Drive assembly and method for determining a torque

Abstract

The invention relates to a drive assembly (10), comprising an electric motor (1), a mechanical system (2) drivingly connected thereto, and a sensor device for determining a torque (M1) of the electric motor (1), characterised in that the sensor device is a virtual sensor (3) which estimates the torque (Mp) on the basis of evaluated signals (S) of the electric motor (1). The invention also relates to a method for determining a torque (Mp) on a drive assembly (10), in which a mechanical system (2) is drivingly connected to an electric motor (1) in order to be driven thereby. The method is characterised in that signals (S) of the electric motor (1) are evaluated by means of a virtual sensor (3) in order to estimate the torque (Mp) of the electric motor (1).

Description

[0001] ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0002] Drive arrangement and method for determining torque

[0003] The invention relates to a drive arrangement according to the preamble of claim 1, and to a method for determining a torque on a drive arrangement according to the preamble of claim 14.

[0004] Drive systems are known from various fields of engineering in which an electric motor is connected to a mechanical system. The electric motor typically serves to drive the mechanical system, enabling it to perform technical tasks. For example, the mechanical system might be a gearbox, or it could be part of a mechatronic actuator used to move one or more (additional) objects. Due to the drive connection between the electric motor and the mechanical system required for this purpose, interactions occur between the two. Depending on the application of the drive system, it may be advantageous to determine the torque of the electric motor, for example, to feed this torque into a control process for the electric motor (feedback).

[0005] It is known from the prior art to equip such drive arrangements with a sensor device for determining the torque of the electric motor. Depending on the specific application and the associated technical, especially design, conditions, so-called torque sensors are used, which measure the torque directly, for example, by detecting deformation of drive train components, such as using strain gauges or, for example, contactlessly by inverse magnetostriction. A disadvantage of direct measurement methods is that the sensor devices required to achieve the necessary accuracy are generally technically complex and / or prone to failure.

[0006] It is an object of the present invention to provide a drive arrangement of the aforementioned type, in which the torque of the electric motor is applied. ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0007] a method for determining the torque of a drive arrangement with reduced effort should be specified. Furthermore, a method for determining the torque of a drive arrangement should be specified, which can be implemented with relatively little design effort.

[0008] The aforementioned problem is initially solved by a drive arrangement according to the features of claim 1. This drive arrangement comprises an electric motor, a mechanical system connected to it for drive purposes, and a sensor device for determining the torque of the electric motor. According to the invention, the drive arrangement is characterized in that the sensor device is a virtual sensor that estimates the torque based on evaluated signals from the electric motor.

[0009] According to the invention, it was first recognized that previously known drive arrangements are equipped with sensor devices for determining the torque of the electric motor, which entail a high degree of design complexity and thus disadvantages, and / or a high susceptibility to malfunctions, for example, due to external influences such as temperature, aging, or the like. The use of a virtual sensor, as provided for in the invention, avoids the design complexity of a conventional torque sensor. Within the scope of the present invention, a virtual sensor is understood to be a sensor device that determines a target variable, in this specific case the torque of the electric motor, by simulating the dependency of one or more proxy measured variables, here: specifically evaluated signals from the electric motor, through an estimation process.Unlike a conventional sensor, the torque to be determined is not measured directly, but estimated using correlated measured variables and a correlation model. It turns out that such an estimation achieves sufficient accuracy for many applications. The problem mentioned at the outset is thus solved advantageously.

[0010] In principle, an estimation performed by the virtual sensor can be based on different measured variables, whereby different measured variables may be suitable depending on the type and design of the drive arrangement. ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0011] According to a preferred embodiment of the drive arrangement, the torque estimation is based solely on an evaluation of signals from the electric motor. In this case, the sensor device used in the drive arrangement requires minimal effort, since signals from the electric motor are often already available without any further design modifications, thus eliminating the need for additional sensors.

[0012] Preferably, the signals from the electric motor are one or more physical state variables of the electric motor.

[0013] Preferably, a signal from the electric motor includes an electrical state variable such as, in particular, a current and / or a voltage applied to the electric motor.

[0014] Alternatively or additionally, a signal from the electric motor includes a mechanical state variable of the electric motor such as, in particular, the angle of rotation and / or the rotational speed (of the motor shaft), wherein the electric motor can advantageously be assigned a sensor for detecting the angle of rotation and / or the rotational speed, and wherein the detected angle of rotation or the detected rotational speed is entered as an input variable into the virtual sensor.

[0015] To achieve a reliable estimate of the torque, the virtual sensor can advantageously be operated to store signals from the electric motor, preferably also their temporal progression, in order to perform an estimate of the torque based on a historical evaluation of the signals.

[0016] The virtual sensor particularly favorably features an artificial neural network that evaluates signals from the electric motor in order to estimate the torque.

[0017] The artificial neural network is particularly well-suited for learning the operational behavior of the mechanical system and / or the electric motor. ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0018] The ability to learn the system's behavior is made possible, in particular, by using historical signals from the motor. This learning process also allows for the learning of more complex processes, especially interactions that arise, for example, in complex mechanical systems, over a longer time horizon.

[0019] Advantageously, the artificial neural network can be operated based on operational processes to gain insights into, in particular, recurring behavioral patterns of the mechanical system and / or the electric motor, and to make these insights available for future purposes. Torque estimation using an artificial neural network is based on the understanding that it is possible to train the behavior of the drive arrangement using histons, since a mechanical system, due to its essentially constant mechanical behavior (e.g., damping behavior, elasticities, etc.), will behave the same way in identical situations. This results in predictability. Advantageously, this predictability also exists despite the possible presence of non-linearities within the mechanical system.Accordingly, torque estimation using a neural network offers particular advantages over sensor devices based on linear calculation models.

[0020] Advantageously, the artificial neural network has a large number of interconnected artificial neurons, which are assigned to an input layer, a hidden layer, and an output layer.

[0021] Advantageously, the signals from the electric motor are linked to artificial neurons in the input layer of the artificial neural network.

[0022] Furthermore, it is advantageously provided that the artificial neural network outputs information describing the estimated torque via its output layer. ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0023] The torque of the electric motor estimated by the virtual sensor according to the invention can be used for various purposes. In an advantageous embodiment of the drive arrangement, feedback of the estimated torque is provided, so that the electric motor can be controlled taking the estimated torque into account.

[0024] As mentioned at the outset, a drive arrangement described according to the invention can be used in various fields of technology. According to one possible embodiment, the mechanical system is a gearbox or a part of a mechatronic actuator with which one or more objects can be set in motion.

[0025] The aforementioned problem is further solved by a method according to the features of claim 14. This method is for determining the torque of a drive arrangement in which a mechanical system is connected to an electric motor in order to be driven by it. According to the invention, signals from the electric motor are evaluated by means of a virtual sensor in order to estimate the torque of the electric motor.

[0026] The invention is explained below with reference to the accompanying drawing. Further advantages and / or application possibilities of the invention also become apparent from this drawing. The drawing shows, in its only aspect:

[0027] Figure shows a schematic representation of a drive arrangement according to the invention.

[0028] The single figure shows a simplified schematic representation of a drive arrangement 10 to illustrate the invention. The drive arrangement 10 comprises, as essential elements, an electric motor 1, a mechanical system 2 connected to it, and a sensor device 3. The electric motor 1 is connected to the mechanical system 2 via a motor shaft 9. The mechanical system 2 can, in principle, be any type of mechanical system. (ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19)

[0029] The device in question is a mechanically driven unit, such as a gearbox in the broadest sense. Due to the drive connection between electric motor 1 and mechanical system 2 via the motor shaft 9, an (actual) torque M1, dependent on the operating state of the drive arrangement 10, is present at the motor shaft 9 of the electric motor 1.

[0030] The drive arrangement 10 is characterized by a sensor device for determining the torque of the electric motor 1, which is a so-called virtual sensor 3 that estimates a torque Mp based on evaluated signals S of the electric motor 1. As shown in the present illustration, various physical state variables of the electric motor 1 are listed as examples of signals S, including in particular electrical state variables such as the applied current l(t) and the applied voltage U(t), as well as mechanical state variables such as the angle of rotation. <p(t) und die Drehgeschwindigkeit cp‘(t) handelt. Durch gestrichelte Pfeile wird angedeutet, dass die genannten physikalischen Zustandsgrößen des Elektromotors 1 dem virtuellen Sensor 3 als Eingangsgrößen zugeführt werden.According to the example shown, the virtual sensor 3 has an artificial neural network 4 that evaluates signals S from the electric motor 1 in order to estimate the torque Mp.

[0031] The artificial neural network 4 comprises a multitude of interconnected artificial neurons assigned to an input layer 5, a hidden layer 6, and an output layer 7. The signals S of the electric motor 1, which are summarized as the exemplary physical state variables current, voltage, angle of rotation, and rotational speed of the electric motor 1, are linked to artificial neurons 8 of the input layer 5 of the artificial neural network 4. Via its output layer 7, the artificial neural network 4 outputs information describing the estimated torque Mp. The estimated torque Mp is based on an estimation made by the virtual sensor 3 through the evaluation of the signals S. Ideally, the estimated torque Mp corresponds to the actual torque M1 of the electric motor.

[0032] Tors 1. At least, however, the estimated torque Mp represents a relatively good approximation to the actual torque M1.

[0033] The mechanical state variables of the electric motor 1, such as the angle of rotation <p(t) und die Drehgeschwindigkeit cp‘(t) werden dadurch erhalten, dass dem Elektromotor 1 ein (hier nicht dargestellter) Sensor zu Erfassung des Drehwinkels <p(t) und der Drehgeschwindigkeit <p‘(t) zugeordnet ist.

[0034] The virtual sensor 3 is capable of storing signals S of the electric motor 1, including their temporal progression, thereby enabling an estimation of the torque Mp based on a historical evaluation of the signals S.

[0035] The artificial neural network 4 of the virtual sensor 3 can be used to learn the operational behavior of the mechanical system 2 and / or the electric motor 1. The ability to use historical signals S allows the system behavior to be learned solely from motor signals S. This offers the advantage that no complex torque sensors, such as those on the motor shaft 9 or other components of the drive assembly 10, are required to determine the torque of the electric motor 1. Thanks to its learning capability, the virtual sensor 3 allows for an increasingly better understanding of the system behavior, particularly based on recurring patterns, so that this knowledge can be applied to future operating situations.

[0036] A particular advantage is that even complex processes within the drive system – for example, due to the presence of non-linearities – can be trained with a time horizon, thus enabling torque estimation with sufficiently good accuracy and speed even for non-linear processes. As the estimator generated by the artificial neural network learns (trains), the quality and speed of the torque estimation can be improved. ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19

[0037] The torque Mp estimated in the described manner can be used for various purposes. According to one advantageous application, the estimated torque Mp is fed back into the system so that the electric motor 1 can be controlled taking the estimated torque Mp into account.

[0038] Finally, it should be noted that the drive arrangement can be used in various fields of technology. According to a preferred application, the drive arrangement is part of a mechatronic actuator that can be used to set one or more objects in motion. ZF Friedrichshafen AG File 305459

[0039] Friedrichshafen 2025-02-19

[0040] Reference mark

[0041] 1 electric motor

[0042] 2 mechanical system

[0043] 3 virtual sensors

[0044] 4 artificial neural network

[0045] 5 Entrance layer

[0046] 6 hidden layer

[0047] 7 Initial layer

[0048] 8 artificial neuron

[0049] 9 Motor shaft

[0050] 10 Drive arrangement

[0051] M1 torque (actual)

[0052] MP estimated torque

[0053] S signals

[0054] U(t) voltage

[0055] l(t) current

[0056] <p(t) Drehwinkel

[0057] cp'(t) rotational speed

Claims

ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19 Patent claims 1. Drive arrangement (10) comprising an electric motor (1), a mechanical system (2) connected to it for drive purposes, and a sensor device for determining a torque (M1) of the electric motor (1), characterized in that the sensor device is a virtual sensor (3) which estimates the torque (Mp) on the basis of evaluated signals (S) of the electric motor (1).

2. Drive arrangement according to claim 1, characterized in that the estimation of the torque (Mp) is based exclusively on an evaluation of signals (S) from the electric motor (1).

3. Drive arrangement according to claim 1 or 2, characterized in that the signals (S) of the electric motor (1) are one or more physical state variables (U(t), l(t), <p(t), cp‘(t)) des Elektromotors (1) handelt.

4. Drive arrangement according to one of the preceding claims, characterized in that a signal (S) of the electric motor (1) includes an electrical state variable such as, in particular, a current (l(t)) applied to the electric motor (1) and / or a voltage (U(t)) applied to the electric motor (1).

5. Drive arrangement according to one of the preceding claims, characterized in that a signal (S) of the electric motor (1) is a mechanical state variable of the electric motor (1) such as, in particular, the angle of rotation (<p(t)) und / oder Drehgeschwindigkeit (q> '(t)) includes, wherein the electric motor (1) is advantageously equipped with a sensor for detecting the angle of rotation ( <p(t)) und / oder der Drehgeschwindigkeit (cp‘(t)) zugeordnet ist, und wobei der erfasste Drehwinkel (cp(t)) bzw. die erfasste Drehgeschwindigkeit (q> '(t)) each as an input variable into the virtual sensor (3).

6. Drive arrangement according to one of the preceding claims, characterized in that the virtual sensor (3) is operable to store signals (S) of the electric motor (1), preferably also their temporal progression, in order to estimate ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19 to perform a determination of the torque (Mp) based on a historical evaluation of the signals (S).

7. Drive arrangement according to one of the preceding claims, characterized in that the virtual sensor (3) has an artificial neural network (4) which evaluates signals (S) of the electric motor (1) in order to estimate the torque (Mp).

8. Drive arrangement according to claim 7, characterized in that the artificial neural network (4) is operable to learn the operational behavior of the mechanical system (2) and / or the electric motor (1), preferably by the artificial neural network (4) gaining insights into, in particular, recurring behavioral patterns based on operational processes and making them available for future purposes.

9. Drive arrangement according to claim 7 or 8, characterized in that the artificial neural (4) network comprises a plurality of interconnected artificial neurons (8) which are assigned to an input layer (5), a hidden layer (6) and an output layer (7).

10. Drive arrangement according to claim 9, characterized in that the signals (S) of the electric motor (1) are linked with artificial neurons (8) of the input layer (5) of the artificial neural network (4).

11. Drive arrangement according to claim 9 or 10, characterized in that the artificial neural network (4) outputs information describing the estimated torque (Mp) via its output layer (7).

12. Drive arrangement according to one of the preceding claims, characterized by a feedback of the estimated torque (Mp) so that the electric motor (1) can be controlled taking into account the estimated torque (Mp).

13. Drive arrangement according to one of the preceding claims, characterized in that the mechanical system (2) is a transmission or part of a mechatro-ZF Friedrichshafen AG File 305459 Friedrichshafen 2025-02-19 A niche actuator is a device that can be used to set one or more objects in motion.

14. Method for determining a torque (Mp) on a drive arrangement (10), in particular according to one of the preceding claims, wherein a mechanical system (2) is in drive connection with an electric motor (1) in order to be driven by it, characterized in that signals (S) of the electric motor (1) are evaluated by means of a virtual sensor (3) in order to estimate the torque (Mp) of the electric motor (1).

Images