Device and method for monitoring a steering system of a vehicle, and a vehicle comprising the device

Abstract

The invention relates to a device (104) and method for monitoring a steering system (102) of a vehicle (100), wherein structure-borne sound is detected in the vehicle (100) by means of an acceleration sensor (108), wherein sensor data of the acceleration sensor (108) represent the detected structure-borne sound, wherein the sensor data are evaluated by means of a classifier (110) based in particular on artificial intelligence, wherein the classifier (110) is designed to classify the structure-borne sound as normal or abnormal for the steering system (102) of the vehicle (100) depending on the sensor data. The invention also relates to a vehicle (100) comprising the device (104).

Description

[0001] Description

[0002] title

[0003] DEVICE AND METHOD FOR MONITORING A STEERING SYSTEM OF A VEHICLE AND VEHICLE COMPRISING THE DEVICE

[0004] State of the art

[0005] The invention relates to a device and a method for monitoring a steering system of a vehicle, and to a vehicle comprising the device.

[0006] The detection of anomalies in a steering system is very complex and must therefore be secured through complex manual test series and applications of the steering system to identify anomalies.

[0007] Disclosure of the invention

[0008] A method for monitoring a vehicle's steering system provides for detecting structure-borne noise in the vehicle using an acceleration sensor. Sensor data from the acceleration sensor represent the detected structure-borne noise. The sensor data are evaluated using a classifier, particularly one based on artificial intelligence. The classifier is configured to classify the structure-borne noise as normal or abnormal for the vehicle's steering system depending on the sensor data. This reliably distinguishes whether or not an anomaly exists in the steering system based on the structure-borne noise in the vehicle.

[0009] The classifier is preferably configured to detect a pattern in the vibrations, depending on vibrations in the sensor data, which classifies the structure-borne sound as abnormal. The classifier is preferably configured to evaluate other sensor data in addition to the sensor data from the acceleration sensor, in particular tire pressure or wheel speed of at least one, preferably steered, wheel of the vehicle, steering angle, steering wheel position, steering torque, vehicle longitudinal acceleration, vehicle lateral acceleration, vehicle vertical acceleration, and vehicle yaw rate. This further improves the quality of the classification using the classifier.

[0010] It may be provided that a user is warned if the structure-borne noise is classified as abnormal for the vehicle's steering system.

[0011] It may be provided that a countermeasure is taken if the structure-borne noise is classified as abnormal for the vehicle's steering system.

[0012] It can be provided that a driving function, in particular a driving situation detection or a route detection, is informed about the structure-borne noise that is abnormal for the steering system of the vehicle if the structure-borne noise is classified as abnormal for the steering system of the vehicle.

[0013] Preferably, the classifier is trained based on training data that includes sensor data from structure-borne sound recorded with an acceleration sensor in a vehicle. This significantly reduces the effort required to provide the monitoring system, as the otherwise required extensive test series and calibration of the steering system to identify anomalies are avoided.

[0014] A device for monitoring a vehicle's steering system is configured to carry out the method. The device has advantages that correspond to those of the method.

[0015] A vehicle comprising the steering system and the device has the advantages of the device.

[0016] Further advantageous embodiments can be seen from the following description and the drawing. In the drawing: Fig. 1 shows a schematic representation of a vehicle,

[0017] Fig. 2 is a flowchart showing steps of a method for monitoring a steering system of the vehicle.

[0018] Figure 1 shows a schematic representation of a vehicle 100.

[0019] The vehicle 100 includes a steering system 102 and a device 104 for monitoring the steering system 102.

[0020] The vehicle 100 includes wheels 106. The vehicle 100 in the example has four wheels 106. The vehicle 100 can also be a two-wheeler or have more than four wheels 106.

[0021] In the example, the steering system 102 is configured to steer the wheels 106 on the front axle of the vehicle 100. It may also be provided to steer the wheels 106 on the rear axle of the vehicle 100 with the steering system 102.

[0022] The vehicle 100 includes an acceleration sensor 108. The acceleration sensor 108 is configured to detect structure-borne sound in the vehicle 100. The acceleration sensor 108 is configured to provide sensor data from the acceleration sensor 108 that represents the detected structure-borne sound.

[0023] It may be provided that the acceleration sensor 108 is arranged in the steering system 102. An installation position of the acceleration sensor 108 is preferred that is close to the source of the structure-borne noise generated by the steering system 102.

[0024] For example, the installation position is close to the wheel or axle on the steering system 102. Particularly for independent wheel steering, the installation position is preferably within the steered wheels 106. For steering the wheels 106 with a steering rack, particularly in hydraulic or electromechanical steering systems, the installation position is preferably on the steering rack. The installation position can also be on a control unit of the steering system 102 or of the vehicle 100, which is coupled to the steering system 102. The device 104 for monitoring the steering system 102 comprises a classifier 110, which is based, in particular, on artificial intelligence. The classifier 110 is designed to evaluate the sensor data.

[0025] The vehicle 100 includes a human-machine interface 112. In the example, the human-machine interface 112 is configured to output speech, text, warning signals, or sound. The device 104 is configured to control the human-machine interface 112.

[0026] The classifier 110 is configured to classify the structure-borne sound as normal or abnormal for the steering system 102 of the vehicle 100 depending on the sensor data.

[0027] It can be provided that the classifier 110 is configured to classify the sensor signals according to the severity of the anomaly. This means that if an anomaly is detected, ie, if structure-borne noise is classified as anomalous for the vehicle's steering system 102, the classifier 110 also outputs the severity of the anomaly.

[0028] It can be provided that the classifier 110 is configured to classify the sensor signals, depending on the frequency of the respective sensor signals, into an abnormal frequency and a normal frequency for the steering system 102. It can be provided that the classifier 110 is configured to determine and delimit a frequency band in order to avoid false triggering due to frequencies outside the specific frequency band. It can be provided that the classifier 110 is configured to classify the duration of the anomaly in order to avoid false triggering if the duration is too short.

[0029] It may be provided that the classifier 110 is configured to classify the structure-borne sound as normal or abnormal for the steering system 102 depending on the sensor data from the acceleration sensor 108 and depending on other sensor data. It may be provided to use multiple acceleration sensors 108 and to evaluate the sensor signals of the multiple acceleration sensors 108 with the classifier 110.

[0030] The other sensor data are, for example, tire pressure or wheel speed of at least one preferably steered wheel 106 of the vehicle 100, steering angle, steering wheel position, steering torque, vehicle longitudinal acceleration, vehicle lateral acceleration, vehicle vertical acceleration, yaw rate of the vehicle 100.

[0031] In the example, the classifier 110 is based on artificial intelligence, which is trained or used in different ways.

[0032] The classifier 110 comprises, for example, a model, in particular a neural network, which is trained by machine learning (ML) to map the sensor signals to a classification that classifies the structure-borne sound as normal or anomalous for the steering system 102.

[0033] The classifier 110 is trained using ML techniques, for example, to recognize patterns and relationships from the sensor data that classify the structure-borne sound as normal or anomalous for the steering system 102.

[0034] ML techniques include supervised learning, unsupervised learning, and reinforcement learning.

[0035] The neural network is, for example, a deep neural network, specifically a convolutional neural network (CNN) or a recurrent neural network (RNN), which is capable of recognizing hierarchical features and temporal dependencies in the sensor signals. This allows complex patterns to be captured and used for classification.

[0036] It can be provided that the classifier 110 provides a time series analysis in which the sensor signals of the acceleration sensor 108 are evaluated over time as time series data. For example, the classifier 110 is configured for a time series analysis using an autoregressive model (AR), moving average (MA), or autoregressive integrated moving average (ARI MA). This allows trends and patterns in the sensor signals to be detected.

[0037] The classifier 110 may be configured to perform a frequency analysis using Fourier transformation or wavelet transformation to convert the sensor signals from the acceleration sensor 108 over time into the frequency domain. This identifies specific frequency patterns that are characteristic of specific events or processes.

[0038] Feature engineering can be performed before training the classifier 110. For example, relevant features are extracted from the sensor signals before the classifier is trained. In the example, these features are used in addition to or instead of the sensor signals as input for the classifier. For example, various statistical, spectral, or time-domain features are used here.

[0039] The sensor signals can be processed by a signal processing unit. The signal processing unit can comprise one or more units that process raw signals from the acceleration sensor 108(s) either close to the sensor, i.e., close to the respective acceleration sensor 108 that provides the raw signal, or in one or more additional control units.

[0040] The monitoring can be designed to run continuously or to start at a specific or predetermined time.

[0041] The model can be determined through transfer learning based on a pre-trained model. The pre-trained model is pre-trained, for example, for a similar task such as classifying sensor signals.

[0042] For example, a model pre-trained for evaluating acceleration signals is transferred to the model of the classifier 110 using transfer learning based on the sensor signals that characterize the structure-borne sound in the vehicle 100. This benefits from the model's already learned properties and reduces training time. The classifier 110 can comprise a combination of several artificial intelligence-based models, i.e., an ensemble of models. This improves the performance and robustness of the evaluation. For example, several models of different architectures work together to make more precise predictions.

[0043] The classifier 110 can be configured to continuously learn in use and adapt its performance to new conditions. This is particularly useful when the characteristics of the sensor signals change over time.

[0044] Figure 2 shows a flowchart with steps of the method for monitoring the vehicle's steering system.

[0045] The method comprises a step 200.

[0046] In step 200, the classifier 110 is trained. The classifier 110 is trained, for example, based on training data that includes the sensor data. The training data includes sensor data from an acceleration sensor that characterizes structure-borne noise in a vehicle. Provision can be made for the training data to also include other sensor data. The training data includes, for example, other sensor data available in the vehicle 100 in which the classifier 100 is used.

[0047] It can be provided that the classifier 110 is pre-trained and is adapted to the vehicle 100 by transfer learning with the training data.

[0048] A step 202 is then executed.

[0049] In step 202, the structure-borne sound in the vehicle 100 is detected using the acceleration sensor 108.

[0050] Subsequently, a step 204 is executed. In step 204, the sensor data is evaluated using the artificial intelligence-based classifier 110.

[0051] The classifier 110 is configured to classify the structure-borne sound as normal or abnormal for the steering system 102 of the vehicle 100 depending on the sensor data.

[0052] The classifier 110 is designed, for example, to evaluate the other sensor data in addition to the sensor data from the acceleration sensor 108.

[0053] The classifier 110 is designed, for example, to detect a pattern in the vibrations depending on vibrations in the sensor data, which classifies the structure-borne sound as anomalous.

[0054] Then a step 206 is executed.

[0055] In step 206, for example, a user is warned if the structure-borne noise has been classified as abnormal for the steering system 102 of the vehicle 100.

[0056] If an anomaly is detected, the user is informed, for example, via the human-machine interface 112 of the vehicle 100, e.g. by outputting speech, text, warning signals, or sound.

[0057] In step 206, for example, a countermeasure is taken if the structure-borne noise has been classified as abnormal for the steering system 102 of the vehicle 100.

[0058] If an anomaly is detected, the countermeasure is requested, for example, via the human-machine interface 112 of the vehicle 100 or sent to the action initiation unit in the vehicle 100.

[0059] The action initiation unit determines, for example, depending on the severity of the anomaly, the maximum speed at which the vehicle 100 may be moved. Depending on the severity of the anomaly, the action initiation unit issues warnings to the driver and / or passengers or recommendations via the human-machine interface 112.

[0060] An example of an output is “Anomaly detected in the steering system”, “Service”, “Reduce speed to ... km / h”).

[0061] In one example, the action initiation unit monitors compliance with the countermeasure. The action initiation unit is trained, for example, to shut down vehicle 100 if significant defects are detected in conjunction with other degradation strategies of the vehicle systems involved.

[0062] In partially autonomous or autonomous driving, where the driver of vehicle 100 is inactive or the passengers are only passengers, the action initiation unit automatically initiates transitions via the vehicle dynamics control systems, for example. Examples of transitions include speed adjustments, lateral dynamics interventions to ensure stability and controllability, e.g., via active interventions by the systems replacing or supporting the defective steering system 102. Examples of these measures include braking individual wheels 106 via a braking system of vehicle 100, controlling an active chassis, requesting individual wheel torques at driven wheels 106, and torque vectoring.

[0063] In step 206, for example, a driving function, in particular a driving situation detection or a route detection, is informed about the structure-borne noise that is abnormal for the steering system of the vehicle 100 if the structure-borne noise has been classified as abnormal for the steering system 102 of the vehicle 100.

[0064] For example, the driving situation detection system is configured to detect straight-ahead driving, cornering, crosswind compensation, oversteering, neutral steering, understeering, acceleration, and deceleration. For example, the route detection system is configured to detect rough road conditions, curb detection, and ramp detection. The countermeasure can be taken in the steering system 102 or in the vehicle 100, e.g., in the driving functions.

[0065] It may be intended that monitoring is carried out only in certain driving conditions, e.g. the situations detected by the driving situation detection or the route detection.

[0066] The classifier 110 improves the robustness of the steering anomaly detection, in particular by avoiding incorrectly identified steering anomalies.

Claims

Claims 1. A method for monitoring a steering system of a vehicle (100), characterized in that structure-borne noise in the vehicle (100) is detected (202) using an acceleration sensor (108), wherein sensor data from the acceleration sensor (108) represent the detected structure-borne noise, wherein the sensor data are evaluated (204) using a classifier (110) based in particular on artificial intelligence, wherein the classifier (110) is designed to classify the structure-borne noise as normal or as abnormal for the steering system (102) of the vehicle (100) depending on the sensor data.

2. Method according to claim 1, characterized in that the classifier (110) is designed to recognize, depending on vibrations in the sensor data, a pattern in the vibrations which classifies the structure-borne sound as anomalous.

3. Method according to one of the preceding claims, characterized in that the classifier (110) is designed to evaluate other sensor data in addition to the sensor data from the acceleration sensor (108), in particular tire pressure or wheel speed of at least one preferably steered wheel of the vehicle (100), steering angle, steering wheel position, steering torque, vehicle longitudinal acceleration, vehicle lateral acceleration, vehicle vertical acceleration, yaw rate of the vehicle (100).

4. Method according to one of the preceding claims, characterized in that a user is warned (206) if the structure-borne noise is classified (204) as abnormal for the steering system (102) of the vehicle (100).

5. Method according to one of the preceding claims, characterized in that a countermeasure is taken (206) if the structure-borne sound is is classified (204) as abnormal for the steering system (102) of the vehicle (100).

6. Method according to one of the preceding claims, characterized in that a driving function, in particular a driving situation detection or a route detection, is informed (206) about the structure-borne noise that is abnormal for the steering system of the vehicle (100) if the structure-borne noise is classified (204) as abnormal for the steering system (102) of the vehicle (100).

7. Method according to one of the preceding claims, characterized in that the classifier (110) is trained based on training data comprising sensor data comprising structure-borne sound detected by an acceleration sensor in a vehicle.

8. Device (104) for monitoring a steering system (102) of a vehicle (100), characterized in that the device (104) is designed to carry out the method according to one of claims 1 to 7.

9. Vehicle (100), characterized in that the vehicle (100) comprises a steering system (102) and the device (104) according to claim 8.

Images