DE102022205676A1 - Determining a clutch temperature of a vehicle clutch using a neural network using a geometric property of the vehicle clutch - Google Patents

Abstract

Disclosed is a method for training a neural network (12), which is designed to determine a clutch temperature (KT) of a vehicle clutch, the method having the following steps: providing (TS1) at least one geometric input variable (26), which is representative for a geometric property of the vehicle clutch, as input data (20b); Providing (TS2) at least one value (18a, 18b, 18c) of an operating parameter of the vehicle clutch as input data (20a); Providing (TS3) values for the clutch temperature (KT) as output data; and training (TS4) the neural network (12) with the input data (20a, 20b) and the output data in order to learn a relationship between the geometric property and the at least one operating parameter and the clutch temperature (KT). Also disclosed is a method for determining a clutch temperature (KT) of a vehicle clutch using a neural network (12) and a control device (10) for determining a clutch temperature (KT) of a vehicle clutch with a neural network (12), each using a geometric input variable .

Description

Technical area

The present invention relates to a method for training a neural network which is designed to determine a clutch temperature of a vehicle clutch using a geometric property of the vehicle clutch. The invention also relates to a method for determining a clutch temperature of a vehicle clutch by means of a neural network using a geometric property of the vehicle clutch. The invention further relates to an associated control device for determining a clutch temperature of a vehicle clutch with a neural network using a geometric property of the vehicle clutch.

State of the art

The temperature of a vehicle clutch can be calculated using a classic rule-based mathematical model. The calculation can take place in a transmission control unit. Machine learning can be used for this, for example a neural network. However, determining the parameters that are relevant for the clutch temperature, especially during a switching process, is very complex or very imprecise, since these can only be determined indirectly.

Presentation of the invention

In a first aspect, the invention relates to a method for training a neural network which is designed to determine a clutch temperature of a vehicle clutch.

The vehicle clutch can be installed in a motor-driven vehicle, such as a motor vehicle, a motorcycle or an at least partially electrically driven two-wheeler. By means of the vehicle clutch, a driving force applied by the engine of the vehicle can be transmitted to a drive axle of the vehicle. The vehicle clutch can have at least two different switching states or gears, which can be defined by a predetermined ratio between an output torque of the engine and a drive torque of the drive axle.

A clutch temperature can be understood to mean the temperature of at least one clutch element, for example the temperature of a clutch disc. Alternatively, a clutch temperature can also be understood as a temperature of the entire vehicle clutch, which can be based, for example, on the determination of an average value of the individual temperatures of the clutch elements. Operating the vehicle may cause the temperature of the vehicle clutch to change. For example, the temperature of the vehicle clutch can increase during a gear shift or a gear change. If the temperature of the vehicle clutch rises above a critical temperature value, this can lead to damage or failure of the vehicle clutch. Determining a clutch temperature of the vehicle clutch therefore contributes to the safety of the vehicle.

A neural network can be understood as a mathematical model that at least partially replicates the structure of the neurons in the human brain. The neural network can be created using a computer. The neural network may include input nodes, output nodes, and a plurality of intermediate nodes disposed between the input nodes and the output nodes. The input nodes can be, for example, data interfaces via which input data can be entered into the neural network. The output nodes can be, for example, data interfaces via which output data from the neural network can be output. The input nodes may be connected to the intermediate nodes and the intermediate nodes may be connected to each other. The intermediate nodes can be connected to the output nodes. The input data can be historical data collected at a specific point in time. Alternatively or additionally, the input data can be synthetic data that was generated by processing recorded or measured data. Analogously, the output data can be historical data or synthetic data.

Information can be stored at least temporarily on the intermediate nodes. It can be provided that at least one arithmetic operation is carried out on the intermediate nodes. The input data can be transferred from the input nodes to the output nodes via the intermediate nodes. During this transfer, the input data can be processed mathematically, for example converted into the output data. The intermediate nodes of the neural network can be arranged in one or more layers or levels. The intermediate nodes can be connected to one another within a layer. In addition, the intermediate nodes of one layer can be connected to the intermediate nodes of other layers be the one. The individual connections of the input nodes, the intermediate nodes and the output nodes can be provided with mathematical weights. Depending on the purpose of the neural network, the individual weights of the connections can be different. The weights can be changed during training of the neural network. By adjusting the mathematical weights of the connections of the individual nodes during training, the neural network can learn a connection between the input data and the output data. While using the neural network for its intended purpose, the learned context from the neural network may be applied to input data to produce output data according to the intended purpose of using the neural network.

To determine a clutch temperature, for example, a multi-layer perceptron, MLP, can be used as a neural network. This neural network has at least one layer of intermediate nodes and uses at least one non-linear mathematical function to calculate the output data. Another exemplary neural network for determining a clutch temperature can be a fully connected layer, FCL, network. In this neural network, all input nodes, intermediate nodes and output nodes are connected to each other. Furthermore, to determine a coupling temperature, a convolutional neural network, CNN, can be used, for example, in which the intermediate nodes of different layers are at least partially connected to one another in the form of a mathematical convolution function.

The method has the step of providing at least one geometric input variable as input data. The geometric input variable is representative of a geometric property of the vehicle clutch. The at least one geometric input variable can be a historically recorded value, a measured value or a synthetic value which was determined from at least one recorded or measured value.

Providing the at least one geometric input variable can be understood as determining, that is, detecting, measuring or synthetically generating at least one value of the geometric input variable. Alternatively or additionally, providing can be understood to mean selecting whether a value of the geometric input variable is suitable as input data for training the neural network. For example, a value of a geometric input variable may not be suitable as input data for training the neural network if it has no or insufficient influence on the clutch temperature. Alternatively or additionally, a value of a geometric input variable may not be suitable as input data for training the neural network if it cannot be processed by the neural network, for example due to its data structure.

Alternatively or additionally, providing can be understood as pre-processing a geometric input variable so that it can be processed faster or at all by the neural network. For example, a geometric input variable can be in the form of a multidimensional matrix. However, it can be provided that the neural network to be trained is designed to process one-dimensional numerical values. In this case, the multidimensional matrix must be broken down into one-dimensional numerical values before it can be used to train the neural network. Furthermore, providing can also be understood as sorting the input data or removing a subset of the input data in order to reduce the amount of data used for training the neural network.

The at least one geometric input variable can be transmitted to at least one of the input nodes of the neural network, for example by means of an input device. Alternatively, the at least one geometric input variable can be transmitted from a control device of the vehicle via a data interface to at least one of the input nodes of the neural network. The at least one input variable can be used as input data for using the neural network to determine the clutch temperature.

A geometric property of the vehicle clutch can be understood to mean, for example, a geometric dimension of the vehicle clutch, a surface quality of the vehicle clutch, a material property of the vehicle clutch or a parameter that arises inherently from the aforementioned properties of the vehicle clutch, for example a thermal conductivity coefficient of the vehicle clutch. The geometric input variable can be specified in the physical unit of the respective geometric property of the vehicle coupling. Alternatively or additionally, the geometric property of the vehicle coupling can be based can be determined on the geometric input size.

The method further comprises the step of providing at least one value of an operating parameter of the vehicle clutch as input data for training the neural network. The at least one value of the at least one operating parameter of the vehicle clutch can be representative of a power supplied to the vehicle clutch. Alternatively or additionally, the at least one value of the at least one operating parameter of the vehicle clutch can be representative of a clutch temperature of the vehicle clutch. The at least one value of the operating parameter can be a historically recorded value, a measured value or a synthetic value which was determined from at least one recorded or measured value. The at least one value of the at least one operating parameter can be provided in a manner analogous to the geometric input variable.

The method further comprises the step of providing values for the clutch temperature as output data of the neural network. The values for the clutch temperature can be historically recorded values, measured values or synthetic values, which were determined from at least one recorded or measured value. The values for the clutch temperature can be provided in a manner analogous to the geometric input variable or the at least one value of the operating parameter. The clutch temperature values can be transmitted to the output nodes of the neural network to be used as output data.

The method further comprises the step of training the neural network with the input data and the output data in order to learn a relationship between the geometric property and the at least one operating parameter and the clutch temperature. The input data can be processed by the neural network, for example on the intermediate nodes, to obtain the output data. When processing the input data, the neural network can change the mathematical weighting of the connections of the individual nodes to obtain the output data. In this way, the neural network can learn the connection between the geometric property and the at least one operating parameter and the clutch temperature.

The proposed method enables a clutch temperature to be determined using a neural network. This makes it possible to avoid the use of classic mathematical models, which are generally very complex and therefore time-consuming to calculate. By using a neural network, the calculation time required to determine the clutch temperature can be reduced. Furthermore, determining the clutch temperature is based on a geometric property of the vehicle clutch. A dependence of the clutch temperature on the geometric properties of the vehicle clutch can therefore also be taken into account by the neural network.

For example, it may be possible for the vehicle clutch to heat up more slowly the larger the geometric expansion of the vehicle clutch. Alternatively, it may be possible that the vehicle clutch heats up faster, the better the heat conduction properties of the materials used for the vehicle clutch are. The clutch temperature can therefore also be determined more precisely by the neural network. Furthermore, due to the consideration of a geometric property of the vehicle clutch, the trained neural network can also be used to determine a clutch temperature of other vehicle clutches that have a similar or comparable geometric property. The neural network therefore does not have to be retrained for each vehicle clutch in order to be able to determine its clutch temperature. Taking into account a geometric property of the vehicle clutch thus enables the trained neural network to be used to determine a clutch temperature for a large number of different vehicle clutches, provided they have a common geometric property.

According to one embodiment, the geometric input variable is determined as a function of at least one measured parameter of the vehicle clutch. The at least one measured parameter can be one of the parameters described above that indicate a geometric property of the vehicle clutch. Alternatively or additionally, the at least one measured parameter can be an operating parameter of the vehicle clutch. The measured parameter of the vehicle clutch can be representative of the geometric input variable. The measured parameter can be calculated mathematically, for example according to a physical formula, to determine the geometric input variable. The at least one parameter of the vehicle clutch can be measured by a device designed for processing electrical signals become. The at least one parameter can be measured taking into account a time course of the same. Alternatively or additionally, the at least one parameter can be measured taking into account its temperature dependence. By measuring the at least one parameter, the geometric input variable can be determined in a particularly simple manner depending on the operating state of the vehicle clutch.

According to a further embodiment, the geometric input variable is determined as a function of at least one predetermined parameter of the vehicle clutch. The at least one predetermined parameter can be one of the parameters described above that indicate a geometric property of the vehicle clutch. Alternatively or additionally, the at least one predetermined parameter can be an operating parameter of the vehicle clutch. The specified parameter of the vehicle clutch can be representative of the geometric input variable. The specified parameter can be calculated mathematically, for example according to a physical formula, to determine the geometric input variable. The at least one predetermined parameter can be predetermined based on structural properties of the vehicle clutch. For example, a maximum installation space available for the vehicle clutch can be specified. Alternatively or additionally, the at least one predetermined parameter can be predetermined based on a type of vehicle clutch. For example, the use of the vehicle clutch can be specified in a motor vehicle or a motorcycle. The at least one predetermined parameter can be predetermined taking into account a time course of the same. Alternatively or additionally, the at least one predetermined parameter can be predetermined taking into account its temperature dependence. By specifying the at least one parameter, the geometric input variable can be determined in a particularly simple manner depending on a design specification or a predetermined type of vehicle in which the vehicle clutch is used.

Alternatively or in addition to the above embodiments, the geometric input variable can be determined depending on at least one measured parameter and depending on at least one predetermined parameter of the vehicle clutch. The individual advantages of the respective embodiments can thus be combined.

According to a further embodiment, the geometric input variable is provided continuously during training of the neural network. For example, a constant value of the geometric input variable for training the neural network can be provided throughout the entire duration of the training process. Alternatively, a time-dependent and/or temperature-dependent value of the geometric input variable for training the neural network can be provided throughout the entire duration of the training process. By providing the geometric input variable over time, the neural network is supplied with the geometric input variable as input data throughout the entire training period. The geometric input variable can therefore be taken into account throughout the entire training process of the neural network to determine a clutch temperature.

According to a further embodiment, the neural network is further trained using at least one specific input value, which is representative of a power supplied to the vehicle clutch. The at least one input value is determined based on processing successive values for the power supplied to the vehicle clutch. The power supplied to the vehicle clutch can be understood as a physical power, i.e. an amount of energy which is supplied to the vehicle clutch during a predetermined period of time. For example, the power supplied to the vehicle clutch can be a switching power supplied to the vehicle clutch during a switching process or a gear change. The power supplied to the vehicle clutch may be representative of a change in the clutch temperature of the vehicle clutch.

The power supplied to the vehicle clutch can be recorded at certain time intervals. The individual values for the power supplied to the vehicle clutch can, for example, be arranged in ascending order according to their detection time. These temporally successive values for the power supplied to the vehicle clutch can be processed mathematically, for example by means of a predetermined arithmetic operation such as forming an average value.

The at least one specific input value can be provided analogously to the geometric input variable, the at least one value of the operating parameter of the vehicle clutch or the values for the clutch temperature. Using the at least one specific input value, the neural network can also take into account a time course of the Power supplied to the vehicle clutch can be trained. The power supplied to the vehicle clutch may be representative of a clutch temperature of the vehicle clutch.

In a second aspect, the invention relates to a method for determining a clutch temperature of a vehicle clutch using a neural network. The method has the following steps: inputting at least one geometric input variable, which is representative of a geometric property of the vehicle clutch, as input data into the neural network; Entering at least one value of at least one operating parameter of the vehicle clutch as input data into the neural network; and determining a clutch temperature by the neural network based on the input data and a relationship of the input data to the clutch temperature learned by the neural network. What has been explained regarding the input data of the training method according to the first aspect applies analogously to the input data of the determination method according to the second aspect.

The clutch temperature determined by the neural network can be used by other components of the vehicle control system. For example, the specific clutch temperature can be used to generate a control value for a control unit of the vehicle, for example a transmission control for an automatic transmission. The control value can also be displayed on a display unit of the vehicle. Alternatively or additionally, the control value based on the clutch temperature can be processed by an evaluation unit, which is used, for example, to monitor the driving safety of the vehicle.

According to one embodiment, the determination method may further comprise the following steps: determining at least one input value which is representative of a power supplied to the vehicle clutch, the at least one input value being determined based on a processing of temporally successive values for the power supplied to the vehicle clutch; and inputting the at least one input value as input data into the neural network.

In this embodiment, determining the clutch temperature is based on a time course of the power supplied to the vehicle clutch. A temporal change in the power supplied to the vehicle clutch can therefore also be taken into account by the neural network. The power supplied to the vehicle clutch may be representative of a clutch temperature of the vehicle clutch. For example, a cooling or heating phase of the vehicle clutch can be recorded and taken into account when determining the clutch temperature. Furthermore, for example, the exceeding of a limit value for the clutch temperature or a deviation from a standard behavior of the clutch temperature during the heating or cooling phase can be taken into account. The clutch temperature can therefore also be determined more precisely by the neural network. Furthermore, the time course of the power supplied to the vehicle clutch is not determined by the neural network itself, but is only transmitted to it in the form of the processed input value. This means that the structure of the neural network is only slightly changed, despite taking additional input data into account. The computing effort required to determine the clutch temperature can thereby be kept low. Therefore, the storage space required for the neural network can also be reduced.

The neural network of the second aspect can be trained according to the method for training a neural network according to the first aspect. The technical effects and advantages of the first and second aspects can thus be combined.

According to one embodiment, the geometric property of the vehicle clutch is determined with respect to at least one clutch element of the vehicle clutch and is selected from at least one of the following: a heat capacity of the clutch element; a coefficient of friction of the coupling element; a heat transfer coefficient of the coupling element; a type of coupling element; a material of the coupling element; a mass of the coupling element; an effective surface of the coupling element; a cross section of the coupling element; a diameter of the coupling element; and a number of equivalent coupling elements. A coupling element of the vehicle clutch can be understood to mean, for example, one of the rotatable clutch disks, which can be brought into mechanical operative connection for torque transmission during operation of the vehicle clutch. The above geometric properties may be representative of a temperature of the vehicle clutch. Alternatively or additionally, the above geometric properties may be representative of power supplied to the vehicle clutch. A clutch temperature can therefore be determined by the neural network in a particularly simple manner from the geometric properties mentioned.

According to one embodiment, the at least one operating parameter of the vehicle clutch is selected from at least one of the following: a torque of a drive axle of a vehicle engine, which is in mechanical operative connection with the vehicle clutch; a speed difference between two rotating clutch elements of the vehicle clutch; a speed of a rotating clutch element of the vehicle clutch; a mechanical pressure acting on a coupling element of the vehicle clutch; a current intensity of an electric current flowing through a coupling element of the vehicle clutch; and a sump temperature of a vehicle transmission at the beginning of a shifting process of the vehicle clutch. Clutch elements can be, for example, the clutch disks of a vehicle clutch, which can be connected to transmit torque from the engine to the drive axle. The above parameters may be representative of power supplied to the vehicle clutch. Alternatively or additionally, the above parameters can be representative of a clutch temperature of the vehicle clutch. By using at least one of these parameters, the determination of the clutch temperature can be made easier.

In a third aspect, the invention relates to a control device for determining a clutch temperature of a vehicle clutch. The control device comprises a computer-readable storage medium on which a neural network for determining the clutch temperature is stored; an input device for inputting input data into the neural network, the input data comprising at least one geometric input variable which is representative of a geometric property of the vehicle clutch and at least one value of at least one operating parameter of the vehicle clutch; and an output device for outputting the determined clutch temperature by the neural network.

The mentioned devices of the control device according to the third aspect can be designed to receive, process and forward electrical signals. The mentioned devices of the control device according to the third aspect can be designed to carry out the method according to the first aspect and/or the second aspect. Analogously, the method according to the first aspect or the second aspect can be carried out by the control device according to the third aspect. The embodiments, technical effects and advantages explained for the first and second aspects therefore also apply analogously to the control device according to the third aspect.

According to one embodiment, the neural network stored on the storage medium has been trained according to the method according to the first aspect. The clutch temperature is determined by the neural network according to the method according to the second aspect. The embodiments, technical effects and advantages explained in the first, second and third aspects can thus be combined.

Brief description of the drawings

• 1 shows a flowchart with steps of a method for training a neural network, which is designed to determine a clutch temperature of a vehicle clutch, according to an embodiment of the invention.
• 2 shows a flowchart with steps of a method for determining a clutch temperature of a vehicle clutch using a neural network, according to a further embodiment of the invention.
• 3 shows schematically a control device for determining a clutch temperature of a vehicle clutch with a neural network according to a further embodiment of the invention.

Detailed Description of Embodiments

1 shows a flowchart with steps of a method for determining a clutch temperature of a vehicle clutch using a neural network according to an embodiment of the invention.

In a first training step TS1, at least one geometric input variable is provided as input data. The at least one geometric input variable is representative of a geometric property of the vehicle clutch. The at least one geometric input variable is in the exemplary embodiment 1 determined depending on the measured steel mass of the vehicle coupling. The at least one geometric input variable is in the exemplary embodiment 1 further determined depending on a predetermined heat capacity of the steel used for the vehicle clutch. The specified heat capacity of the steel used for the vehicle clutch is in the exemplary embodiment 1 temperature dependent. A synthetic value is generated from the measured steel mass and the specified heat capacity, which represents a heat conductivity of the vehicle clutch is representative as a geometric property of the same.

In an embodiment not shown, the at least one geometric input variable is determined as a function of a measured diameter and a measured number of clutch linings of the vehicle clutch. Furthermore, the at least one geometric input variable in this exemplary embodiment is determined depending on a predetermined coefficient of friction of the clutch linings. A synthetic value is generated from the measured diameter, the number of clutch linings and the specified coefficient of friction, which is representative of the friction of the vehicle clutch as a geometric property of the same.

In a second training step TS2, at least one value of at least one operating parameter of the vehicle clutch is provided as input data for the neural network. The provided value of the at least one operating parameter is transmitted to at least one input node of the neural network.

In a third training step TS3, values for a clutch temperature are provided as output data. These values are measured values for the clutch temperature. In an exemplary embodiment, not shown, the values for the clutch temperature are generated synthetically. These values are transmitted to at least one output node of the neural network.

In a fourth training step TS4, the neural network is trained with the input data and the output data in order to learn a connection between the geometric property and the at least one operating parameter and the clutch temperature. Based on this relationship learned by the neural network, the trained neural network can be used to determine a clutch temperature of the vehicle clutch during operation of the vehicle.

For training, the input data transmitted to the input nodes of the neural network are transmitted to intermediate nodes of the neural network via connections mathematically weighted by the neural network. The transmitted input data is processed on the intermediate nodes to obtain the output data provided on the output nodes. The output nodes are connected to the intermediate nodes via connections mathematically weighted by the neural network. The mathematical weighting of the respective connections is adjusted by the neural network during the training process.

By training the neural network with the geometric input variable, the geometric property of the vehicle clutch is taken into account when determining the clutch temperature. As a result, the trained neural network can also be used to determine the clutch temperature of other vehicle clutches that have the same or a comparable geometric property. This means that a new neural network does not have to be trained for every vehicle clutch. The calculation effort for determining the clutch temperature of various vehicle clutches can therefore be reduced.

2 shows a flowchart with steps of a method for determining a clutch temperature of a vehicle clutch using a neural network, according to a further embodiment of the invention.

In a first determination step BS1, at least one geometric input variable is entered into the neural network as input data, which is representative of a geometric property of the vehicle clutch. The geometric input variable is in the exemplary embodiment 2 analogous to the exemplary embodiment of 1 based on the measured steel mass and the specified heat capacity of the vehicle clutch and is representative of the thermal conductivity of the vehicle clutch as a geometric property of the same.

In a second determination step BS2, at least one value of at least one operating parameter of the vehicle clutch is entered into the neural network as input data. The at least one value of the at least one operating parameter is transmitted to at least one input node of the neural network.

In a third determination step BS3, a clutch temperature is determined by the neural network. The clutch temperature is determined based on the input data and a relationship of the input data to the clutch temperature learned by the neural network. The clutch temperature determined by the neural network is transmitted to a transmission control of the vehicle and used by it.

This is used to carry out the method of the embodiment 2 The neural network used can be according to the method of the embodiment 1 be trained.

3 schematically shows a control device 10 for determining a clutch temperature of a vehicle clutch, not shown, with a neural network 12 according to a further embodiment of the invention.

The control device 10 includes a computer-readable storage medium 14 on which the neural network 12 is stored. The control device 10 further comprises an input device 16, which is designed to receive values 18a, 18b, 18c of operating parameters of the vehicle clutch. The input device 16 transmits the values 18a, 18b, 18c of the operating parameters of the vehicle clutch as input data 20a to the neural network 12.

The control device 10 further comprises a determination device 22 which receives a measured parameter 24a and a predetermined parameter 24b of the vehicle clutch. The measured parameter 24a and the predetermined parameter 24b are representative of a geometric property of the vehicle clutch. From the measured parameter 24a and the predetermined parameter 24b, the determination device 22 determines a geometric input variable 26, which is representative of the geometric property of the vehicle clutch. Analogous to the embodiments of 1 and 2 The measured parameter 24a is a steel mass of the vehicle clutch and the predetermined parameter 24b is a heat capacity of the vehicle clutch. In this exemplary embodiment, the geometric input variable is representative of a thermal conductivity of the vehicle clutch. In an alternative exemplary embodiment, not shown, the measured parameter 24a is a diameter and a number of clutch linings of the vehicle clutch. The specified parameter 24b is a coefficient of friction of the clutch linings. A synthetic value is generated from the measured diameter, the number of clutch linings and the specified coefficient of friction, which is representative of the friction of the vehicle clutch as a geometric property of the same. The geometric input variable 26 is transmitted from the determination device 22 to the input device 16. The input device 16 in turn transmits the geometric input variable 26 as input data 20b to the neural network.

The control device 10 further comprises detection devices 28a, 28b, which detect successive values 30a, 30b, or 30c, 30d for a power supplied to the vehicle clutch. The detection devices 28a, 28b are low-pass filters and the values 30a, 30b, 30c, 30d for a power supplied to the vehicle clutch are a switching power supplied to the vehicle clutch during a switching process or gear change. The detected values 30a, 30b, 30c and 30d for the switching power are transmitted from the detection devices 28a, 28b to a further determination device 32. The further determination device 32 determines at least one average value for the switching power by processing the temporally successive values 30a, 30b, 30c, 30d. The at least one average value for the switching power is transmitted from the determining device 32 to the input device 16 as an input value 34. The input device 16 transmits the input value 24 as input data 20c to the neural network 12.

A clutch temperature KT of the vehicle clutch is determined by the neural network 12 from the input data 20a, 20b, 20c. The determined clutch temperature KT is transmitted from the neural network 12 to an output device 36. The output device 30 transmits the specific clutch temperature KT to other vehicle control devices.

The ones in the 3 The control device 10 shown can be used to train the neural network according to the exemplary embodiment 1 be used.

Reference symbols

10

Control device

12

Neural network

14

storage medium

16

Input device

18a, 18b, 18c

Values of operating parameters

20a, 20b, 20c

Input data

22

Determination device

24a

measured parameter

24b

specified parameter

26

geometric input size

28a, 28b

Detection devices

30a, 30b,

successive values for one of the vehicle clutches

30c, 30d

power supplied

32

further determination device

34

Input value

36

Dispensing device

KT

Clutch temperature

TS1

first training step

TS2

second training step

TS3

third training step

TS4

fourth training step

BS1

first determination step

BS2

second determination step

BS3

third determination step

Claims (12)

Method for training a neural network (12) which is designed to determine a clutch temperature (KT) of a vehicle clutch, the method having the following steps: - Providing (TS1) at least one geometric input variable (26), which is representative of a geometric property of the vehicle clutch, as input data (20b); - Providing (TS2) at least one value (18a, 18b, 18c) of an operating parameter of the vehicle clutch as input data (20a); - Providing (TS3) values for the clutch temperature (KT) as output data; and - Training (TS4) of the neural network (12) with the input data (20a, 20b) and the output data in order to learn a connection between the geometric property and the at least one operating parameter and the clutch temperature (KT). Procedure according to Claim 1 , wherein the geometric input variable (26) is determined depending on at least one measured parameter (24a) of the vehicle clutch. Procedure according to Claim 1 or 2 , wherein the geometric input variable (26) is determined depending on at least one predetermined parameter (24b) of the vehicle clutch. Method according to one of the preceding claims, wherein the geometric input variable (26) is provided continuously during training of the neural network (12). Method according to one of the preceding claims, wherein the neural network (12) is further trained by means of at least one specific input value (34) which is representative of a power supplied to the vehicle clutch, the at least one input value (34) being based on a processing of temporally successive Values (30a, 30b, 30c, 30d) are determined for the power supplied to the vehicle clutch. Method for determining a clutch temperature (KT) of a vehicle clutch using a neural network (12), the method having the following steps: - Input (BS1) of at least one geometric input variable (26), which is representative of a geometric property of the vehicle clutch, as input data (20b) into the neural network (12); - Input (BS2) of at least one value (18a, 18b, 18c) of at least one operating parameter of the vehicle clutch as input data (20a) into the neural network (12); and - Determining (BS3) a clutch temperature (KT) by the neural network (12), based on the input data (20a, 20b) and a connection between the input data and the clutch temperature (KT) learned by the neural network (12). Procedure according to Claim 6 , wherein the method further comprises the following steps: - determining at least one input value (34) which is representative of a power supplied to the vehicle clutch, the at least one input value (34) being based on a processing of temporally successive values (30a, 30b, 30c , 30d) is determined for the power supplied to the vehicle clutch; and - inputting the at least one input value (34) as input data (20c) into the neural network (12). Procedure according to Claim 6 or 7 , wherein the neural network (12) according to the method for training a neural network (12) according to one of Claims 1 until 5 was trained. Method according to one of the preceding claims, wherein the geometric property of the vehicle coupling is determined with respect to at least one coupling element of the vehicle coupling and is selected from at least one of the following: a heat capacity of the coupling element; a coefficient of friction of the coupling element; a heat transfer coefficient of the coupling element; a type of coupling element; a material of the coupling element; a mass of the coupling element; an effective surface of the coupling element; a cross section of the coupling element; a diameter of the coupling element; and a number of equivalent coupling elements. Method according to one of the preceding claims, wherein the at least one operating para meter of the vehicle clutch is selected from at least one of the following: a torque of a drive axle of a vehicle engine, which is in mechanical operative connection with the vehicle clutch; a speed difference between two rotating clutch elements of the vehicle clutch; a speed of a rotating clutch element of the vehicle clutch; a mechanical pressure acting on a coupling element of the vehicle clutch; a current intensity of an electric current flowing through a coupling element of the vehicle clutch; and a sump temperature of a vehicle transmission at the beginning of a shifting process of the vehicle clutch. Control device (10) for determining a clutch temperature (KT) of a vehicle clutch, the control device comprising: - a computer-readable storage medium (14) on which a neural network (12) for determining the clutch temperature (KT) is stored; - an input device (16) for inputting input data (20a, 20b, 20c) into the neural network (12), the input data (20a, 20b, 20c) being at least one geometric input variable (26), which is representative of a geometric property of the vehicle clutch, and include at least one value (18a, 18b, 18c) of at least one operating parameter of the vehicle clutch; and - an output device (36) for outputting the determined clutch temperature (KT) by the neural network (12). Control device (10). Claim 11 , wherein the neural network (12) according to the method for training a neural network (12) according to one of Claims 1 until 5 was trained; and wherein the clutch temperature (KT) is determined by the neural network (12) according to the method according to one of Claims 6 until 10 is determined.

Images