Method for determining torque

Abstract

The present invention relates to a method for determining a torque (T) of a drivetrain (1) by means of a mathematical model (9). The method may include taking in input data (E). The input data (E) may include a voltage (U) measured using a torque sensor (5). The input data (E) may also include at least one signal value (Ω, Ψ, I) of a motor controller (4). Furthermore, the method may include determining the torque (T) by means of the mathematical model (9) based on the voltage (U) and the at least one signal value (Ω, Ψ, I). The present invention also relates to a machine learning model for determining the torque (T) of a drivetrain (1).

Description

[Technical field]

[0001] The present invention relates generally to the technical field of methods for determining the torque of a driven device, in particular of a drive train. [Background technology]

[0002] Methods for determining torque have many applications in the field of drive technology. For example, the torque of a drive train is measured or determined by calculation in order to improve the driving behavior of a car for passenger transport. A drive train typically comprises a motor with a motor shaft, a motor controller, a torque sensor with a microcontroller, and a transmission with an output shaft. The term "torque" in the sense of the present invention is understood by those skilled in the art as referring to the torque applied to the output shaft of the drive train. In addition to the field of automotive engineering, torque is also an important parameter in the field of robotics. Industrial robots in particular often require high torques to move the joints of multi-element manipulators with sufficient force.

[0003] Methods known to date for determining the torque of a drive train typically consist of mounting a torque sensor on the output shaft of the drive train. Such torque sensors are generally formed by mounting strain gauges on components of the transmission device, in particular on elastic transmission elements with external toothing, and determining the torque on the basis of input data, which are measured by the strain gauges. Such methods are described, for example, in the following scientific publications:

[0004] Hashimoto, Minoru, Yoshihide Kiyosawa, and Richard P. Paul. “A torque sensing technique for robots with harmonic drives.” IEEE Transactions on Robotics and Automation 9.1 (1993): 108-116.

[0005] Hashimoto, Minoru, and Yoshihide Kiyosawa. “Experimental study on torque control using Harmonic Drive built-in torque sensors.” Journal of Robotic Systems 15.8 (1998): 435-445.

[0006] Other known methods for determining the torque are based on mathematical formulas derived from the physical behavior of the drive train. Such a method for determining the torque is disclosed, for example, in DE 19963279 B4, entitled "Method for determining a output torque of a drive device and drive device". Here, the output torque is derived by means of a mathematical formula in which the lubricant temperature of the transmission device, the rotational speed of the motor and the output torque of the motor are used as input data. No torque sensor is intended to be used here.

[0007] Although the above-described approach allows the torque of the drivetrain to be determined at least as a rough approximation, there are various interfering variables acting on the drivetrain, such as inertial and temperature effects, that are not accounted for by known methods and that result in an inaccurate determination of the torque. Summary of the Invention [Problem to be solved by the invention]

[0008] It is therefore an object of the present invention to provide an improved method for determining torque in a drive train, thereby at least partially overcoming the above-mentioned shortcomings of the prior art. It is a further object of the present invention to provide an improved mathematical model for determining torque. [Means for solving the problem]

[0009] These objects are achieved by the features of the independent claims. In its most general form, the invention relates to a method for determining a torque of a driven device, in particular of a drive train. In this respect, the method comprises obtaining input data. The input data may comprise a voltage measured using a torque sensor. The input data may also comprise at least one signal value of a motor controller. Furthermore, the method may comprise determining the torque by means of a mathematical model based on the voltage and the at least one signal value.

[0010] The drive train is preferably part of a vehicle or a robot and may comprise a motor having a motor shaft, a motor controller, a torque sensor having a microcontroller, and a transmission having an output shaft. The motor may be an internal combustion engine, an electric motor, a fuel cell, a hybrid drive, or another type of energy converter.

[0011] Here, drive train torque is understood to be the torque available to drive a vehicle, a robot, or a manipulator element of a robot. The term "torque" is understood by those skilled in the art to refer to the torque applied to the output shaft of the drive train. Thus, drive train torque is the output torque of a transmission device. For example, drive train torque can be directly transmitted to the wheels of a vehicle.

[0012] "Determination of the torque by means of a mathematical model" is typically understood by those skilled in the art as any kind of checking and / or calculation in a mathematical sense. For this purpose, any type of mathematical operation can be used, in particular iterative methods or the solution of differential equations. It is also possible to apply a reading from a table to determine the torque.

[0013] Some or all of the method steps may be performed by (or using) a hardware device, such as a microcontroller of the torque sensor. The hardware device may also comprise a processor, a microprocessor, a programmable computer, or an electronic circuit. In some exemplary embodiments, one or more of the key method steps may be performed by such a device.

[0014] Naturally, the method can be fully automated, it is also possible to carry out the method continuously in real time, as is common in the field of motor technology.

[0015] Typically, those skilled in the art will understand that input data is data that is made available to a mathematical model for determining torque.

[0016] A motor controller is typically used for open-loop and / or closed-loop control of a motor. In this regard, the motor controller can read, measure and / or store all signal values ​​related to the motor. The motor controller can have means for reading, measuring and / or storing the signal values. The signal values ​​can be electrical signals such as voltage and / or current. The signal values ​​can also be pressure, temperature, rotational speed, acceleration and / or speed. Depending on the application, the motor controller can have corresponding access to sensors such as temperature and / or pressure sensors or can communicate the signal values ​​from the sensors to the motor controller. It is also possible for the motor controller to measure multiple signal values ​​simultaneously and communicate them to other components of the drive train.

[0017] The motor controller may also be configured to control the motor and / or other components of the driven device. In this regard, the motor controller may, for example, control the speed of the vehicle. This type of closed loop control may be achieved here using a control loop implemented as an algorithm.

[0018] The torque sensor is preferably equipped with a microcontroller, in which respect the mathematical model or at least a part of the mathematical model may be implemented on the microcontroller.

[0019] The transmission or gear used may be a spur gear, a bevel gear, a helical gear, a planetary gear or a harmonic drive transmission.

[0020] The method according to the invention advantageously results in an improvement of the measurement method implemented by the torque sensor. In addition to the voltage measured by the torque sensor, additional data from the motor controller can now be made available to the method in order to determine the torque of the drive train even more accurately. Furthermore, the method according to the invention has the advantage that input data from various components of the drive train, in particular the motor controller and the torque sensor, can be made available to the mathematical model simultaneously. In other words, the measured signal values ​​from different sensors can be merged (multi-sensor fusion) and further processed by the mathematical model.

[0021] In a further aspect of the invention, the at least one signal value of the motor controller includes a motor position and / or a direction of rotation. In this regard, the motor position refers to the rotation angle of the motor shaft exiting the motor. For example, the motor shaft may initially have a rotation angle of about 0°. After half a rotation of the motor, the rotation angle is essentially 180°. In this regard, the direction of rotation refers to the direction of rotation of the motor shaft. The direction of rotation may be clockwise or counterclockwise. The motor position and the direction of rotation have the advantage that they are preferably available as stored data in the motor controller or are measured by the motor controller.

[0022] In a further aspect, the input data includes temperature. In this regard, the temperature may be a transmission lubricant temperature or a temperature measured in a motor. Advantageously, the method is thereby able to detect temperature effects that may affect the torque of the drive train.

[0023] In a further aspect, the mathematical model comprises a finite element model, which has the advantage that any complex geometry of the drivetrain and its physical behavior can be represented based on differential equations and continuum theory, and nonlinearities related to materials and / or geometry can be easily represented.

[0024] In a further aspect, the mathematical model comprises a machine learning model. In this regard, the machine learning model may in particular be an artificial neural network.

[0025] An artificial neural network (ANN) is a system inspired by biological neural networks such as those found in the retina or the brain. An ANN contains multiple interconnected nodes and multiple connections between the nodes, so-called edges. There are usually three types of nodes: input nodes that receive input values, hidden nodes that are connected (only) to other nodes, and output nodes that provide output values. Every node can represent an artificial neuron. Every edge can transmit information from one node to another. The output of a node can be defined as a (non-linear) function of its inputs (e.g. the sum of its inputs). The inputs of a node can be used in a function based on the "weights" of the edges or nodes that provide the inputs. The weights of the nodes and / or edges can be adjusted as part of the learning process. In other words, training an artificial neural network can include adjusting the weights of the nodes and / or edges of the artificial neural network, i.e. to achieve a desired output for a particular input.

[0026] Alternatively, the machine learning model may be a support vector machine, a random forest model, or a gradient boosting model. A support vector machine (i.e., a support vector network) is a supervised learning model with an associated learning algorithm that can be used to analyze data (e.g., in classification or regression analysis). A support vector machine can be trained by providing inputs with multiple training input values ​​that belong to one of two categories. A support vector machine can be trained to assign new input values ​​to one of two categories. Alternatively, the machine learning model may be a Bayesian network, which is a probabilistic directed acyclic graphical model. A Bayesian network can use a directed acyclic graph to represent a set of random variables and their conditional dependencies. Alternatively, the machine learning model can be based on a genetic algorithm, which is a search algorithm and heuristic technique that mimics the process of natural selection.

[0027] In a further aspect, the machine learning model is trained using at least one training data set, which can include a voltage measured using the torque sensor, at least one signal value of the motor controller, and a reference torque.

[0028] The training data set can be created, for example, on a test bench, in which the voltage measured using the torque sensor, at least one signal value of the motor controller, and the reference torque are measured on the test bench at various operating conditions of the drive train.

[0029] For example, supervised learning can be used as a learning method for machine learning models. In supervised learning, the machine learning model is trained using multiple training sample values, each of which may include multiple input data values ​​and multiple desired output values, i.e., each training sample value is associated with a desired output value. By specifying both the training sample values ​​and the desired output value (here, the reference torque mentioned above), the machine learning model "learns" which output value should be provided based on input sample values ​​that are similar to the sample values ​​provided as part of the training. In addition to supervised learning, semi-supervised learning can also be used. In semi-supervised learning, some of the training sample values ​​lack the desired output value. Supervised learning can be based on supervised learning algorithms (e.g., classification algorithms, regression algorithms, or similarity learning algorithms). Classification algorithms can be used when the output is restricted to a limited set of values ​​(categorical variables), i.e., when the input is classified as one value out of a limited set of values. Regression algorithms can be used when the output indicates any numerical value (within a range). Similarity learning algorithms can be similar to both classification and regression algorithms, but are based on learning from examples using a similarity function that measures how similar or related two objects are. In addition to supervised or semi-supervised learning, unsupervised learning can be used to train machine learning models. In unsupervised learning, input data (only) may be provided, and unsupervised learning algorithms can be used to find structure in the input data (e.g., by grouping or clustering the input data and finding commonalities in the data). Clustering is the assignment of input data containing multiple input values ​​into subsets (clusters) such that input values ​​in the same cluster are similar according to one or more (predetermined) similarity criteria, but dissimilar to input values ​​contained in other clusters.

[0030] In a further aspect, the mathematical model is implemented on a microcontroller of the torque sensor. One advantage of this embodiment is that determining the torque by means of the mathematical model can be performed entirely on the microcontroller and no additional components are required to implement the mathematical model.

[0031] In a further aspect, the torque determined by the mathematical model is transferred to a control loop. In this regard, a control loop may be provided for controlling the current and / or the speed of the motor. Advantageously, a closed-loop control based on the control loop may be improved in this way.

[0032] In a further aspect, the method for plausibility check comprises further method steps. In this respect, in a further step, an estimated torque can be checked based on the motor current. Furthermore, an upper torque limit and / or a lower torque limit can be checked based on the estimated torque. The torque determined by the mathematical model can then be evaluated based on the upper torque limit and / or the lower torque limit. Here, the motor current can be read out as a signal value by the motor controller.

[0033] The estimated torque can be ascertained, for example, using the following formula, where the torque constant and the transmission ratio of the transmission device have been previously ascertained by the motor controller:

[0034] Estimated torque = motor current * torque constant * transmission ratio of the transmission device

[0035] The torque estimate determined by the mathematical model can be checked, for example, by the following conditions:

[0036] Estimated torque - Δ < torque determined by mathematical model < estimated torque + Δ

[0037] where Δ is a calibration parameter that takes into account the current accuracy level of the estimated torque. The parameter Δ can be predefined and stored in the microcontroller of the torque sensor. As long as the torque determined by the mathematical model is within the range of "estimated torque (current based) ± Δ", it is possible to evaluate whether the torque determined by the mathematical model is reasonable. If the torque determined by the mathematical model is out of range, an "error" condition is output as a diagnostic condition or status.

[0038] In a further aspect, there is provided an apparatus for data processing, in particular a microcontroller of a torque sensor, comprising means for performing one of the methods described herein.

[0039] An apparatus for data processing may comprise a storage medium (or a data carrier or a computer readable medium) on which a computer program for performing one of the methods described herein when executed by a processor is stored. The data carrier, digital storage medium or recording medium is typically tangible and / or non-transitory. A further exemplary embodiment of the invention is an apparatus as described herein, comprising a processor and a storage medium.

[0040] In a further embodiment, the present invention provides a computer program comprising instructions which, when the program is executed by a computer, cause the computer to perform one of the methods described herein.

[0041] According to the present invention, there is also provided a machine learning model for determining a torque of a drive train. In this regard, the machine learning model may have input means for receiving a voltage measured using a torque sensor and at least one signal value of a motor controller. Furthermore, the machine learning model may have means for determining a torque based on the voltage and the at least one signal value.

[0042] A machine learning model can refer to algorithms and statistical models that a computer system can use to perform a certain task without explicit instructions, rather than relying on models and inference. Machine learning can use, for example, transformations of data that can be derived from analysis of historical data and / or training data, instead of rule-based transformations of data. For example, the physical behavior of a drivetrain can be analyzed using a machine learning model or using a machine learning algorithm. To enable the machine learning model to analyze the physical behavior of the drivetrain, the machine learning model can be trained using a training dataset as input and training content information as output. By training the machine learning model with multiple training datasets and / or training sequences (e.g., signal values) and associated training content information (e.g., labels or annotations), the machine learning model "learns" to recognize the physical behavior of the drivetrain, such that the machine learning model can be used to predict physical behaviors not included in the training data. The same principle can be used for other types of sensor data. By training the machine learning model using training sensor data and a desired output, the machine learning model "learns" the transformation between sensor data and output, and this transformation can be used to provide an output based on the non-training sensor data provided to the machine learning model. The provided data (e.g., sensor data and / or metadata) can be pre-processed to obtain feature vectors, which are used as inputs for machine learning models.

[0043] In a further embodiment, the present invention provides a training data set for training a machine learning model, wherein the at least one training data set can include a voltage measured using a torque sensor, at least one signal value of a motor controller, and a reference torque.

[0044] Preferred embodiments of the present disclosure are described below with reference to the following figures. [Brief description of the drawings]

[0045] [Figure 1] 1 shows a drive train 1 according to an embodiment of the invention. [Diagram 2] 1 shows an overview of a method according to an embodiment of the present invention. [Figure 3A] 1 shows a mathematical model 9 according to an embodiment of the present invention. [Figure 3B] A motor controller 4 and a torque sensor 5 according to an embodiment of the present invention are shown. [Figure 4] 1 illustrates a method for validity checking according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0046] Below, an embodiment of the invention is described in which a method for determining the torque of a drivetrain is used to make the drivetrain particularly efficient.

[0047] Figure 1 shows a drive train 1 according to an embodiment of the invention. In the illustrated embodiment, the drive train 1 is used in a passenger vehicle and comprises a motor 2 having a motor shaft 3, a motor controller 4, a torque sensor 5 having a microcontroller, a data transmission system 6 and a transmission 7 having an output shaft 8. The motor shaft 3 of the motor 2 is connected to the transmission 7, thereby transmitting the motor torque generated by the motor 2 to the transmission 7. The transmission 7 is a harmonic drive transmission with an elastic transmission element characterized by a high transmission ratio and stiffness. The elastic transmission element is also called a flexspline. The transmission 7 further comprises an output shaft 8, the torque of which is transmitted to the wheels of the passenger vehicle.

[0048] The torque sensor 5 comprises a strain gauge, which consists of a semiconductor and measures the voltage applied to a resistor based on the piezoresistive effect, where a change in resistance and voltage is caused by a deformation of the strain gauge. The advantage of the strain gauge used compared to conventional metal strain gauges is its high sensitivity.

[0049] The strain gauges of the torque sensor 5 are arranged on an elastic transmission element of the transmission device 7. The torque sensor 5 is thus integrated in the transmission device 7. The torque sensor 5 is connected to the motor controller 4 via a data transmission system 6. The motor controller 4 measures signal values ​​related to the motor 2 and transmits these signal values ​​to the torque sensor 5.

[0050] The method for determining the torque is implemented in the form of a computer program on the microcontroller of the torque sensor 5. Once the torque has been determined, it is transmitted to the motor controller 4 via a data transmission system 6.

[0051] 2 shows a schematic diagram of a method according to an embodiment of the present invention. The method according to this embodiment includes the following steps:

[0052] First step S1.1: The torque sensor 5 measures the voltage and passes it to the microcontroller.

[0053] Second step S1.2: The motor controller 4 passes the signal value to the microcontroller.

[0054] Third step S1.3: A mathematical model 9 implemented on a microcontroller determines the torque based on the voltage and signal values.

[0055] It is understood that the order of the mentioned steps S1.1, S1.2 and S1.3 may vary depending on the application. For example, the signal value of the motor controller 4 may be sent to the microcontroller first, and then the voltage may be measured by the torque sensor 5 and sent to the microcontroller. It is also possible to perform the first step S1.1 and the second step S1.2 simultaneously.

[0056] After determining the torque, it can be transmitted to the motor controller 4 in a further step not shown in the exemplary embodiment.

[0057] Fig. 3A shows a mathematical model 9 according to an embodiment of the present invention. Various signal values ​​are transmitted to the mathematical model 9 as input data E, which in the illustrated embodiment are the voltage U measured by the torque sensor 5, the temperature Θ of the transmission device, the position Ω of the motor, and the direction of rotation Ψ of the motor. The mathematical model 9 calculates the torque T of the output shaft 8 of the drive train 1 based on the four signal values ​​U, Θ, Ω, Ψ. The calculated torque T is provided as a signal value for further use. Here, the mathematical model 9 is an artificial neural network trained using training data before being used in the drive train 1. The calculation of the torque T by the artificial neural network and the components of the drive train 1 is optimally adapted to the stringent real-time requirements required in the field of motor technology. In this respect, the calculation times are within the usual range of the application of the motor controller 4.

[0058] 3B shows a motor controller 4 and a torque sensor 5 according to a further embodiment of the invention. The motor controller 4 and the torque sensor 5 are connected to each other via a data transmission system 6 and can therefore transmit signal values ​​Ω, Ψ, I, T and Z to each other. In the shown example, the motor controller 4 transmits the motor position Ω, the direction of motor rotation Ψ and the motor current I to the torque sensor 5. The torque sensor 5 transmits the torque T of the output shaft 8 of the drive train 1 and the status Z of the torque sensor 5 to the motor controller 4.

[0059] In the illustrated embodiment, the data transmission system 6 is an EtherCAT system. The data transmission system 6 thereby advantageously meets real-time requirements. The motor controller 4 and the torque sensor 5 can transmit the signal values ​​Ω, Ψ, I, T and Z to each other or exchange them synchronously. Furthermore, the signal values ​​Ω, Ψ and I can be transmitted from the motor controller 4 to the torque sensor 5, where they can be used directly, for example, to calculate the torque T. The torque T can then be transmitted directly back to the motor controller 4.

[0060] 4 shows a method for plausibility checking according to an embodiment of the present invention. In this respect, all method steps S2.1, S2.2, S2.3, S2.4 and S2.5 are executed in a microcontroller belonging to the torque sensor 5.

[0061] In a first step S2.1, the microcontroller receives signal values ​​Ω, Ψ and I from the motor controller 4 via the data transmission system 6, where the signal values ​​include the motor current I. In a second step S2.2, the microcontroller calculates the torque T by means of a mathematical model 9 implemented on the microcontroller and based on the signal values ​​Ω, Ψ and I. Furthermore, in the second step S2.2, the microcontroller checks the estimated torque based on the motor current I. The microcontroller also checks an upper torque limit and a lower torque limit based on the estimated torque. In a third step S2.3, the microcontroller evaluates the torque T determined by the mathematical model 9 based on the upper torque limit and the lower torque limit.

[0062] If the torque T ascertained by the mathematical model 9 is outside the torque limits, i.e. is greater than the value defined by the upper torque limit or less than the value defined by the lower torque limit, the microcontroller sets the status Z to "error" in a further step S2.4.

[0063] If the torque T as checked by the mathematical model 9 is within the torque limits, i.e. smaller than the value defined by the upper torque limit and at the same time larger than the value defined by the lower torque limit, the microcontroller sets the status Z to "valid" in a further step S2.5.

[0064] In a further step, the status Z and the torque T can be transmitted by a data transmission system 6 to the motor controller 4. Thereby, the method shown in Fig. 4 is used for a technical plausibility check of the torque T verified by the mathematical model 9. If the status Z is output as "error", the torque T is not used for further calculations in the motor controller 4.

[0065] It should be mentioned that although embodiments of the present invention are primarily described for use in passenger cars, the principles disclosed herein may nevertheless be implemented in other fields of application, for example in heavy goods vehicles, trucks, motorcycles, axle drives for robots, and drives for flight simulators.

[0066] The algorithms or computer programs required for the above functionality may conveniently be implemented in whole or in part in the torque sensor 5, the microcontroller, the motor controller 4, and / or in another computer system connected to these devices. The computer system may be a local computing device (e.g. a personal computer, laptop, tablet computer, or mobile phone) having one or more processors and one or more storage devices, or may be a distributed computing system (e.g. a cloud computing system having one or more processors or one or more storage devices distributed across various locations, e.g. local clients and / or one or more remote server farms and / or data centers).

[0067] Depending on certain implementation requirements, the exemplary embodiments of the invention can be implemented using hardware or software. Some or all of the method steps can be executed by (or using) hardware apparatuses such as processors, microprocessors, programmable computers, or electronic circuits.

[0068] Although some aspects have been described in the context of an apparatus, it will be apparent that these aspects also constitute a description of a corresponding method, where a block or apparatus corresponds to a method step or a function of a method step. Similarly, aspects described in the context of a method step also represent a description of a corresponding block or element, or feature of a corresponding apparatus.

Claims

1. A method for determining the torque (T) of a driven device (1), in particular of a drive train, said method comprising the following method steps: and acquiring input data (E), the input data (E) comprising: A voltage (U) measured using a torque sensor (5); At least one signal value (Ω, Ψ, I) of the motor controller; determining said torque (T) by a mathematical model (9) based on said voltage (U) and said at least one signal value (Ω, Ψ, I).

2. 2. The method of claim 1, wherein the at least one signal value (Ω, Ψ, I) of the motor controller (4) comprises a motor position (Ω) and / or a direction of rotation (Ψ).

3. The method according to claim 1 or 2, wherein the input data (E) further comprises a temperature (Θ).

4. The method according to any one of claims 1 to 3, wherein the mathematical model (9) comprises a finite element model.

5. The method according to any one of claims 1 to 3, wherein the mathematical model (9) comprises a machine learning model, in particular an artificial neural network.

6. The machine learning model is trained using at least one training dataset, the at least one training dataset comprising: A voltage (U) measured using a torque sensor (5); At least one signal value (Ω, Ψ, I) of the motor controller; A reference torque.

7. The method according to any one of the preceding claims, wherein the mathematical model (9) is implemented on a microcontroller of the torque sensor (5).

8. The method according to any one of the preceding claims, wherein the torque (9) determined by the mathematical model (9) is transmitted to a control loop, which controls the current (I) and / or the speed of a motor.

9. The method for validity checking comprises the following further method steps: Ascertaining an estimated torque based on a motor current (I); determining an upper torque limit and a lower torque limit based on the estimated torque; Evaluating the torque (T) determined by the mathematical model (9) based on the upper torque limit and the lower torque limit.

10. A device for data processing, in particular a microcontroller of a torque sensor (5), comprising means for carrying out the method according to any one of the claims 1 to 9.

11. A computer program comprising commands which, when said program is executed by a computer, cause said computer to carry out a method according to any one of claims 1 to 10.

12. 1. A machine learning model for determining a torque (T) of a drivetrain (1), the machine learning model comprising: A voltage (U) measured using a torque sensor (5); input means for receiving at least one signal value (Ω, ψ, I) of the motor controller; and means for determining the torque (T) based on the voltage (U) and the at least one signal value (Ω, Ψ, I).

13. A training data set for training a machine learning model according to claim 12, wherein at least one of the training data sets comprises: A voltage (U) measured using a torque sensor (5); At least one signal value (Ω, Ψ, I) of the motor controller; A training data set including:

Images