the additional operating data have been recorded, the respective wear of the respective additional vehicle component is determined in the form of a respective training value 21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m of the wear index. The training values 21 of the wear index can be determined, for example, by examining the respective additional vehicle component. During such an examination, for example, respective traces of wear can be evaluated in order to determine the respective training value 21 ₁ , 21 ₂ , 21 _j , 21 _m of the wear index.

In order to be able to train the AI module 5, the further operating data of the respective additional time series 20 ₁ , 20 ₂ , 20 _j and 20 _m are preferably assigned to the respective training value 21 ₁ , 21 ₂ , 21 _j or 21 _m of the wear index. The further operating data of the respective additional time series 20 ₁ , 20 ₂ , 20 _j and 20 _m are each combined with the respective training value 21 ₁ , 21 ₂ , 21 _j or 21 _m of the wear index to form a respective training data set 22 ₁ , 22 ₂ , ... 22 _j , ... 22 _m .

When training the AI module 5, the respective training data set 22 ₁ , 22 ₂ , 22 _j , 22 _m can be presented to the network 30. In this case, the respective further time series 20 ₁ , 20 ₂ , 20 _j , 20 _m of the further operating data each form an input data set of the respective training data set 22 ₁ , 22 ₂ , 22 _j or 22 _m . The respective training value 21 ₁ , 21 ₂ , 21 _j , 21 _m of the respective training data set 22 ₁ , 22 ₂ , 22 _j or 22 _m forms an initial data set of the respective training data set 22 ₁ , 22 ₂ , 22 _j or 22 _m . The training of the AI module 4 can be carried out by determining the sum of squares of the errors, as described above.

As described above, a respective training output data set is calculated depending on the input data set of the respective training data set 22 ₁ , 22 ₂ , 22 _j and 22 _m . The respective training output data set is simply a respective training output value. The above-mentioned respective deviations can each be calculated in the form of a respective difference between the respective training output value and the respective training value 21 ₁ , 21 ₂ , 21 _j and 21 _m of the wear index. The respective deviations can be squared to form the error sum of squares. In order to change the values of the weights 41, 42, 43, the backpropagation algorithm can be used, for example, to calculate respective changes in the values of the weights 41, 42, 43 depending on partial derivatives of the error sum of squares with respect to the respective values of the weights 41, 42, 43.

According to a further embodiment, the AI module 4 can be trained as a function of further second time series 70 of further operating data. The further second time series 70 of the further operating data comprise, as in 4 shown, a first further second time series 70 ₁ of further operating data, a second further second time series 70 ₂ of further operating data, ... a j-th further second time series 70 _j of further operating data up to an m-th further second time series 70 _m of further operating data. In the 7 In the embodiment shown, the operating data of the further second time series 70 are each in the form of vectors, each of which represents at least two further operating data. Entries of the vectors are shown in the form of a rectangle.

7 shows a first operating vector 70 ₁₁ , a second operating vector 70 ₁₂ , an i-th operating vector 70 _1i and an n-th operating vector 70 _1n of the first further second time series 70 ₁ of the further operating data. Furthermore, 4 a first operating vector 70 ₂₁ , a second operating vector 70 ₂₂ , an i-th operating vector 70 _2i and an n-th operating vector 70 _2n of the second further second time series 70 ₂ of the further operating data. Accordingly, in 4 a first operating vector 70 _j1 , a second operating vector 70 _j2 , an i-th operating vector 70 _ji and an n-th operating vector 70 _jn of the j-th further second time series 70 _j of the further operating data are shown. For the sake of clarity, 7 Furthermore, a first operating vector 70 _m1 , a second operating vector 70 _m2 , an i-th operating vector 70 _mi and an n-th operating vector 70 _mn of the m-th further second time series 70 _m are shown.

The operating vectors of the respective further second time series 70 ₁ , 70 ₂ , ... 70 _j , ... 70 _m are advantageously recorded during the first, second, ... j-th, ... or m-th time interval. In the 7 In the embodiment shown, a second time series of operating vectors is acquired instead of the above-mentioned time series 10 of the operating data. The operating vectors each preferably comprise at least one respective sensor value and one respective user input, which are each generated at the respective point in time.

The respective time interval therefore has the n different points in time at which the respective operating vector of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m is recorded. For example, the first operating vector 70 ₁₁ , the second operating vector 70 ₁₂ , the i-th operating vector 70 _1i and the n-th operating vector 70 _1n of the first further second time series 70 ₁ of the operating data are recorded at the first point in time, the second point in time, the i-th point in time and the n-th point in time of the first time interval, respectively. Analogously, the first operating vector 70 ₂₁ , the second operating vector 70 ₂₂ , the i-th operating vector 70 _2i and the n-th operating vector 70 _2n of the second further second time series 70 ₂ of the further operating data are recorded at the first time, the second time, the i-th time and the n-th time of the second time interval, respectively.

Analogously, the first operating vector 70 _j1 , the second operating vector 70 _j2 , the i-th operating vector 70 _ji and the n-th operating vector 70 _jn of the j-th further second time series 70 _j are recorded at the first point in time, the second point in time, the i-th point in time or at the n-th point in time of the j-th time interval. Analogously, the first operating vector 70 _m1 , the second operating vector 70 _m2 , the i-th operating vector 70 _mi and the n-th operating vector 70 _mn of the m-th further second time series 70 _m are recorded at the first point in time, the second point in time, the i-th point in time or at the n-th point in time of the m-th time interval. Recording the respective operating vector can mean, for example, measuring or determining the respective operating vector with the aid of respective measured values.

For the sake of simplicity, the second time series of the operating vectors also represents the period of use of vehicle 1, in especially component 2, from the user side.

The further operating vectors of the respective further second time series 70 ₁ , 70 ₂ , ... 70 _j , ... 70 _m are related to the respective further preceding operation of the respective further vehicle component functioning similarly to the vehicle component 2 at the respective different further points in time. The respective different further points in time comprise the above-mentioned points in time of the respective time intervals.

The further operating data of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m can, for example, be related to the respective further previous operation of the respective further component by recording the respective operating vector of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m at the respective point in time of the first time interval, the second time interval, the j-th time interval or the m-th time interval.

The operating vectors of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m can comprise the respective further time series of the respective further user inputs of the respective user of the respective further vehicle component. The respective further user input of the respective operating vector of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m is in 7 shown in the form of a rectangle. The respective additional user inputs are generated during operation of the respective additional vehicle component by the respective additional user at the respective additional point in time of the respective first time interval, second time interval, j-th time interval or m-th time interval.

In addition, the further time series 70 of the further operating data preferably comprise the respective time series of the respective further sensor values. In this variant, the respective further sensor values are generated in each case as a function of the respective further previous operation of the respective further vehicle component at the respective further point in time of the respective first time interval, second time interval, j-th time interval or m-th time interval using a respective sensor of the respective further vehicle. The respective further sensor value of the respective operating vector of the respective further second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m is in 7 shown in the form of a rectangle.

After the respective additional vehicle component has been operated for a duration of the respective time interval and the further operating vectors of the respective additional second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m of the additional operating data have been recorded, the respective wear of the respective additional vehicle component is determined in the form of the respective training value 21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m of the wear index as described above.

For training the AI module 5 using the further second time series, the further operating data of the respective further second time series 70 ₁ , 70 ₂ , 70 _j and 70 _m are preferably assigned to the respective training value 21 ₁ , 21 ₂ , 21 _j or 21 _m of the wear index. The further operating data of the respective further second time series 70 ₁ , 70 ₂ , 70 _j and 70 _m are each combined with the respective training value 21 ₁ , 21 ₂ , 21 _j or 21 _{m of} the wear index to form a respective further training data set 72 ₁ , 72 ₂ , ... 72 _j , ... 72 _m .

When training the AI module 5, the respective additional training data set 72 ₁ , 72 ₂ , 72 _j , 72 _m can be presented to the network 30. In this case, the respective additional second time series 70 ₁ , 70 ₂ , 70 _j , 70 _m of the additional operating data each forms an input data set of the respective additional training data set 72 ₁ , 72 ₂ , 72 _j or 72 _m . The respective training value 21 ₁ , 21 ₂ , 21 _j , 21 _m of the respective additional training data set 72 ₁ , 72 ₂ , 72 _j or 72 _m each forms an output data set of the respective additional training data set 72 ₁ , 72 ₂ , 72 _j or 72 _m . The training of the AI module 4 can be carried out, as described above, by determining the sum of squares of the errors.

As described above, a respective further training output data set is calculated depending on the input data set of the respective further training data set 72 ₁ , 72 ₂ , 72 _j and 72 _m . The respective further training output data set is simply a respective further training output value. The respective deviations mentioned above can each be calculated in the form of a respective difference between the respective further training output value and the respective training value 21 ₁ , 21 ₂ , 21 _j and 21 _m of the wear index. The respective deviations can be squared to form the error sum of squares. In order to change the values of the weights 41, 42, 43, the backpropagation algorithm can be used, for example, to calculate respective changes in the values of the weights 41, 42, 43 depending on partial derivatives of the error sum of squares with respect to the respective values of the weights 41, 42, 43.

5 shows steps of a method for training the AI module 4. In a first step 501, the further time series 20 of the further Operating data is recorded. In a second step 502, the training values 21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m of the wear index are determined. In a third step 503, the AI module 4 is trained using the further time series 20 of the further operating data and the training values 21 _1, 21 ₂ , ... 21 _j , ... 21 _m of the wear index. The above-mentioned training data sets 22 ₁ , 22 ₂ , ... 22 _j , ... 22 _m are formed from the further time series 20 and the training values 21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m of the wear index. With the help of the training data sets 22 ₁ , 22 ₂ , ... 22 _j , ... 22 _m , a learning process for adapting values of parameters of the AI module 4 can be carried out. The parameters can be, for example, the connection weights 41, 42 and 43 if the AI module 4 has the neural network 30. The learning method can be, for example, the backpropagation method mentioned above.

The training of the AI module 4 is preferably carried out on a server not shown in the figures that is external to the vehicle, in particular a mainframe computer.

6 shows a further method for determining the value 11 of the wear index of the vehicle component 2. In a first step 601 of the further method, the time series 10 of the operating data is recorded. In this case, the respective operating data 10 ₁ , 10 ₂ , ... 10i, ... 10 _n is recorded at regular intervals at different times during the period of use of the component 2 by the user. In a second step 602 of the further method, the value 11 of the wear index is calculated with the help of the trained AI module 4 and depending on the time series 10 of the operating data. The trained values of the connection weights 41, 42, 43 are used to calculate the value of the wear index. In a third step 603 of the further method, the value 11 of the wear index is output, for example on a display of a dashboard of the vehicle 1. Alternatively or additionally, the value 11 of the wear index can be sent from the AI module 4 to a central control unit of the vehicle 1. The value 11 of the wear index can be read out by the central control unit, in particular during maintenance of the vehicle 1. Outputting the value of the wear index therefore also includes forwarding the value of the wear index from the AI module 4, in particular the evaluation unit 3, to the central control unit.

Claims (10)

Vehicle (1) comprising at least one vehicle component (2) and an evaluation unit (3) with an artificial intelligence module (AI module) (4), wherein the evaluation unit (3) is set up to determine a value (11) of a wear index that is related to a degree of wear of the vehicle component (2) as a function of a time series (10) of operating data that is related to an operation of the vehicle component (2) at different points in time and with the aid of the AI module (4), wherein the AI module (4) is set up to determine a value (11) of a wear index that is related to a degree of wear of the vehicle component (2) as a function of further time series (20) of further operating data, wherein the further operating data of the respective further time series (20 ₁ , 20 ₂ , ... 20 _j , ... 20 _m ) are related to a respective further previous operation of a respective further vehicle component that functions similarly to the vehicle component (2) at respective different further points in time, and as a function of training values (21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m ) of the wear index, which are each assigned to the respective further time series (20 ₁ , 20 ₂ , ... 20 _j , ... 20 _m ) of the further operating data, and the evaluation unit (3) is set up to output the value (11) of the wear index. Vehicle (1) to Claim 1 , wherein the time series (10) of the operating data comprises a time series of user inputs for controlling the vehicle component (2), which are generated by a user of the vehicle (1) at the different points in time, and the further time series (20) of the further operating data comprise further time series of further user inputs for controlling the respective further vehicle component (2), wherein the further user inputs are generated during the operation of the respective further vehicle component (2) by a respective further user at the respective further points in time. Vehicle (1) to Claim 1 or 2 , wherein the vehicle (1) has a sensor for detecting a state of the vehicle component (2), and the time series (10) of the operating data comprises a time series of sensor values generated with the aid of the sensor, which are generated as a function of the operation of the vehicle component (2) at the different points in time, wherein the further time series (20) of the further operating data comprise further time series of further sensor values, which are each generated as a function of the respective further previous operation of the respective further vehicle component (2) at the respective further points in time. Vehicle (1) according to one of the preceding claims, wherein the vehicle (1) has a further sensor for detecting a state of an environment of the vehicle (1), and the time series (10) of the operating data is a Sensors generated time series of environmental data which describe the state of the environment during operation of the vehicle component (2) at the different points in time, wherein the further time series (20) of the further operating data comprise further time series of further environmental data which each describe the state of the environment during the respective further preceding operation of the respective further vehicle component (2) at the respective further points in time. Vehicle (1) according to one of the preceding claims, wherein the time series (10) of the operating data comprises a time series of data for describing a respective state of a respective roadway at the different points in time, and the further operating data comprise further time series (20) of further data which describe a respective state of a respective further roadway at the respective further points in time. Vehicle (1) according to one of the preceding claims, wherein the time series (10) of the operating data comprises a time series of control values which describe a respective control of the vehicle component (2) by means of a control unit of the vehicle (1) at the different points in time, and the further time series (20) of the further operating data comprise further control values which each describe a control of the respective further vehicle component (2) by means of a respective control unit of a respective further vehicle (1) in which the respective further vehicle component (2) is installed, at the respective further points in time. Vehicle (1) according to one of the preceding claims, wherein the AI module (4) is trained as a function of simulation data describing results of a simulation of an operation of the vehicle component (2). Vehicle (1) according to one of the preceding claims, wherein the AI module (4) is trained as a function of test bench data describing the results of tests with the vehicle component (2) on a test bench. Method for determining a value (11) of a wear index of a vehicle component (2), wherein the value (11) of the wear index is related to a degree of wear of the vehicle component (2), with the aid of an artificial intelligence module (AI module) (4), the method comprising the following steps: - recording a time series (10) of operating data which is related to operation of the vehicle component (2) at different points in time; - Determining the value (11) of the wear index using the AI module (4) and depending on the time series (10) of the operating data, wherein the AI module (4) is trained depending on further time series (20) of further operating data, wherein the further operating data of the respective further time series are related to a respective further previous operation of a respective further vehicle component functioning similarly to the vehicle component (2) at respective different further points in time, and depending on training values (21 ₁ , 21 ₂ , ...21 _j , ...21 _m ) of the wear index, which are each assigned to the respective further time series of the further operating data; - Outputting the value (11) of the wear index. Method for training an artificial intelligence module (AI module) (4) to determine a value (11) of a wear index of a vehicle component (2), wherein the value (11) of the wear index is related to a degree of wear of the vehicle component (2), the method comprising the following steps: - recording further time series (20) of further operating data, wherein the further operating data of the respective further time series are related to a respective further previous operation of a respective further vehicle component functioning similarly to the vehicle component (2) at respective different further points in time; - recording training values (21 ₁ , 21 ₂ , ... 21 _j , ... 21 ) of the wear index, which are each assigned to the respective further time series of the further operating data; - training the AI module (4) using the further time series (20) of the further operating data and the training values (21 ₁ , 21 ₂ , ... 21 _j , ... 21 _m ) of the wear index.

Images