Control of a manual transmission

Abstract

A manual transmission (115) is connected to an internal combustion engine (110). A method (500) for controlling the manual transmission (115) comprises steps of acquiring (510) at least one parameter that influences the torque actually provided by the internal combustion engine (110); and determining (520) a torque signal (815) that indicates the torque provided by the internal combustion engine (110), based on the at least one parameter. The torque signal (815) is determined using an artificial neural network (230).

Description

[0001] The present invention relates to the control of a manual transmission. In particular, the invention relates to the control of a manual transmission that converts torque provided by an internal combustion engine.

[0002] A motor vehicle includes a drivetrain with a manual transmission. An internal combustion engine can be configured to provide torque, and the manual transmission can transmit the torque to a drive wheel. Different gear ratios can be engaged in the manual transmission to achieve different reductions between a speed of the drive engine and a speed of the drive wheel.

[0003] A gear change in the transmission may involve disengaging a first clutch (or brake) and simultaneously engaging a second clutch (or brake) of the manual transmission. To ensure the smoothest gear change possible, it is necessary to know the torque provided by the engine as accurately as possible. A control unit for the internal combustion engine provides a torque signal to a control unit of the manual transmission.

[0004] It has been shown that such a torque signal may deviate significantly from the actual torque provided by the internal combustion engine. In some cases, a deviation of approximately 30% of the true value has been observed.

[0005] State of the art is disclosed, for example, in DE 10 2023 207 025 B3, JP 2000 - 104 813 A and DE 11 2019 007 222 T5.

[0006] One object underlying the present invention is to provide a torque signal that is more accurately indicative of the torque actually provided by an internal combustion engine. The invention achieves this object by means of the subject matter of the independent claims. Subclaims specify preferred embodiments.

[0007] A manual transmission is connected to an internal combustion engine. A method for controlling the manual transmission comprises the steps of detecting at least one parameter that influences the torque actually delivered by the internal combustion engine; and determining a torque signal indicative of the torque delivered by the internal combustion engine based on the at least one parameter. The torque signal is determined using an artificial neural network (ANN).

[0008] The invention can preferably be used in an automated transmission or an automatic transmission. The transmission can comprise a simple transmission with an input side and an output side. Alternatively, the transmission can be designed as a hybrid transmission, having two input sides for two different drive motors and one output side.

[0009] The determined torque signal can be used to control the manual transmission. In particular, a gear change between different gear stages of the manual transmission can be controlled depending on the determined torque signal. The transmission can comprise a torque converter transmission with a hydrodynamic torque converter. Optionally, the torque converter can also be controlled, for example, by influencing flow conditions in the torque converter or by opening or closing a lock-up clutch. The internal combustion engine further preferably comprises a reciprocating piston engine, in particular with multiple cylinders. The internal combustion engine and the manual transmission can be part of a drivetrain, which can in particular be provided on board a motor vehicle to drive the motor vehicle.

[0010] The ANN can be trained using methods similar to those typically used to match a manual transmission to an internal combustion engine. In particular, the internal combustion engine can be operated under different conditions, for example, different speeds or torque requirements. However, key training steps can be automated, eliminating the need for manual adjustment. The ANN can be specifically trained at those operating points of the internal combustion engine where the actual torque cannot yet be determined sufficiently well. The trained ANN can easily be used on a large number of manual transmissions of the same type or on control devices designed for them. The strength of an ANN, namely mapping a large number of input parameters to a small number of output parameters, can be advantageously exploited.

[0011] The trained ANN can be implemented or realized with minimal resources. A control device for the manual transmission can be implemented with the ANN at a reasonable cost. The control device can be configured to control the transmission, in particular to change an engaged gear. Alternatively, the control device can be constructed separately and configured to provide a torque correction signal or a corrected torque signal to a control device controlling the transmission or the gear change.

[0012] The ANN can support the determination of the delivered torque even at operating points for which it has not been specifically trained. The trained ANN can be used advantageously at different operating points, with different degrees of wear on the manual transmission or internal combustion engine, or with different individual deviations of the parameters of an individual manual transmission from the specified series parameters.

[0013] An external torque signal is detected, indicating a torque delivered by the internal combustion engine; the torque signal is determined based on the external torque signal. Even if the external torque signal deviates significantly from the actual torque delivered, it can be used as the basis for determining the torque signal. The ANN can be trained to support an improved determination of the actual torque delivered. The determined torque can be significantly closer to the actual torque delivered than the external torque signal.

[0014] The external torque signal is provided by a control device for the internal combustion engine. This is generally not a measured signal derived, for example, from sensor signals from a torque sensor, but rather a synthetic signal that represents an approximation of the provided torque. The method can essentially be carried out using a control device for the manual transmission. This allows system separation between the internal combustion engine and the manual transmission to be maintained, while simultaneously enabling the provided torque to be determined with a level of accuracy that may exceed that with which the external torque signal is determined by the control device for the internal combustion engine. The manual transmission can thus be controlled independently of any determination error in the external torque signal.

[0015] In a first variant, the ANN is trained to provide the torque signal directly based on its input values. In this case, the manual transmission can be controlled directly based on the determined torque signal.

[0016] In a second variant, the ANN determines a difference between the external torque signal and the actual torque. The determined torque signal can then be calculated based on the external torque signal and the determined difference between the external torque signal and the actual torque. This advantageously provides better insight into the extent to which the ANN compensates for the external torque signal, or how far the determined torque signal deviates from the external one. The external torque signal is available unchanged for various applications.

[0017] In this variant, the relative influence of the difference in the generation of the specific torque signal can be more easily limited. For example, it can be determined that the specific torque signal does not deviate from the external torque signal by more than approximately 15%. The specific difference can be limited relatively or absolutely. This prevents the use of a value that deviates significantly from the torque actually provided. This can prevent poor shifting comfort when changing gears in the manual transmission or even damage to the manual transmission due to an incorrect torque value being used as a basis.

[0018] The specific difference is typically only a fraction of the external torque signal. A constant shift (offset) across the torque range can be represented in the same way as a relative error. The specific error can be dependent on the torque or another parameter.

[0019] For example, if the torque provided according to the external torque signal is 300 Nm and the ANN determines a value of 325 Nm, the first variant can provide a torque signal indicating 325 Nm. The second variant provides a torque signal indicating a deviation of 25 Nm, so that the sum of this torque signal and the external torque signal again results in 325 Nm.

[0020] It is further preferred that the specific torque signal be determined based on a plurality of parameters. The parameters used can be provided by a control device for the internal combustion engine or by another control device. When the method is carried out on board a motor vehicle, various systems of the motor vehicle can be interconnected via data technology, for example, via a data bus. Parameters can be transmitted via the data bus. Thus, even a parameter that is not used to control the internal combustion engine can be taken into account.

[0021] A parameter used as an input signal for the artificial neural network preferably comprises a physical parameter that was recorded or sampled in the area of ​​the internal combustion engine or the manual transmission. The parameter can be processed, for example by smoothing it or limiting its value range. In a further embodiment, a time derivative of a parameter can also be formed. In a corresponding manner, time derivatives of several parameters can also be formed. A time derivative and / or the parameter can be provided as an input signal to the ANN. However, a parameter that is determined on the basis of several physical parameters that were sampled in the area of ​​the internal combustion engine or the manual transmission is less preferred as an input parameter for the ANN.

[0022] In a further embodiment, the torque signal is determined based on the additional temporal dependence of a parameter. For this purpose, a temporal filter can be provided that determines the temporal correlation of the parameter. The relationship or correlation can also be used as an input parameter of the ANN in addition to one or more of the parameters under consideration. Optionally, a correlation between two or more parameters can also be determined and used as the basis for providing the torque signal.

[0023] A parameter can, for example, include a torque request to the internal combustion engine, a speed, an ignition timing, a speed gradient, a gearshift signal, an internal combustion engine temperature, a prevailing air pressure, a cylinder deactivation, an air mass introduced into the internal combustion engine, a fuel quantity introduced into the internal combustion engine, an adjustment angle of a camshaft of the internal combustion engine, or a lambda value. Other or further parameters that directly or indirectly influence the torque provided by the drive engine can also be used. Virtually any combination is possible. Such a parameter can preferably be sensed directly by a sensor in the area of ​​the drive engine or the manual transmission.If the manual transmission includes a hydrodynamic torque converter, a parameter on the torque converter, for example a fluid temperature or torque converter slip, can also be evaluated as a parameter.

[0024] In another embodiment, an electric drive motor is additionally connected to the transmission and the internal combustion engine. The electric drive motor can be integrated with the transmission or implemented separately. Another parameter can include a speed or a torque provided by the electric drive motor. The torque of the drive motor can be easily determined based on one or more electrical parameters of the drive motor. In one embodiment, the electric drive motor is controlled using a field-oriented controller, which can provide a good value for the provided torque.

[0025] Controlling the manual transmission may involve changing an engaged gear. This process is preferably controlled in response to the specific torque signal.

[0026] According to a further aspect of the present invention, a first method for training an artificial neural network comprises steps of operating an internal combustion engine configured to provide a torque in different operating states; determining a torque actually provided by the internal combustion engine; detecting at least one parameter that influences the torque actually provided by the internal combustion engine; and training the artificial neural network to minimize a deviation between a signal provided by the network and the torque actually provided.

[0027] Accordingly, a second method for training an artificial neural network comprises steps of operating an internal combustion engine configured to provide torque in different operating states; detecting an external torque signal indicative of the torque provided by the internal combustion engine; determining a torque actually provided by the internal combustion engine; detecting at least one parameter influencing the torque actually provided by the internal combustion engine; determining a sum of the external torque signal and a signal provided by the network; and training the artificial neural network to minimize a deviation between the sum and the torque actually provided.

[0028] According to yet another aspect of the present invention, a third method for training an artificial neural network comprises the steps of operating a first internal combustion engine configured to provide torque in different operating states; detecting an external torque signal indicative of the torque provided by the internal combustion engine; determining a torque actually provided by the internal combustion engine; detecting at least one parameter influencing the torque actually provided by the internal combustion engine; determining a sum of the external torque signal and a signal provided by the network; and training the artificial neural network to minimize a deviation between the sum and the torque actually provided.It further comprises steps of operating a second internal combustion engine, which is configured to provide a torque, in different operating states; detecting a second external torque signal indicative of the torque provided by the second internal combustion engine; determining a torque actually provided by the second internal combustion engine; detecting at least one parameter that influences the torque actually provided by the second internal combustion engine; and training the artificial neural network to minimize a deviation between the external torque signal of the first internal combustion engine and an external torque signal of the second internal combustion engine.

[0029] The external torque signal is preferably provided by a control device for the internal combustion engine. A corresponding sensor can be used to determine the torque actually provided by the internal combustion engine. Such a sensor can be installed in a measuring shaft that is inserted between the internal combustion engine and the manual transmission. The internal combustion engine can be operated under predetermined conditions or at predetermined operating points on a test bench or in a vehicle. Torque provided by the internal combustion engine can be picked up, for example, by means of an electric or hydraulic brake. The data is preferably collected while the manual transmission is arranged between the internal combustion engine and the brake. This can control gear changes in which a gear position engaged in the manual transmission is changed.Further data may be collected while the internal combustion engine is driving a predetermined device. For example, the internal combustion engine and the manual transmission may be part of a motor vehicle's drivetrain. Data may be recorded while the motor vehicle is driven by the internal combustion engine and, if applicable, using the manual transmission.

[0030] Data used to train the artificial neural network can be collected essentially in the same way as is used to conventionally adapt a manual transmission to an internal combustion engine. The ANN is preferably trained using back propagation. Training can be considered complete when the deviation consistently falls below a predetermined threshold or no further improvement in the quality of the determination is observed.

[0031] The trained artificial neural network can be used in a method described herein for determining a torque provided by an internal combustion engine or for controlling a manual transmission based on this torque.

[0032] According to yet another aspect of the present invention, a control device for a manual transmission connected to an internal combustion engine comprises an interface for detecting at least one parameter that influences the torque actually provided by the internal combustion engine; and a processing device configured to determine, based on the at least one parameter, a torque signal indicative of the torque provided by the internal combustion engine. The processing device comprises or implements an artificial neural network.

[0033] Further preferably, the control device comprises an interface for detecting an external torque signal that indicates the torque provided by the internal combustion engine. The processing device is configured to determine the torque signal based on the additional external torque signal.

[0034] Furthermore, a manual transmission is proposed which comprises a control device as described herein. The manual transmission can, in particular, comprise a torque converter transmission with a hydrodynamic torque converter. In a preferred embodiment, the manual transmission is designed as a multi-stage, cascaded planetary gear. A gear engaged in the manual transmission can be controlled by actuating a brake or a clutch. Furthermore, a lock-up clutch of the torque converter can be controlled.

[0035] A further proposed system comprises a manual transmission as described herein and an internal combustion engine connected to the manual transmission.

[0036] The processing device can be configured to carry out a method described herein in whole or in part. For this purpose, the processing device can be implemented electronically and comprise a programmable microcomputer or microcontroller, and the method can be in the form of a computer program product with program code means. The computer program product can also be stored on a computer-readable data carrier. Features or advantages of the method can be transferred to the device, or vice versa.

[0037] According to yet another aspect of the present invention, a motor vehicle comprises a system described herein. The motor vehicle may, in particular, comprise a passenger car. In further embodiments, the motor vehicle may also comprise a motorcycle, a truck, a bus, or a similar vehicle.

[0038] The invention will now be described in more detail with reference to the accompanying figures, in which: Fig. 1 a system on board a motor vehicle; Fig. 2 and Fig. 3 different embodiments of a control device for a manual transmission; Fig. 4 a schematic representation of a torque determination using an artificial neural network; Fig. 5 a flowchart of a procedure Fig. 6 and Fig. 7 approaches to training an artificial neural network; and Fig. 8 and Fig. 9 exemplary torque curves represents.

[0039] Fig. 1 shows a schematic representation of a system 100 on board a motor vehicle 105. The system 100 comprises an internal combustion engine 110 and a manual transmission 115. Optionally, an electric drive motor 120 can be provided, which is preferably integrated with the manual transmission 115. Both the internal combustion engine 110 and the electric drive motor 120 can each provide a torque, which can be transmitted to a drive wheel 125 of the motor vehicle 105 via the manual transmission 115.

[0040] A first control device 130 is provided for controlling the internal combustion engine 110. A second control device 135 can be provided for controlling the manual transmission 115. Furthermore, a third control device 140 can be used to control the electric drive motor 120. The control devices 130 to 140 can be communicatively connected to exchange parameters, measured values, or requirements. In the illustration of Fig. 1, the control devices 130 to 140 are connected to a data bus 145. Another control unit on board the motor vehicle 105 can also be connected to the data bus 145. In a further embodiment, several different data buses 145 can be used.

[0041] A gear engaged in the manual transmission 115 can be changed depending on a torque provided by the internal combustion engine 110. Since the actual torque cannot be measured on board a conventional motor vehicle 105 for cost reasons, control can be based on a first torque signal provided by the control device 130 of the internal combustion engine 110. To improve the torque signal against errors or deviations from the actual torque, it is proposed to use a machine learning technique on the part of the control device 135 of the manual transmission 115 or in another control device. In one embodiment, the control device 130 of the internal combustion engine 110 can be used for this purpose.

[0042] Fig. Figure 2 shows a schematic representation of a control device 135 of a manual transmission 115 in a first embodiment. Only elements of the control device 135 that serve to provide a torque signal are shown.

[0043] A first interface 205 is configured to receive a first torque signal, which can be provided by the first control device 130 of the internal combustion engine 110. A second interface 210 and a third interface 215 are configured to receive parameters that can influence the first torque signal or its deviation from a torque actually provided by the internal combustion engine 110. A parameter can be provided by the first control device 130 of the internal combustion engine 110 or another control unit.

[0044] Parameters received via the third interface 215 can be derived over time using a gradient generator 220. The derived parameters, as well as parameters received via the second interface 210, can be passed along with the first torque signal from the first interface 205 to a filter 225, which can establish a temporal relationship between individual parameters. Such a filter is described, for example, in DE 10 2021 205 835 A1.

[0045] In this way, a plurality of signals or parameters can be provided that relate to an operating point of the internal combustion engine 110. These signals and parameters can be provided to an artificial neural network 230, which is trained to provide a difference or compensation signal, which is additively combined with the first torque signal of the first interface 205 in a summing element 235 to provide a second torque signal that is an approximation of the torque provided by the internal combustion engine 110. The second torque signal can be provided to an interface 240.

[0046] In the illustrated embodiment, the compensation signal can be processed. For example, a value range of the compensation signal can be limited, or the compensation signal can be limited to a predetermined fraction of a value of the first torque signal. For example, in this way, it can be achieved that a correction of the first torque signal does not exceed + / - 15% of a current value of the first torque signal.

[0047] Fig. 3 shows a control device 135 for the manual transmission 115 in a further embodiment. In contrast to the Fig. In the embodiment shown in Figure 2, the artificial neural network 230 is trained to provide the second torque signal directly. A compensation signal for the external torque signal is not provided. It should be noted that the determination of the torque signal does not necessarily have to be based on the torque signal at the first interface 205, but can also be performed independently. In this case, the first interface 205 can also be omitted.

[0048] Fig. 4 shows a schematic representation of a torque determination using an artificial neural network 230. With reference to the representations of Fig. 2 and Fig. 3 shows various parameters 405 that are to be processed by the ANN 230. Example parameters include a first torque signal, a temperature of the internal combustion engine 110, or a speed at the output shaft.

[0049] The parameters are provided to different filters 225, each of which can determine a temporal correlation between a parameter 405 and time. In other words, a filter maps the past of a signal in different time windows. Multiple filters can have different time constants and thus process histories of different lengths.

[0050] By way of example only, the ANN 230 in the illustrated embodiment comprises an input layer 410, a first intermediate layer (hidden layer) 415, a second intermediate layer 420, and an output layer 425. Each of the layers 410 to 425 comprises one or more neurons 430.

[0051] Outputs of filters 225 are each connected to a neuron 430 of input layer 410 of ANN 230. In the illustrated embodiment, output layer 425 comprises only a single neuron 430, which provides a correction signal or the second torque signal. In another embodiment, multiple neurons 430 can be located in output layer 425 and provide different output signals. For example, one output signal can take into account a flywheel mass of the manual transmission 115 and / or a drive motor 110, 120, while another can not.

[0052] The neurons 430 are interconnected in layers as shown. If a neuron 430 is stimulated by a signal, it can provide an output signal to one or more other neurons 430. Incoming and / or outgoing signals from the neurons 430 can be weighted differently. Determining the weights is known as training the ANN 230. It is proposed to train the ANN 230 to provide the second torque signal or a compensation signal for the first torque signal to determine the second torque signal based on the parameters presented to it.

[0053] Fig. 5 shows a flowchart of an exemplary method 500 for controlling a manual transmission 115. The method 500 is preferably carried out by means of the control device 135.

[0054] In a step 505, the first torque signal can be detected. The first torque signal is preferably provided by the control device 130 of the internal combustion engine 110.

[0055] In a step 510, one or more parameters can be detected that can influence either the torque actually provided by the internal combustion engine 110 or a relationship between the first torque signal and the torque actually provided. Such parameters can, for example, relate to a rotational speed, a temperature, or an operating point of the internal combustion engine 110.

[0056] In a step 515, one or more acquired parameters can be processed. The processing can, in particular, include forming a gradient or filtering. Filtering can include determining a temporal relationship of a parameter.

[0057] In a step 520, the second torque signal can be determined. For this purpose, the processed parameters can be provided to an appropriately trained ANN 230 as input signals. An output signal of the ANN 230 can comprise either the second torque signal or a correction signal, which can be processed additively with the first torque signal to determine the second torque signal. The second torque signal typically reflects the torque actually provided by the internal combustion engine 110 at least as well as the first torque signal. In many cases, a deviation between the second torque signal and the actually provided torque can be significantly smaller than that of the first torque signal.

[0058] Based on the second torque signal, the manual transmission 115 can be controlled. In particular, a gear change, i.e., the disengagement of a first gear and the engagement of a second gear, can be controlled at the manual transmission 115. For this purpose, one or more input signals can be processed in a step 525. An input signal can be determined in the manual transmission 115 or can include a request to change the gear.

[0059] In a step 530, the manual transmission 115 can be controlled accordingly. For example, a clutch or brake of the manual transmission 115 can be opened or closed to control the gear change. It is preferred that the gear change be controlled in an overlapping manner so that the engagement of the first gear and the engagement of the second gear take place with as little jerking as possible.

[0060] Fig. 6 shows a first exemplary procedure for training an artificial neural network 230. A torque support 605 is mounted between the internal combustion engine 110 and the manual transmission 115, which determines a torque actually provided by the internal combustion engine 110. The determined torque can be provided as a torque signal. Furthermore, the first torque signal can be provided by the internal combustion engine 110 or its control device 130.

[0061] Additionally, one or more parameters may be provided that control the provision of torque by the internal combustion engine 110. Such a parameter may, in particular, relate to a rotational speed of the internal combustion engine 110.

[0062] The parameters mentioned can be forwarded to the KNN 230, see Fig. 4. Weights between the neurons 430 can then be adjusted such that the determined second torque signal follows the measured torque signal as closely as possible.

[0063] Fig. 7 shows a further procedure for training an artificial neural network 230 for a system 100 when a first compensation model 705 is already available, which can be used to control a similar system 100 or transmission 115. Such a case may arise, for example, if a control system for an internal combustion engine 110 is present and a control system for a variant of the internal combustion engine 110 is to be provided. In another example, the transmission 115, which is adapted to a first internal combustion engine 110, is to be adapted to a second internal combustion engine 110.

[0064] The first compensation model 705 can be created in any desired manner, in particular classically or by manual tuning, optionally also by means of machine learning, for example according to a method described herein. The first compensation model 705 can comprise compensation for a torque error between a first torque signal provided by the control device 130 of the internal combustion engine 110 and the actually provided torque. In other words, the first compensation model 705 can provide a torque signal based on a measured torque signal and at least one parameter of the existing system indicative of the actual torque, which reflects the external torque signal of the first system. This external torque signal is typically provided by a control unit of the internal combustion engine 110.The first compensation model 705 can also be called an inverted compensation model with respect to the solution proposed herein.

[0065] For example, an ANN can be operated with the following input variables: - Speed - Temperature - Torque signal of the measuring technology (measured true value) - Speed ​​gradient

[0066] The torque signal from the engine control unit, which corresponds to the external torque signal, is used as the output signal and thus as ground truth. This intentionally replicates the "false" torque signal of an "old" setup. The measurement data and everything else usually come from a vehicle from the old project.

[0067] The following procedure is proposed for training the compensation model 710. First, the present system 100 can be equipped with a torque sensor 605 that reflects the torque actually provided by the internal combustion engine 110. An inverted compensation model 705 is then created and trained to provide a torque signal based on the external torque signal of the present internal combustion engine 110 that reflects an external torque signal from the other system with improved accuracy. This signal forms the ground truth for the subsequent training process.

[0068] Subsequently, measurement data is generated using the vehicle 105 of the new system 100 and the transmission parameters of the other system. By nesting the inverted compensation model 705 and the new compensation model 710 during the training process, new parameters for the torque compensator are generated. The torque compensator now creates a relationship that aligns the torque signal of the existing system with the torque signal of the other system. Finally, fine-tuning can be performed.

[0069] Fig. 8 and Fig. 9 show exemplary torque curves on a system 100. Fig. 8 is based on operation without gear change in the time range shown, while Fig. 9 illustrates operation with gear changes. Both illustrations show a torque signal 805, a first torque signal 810, and a second torque signal 815. The torque signal 805 originates from a measurement, for example, using a torque sensor 605. The first torque signal 810 is provided by the control device 130 of the internal combustion engine 110. The second torque signal 815 is generated by the control device 135 of the manual transmission 115 using a trained ANN 230. A time is plotted horizontally, and a torque is plotted vertically.

[0070] It can be seen that the second torque signal 815 deviates significantly less from the torque signal 805 than the first torque signal 810.

[0071] In Fig.9 also shows a speed signal 905 relating to a shaft between the internal combustion engine 110 and the manual transmission 115. Shortly before second 71 and shortly after second 73, the speed signal 905 drops significantly in each case. At this time, a gear change from a lower to a higher gear occurs. At these times, the second torque signal 815 indicates a temporary drop. A corresponding drop can also be seen in the first torque signal 810. The temporary drops are due to a torque reduction during the gear change, which is controlled by a control system of the internal combustion engine 110 in order to compensate for the rotating inertia of the manual transmission 115. This control system is also referred to as engine torque intervention. Overall, a high degree of alignment of the second torque signal 815 with the torque signal 805 can be achieved even in the dynamic case.

[0072] It should be noted that some parts of the transmission control system want to "see" this engine torque intervention, while others do not. In other words, there are parts that expect a signal similar to the first torque signal 805 and others that expect a signal similar to the second torque signal 815. The torque compensator is preferably configured to provide a first signal representing the engine torque intervention and / or a second signal that does not represent it. Reference symbol 100 systems 105 Motor vehicle 110 internal combustion engine 115 manual transmissions 120 electric drive motor 125 drive wheel 130 Control device of the internal combustion engine 135 Manual transmission control unit 140 Control device of the electric drive machine 145 Data bus 205 first interface for the first torque signal 210 second interface for parameters 215 third interface for parameters 220 gradient images 225 filters 230 artificial neural network, ANN 235 summation term 240 interface 405 parameters 410 Input layer 415 first intermediate layer (hidden layer) 420 second intermediate layer 425 Output layer 430 neurons 500 procedures 505 capture first torque signal Record 510 parameters 515 Prepare parameters 520 determine second torque signal Process 525 input signals 530 Control manual transmission 605 torque sensor 705 first (inverted) compensation model 710 second compensation model 805 Torque signal 810 first torque signal 815 second torque signal 905 Speed ​​signal

Claims

[1] Method (500) for controlling a manual transmission (115) connected to an internal combustion engine (110), the method (500) comprising the following steps: - detecting (510) at least one parameter that influences the torque actually provided by the internal combustion engine (110); and - determining (520) a torque signal (815) indicative of the torque provided by the internal combustion engine (110) based on the at least one parameter; - wherein the torque signal (815) is determined by means of an artificial neural network (230); wherein an external torque signal (810) is detected, which indicates a torque provided by the internal combustion engine (110); and wherein the torque signal (815) is determined on the additional basis of the external torque signal (810), wherein the external torque signal (810) is provided by a control device (130) for the internal combustion engine (110). [2] The method (500) of claim 1, wherein the network (230) determines a difference between the external torque signal (810) and the actual torque, and the torque signal (815) is determined based on the external torque signal (810) and the difference. [3] The method (500) of claim 2, wherein a relative influence of the difference in the formation of the determined torque signal (815) is limited. [4] The method (500) of any preceding claim, wherein the torque signal (815) is determined based on a plurality of parameters. [5] The method (500) of claim 4, wherein the torque signal (815) is determined on the additional basis of a time derivative of one of the parameters. [6] Method (500) according to claim 4 or 5, wherein the torque signal (815) is determined on the additional basis of a time dependence of a parameter. [7] Method (500) according to one of the preceding claims, wherein a parameter comprises a torque request, a rotational speed, an ignition timing, a rotational speed gradient, a gear change signal, a temperature of the internal combustion engine (110), a prevailing air pressure, a cylinder deactivation, an air mass introduced into the internal combustion engine (110), a fuel quantity introduced into the internal combustion engine (110), an adjustment angle of a camshaft of the internal combustion engine (110) or a lambda value. [8] Method (500) according to one of the preceding claims, wherein an electric drive machine is connected to the transmission and the internal combustion engine (110); wherein a parameter comprises a speed or a torque provided by the electric drive machine. [9] Method (500) according to one of the preceding claims, wherein a change of a gear stage engaged in the manual transmission (115) is controlled on the basis of the determined torque signal (815). [10] A method for training an artificial neural network, the method (500) comprising the following steps: - operating an internal combustion engine (110) configured to provide a torque in different operating states; - detecting an external torque signal indicative of the torque provided by the internal combustion engine (110); - determining a torque actually provided by the internal combustion engine (110); - detecting at least one parameter that influences the torque actually provided by the internal combustion engine (110); - Determining a sum of the external torque signal and a signal provided by the network; and - Training the artificial neural network to minimize any deviation between the sum and the actual torque delivered. [11] A method for training an artificial neural network, the method comprising the following steps: - operating a first internal combustion engine (110) configured to provide a torque in different operating states; - detecting an external torque signal indicative of the torque provided by the internal combustion engine (110); - determining a torque actually provided by the internal combustion engine (110); - detecting at least one parameter that influences the torque actually provided by the internal combustion engine (110); - Determining a sum of the external torque signal and a signal provided by the network; and - Training the artificial neural network to minimize any deviation between the sum and the actual torque delivered - operating a second internal combustion engine (110) (110) which is configured to provide a torque in different operating states; - detecting a second external torque signal indicative of the torque provided by the second internal combustion engine (110); - determining a torque actually provided by the second internal combustion engine (110); - detecting at least one parameter that influences the torque actually provided by the second internal combustion engine (110); and - training the artificial neural network to minimize a deviation between the external torque signal of the first internal combustion engine (110) and an external torque signal of the second internal combustion engine (110).

Images