Method and device for identifying a cause of a fault event in a vehicle based on vehicle events with the aid of machine learning

Abstract

The invention relates to a method for predicting a fault of a particular fault type in a vehicle, said method comprising: - recording (S1) vehicle events in a temporal sequence during the service life of the vehicle; - in particular after the occurrence of a fault, creating (S2) a temporal sequence of input vectors (E) in which the recorded vehicle events and a description of the fault that has occurred are encoded; - providing a recurrent data-based classification model that is trained to assign the temporal sequence of input vectors (E) to an output vector with information about a probability of the cause of the fault that has occurred; - evaluating (S3, S4) the recurrent data-based classification model depending on the created sequence of input vectors (E) in order to determine the most probable cause of a fault that has occurred; - signalling (S5) the most probable cause of the fault that has occurred.

Description

[0001] Description

[0002] title

[0003] Method and device for detecting a fault cause of a fault event in a vehicle based on vehicle events using machine learning

[0004] Technical area

[0005] The invention relates to the detection of fault causes in vehicles after a fault has occurred, in particular depending on a time series of past vehicle events. The invention further relates to methods for diagnosing vehicles depending on a time series of vehicle events and diagnostic events from consecutive and / or simultaneous examination interventions.

[0006] Technical background

[0007] When a fault occurs in a vehicle, the cause of the fault is often unclear. For this purpose, a series of tests is usually conducted in the workshop, and the corresponding test results are evaluated. The test results are derived from a series of diagnostic steps in a workshop. The sequence of tests is determined intuitively by the workshop employee or by the diagnostic software as part of the guided troubleshooting. Due to the intuitive approach in the workshop, the sequence and selection of tests and the narrowing down of the fault cause do not always represent an optimal path to finding the cause of the fault.

[0008] Disclosure of the invention According to the invention, a computer-implemented method for finding a future fault event in a vehicle according to claim 1 and a corresponding device according to the independent claim are provided.

[0009] Further embodiments are specified in the dependent claims.

[0010] According to a first aspect, a method for detecting a cause of a fault in a vehicle is provided, comprising the following steps:

[0011] Recording vehicle events in a chronological sequence throughout the vehicle’s lifetime;

[0012] In particular, after an error occurs, creating a temporal sequence of input vectors in which the detected vehicle events and a description of the error that occurred are encoded;

[0013] Providing a recurrent data-based classification model that is trained to map the temporal sequence of input vectors to an output vector containing information about a probability of the error cause of the error that occurred;

[0014] - Evaluating the recurrent data-based classification model depending on the created sequence of input vectors to determine the most likely cause of an error that has occurred;

[0015] Signaling the most likely cause of the error that occurred.

[0016] During the operation of a motor vehicle, data on vehicle events that occur are continuously recorded. Vehicle events include operational events, repair events, fault events, and diagnostic events at a specific point in time.

[0017] The vehicle events recorded in this way during the vehicle's lifetime can be transmitted, each with a time stamp, to a central unit and / or stored in a memory within the vehicle. In addition, events for the respective vehicle are stored by the vehicle manufacturer or other institutions, such as insurance companies.

[0018] The operating events include all events that characterize the driver-specific and / or vehicle-specific operation, such as driving at high speed above a speed threshold and for more than a predetermined driving time, an impact due to driving on uneven ground or due to hitting an object in the environment with an acceleration effect above a predetermined threshold, an environmental condition such as rain, a very low or very high outside temperature (each defined by exceeding or falling below limit values) for a predetermined minimum period of time, the times and extent of software updates, the input and modification of presets by the driver, the triggering of the alarm system, a restricted view of an outside camera and the like.

[0019] Error events can be errors detected by an automatic monitoring function in the control units. These are stored in an error log and, if necessary, transmitted to a central unit, where the chronological sequence of errors that occurred is logged. Furthermore, error events can also be errors detected and logged through driver observations.

[0020] Repair events correspond to events of a specific repair, replacement of a vehicle part or the like, which are carried out, for example, when the vehicle is taken to a workshop for repair or routine maintenance.

[0021] In addition, diagnostic steps can be performed during a workshop visit, leading to results that can be provided as diagnostic events or diagnostic results. The diagnostic results can include, for example, battery voltage measurements, error log readings, and the like. The repair measures are usually also stored electronically and are thus available as vehicle events for the individual vehicle.

[0022] All vehicle events are assigned a timestamp and / or mileage reading, which indicates the time and thus the order in which the events occurred. Thus, a vehicle has a history of vehicle events, which are arranged chronologically according to the timestamps. In complex systems such as vehicles, events in the vehicle affect subsequent operation and behavior and can thereby influence, accelerate, or delay the development of errors, and possibly even encourage the occurrence of errors. However, the very different vehicle events and their sequences have so far been difficult to link with one another, as they have different effects on the vehicle's overall system. However, vehicle events can influence one another without the connection being apparent at first glance.However, it is likely that past vehicle events can trigger the development of future fault events. It is also conceivable that a developing fault is always preceded by a similar sequence of vehicle events.

[0023] It is therefore proposed to evaluate all vehicle events that have occurred using a recurrent data-based classification model in order to determine the cause of an error that has occurred.

[0024] This can include a description of the fault that occurred, e.g., a specification of the defective component, such as a faulty alternator, or a specification of the fault type, such as a specification of the defective function, e.g., "no display on the screen," and the vehicle events are coded according to a mapping table. All vehicle events within a specified time window are summarized and coded into an input vector. The time windows are consecutive and cover the entire service life of the vehicle, or parts of it.

[0025] Using the recurrent data-based classification model, vehicle events can be evaluated in consecutive time windows. This classifier makes it possible to identify the root cause of the error that occurred.

[0026] The data-based classification model can be implemented as a recurrent neural network, such as an LSTM (Long Short-Term Memory), GRU, or transformer. The timestamps of the vehicle events allow them to be specified as input during the evaluation of the recurrent model in consecutive time windows according to the timeline, so that the temporal relationship between the vehicle events can also be determined using the data-based model.

[0027] A preferred data-based model is an LSTM (Long Short-Term Memory), in which a state vector plus an input vector of the respective time step is recursively evaluated. The input vector represents a code of the vehicle events that have occurred so far or within a past time window. If multiple vehicle events occur simultaneously, these can be linked element-specifically by vector addition, averaging, or maximum value selection to generate each element of the input vector. Such linking can be performed element-by-element in such a way that vehicle events within a time window that are assigned to the same element of the input vector are linked or aggregated. Vehicle events that are assigned to different elements can be integrated into the input vector by simple vector addition.

[0028] In particular, an internal state vector of the data-based classification model can be mapped to an output vector using at least one fully connected layer, the elements of which each indicate a probability of a possible error cause (from a plurality of possible error causes) of the error that has occurred, wherein the most probable error cause is determined as the error cause whose associated element of the output vector has the highest value using a softmax function applied to the output vector.

[0029] It can be provided that the most probable cause of the error is signaled by outputting a description of the cause of the error to a user. In particular, depending on the cause of the error, a user can be informed whether a component or function for which the error is described is the source of the error, or whether the cause of the error lies in a different component or function. According to a further aspect, a device can be provided that is configured to carry out the above method.

[0030] Brief description of the drawings

[0031] Embodiments are explained in more detail below with reference to the attached drawings. They show:

[0032] Figure 1 is a schematic representation of a system of a motor vehicle and a central unit for recording vehicle events;

[0033] Figure 2 shows a schematic representation of the evaluation of a time series of vehicle events using a recurrent neural network; and

[0034] Figure 3 is a flowchart illustrating a method for detecting a cause of a fault that has occurred in a vehicle based on predictive diagnostic vehicle events.

[0035] Description of embodiments

[0036] Figure 1 shows a vehicle 1 with a plurality of components and with one or more control units 2, which are designed to detect and control vehicle functions and to detect vehicle events.

[0037] Vehicle events may include operational events indicating special operating modes or conditions that place a greater-than-normal load on the vehicle 1, such as driving at high speed above a speed threshold and for more than a certain duration of travel, an impact due to driving on uneven surfaces or striking an environmental object, a driving condition such as rain of a predetermined intensity for a predetermined minimum period of time, and the like. Other vehicle events may, for example, indicate times of software updates, the input of driver preferences, and the like.

[0038] Furthermore, a vehicle event may include an error event that is detected, for example, by an automatic monitoring function in one of the control units.

[0039] Furthermore, vehicle events can also include diagnostic results obtained, for example, during a vehicle visit to the workshop for repair or maintenance. These diagnostic results can include, for example, voltage measurements, the reading of an error log, and the like.

[0040] The detected vehicle events can be transmitted to a central unit 3 with a corresponding timestamp indicating the time of occurrence of the corresponding vehicle event. This central unit logs the chronological sequence of vehicle events that have occurred. Alternatively, any error events that have occurred can also be appropriately recorded in an internal vehicle error log.

[0041] Vehicle events are stored with appropriate timestamps that indicate the occurrence of the corresponding event from the beginning of the vehicle's lifetime. The timestamp can indicate or correspond to a current date and time, a time indication of the time elapsed since the beginning of the vehicle's lifetime, or the vehicle's mileage.

[0042] A classification model 31 can be implemented in the central unit 3, which is designed with a recurrent data-based model, such as an LSTM, GRU, transformer model, or the like. The classification model 31 makes it possible to evaluate the time series of vehicle events and provide an output vector whose elements are each assigned to a possible fault cause. The respective element value indicates a probability for the respective fault cause and thus makes it possible to identify probable fault causes. The classification model 31 is accordingly trained as a recurrent model and can have a structure as shown, for example, in Fig. 2. The method for evaluating the data-based classification model 31 is described in conjunction with Fig. 3.

[0043] In step S1, the vehicle events are recorded and stored as described above.

[0044] In step S2, the vehicle events F1 ... Fn can be saved with time stamps.

[0045] Alternatively, the time series of vehicle events F1 ... Fn can be summarized in consecutive time windows Z1 ... Zm, whose duration can, for example, range from one second to one hour. The time windows serve to aggregate the vehicle events F1 ... Fn occurring within them and map them to an input vector E or an input matrix. The mapping is performed using an assignment table that assigns the different vehicle events to a code or a corresponding element of the input vector.

[0046] The mapping can be done in such a way that the occurrence or non-occurrence of vehicle events is marked with an element value of 0 or 1, and measured values ​​are assigned accordingly as an element value to an element of the input vector. If several vehicle events within a time window affect the same element of the input vector E, the corresponding values ​​can be aggregated as a sum, mean, or maximum value, depending on the type of element. Furthermore, the input vector can also specify a description and / or the type of error that has occurred, for example in the form of information about a defective component, such as an alternator, a display unit, and the like, or an information about the type of error, such as a failure of a function, such as a missing or stuck speedometer, etc.

[0047] The recurrent data-based classification model 31 now evaluates the time series of the successive input vectors E one after the other in step S3, whereby an internal state of the classification model 31 is updated with each evaluation.

[0048] At the end of the sequence of input vectors E1...Em, the internal state Z can be used in step S4 to perform a classification. For example, using at least one fully connected layer 32, the internal state can be mapped to a number of labels, each of which indicates a possible cause of the error that has occurred. Using a softmax function 33, the probabilities of the classification result can be determined from the labels. The probabilities are specified for the possible, classified error causes.

[0049] Furthermore, the number of diagnostic events during a workshop visit can vary considerably, as some vehicles can be repaired very quickly after very few examinations, while others require considerable effort.

[0050] For classification models trained as LSTMs, a technique known as truncated backpropagation through time can be applied. Backpropagation through time is a version of the backpropagation method used to train modern recurrent neural networks such as the LSTM. Truncated backpropagation through time is a modified version of this approach that is more efficient for sequence prediction problems with very long sequences. It divides long sequences of diagnostic events into consecutive subsequences and treats them separately.

[0051] The training of the classification model 31 can be carried out accordingly using training data sets from vehicles that, after a period of operation, exhibit a fault with a specific known fault cause. By recording sequences of vehicle events in these vehicles, these can be assigned to the corresponding fault cause, resulting in a corresponding sequence of input vectors for the time windows that can be assigned to the fault cause. The training data sets are thus then determined or defined by the time series of the input vectors and the fault cause (as a label). When providing the training data sets in this form, it is assumed that the previously occurred vehicle events lead to the corresponding fault event with a certain probability. Thus, such training data sets can be used to train the classification model.

[0052] In a subsequent step S5, a cause of the error can be output for the described errors. This can, for example, lead to a replaced component of the vehicle, the function of which was impaired according to the error description,

Claims

Claims 1. Computer-implemented method for detecting the cause of a fault that has occurred in a vehicle, comprising the following steps: recording (S1) vehicle events in a chronological sequence during the lifetime of the vehicle; In particular, after the occurrence of an error, creating (S2) a temporal sequence of input vectors (E) in which the detected vehicle events and a description of the error that occurred are coded; Providing a recurrent data-based classification model that is trained to assign the temporal sequence of input vectors (E) to an output vector with information on a probability of the error cause of the error that occurred; Evaluating (S3, S4) the recurrent data-based classification model depending on the created sequence of input vectors (E) in order to determine the most probable cause of an error that has occurred; Signaling (S5) the most probable cause of the error that occurred.

2. The method of claim 1, wherein the vehicle events include operational events, repair events, fault events, and diagnostic events.

3. Method according to claim 1 or 2, wherein the input vectors (E) further contain vehicle master data, in particular information about the date of manufacture, a component, special equipment and / or about a manufacturing plant.

4. The method according to claim 2, wherein the operating events indicate the events characterizing the driver-individual operation, in particular over-the-air updates, triggering of the alarm system, a detected temperature and / or a detected voltage, wherein fault events indicate faults detected by an automatic monitoring function in the control units, wherein repair events correspond to events of a specific repair or replacement of a vehicle part, and wherein diagnostic events correspond to examination results of workshop examinations.

5. Method according to one of claims 1 to 4, wherein the vehicle events recorded during the lifetime, each provided with a time stamp, a mileage indication or an incrementing event counter according to the sequence of occurrence, are stored in a vehicle-external central unit and / or in a memory in the vehicle.

6. The method according to any one of claims 1 to 5, wherein the sequence of input vectors (E) is created by assigning vehicle events in successive time windows to an input vector (E), wherein the vehicle events are each coded to at least one element of the input vector (E), wherein element values determined by a plurality of vehicle events within the respective time window are linked to one another.

7. The method according to one of claims 1 to 6, wherein the data-based classification model (31) is provided with a recurrent neural network, in particular in the form of an LSTM (Long Short-Term Memory), GRU or Transformer.

8. The method according to claim 7, wherein an internal state vector of the data-based classification model (31) is mapped to an output vector using at least one fully connected layer (32), the elements of which each indicate a probability of a possible error cause of the error that has occurred, wherein the most probable error cause is determined using a softmax function (33) applied to the output vector.

9. Method according to one of claims 1 to 8, wherein the signaling of the most likely cause of the error is determined by issuing a description of the cause of the error to a user, whereby, depending on the cause of the error, it is stated whether a component or a function for which the error is described is the source of the error that has occurred or not.

10. Device (3) designed to carry out the method according to one of claims 1 to 9.

11. A computer program comprising instructions which, when executed in a device according to claim 9, cause the device to perform the steps recited therein.

12. A machine-readable storage medium on which a computer program according to claim 11 is stored.

Images