Method, apparatus, and computer program for determining the wear condition of vehicle brake linings

Abstract

A method for determining a wear state of brake linings of a vehicle (F), comprising the steps of: receiving (S10) time series data (Dt), said time series data (Dt) comprising a time series of data relating to a braking system of said vehicle (F); and identifying (S20) at least one brake event (B1, B2) in said time series data (Dt), wherein each brake event (B1, B2) identified in said time series data (Dt) falls within a temporal data window of brake event data (Db) of said time series data. Correspondingly, the method comprises a step (S20) of identifying, wherein the data window relates to a real braking event of the vehicle (F), a step (S30) of determining features (M) from the braking event data (Db) using a predetermined operator for each identified braking event (B1, B2), and a step (S40) of classifying, using the features (M) determined therefor, the at least one braking event (B1, B2), wherein a classification (K) is assigned to a wear state of the brake linings of the vehicle (F).

Description

Technical Field

[0001] The present invention relates to a method, particularly for continuous observation, and an apparatus and a computer program for determining the wear state of a brake lining of a vehicle.

Background Art

[0002] For example, a hydraulic brake system used in a vehicle such as a passenger car or a commercial vehicle is designed to decelerate the wheels with torque generated by friction. The brake caliper is usually moved towards the rotating rotor, particularly towards the brake disc fixedly coupled to the wheel, with the brake lining. Thus, a surface contact due to friction is formed between the brake lining and the brake disc. The brake lining with thermal conductivity is designed on the premise of wear to ensure a long-life brake system. It is necessary to monitor the wear of the brake lining to meet the requirements for vehicle safety, error avoidance, maximization of the life of the brake caliper-rotor system, vehicle monitoring and maintenance, as well as fleet management and supply chain management.

[0003] The wear certification of the brake lining is substantially based on a combination of a direct certification approach and an indirect certification approach. The thickness of the lining material can be directly measured or monitored with a hardware sensor (direct scanning). The indirect measurement method derives the lining thickness or the wear state of the lining from the pre-set system parameters and the data of the peripheral sensors.

[0004] Conventional brake lining wear sensors (BPWS) include an electrical circuit embedded in the brake lining friction material perpendicular to the direction of lining wear. The sensor is typically mounted on or very close to the back plate of the brake lining. BPWS can have multiple stages to classify the lining wear state, resulting from stepped resistance changes as the electrical circuit is destroyed by brake lining wear. These sensors themselves are destroyed as lining wear progresses (destructive scanning).

[0005] Non-destructive scanning approaches include sensor systems that utilize methods different from direct or indirect measurement of lining thickness. Examples include position or distance sensors based on ultrasonic technology.

[0006] The evaluation of sensor signals is sometimes combined with relatively complex software algorithms. The actual algorithms typically involve either auxiliary hardware sensors or software algorithms that rely on measurements from other sensors and quantities provided by the brake system, relating to the brake disc surface temperature. The brake disc temperature model (BTM) is used to derive the disc temperature from the physical work performed by the brake lining and the radiative cooling of the disc. The main parameters of the BTM are pressing force, wheel speed, and ambient temperature, excluding coefficients related to wheel characteristics, lining characteristics, and brake system characteristics.

[0007] Brake lining wear ΔWz is derived from physical modeling. In the first-order approximation, ΔWz depends linearly on the energy Eb dispersed during a braking event, i.e., ΔWz∝Eb for each braking. The proportionality constant K in this model typically depends on pressure, wheel speed, and disc temperature (given by BTM), and is usually approximated by a polynomial. The total wear of a single brake lining at the current time t0 is derived by the sum of the ΔWz evaluations of all braking events performed up to that point (integral approach). Multistage BPWS is used both to stepwise recalibrate model predictions, which are subject to considerable uncertainty, and as a safety unit (driver warning) for when the brake lining is completely worn out.

[0008] However, the integral approach to calculating total brake lining wear is based on the assumption that the brake lining is replaced with a new one of the same type (brand, model). If a different type of already worn brake lining is used, the installed model would need to be recalibrated first, which is not a solution that current mounting equipment can overcome. Secondly, several systemic uncertainties in the model prediction or unmodeled brake lining behavior combine to continuously increase measurement uncertainty until the model is recalibrated by BPWS measurement. In addition, current BTM mounting equipment has high uncertainty (≤100k) regarding the brake disc temperature being evaluated. Therefore, BTM mounting equipment only meets the ASIL-A-Standard (Automotive Safety Integration Level). Using a hardware sensor for brake disc temperature instead of BTM would significantly increase production costs. Finally, the internal BPWS embedded in the brake lining material wears out along with the brake lining wear. Therefore, replacing the lining requires replacing the sensor, which consequently increases maintenance costs. [Overview of the Initiative]

[0009] According to one aspect of the present invention, a method for determining the wear state of a vehicle's brake lining comprises the following steps: In one step, time-series data is received, in which case the time-series data includes a time series of data relating to the vehicle's brake system. In a further step, at least one brake event is identified within the time-series data, in which case each brake event identified within the time-series data corresponds to a temporal data window of brake event data in the time-series data, where the data window relates to an actual brake event of the vehicle. In a further step, a feature is determined from the brake event data using a predetermined operator for each identified brake event. In a further step, the at least one brake event is classified using the feature determined for that purpose, in which case the classification is assigned to the wear state of the vehicle's brake lining.

[0010] The concept of "brake lining wear" used here specifically includes the thickness of the brake lining.

[0011] Thus, time-series data contains numerous datasets from a single source across multiple time steps.

[0012] In other words, the determined characteristics describe the system behavior of the vehicle, particularly the brake system, during a single braking event. Based on the determined characteristics, the analyzed braking events can be classified, or in other words, statements can be made about the characteristics of the braking events. Assuming that the characteristics of the braking events are directly related to the wear state of the brake lining, statements about the wear state of the brake discs are made based on the classification of the braking events.

[0013] Preferably, the classification of at least one brake event includes a classification of at least one brake event by a machine learning model, particularly a pre-trained machine learning model, utilizing the features determined for that purpose, in which case the machine learning model identifies a probability for each classification of at least one brake event and assigns a rank to at least one brake event using the identified probabilities.

[0014] Preferably, the machine-learning model is automatically trained during the initial riding period, for example, during the first 10,000 km. More preferably, the machine-learning model is automatically trained after the brake linings are replaced, for example, during the first 10,000 km after the brake linings are replaced.

[0015] The machine learning algorithms of the machine learning model can use methods such as logistic regression, neural networks, and random forests. The model implemented in the vehicle's active brake lining condition monitoring system may be pre-trained before the vehicle system is brought to market, or it may self-learn during the initial stages of vehicle operation.

[0016] Time-series data, i.e., input data, is represented by the time series Sk(t) using individually different scans. The index k∈N indicates the k-th signal source or data source. The variable t represents time. All data types of time-series data are provided by hardware or control software related to the brake system, including standard vehicle state data sources or peripheral data sources such as inertial measurement units.

[0017] Preferably, each braking event has a length of 10 seconds or less, and more preferably, a length of 5 seconds or less.

[0018] Time series data are characterized by statistical features derived for all time series data within each data window. Statistical evaluations or mathematical operators considered to determine these features include, for example, quantiles, standard deviation, mean, minimum, and / or maximum values.

[0019] For each brake event, i.e., for a data window, features, or a particularly optimized set of features, are provided to a machine learning model pre-trained for event classification. The machine learning model preferably uses a supervised logistic regression classifier, which is a standard method for classification problems. Because the trained logistic regression model has low numerical and computational complexity, it thus becomes easy to implement embedded in the controller. An alternative version of the classification algorithm may use other supervised machine learning classifiers, such as a random forest, whether in an in-vehicle implementation or an external, cloud-based implementation. To avoid general pre-training of the classifier, i.e., to avoid pre-training of brake event classification, or to avoid vehicle-specific training, an unsupervised machine learning model may be used.

[0020] Machine learning models also preferably use logistic regression. Logistic regression is a linear, stochastic, discriminant model for classification. Discriminant means that the model learns a mapping function, commonly called a discriminant function, that maps input data to classes. Stochastic means that the model learns the discriminant function based on statistically allocated input data and their respective classes.

[0021] Event classification, that is, formally assigning one brake lining wear condition class to one given brake event, is based on a probabilistic approach. In the basic approach, individual events are ranked into a class with the highest probability. For example, in the case of two wear condition classes, C ∈ {good, bad}, the classification limit is given by a probability of 0.5. More classes than this can be handled correspondingly, for example, by generalizing logistic regression to two or more wear condition classes.

[0022] Therefore, the above approach enables classification based on vehicle sensors and brake system sensors, or on individual, independent brake events indicated by data and status, and thus also enables measurement of brake lining thickness. In other words, here, a differential approach is selected to assess the wear state of the brake lining compared to the integral approach described in the background section. Therefore, it is not necessary to observe the wear state of the brake lining using the integral method over a relatively long period of time.

[0023] By observing individual braking events and the occurrence of time-series data characteristics assigned to each, the overall behavior of the braking system and vehicle is observed, and the wear of the brake lining is assessed based on this observation.

[0024] In this way, a more accurate method is provided for determining the wear condition of brake linings. The uncertainty is significantly reduced compared to the integral method. Therefore, in addition, (for example, when replacing linings) it is also possible to use and monitor alternative linings from sources other than the manufacturer.

[0025] A purely model-based approach, or at least primarily a model-based approach, eliminates the need for auxiliary direct sensors on the brake lining, which would typically measure the brake lining thickness directly, a process that is usually relatively time-consuming and costly.

[0026] In this way, an improved method for determining the wear state of the brake lining is provided.

[0027] According to an advantageous embodiment, the data related to the brake system includes vehicle sensor data, controller data and / or brake system data.

[0028] The time series data preferably includes a large number of data related to the brake over time, especially raw data.

[0029] According to an advantageous embodiment, the sensor data is provided by a brake master cylinder pressure sensor, a tire rotation speed sensor, a vehicle inertia sensor and / or a brake system sensor.

[0030] Preferably, the sensor data includes measurements of physical quantities that depend on time. More preferably, the measurements of the vehicle sensors for determining the sensor measurements are carried out at a predetermined frequency. In other words, signal scanning, also called sampling, is performed.

[0031] Preferably, the brake system sensor includes a sensor for determining the brake fluid volume displaced by the brake system.

[0032] Preferably, the control data includes data generated from the sensor data.

[0033] Preferably, the controller data detects various quantities, especially physical quantities, derived from the sensor data, system data or software data. These data are provided especially as a function of the scanned time.

[0034] According to an advantageous embodiment, the brake system data includes the brake system status and / or brake system flags.

[0035] Preferably, the brake system data includes brake system conditions, brake system settings, and the functional operating state or operating mode of the brake system as a function of time.

[0036] According to an advantageous embodiment, the identification of at least one brake event includes the steps of: receiving at least one brake trigger, wherein the brake trigger is related to an actual brake event of the vehicle; and identifying at least one brake event using the received at least one brake trigger.

[0037] Preferably, the identification of at least one brake event includes, for example, the selection of time-series data filed in memory, in which case the selected time-series data is assigned to one brake event.

[0038] Preferably, the brake trigger is received externally.

[0039] Essentially, brake lining wear is evaluated based on braking events. Therefore, time-series data is allocated to time intervals, also known as data windows. The length of these time intervals is selected to fully cover individual, distinct braking events, thus providing a unique dataset for the analysis of those events.

[0040] For example, a brake trigger includes a brake light switching signal, i.e., a signal indicating whether the brake lights are activated or not; a speed signal, i.e., a signal indicating the speed at which the vehicle is moving; and / or an engine status signal, i.e., a signal indicating the state of the engine.

[0041] Preferably, the start of a brake event is determined if the brake light signal indicates "operated," the speed signal exceeds a predetermined value, for example, 0.1 m / s, and the engine status signal indicates "operated."

[0042] Since the braking time for each braking event varies, it is important to select a portion of a single braking event that provides meaningful information for classification. For example, a window fixing amount is set that applies to all braking events and is optimized during training of the machine learning model. The time window is characterized by a tuple (ts, Δtw), where ts is the relative activation time of the window and Δtw is the window length. All data points of time series data outside the interval [ts, ts + Δtw] are cut off. Brake events that do not meet the minimum requirements of a given window, such as short events, are not considered for analysis. Alternatively, various time windows can be defined that are mathematically handled by event normalization methods.

[0043] In other words, a brake event includes a buffering time. That is, a brake event includes time-series data before and after the actual brake event. Therefore, time-series data with a pre-defined buffering time is assigned to each brake event before and after the brake trigger.

[0044] According to an advantageous embodiment, at least one brake trigger includes the state of the brake light switch, the vehicle speed, and / or the engine state.

[0045] According to a favorable embodiment, extraneous time-series data that cannot be related to braking events is eliminated.

[0046] In other words, it eliminates intervals in time-series data that only contain data recorded in the gap between two consecutive braking events.

[0047] According to a favorable embodiment, time series data unsuitable for feature determination is eliminated.

[0048] Not all braking events are eligible for data analysis regarding brake lining wear. In particular, time series data that are classified as invalid are not suitable for determining features. For example, incomplete or inaccurate time series data should be classified as invalid. In addition, data selection may be performed based on criteria for signal parameters, such as vehicle motion, braking strength and duration parameters, or other parameter criteria. Furthermore, restricting classification with respect to sensitivity, systematics, or limitations within the model, i.e., restricting the classification algorithm in particular, may result in individual data packets being excluded from data analysis.

[0049] Therefore, event selection imposes restrictions on data selection.

[0050] In a favorable embodiment, the method includes relating suitability to each of the determined features and using a predetermined number of features with the highest suitability to classify at least one brake event.

[0051] During the training of a machine learning model, a large number of features are first considered and then iteratively arranged according to their relevance, i.e., according to their impact on the classification probability derived by the classifier. For example, 15 of the most important features are imposed on the classification algorithm, i.e., feature selection and optimization are performed according to the method of inductive feature elimination. Further optimization criteria include prioritizing uncorrelated time series to avoid nearly constant signals, and from there setting upper bounds on statistical correlation and lower bounds on the variance of the time series. According to alternative algorithms, fewer or more features can be considered as a result of model optimization.

[0052] According to a favorable embodiment, receiving time-series data includes the following steps: recording the received time-series data into memory, in which case the time-series data is retained in memory unless the memory is exhausted or the characteristics of the corresponding time-series data have not been determined.

[0053] According to a favorable embodiment, at least one braking event is classified, taking into account the vehicle's braking history.

[0054] Preferably, the brake history includes the assumption that it incorporates a continuous wear process. More preferably, the brake history includes the characteristics of a series of consecutive brake events. Even more preferably, the brake history includes potential brake abnormalities identified in previous brake events.

[0055] According to an advantageous embodiment, the method includes receiving temperature data, wherein the temperature data includes the temperature of a brake lining, the temperature of a brake disc of a vehicle, in particular the temperature of the brake disc to which the brake lining is attached, and / or the ambient temperature of a vehicle; and classifying at least one brake event using a characteristic determined for that purpose and the corresponding temperature data.

[0056] Preferably, the brake lining temperature data includes the temperature of the brake lining on the side opposite to the brake disc. Based on the temperature on the side opposite to the brake disc, the thermal conductivity of the brake lining can be inferred, and therefore, in particular, the thickness of the brake lining can be inferred.

[0057] Preferably, the temperature data further includes the ambient temperature of the vehicle, and more preferably, the ambient temperature of the brake disc.

[0058] For the sake of accuracy in temperature data, heating and cooling processes should be considered in relation to the ambient temperature of the vehicle.

[0059] Preferably, temperature data is of particularly high importance compared to features, i.e., it correlates with the condition of the brake lining.

[0060] By using temperature data, and working in conjunction with determined features, machine learning models can perform particularly accurate classifications.

[0061] In another aspect of the present invention, an apparatus is proposed for carrying out a method for determining the wear condition of a brake lining as described herein.

[0062] In another aspect of the present invention, a computer program is proposed that includes instructions causing a computer to perform the method described herein when the computer executes the computer program.

[0063] Next, further measures to improve the present invention will be described in more detail with reference to the figures, along with a description of advantageous embodiments of the present invention. [Brief explanation of the drawing]

[0064] [Figure 1] This figure shows a device for determining the wear condition of brake linings. [Figure 2] This figure shows a vehicle equipped with a device for determining the wear condition of brake linings. [Figure 3] This figure shows a method for determining the wear condition of brake linings. [Figure 4] This figure shows time-series data in data memory. [Modes for carrying out the invention]

[0065] Figure 1 shows a device 100, in particular an electronic control unit, for determining the wear condition of the brake linings of a vehicle F. Preferably, the device includes components that can be used within a normal controller for the brake system of the vehicle F, but is fully functional for carrying out the determination of the wear condition of the vehicle F. The device 100 includes an electronic processor 30, such as a programmable microprocessor, microcontroller, or other processor unit, a memory 20, such as a non-transient, machine-readable memory, and a communication interface 10. The processor 30 is installed to implement software instructions related to determining the wear condition of the brake linings of the vehicle F. Auxiliarily, the processor 30 can carry out other brake system processes. The processor 30 can read from and write to the memory 20. The communication interface 10 forms a connection between the device 100 and the vehicle communication bus of the vehicle F, which is coupled with other systems inside the vehicle. In particular, the vehicle communication bus can be used for data exchange with the vehicle computer 200 or the vehicle communication unit 300. The vehicle communication unit 300 enables the vehicle F to connect with external entities. The device 100 is connected to, for example, a brake master cylinder pressure sensor, a wheel rotation speed sensor, a vehicle inertia sensor, and various internal sensors of the vehicle F's brake system. Optionally, external sensors may also be connected, for example, via a vehicle communication bus.

[0066] The communication interface 10 is installed to receive time-series data Dt from various sensors of the vehicle F, the brake system itself, or the vehicle communication bus, in which case the time-series data includes a time series of data related to the vehicle's brakes. The memory 20 is installed to store the time-series data Dt received by the communication interface 10. The processor 30 includes a data detection unit 31, which is installed to identify at least one brake event B1, B2 in the time-series data Dt. Each brake event B1, B2 identified in the time-series data Dt corresponds to a temporal data window of the brake event data Db in the time-series data Dt, in which case the data window relates to one actual brake event of the vehicle F. The processor 30 also includes a pre-processing unit 32, which is installed to select brake events based on predefined criteria and to determine features M from the brake event data Db using pre-determined operators for each identified brake event B1, B2. Brake event data Db is raw data that is preprocessed by time filtering and the calculation of signal features using predetermined operators for further processing, such as minimum, maximum, mean, standard deviation, absolute and / or quantiles. The processor 30 includes a machine learning model unit 33, which is set up to classify at least one brake event using a predetermined feature M. Classification K is assigned to the wear status of the brake lining of the vehicle F.

[0067] The device detects sensor signals and brake system software signals in the form of time-series data Dt. The time-series data Dt is stored in intermediate storage in memory 20. The scanning frequency for receiving the time-series data Dt is pre-set and may, in particular, follow the default settings of the brake system.

[0068] The data detection unit 31 detects brake event data Db that corresponds to the temporal data window of the time series data Dt. In other words, the brake event data Db relates to the time series data Dt associated with one brake event of vehicle F. The data detection unit 31 identifies brake events B1 and B2 of vehicle F and detects the corresponding brake event data Db for the identified brake events B1 and B2 from the time series data Dt.

[0069] Identification of brake events B1 and B2 is performed by the data detection unit 31 using a brake trigger T. The brake trigger T is generated, for example, by the driver of vehicle F, the brake system, or the autonomous vehicle computer 200 of vehicle F. Once the brake request is fulfilled and the brake event ends, data detection ends. Alternatively, data detection may include a buffering time, i.e., it may also include detection of data before and after the identified brake events B1 and B2. In this case, the brake event data Db is held in memory 20 as long as memory 20 is not exhausted and data preprocessing of the brake event data Db has not yet been completed. Alternatively, the brake event data Db may be transmitted to other systems, particularly the vehicle connection unit 300, via the communication interface 10 of the device 100.

[0070] Brake lining wear detection and monitoring are performed based on individual brake events B1, B2, in which case the corresponding time-series data Dt is analyzed. In the case of purely model-based brake lining wear detection (also known as BPWD), a machine learning model 33 is used to classify the wear state of the brake lining. The wear state can be defined through intervals of remaining brake lining material thickness. A simple embodiment is to observe two or three sets of states as the basis for classification, namely (good, bad) or (new, used, worn). Alternatively, three or more or other states may be defined.

[0071] BPWD uses raw data related to the brake system, specifically data from the vehicle software that is normally provided and does not need to be added separately. A special BPWD algorithm is used as the sensor input signals, namely brake master cylinder pressure, wheel speed, vehicle acceleration, and sensors within the brake system, particularly rod stroke and tappet motion. The brake system software provides auxiliary quantities derived from the raw sensor data, vehicle characteristics, and brake request characteristics, particularly wheel torque and characteristics related to brake requests. Alternatively, other characteristic data, similarly invoked by the vehicle communication bus, may also be considered.

[0072] Instead of a model-based approach, a temperature sensor St located on the back plate of the brake lining may be used as an auxiliary input to the machine learning unit 33. Alternatively, data from a brake disc temperature sensor may also be considered. Due to the influence of brake lining temperature measurement, the machine learning model 33 can measure brake lining strength and thus brake lining condition with greater accuracy.

[0073] A series of analyses on brake event data Db includes the following main tasks: First, event selection. Not all brake events B1 and B2 are included in the analysis. Data selection can be based on criteria such as data validity, vehicle inertia, and brake intensity. Second, data preprocessing. Raw data is preprocessed for analysis by operators such as predetermined time filtering and by calculating features M (minimum, maximum, mean, standard deviation, module, quantile, etc.). Third, data analysis. The preprocessed features M are analyzed using a machine learning model 33. Fourth, classification. Based on the analysis results, brake events B1 and B2 are classified, for example, based on the wear status assigned by the machine learning model 33. The wear status label (also referred to as classification K) corresponds to the assessment of the brake lining wear status.

[0074] Figure 2 shows a vehicle F equipped with a device 100 for determining the wear condition of the brake lining.

[0075] Vehicle F includes a vehicle computer 200 and a vehicle communication unit 300, in addition to a device 100 in the form of an electronic control unit 100 for determining the wear condition of the vehicle's brake linings. Required time-series data Dt is provided to the electronic control unit 100 via direct sensor connections, via a vehicle communication bus, or via the vehicle computer 200. The vehicle computer 200 can also be used to enable the vehicle driver to access the results of the determination of the brake lining wear condition, or to present said results. In this example, each brake lining has a temperature sensor St, which provides temperature data Dtemp to the electronic control unit 100. Furthermore, vehicle F has a vehicle communication unit 300, which is installed to transmit brake event data Db from the electronic control unit 100 to an external cloud or data bank 400. In this example, the cloud 400 has a machine learning model installed to detect the brake lining wear condition from the provided brake event data. Compared to the machine learning model within the electrical control unit 100, the external cloud 400 can accommodate a relatively more complex machine learning model, including more complex pre-processing or post-processing algorithms. In this case, the resulting classification of corresponding brake events assigned to the brake event data is returned from the cloud 400 to the electronic control unit 100 via the vehicle communication unit 300.

[0076] Figure 3 shows a method for determining the wear condition of the brake lining.

[0077] In the first step S10, time-series data Dt is detected, in which case the time-series data Dt contains a time series of data related to the brakes of vehicle F. In the second step S20, at least one brake event B1, B2 is identified within the time-series data Dt, in which case each brake event B1, B2 identified within the time-series data Dt corresponds to a temporal data window of the brake event data Db in the time-series data, in which case the data window is related to an actual brake event of vehicle F. In the third step S30, a feature M is determined from the brake event data Db using a predetermined operator for each identified brake event B1, B2. In the fourth step S40, at least one brake event B1, B2 is assessed using the feature M determined for that purpose, in which case the classification K is assigned to the wear state of the brake lining of vehicle F.

[0078] Figure 4 shows time-series data Dt temporarily filed in data memory. In this example, time-series data Dt includes vehicle speed Vveh, brake light switch state Swl, brake master cylinder pressure data Dp, and tire rotation speed data Dd. Time-series data Dt contains data over nine time steps t0-t8, which have their place in the shown memory portion. In other words, the brake master cylinder pressure data Dp contains multiple datasets Dp0, Dp1, Dp2, Dp3, Dp4, Dp5, Dp6, Dp7, Dp8 over this time. Similarly, the tire rotation speed data Dd contains multiple datasets Dd0, Dd1, Dd2, Dd3, Dd4, Dd5, Dd6, Dd7, Dd8 over this time. The individual datasets of brake master cylinder data Dp and tire rotation speed data Dd are shown here as wildcards because their exact values ​​are not important here. Similarly, vehicle speed includes the vehicle's speed at any given moment over the aforementioned time (here in meters per second). The brake light switch state Swl indicates whether the vehicle's brake lights are activated or not at each time step, with A indicating activated and D indicating deactivated.

[0079] Using Figure 4, we will explain how to identify the first brake event B1 and the second brake event B2 from the time series data Dt. On the one hand, we take into account the brake light switch state Swl. We start with one brake event at each time step where value A is notified, i.e., we start at time steps t0-t2, t4, and t6-t7. However, at time step t4, the vehicle speed Vveh is only 3m per second, and therefore below the pre-set limit for one brake event. In this respect, only time steps t0-t2 are identified as the first brake event B1, and time steps t6-t7 are identified as the second brake event B2. As a result, the respective datasets Dp0, Dp1, Dp2 and Dd0, Dd1, Dd2 are identified as brake event data Db for the first brake event B1, and the datasets Dp6, Dp7 and Dd6, Dd7 are identified as brake event data Db for the second brake event B2. The brake event data Db remains in memory until features M for the machine learning model 33 are determined from it. Other datasets are discarded, thus freeing up space in memory for new time-series data. [Explanation of Symbols]

[0080] 20 memory 100 devices B1, B2 Brake Event Db Brake Event Data Dt time series data Dtemp temperature data F Vehicle K classification M Features S10 Step to receive time-series data Steps to identify S20 brake events S30 Steps to determine features Steps to classify S40 brake events T Brake Trigger

Claims

1. A method for determining the wear condition of a vehicle's brake lining includes the following steps: Step (S10) of receiving time-series data (Dt), wherein the time-series data (Dt) includes a time-series of data relating to the brake system of the vehicle (F), Step (S20) of identifying at least one brake event (B1, B2) in the time series data (Dt), wherein each brake event (B1, B2) identified in the time series data (Dt) corresponds to a temporal data window of brake event data (Db) in the time series data, and the data window is related to an actual brake event of the vehicle (F), Step (S30) of determining a feature (M) from the brake event data (Db) using a predetermined operator for each identified brake event (B1, B2), A method comprising the step (S40) of classifying at least one of the brake events (B1, B2) based on the wear state assigned by analyzing the aforementioned feature (M) using a machine learning model (33).

2. The method according to claim 1, wherein the data relating to the brake system includes sensor data, controller data and / or brake system data of the vehicle (F).

3. The method according to claim 2, wherein the sensor data is provided from a brake master cylinder pressure sensor, a tire rotation speed sensor, a vehicle inertia sensor, and / or a brake system sensor.

4. The method according to claim 2 or 3, wherein the brake system data includes brake system status and / or brake system flags.

5. The step of identifying at least one brake event is, Reception of at least one brake trigger (T), wherein the brake trigger (T) is associated with an actual brake event of the vehicle (F), Identification of the at least one brake event (B1, B2) using the at least one brake trigger (T) received, The method according to any one of claims 1 to 3, comprising:

6. The method according to any one of claims 1 to 3, wherein the at least one brake trigger (T) includes the state of the brake light switch, the longitudinal acceleration of the vehicle and / or the engine state.

7. The method according to any one of claims 1 to 3, comprising the elimination of extraneous time-series data (Dt) that cannot be related to the aforementioned brake events (B1, B2).

8. The method according to any one of claims 1 to 3, comprising excluding time series data (Dt) that are not suitable for determining the aforementioned feature (M).

9. The relationship of suitability to each of the determined features (M), To classify the at least one brake event (B1, B2), use a predetermined number of the feature (M) that has the highest relevance, The method according to any one of claims 1 to 3, comprising:

10. The step of receiving time-series data (Dt) is, This includes recording the received time-series data (Dt) into memory (20), The method according to any one of claims 1 to 3, wherein the time series data (Dt) is kept in memory as long as the memory (20) is not exhausted, or as long as the characteristics of the corresponding time series data (Dt) have not been determined.

11. The method according to any one of claims 1 to 3, which classifies the at least one brake event (B1, B2) taking into consideration the brake history of the vehicle (F).

12. Reception of temperature data (Dtemp), wherein the temperature data (Dtemp) includes the temperature of the brake lining, the temperature of the vehicle's brake discs and / or the ambient temperature of the vehicle. Classifying the at least one brake event (B1, B2) using the characteristic (M) determined for that purpose and the corresponding temperature data (Dtemp), The method according to any one of claims 1 to 3, comprising:

13. A device (100) for determining the wear condition of the brake lining of a vehicle (F), which is installed to carry out the method according to any one of claims 1 to 3.

14. A computer program that includes an instruction to cause the computer to perform the method described in any one of claims 1 to 3 when the computer executes the computer program.

Images