Fault early warning method and model training method, device, vehicle and storage medium

Abstract

The application relates to the field of intelligent automobiles, in particular to a fault early warning method and model training method and device, a vehicle and a storage medium. The method is applied to a vehicle terminal; the vehicle terminal comprises a fault early warning model; the method comprises the following steps: acquiring state data of the vehicle in a first preset time period; determining M start-stop time intervals of the vehicle in the first preset time period according to the change trend of the battery current; dividing the state data in the first preset time period according to the M start-stop time intervals to obtain M state data sets; aggregating the state data in the M state data sets to obtain a target state data set; and determining a fault early warning result according to the target state data set and the fault early warning model. Therefore, the problem that the accuracy of fault early warning is not high due to poor representativeness of the division of the state data of the vehicle when the parts of the automobile are subjected to fault early warning can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a fault early warning method, a model training method, a device, a vehicle and a storage medium. BACKGROUND

[0002] In recent years, with the rapid development of the automobile industry, the integration and automation of automobile parts are becoming higher and higher, and the complexity of the structure is also increasing. Under this situation, the reliability, safety and maintainability of automobile parts are becoming more and more important. Therefore, the fault early warning method based on automobile parts has become a research hotspot in the automobile industry. In the related technology, the fault early warning method of automobile parts mainly relies on whether the signal value collected by the vehicle terminal reaches the set early warning threshold to make a judgment. However, if the vehicle sends an alarm signal after judging that the signal value collected by the vehicle terminal reaches the set early warning threshold, there will be a problem of insufficient reaction time of the driver or the vehicle has already failed, which brings a poor user experience.

[0003] To solve the above problems, the related technology proposes a cloud-based intelligent part fault early warning algorithm constructed based on historical data and real-time data of the automobile. However, this algorithm uses a fixed time threshold to determine the start-stop time interval of the vehicle, and then divides the state data of the vehicle according to the start-stop time interval. There will be a problem that the size of the fixed time threshold is difficult to determine or the state data of the vehicle in one start-stop time interval is incomplete, which leads to poor representativeness of the state data when the state data of the vehicle is divided, and further leads to the problem of low accuracy of fault early warning. SUMMARY

[0004] The present application provides a fault early warning method, a model training method, a device, a vehicle and a storage medium to at least solve the problem that if the division of the state data of the vehicle is not representative when the fault early warning of the parts of the automobile is performed, the accuracy of the fault early warning is low. The technical solutions of the present application are as follows:

[0005] According to a first aspect of the present application, a fault early warning method is provided, applied to a vehicle terminal; the vehicle terminal comprises a fault early warning model; the method comprises: acquiring state data of the vehicle in a first preset time period; the state data comprises a change trend of a battery current of the vehicle; determining M start-stop time intervals of the vehicle in the first preset time period according to the change trend of the battery current; M is an integer greater than or equal to 1; dividing the state data in the first preset time period according to the M start-stop time intervals to obtain M state data sets; wherein one state data set is the state data of the vehicle in one start-stop time interval; aggregating the state data in the M state data sets to obtain a target state data set; determining a fault early warning result according to the target state data set and the fault early warning model.

[0006] According to the above technical means, compared with using a fixed time threshold to determine the start-stop time interval of the vehicle in the related art, and then dividing the state data of the vehicle according to the start-stop time interval, there is a problem that the size of the fixed time threshold is difficult to determine or the state data of the vehicle in one start-stop time interval is incomplete. The method provided in the present application determines the start-stop time interval of the vehicle through the change trend of the battery current of the vehicle and divides the state data of the vehicle, which can avoid dividing the state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the state data, and further improve the accuracy of fault warning and the user's use experience.

[0007] At the same time, compared with inputting the state data in a single start-stop time interval into the fault warning model to perform fault warning on the vehicle in the related art, the influence of the state data in the historical start-stop time interval of the vehicle on the fault warning of the vehicle is not considered. The method provided in the present application aggregates the state data in the M state data sets to obtain a target state data set, and inputs the target state data set into the fault warning model to obtain a fault warning result, which can make the result of the fault warning of the vehicle more accurate and reliable.

[0008] In a possible implementation, the M start-stop time intervals of the vehicle in the first preset time period are determined according to the change trend of the battery current, including: determining M+1 time nodes at which the battery current is suddenly changed in the first preset time period according to the change trend of the battery current; and dividing the first preset time period into M start-stop time intervals according to the M+1 time nodes at which the battery current is suddenly changed.

[0009] According to the above technical means, compared with dividing the start-stop time interval of the vehicle according to the time of starting and extinguishing the vehicle, or dividing the start-stop time interval of the vehicle through a fixed time threshold in the related art, the state data of the vehicle is not accurately divided. The method provided in the present application obtains the change trend of the battery current, and divides the start-stop time interval according to the time nodes at which the battery current is suddenly changed, which can make the division of the state data of the vehicle more accurate, and further improve the accuracy of fault warning.

[0010] In another possible implementation, the state data in the M state data sets is aggregated to obtain a target state data set, including: respectively aggregating each type of state data in the M state data sets to obtain target state data of each type of state data in the M state data sets; and obtaining the target state data set according to the target state data of each type of state data in the M state data sets.

[0011] According to the above technical means, it can be understood that the fault state of the parts of the vehicle depends not only on the state data in the current start-stop time interval of the vehicle, but also on the state data in the historical start-stop time interval of the vehicle. If the state data in a single start-stop time interval is input into the fault early warning model to perform fault early warning on the vehicle, the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle is ignored, which may cause the accuracy of fault early warning to be low. Therefore, the method provided in the present application considers the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle, and by aggregating the state data in the plurality of state data sets, the fault features of the obtained target state data are more obvious and can better reflect the current state of the vehicle. By taking the target state data as the input of the fault early warning model, the input state data is more representative, and the accuracy of fault early warning is improved.

[0012] According to a second aspect of the present application, a model training method is provided. The method comprises: obtaining state data of a vehicle in a second preset time period; the second preset time period includes a period of time before vehicle maintenance and a period of time after vehicle maintenance; the state data includes a change trend of a battery current of the vehicle; determining N start-stop time intervals of the vehicle in the second preset time period according to the change trend of the battery current; N is an integer greater than or equal to 1; dividing the state data in the second preset time period according to the N start-stop time intervals to obtain N state data sets; wherein one state data set is the state data of the vehicle in one start-stop time interval; constructing a training sample according to the N state data sets; training a fault early warning model according to the training sample to obtain a trained fault early warning model; and the fault early warning model is used to issue a fault early warning result to a user.

[0013] According to the above technical means, unlike in the related art, when training the fault early warning model, a fixed time threshold is usually used to determine the historical start-stop time interval of the vehicle, and the historical state data of the vehicle is divided according to the historical start-stop time interval of the vehicle. There may be a problem that it is difficult to determine the fixed time threshold, or the historical state data of the vehicle in one start-stop time interval is incomplete. The method provided in the present application determines the historical start-stop interval of the vehicle and divides the historical state data of the vehicle by the change trend of the current of the battery of the vehicle, which can avoid dividing the historical state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the historical state data, make the state data in each training sample more accurate and representative, and further make the fault early warning model trained according to the training sample more reliable, and improve the accuracy of the fault early warning result.

[0014] In a possible implementation, the N start-stop time intervals of the vehicle in the second preset time period are determined according to the change trend of the battery current, including: determining N+1 time nodes at which the battery current changes abruptly in the second preset time period according to the change trend of the battery current; and dividing the second preset time period into N start-stop time intervals according to the N+1 time nodes at which the battery current changes abruptly.

[0015] According to the technical means described above, compared with the related art, the historical start-stop time intervals of the vehicle are divided according to the time at which the vehicle starts and is turned off, or the historical start-stop time intervals of the vehicle are divided by a fixed time threshold, which may cause inaccurate division of the historical state data of the vehicle. The method provided in the present application determines the time nodes at which the battery current changes abruptly according to the change trend of the battery current, and divides the historical start-stop intervals according to the time nodes at which the battery current changes abruptly, which can make the division of the historical state data more accurate, and thus make the training of the fault warning model more accurate and reliable.

[0016] In another possible implementation, the training sample is constructed according to the N state data sets, including: using a sliding window algorithm to divide the N state data sets into a plurality of sample data according to a preset window length; wherein the state data set in one preset window length is aggregated as one sample data; the preset window length is the length of K start-stop time intervals; the label of each sample data is determined according to the fault condition of the vehicle in each preset window length; the label includes: fault and no fault; and the training sample is constructed according to the plurality of sample data and the label of each sample data.

[0017] According to the technical means described above, compared with the related art, the historical state data in one historical start-stop time interval and the label thereof are taken as one sample data, and the method provided in the present application aggregates the state data in the N state data sets each time by using a sliding window algorithm, and takes the aggregated state data and the label thereof as one sample data, which can improve the representativeness of the sample data, make the sample data better reflect the state of the vehicle at that time, and make the training sample more accurate and reliable.

[0018] According to a third aspect of the present application, a fault early warning device is provided, which is applied to a vehicle terminal; the vehicle terminal comprises a fault early warning model; the fault early warning device comprises: an acquisition module, configured to acquire state data of the vehicle in a first preset time period; the state data comprises a variation trend of a battery current of the vehicle; a determination module, configured to determine M start-stop time intervals of the vehicle in the first preset time period according to the variation trend of the battery current; M is an integer greater than or equal to 1; a division module, configured to divide the state data in the first preset time period according to the M start-stop time intervals to obtain M state data sets; wherein one state data set is state data of the vehicle in one start-stop time interval; an aggregation module, configured to aggregate the state data in the M state data sets to obtain a target state data set; and the determination module is further configured to determine a fault early warning result according to the target state data set and the fault early warning model.

[0019] In a possible implementation, the determination module is specifically configured to determine M+1 time nodes at which the battery current is mutated in the first preset time period according to the variation trend of the battery current; and divide the first preset time period into the M start-stop time intervals according to the M+1 time nodes at which the battery current is mutated.

[0020] In a possible implementation, the aggregation module is specifically configured to aggregate each type of state data in the M state data sets respectively to obtain target state data of each type of state data in the M state data sets; and obtain the target state data set according to the target state data of each type of state data in the M state data sets.

[0021] According to a fourth aspect of the present application, a model training device is provided, which comprises: an acquisition module, configured to acquire state data of a vehicle in a second preset time period; the second preset time period comprises a period of time before vehicle maintenance and a period of time after vehicle maintenance; the state data comprises a variation trend of a battery current of the vehicle; a determination module, configured to determine N start-stop time intervals of the vehicle in the second preset time period according to the variation trend of the battery current; N is an integer greater than or equal to 1; a division module, configured to divide the state data in the second preset time period according to the N start-stop time intervals to obtain N state data sets; wherein one state data set is state data of the vehicle in one start-stop time interval; a construction module, configured to construct a training sample according to the N state data sets; and a training module, configured to train a fault early warning model according to the training sample to obtain a trained fault early warning model; and the fault early warning model is configured to send a fault early warning result to a user.

[0022] In a possible implementation, the determining module is specifically configured to determine N+1 time nodes at which the battery current is mutated in the second preset time period according to a change trend of the battery current; and divide the second preset time period into N start-stop time intervals according to the N+1 time nodes at which the battery current is mutated.

[0023] In a possible implementation, the constructing module is specifically configured to divide the N state data sets into a plurality of sample data according to a preset window length by using a sliding window algorithm; wherein, a state data set in one preset window length is aggregated into one sample data; the preset window length is the length of K start-stop time intervals; determine a label of each sample data according to a fault condition of the vehicle in each preset window length; the label includes: fault and no fault; and construct the training sample according to the plurality of sample data and the label of each sample data.

[0024] According to a fifth aspect of the present application, a vehicle is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method of the first aspect or the second aspect or any possible implementation thereof.

[0025] According to a sixth aspect of the present application, a computer readable storage medium is provided, when the instructions in the computer readable storage medium are executed by the processor of the electronic device, the electronic device can execute the method of the first aspect or the second aspect or any possible implementation thereof.

[0026] Therefore, the above technical features of the present application have the following beneficial effects:

[0027] (1) Compared with the related art which uses a fixed time threshold to determine the start-stop time interval of the vehicle, and then divides the state data of the vehicle according to the start-stop time interval, there is a problem that the size of the fixed time threshold is difficult to determine or the state data of the vehicle in one start-stop time interval is incomplete. The method provided by the present application determines the start-stop time interval of the vehicle and divides the state data of the vehicle by the change trend of the battery current of the vehicle, which can avoid dividing the state data originally in one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the state data, and further improve the accuracy of fault warning and the user experience. At the same time, compared with the related art which inputs the state data in a single start-stop time interval into a fault warning model to perform fault warning on the vehicle, without considering the influence of the state data in the historical start-stop time interval of the vehicle on the fault warning of the vehicle, the method provided by the present application aggregates the state data in the plurality of state data sets to obtain a target state data set, and inputs the target state data set into the fault warning model to obtain a fault warning result, which can make the result of the fault warning of the vehicle more accurate and reliable.

[0028] (2)It can be understood that the fault state of the parts of the vehicle is not only related to the state data in the current start-stop time interval of the vehicle, but also related to the state data in the historical start-stop time interval of the vehicle. If the state data in a single start-stop time interval is input into the fault early warning model to perform fault early warning on the vehicle, the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle is ignored, which may cause the problem of low accuracy of fault early warning. Therefore, the method provided in the present application considers the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle. By aggregating the state data in the plurality of state data sets, the fault features of the obtained target state data are more obvious, and the current state of the vehicle is better reflected. By taking the target state data as the input of the fault early warning model, the input state data is more representative, and the accuracy of fault early warning is improved.

[0029] (3)Unlike the related art, when training the fault early warning model, a fixed time threshold is usually used to determine the historical start-stop time interval of the vehicle, and the historical state data of the vehicle is divided according to the historical start-stop time interval of the vehicle. There may be a problem that the fixed time threshold is difficult to determine, or the historical state data of the vehicle in one start-stop time interval is incomplete. The method provided in the present application determines the historical start-stop interval of the vehicle and divides the historical state data of the vehicle by the change trend of the current of the battery of the vehicle. This can avoid dividing the historical state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the historical state data, make the state data in each training sample more accurate and representative, and thus make the fault early warning model trained according to the training sample more reliable and improve the accuracy of the fault early warning result.

[0030] (4)Compared with the related art, the historical start-stop time interval of the vehicle is divided according to the time of starting and extinguishing the vehicle, or the historical start-stop time interval of the vehicle is divided by a fixed time threshold, which may cause the problem of inaccurate division of the historical state data of the vehicle. The method provided in the present application determines the time node at which the current of the battery changes abruptly by the change trend of the current of the battery, and divides the historical start-stop interval according to the time node at which the current of the battery changes abruptly. This can make the division of the historical state data more accurate, and thus make the training of the fault early warning model more accurate and reliable.

[0031] (5) Compared with the related art, the historical state data and the label thereof in a historical start-stop time interval are taken as a piece of sample data, the method provided in the present application adopts a sliding window algorithm to aggregate the state data in N state data sets each time, and takes the aggregated state data and the label thereof as a piece of sample data, so that the representativeness of the sample data can be improved, the sample data can better reflect the state of the vehicle at the time, and the training sample is more accurate and reliable.

[0032] It should be noted that the technical effects brought by any implementation manner in the third aspect to the sixth aspect can refer to the technical effects brought by the corresponding implementation manner in the first aspect or the second aspect, which will not be repeated here.

[0033] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0034] The accompanying drawings, which are incorporated into and form part of the specification, illustrate an embodiment consistent with the present application and, together with the specification, serve to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0035] Figure 1 is a structural diagram of a fault early warning system according to an exemplary embodiment;

[0036] Figure 2 is a flowchart of a fault early warning method according to an exemplary embodiment;

[0037] Figure 3 is a flowchart of another fault early warning method according to an exemplary embodiment;

[0038] Figure 4 is a flowchart of another fault early warning method according to an exemplary embodiment;

[0039] Figure 5 is a flowchart of another fault early warning method according to an exemplary embodiment;

[0040] Figure 6 is a flowchart of another fault early warning method according to an exemplary embodiment;

[0041] Figure 7 is a flowchart of a model training method according to an exemplary embodiment;

[0042] Figure 8 is a flowchart of another model training method according to an exemplary embodiment;

[0043] Figure 9is a flow chart of another model training method according to an example embodiment;

[0044] Figure 10 is a flow chart of another model training method according to an example embodiment;

[0045] Figure 11 is a flow chart of a fault training method and a model training method according to an example embodiment;

[0046] Figure 12 is a structural diagram of a fault early warning device according to an example embodiment;

[0047] Figure 13 is a structural diagram of a model training device according to an example embodiment;

[0048] Figure 14 is a structural diagram of a vehicle according to an example embodiment.

[0049] Among them, the sensor 100, the processor 200, the training device 300, the execution unit 400, the fault early warning device 500, the acquisition module 501, the determination module 502, the division module 503, the aggregation module 504, the model training device 600, the acquisition module 601, the determination module 602, the construction module 603, the training module 604, the vehicle 700, the processor 701, the memory 702. DETAILED DESCRIPTION

[0050] In order to make the ordinary person skilled in the art better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings.

[0051] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. The embodiments described in the following example embodiments do not represent all the embodiments consistent with the present application. On the contrary, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0052] The fault early warning method, model training method, device, vehicle and storage medium are described below with reference to the accompanying drawings. With the development of artificial intelligence, machine learning and deep learning technologies have made great progress. The state data of the vehicle is obtained based on the cloud, and an intelligent component fault early warning algorithm is constructed according to the state data of the vehicle. The vehicle can be warned in time before the vehicle fails, and the accuracy and speed of fault early warning can be effectively improved. However, in related technologies, when extracting the state data of the vehicle in the cloud, a fixed window is usually used to divide the state data of the vehicle. However, the state data of the vehicle is strongly associated with each start and stop of the vehicle. If a fixed window is used, the data across the start and stop intervals will be fragmented, and the size of the fixed window is difficult to determine. This will result in poor representativeness of the division of the state data and affect the accuracy of the fault early warning.

[0053] To solve the above problems, related technologies propose a data aggregation method for dividing the state data of each vehicle according to a stop time threshold of the vehicle. This method solves the problem of fragmented data across the start and stop intervals to some extent. However, after the vehicle is turned off, it may not be in a dormant state or may be abnormally awakened. After the vehicle is turned off for a period of time, there will still be state data transmitted to the cloud. If the start and stop time intervals of the vehicle are determined according to the stop time threshold of the vehicle, and the state data of the vehicle is divided according to the start and stop time intervals, the state data of the vehicle after being turned off for a period of time may be divided into the next start and stop time interval, which will affect the accuracy of the fault early warning.

[0054] Therefore, to solve the above problems, the present application provides a fault early warning method. The method provided by the present application determines the start and stop time intervals of the vehicle and divides the state data of the vehicle according to the change trend of the battery current of the vehicle. This can avoid dividing the state data originally belonging to one start and stop time interval into different start and stop time intervals, improve the accuracy of the division of the state data, and further improve the accuracy of the fault early warning and the user experience.

[0055] Meanwhile, compared with the related technologies in which the state data in a single start and stop time interval is input into a fault early warning model to perform fault early warning on the vehicle, the present application does not consider the influence of the state data in the historical start and stop time interval of the vehicle on the fault early warning of the vehicle. The method provided by the present application aggregates the state data in multiple state data sets to obtain a target state data set, and inputs the target state data set into a fault early warning model to obtain a fault early warning result. This can make the result of the fault early warning on the vehicle more accurate and reliable.

[0056] For ease of understanding, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0057] Figure 1 A fault early warning system is provided in the embodiments of the present application, and the fault early warning system comprises a sensor 100, a processor 200, a training device 300 and an execution unit 400. The sensor 100, the processor 200 and the execution unit 400 are in communication connection, and the processor 200 and the training device 300 are in communication connection.

[0058] In some embodiments, the processor 200 can be integrated in the sensor 100 as a sub-module of the sensor 100, or the sensor 100 and the processor 200 can be two independent devices.

[0059] The sensor 100 is configured to collect state data of a vehicle.

[0060] The state data of the vehicle comprises a variation trend of a battery current of the vehicle.

[0061] In some embodiments, the categories of the state data comprise a voltage of the battery, a current of the battery, a temperature of the battery, an outside temperature of the vehicle, an inside temperature of the vehicle, a power supply state feedback, an engine speed, an engine temperature and a sensor temperature.

[0062] The battery sensor is configured to collect the state data of the voltage of the battery, the current of the battery and the temperature of the battery of the vehicle; the power supply sensor is configured to collect the state data of the power supply feedback of the vehicle; the temperature sensor is configured to collect the state data of the inside temperature of the vehicle, the outside temperature of the vehicle, the engine temperature and the sensor temperature; and the speed sensor is configured to collect the state data of the engine speed of the vehicle.

[0063] It can be understood that the number of sensors installed in different vehicles is different, and in the method provided in the embodiments of the present application, the sensor 100 can be one or more, which can be selected according to actual needs by those skilled in the art, and the embodiments of the present application do not limit this.

[0064] The processor 200 is configured to acquire the state data of the vehicle collected by the sensor 100.

[0065] In some embodiments, the processor 200 is further configured to send the state data of the vehicle collected from the sensor 100 to the training device 300.

[0066] The training device 300 is configured to train a fault early warning model. For example, the training device 300 can be a server, which can be a single server or a server cluster composed of multiple servers. In some embodiments, the server cluster can also be a distributed cluster.

[0067] Further exemplary, the training device 300 can be a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, and a cellular phone, a personal digital assistant (PDA), an augmented reality (AR) \ virtual reality (VR) device, etc. The specific form of the training device 300 is not specially limited in the present application. The training device 300 can interact with the user through one or more of a keyboard, a touchpad, a touch screen, a remote controller, voice interaction, or a handwriting device.

[0068] In some embodiments, the training device 300 can send the fault early warning result of the fault early warning model to the processor 200.

[0069] The execution unit 400 is configured to receive the air conditioner running parameter sent by the processor 200, and remind the user according to the fault early warning result.

[0070] Figure 2 A fault early warning method provided by the embodiments of the present application is applied to a vehicle terminal, and the vehicle terminal comprises a fault early warning model.

[0071] The present application will take the battery in the vehicle parts as an example to illustrate the specific implementation of the fault early warning method. The method comprises the following steps:

[0072] S101, acquiring state data of the vehicle in a first preset time period.

[0073] The first preset time period is a period of time before the current time. Exemplarily, the first preset time period can be 2 days or 3 days. Optionally, the first preset time period includes the current time.

[0074] The state data comprises the change trend of the battery current of the vehicle.

[0075] In some embodiments, the state data of the vehicle is uploaded to the cloud in real time, so that the state data of the vehicle in the first preset time period can be extracted from the cloud server, for example, Figure 3 The above step S101 can be implemented as the following steps S1011-S1012.

[0076] S1011, determining the category of the state data to be extracted.

[0077] In some embodiments, the cloud server determines the category of the state data to be extracted according to the state data list uploaded by the CAN and the state data with high correlation to the battery failure in the past state data. Alternatively, the category of the state data to be extracted can be determined according to the suggestion of a battery expert.

[0078] The category of the state data includes the voltage of the battery, the current of the battery, the temperature of the battery, the temperature outside the vehicle, the temperature inside the vehicle, the power state feedback, the engine speed, the engine temperature, and the sensor temperature.

[0079] For example, in this embodiment, the cloud server extracts 13 categories of state data according to the state data list uploaded by the CAN and the state data with high correlation to the battery failure. The 13 categories of state data include the voltage of the battery, the current of the battery, the temperature of the battery, the temperature outside the vehicle, the power state feedback, and the engine speed.

[0080] S1012, extracting the corresponding state data from the cloud server according to the category of the state data to be extracted.

[0081] In some embodiments, after extracting the corresponding state data from the cloud server, the state data needs to be processed. It can be understood that, since the state data of the vehicle is transmitted to the cloud server through the CAN network in the vehicle, and a series of processes such as processing and storage are performed by the cloud server, and the state data is complex, the state data of the vehicle may be lost, repeated, and abnormal during storage in the cloud server, so the vehicle data needs to be preprocessed.

[0082] For example, it is checked whether there are repeated items, null values, and abnormal values in the extracted state data. If there are repeated items, only the first data is retained in the multiple repeated data. If there are null values, the next data or the previous data is selected to fill in to avoid the first data and the last data being null values. If there are abnormal values, the abnormal values are deleted.

[0083] S102, determining M start-stop time intervals of the vehicle in the first preset time period according to the change trend of the battery current.

[0084] Wherein, M is an integer greater than or equal to 1.

[0085] It can be understood that when the vehicle starts, the battery provides a current to the starter, and at the moment when the vehicle starts, the current provided by the battery is high, and after the vehicle starts, the current provided by the battery tends to be stable, so the start-stop time interval of the vehicle can be divided according to the change trend of the battery current.

[0086] In some embodiments, as shown in Figure 4 The above step S102 can be specifically implemented as steps S1021-S1022.

[0087] S1021, according to the change trend of the battery current, determining M+1 time nodes at which the battery current mutates in the first preset time period.

[0088] It can be understood that when the vehicle starts, the battery provides a higher current to the starter, at this time the change trend of the battery current is that the current mutates, and after the starter is started, the battery provides a current to the starter, which tends to be stable, at this time the current of the battery is small and the change trend of the current is stable. Therefore, according to the change trend of the battery current, M+1 time nodes at which the battery current mutates in the first preset time period can be determined.

[0089] S1022, according to the M+1 time nodes at which the battery current mutates, dividing the first preset time period into M start-stop time intervals.

[0090] In some embodiments, the time interval between two adjacent time nodes in the M+1 time nodes is taken as a start-stop time interval. Specifically, the time node at which the battery current mutates is taken as the starting point of a start-stop time interval, and the next time node at which the battery current mutates is taken as the end point of the start-stop time interval.

[0091] Optionally, when the vehicle starts, the battery generally provides a current of 150A-200A to the starter. Therefore, the size of the current at which the battery current mutates can be set to 150A, and when the battery current reaches 150A, it is considered that the battery current mutates.

[0092] For example, the time nodes at which the battery current mutates are arranged in time sequence, the first time node at which the battery current reaches 150A is taken as the starting point of the first start-stop time interval, the second time node at which the battery current reaches 150A is taken as the end point of the first start-stop time interval, and the above-mentioned second time node is also the starting point of the second start-stop time interval. In this way, the first preset time period can be divided into M start-stop time intervals.

[0093] It can be understood that, in actual division of the first preset time period, since the time interval of each start and stop of the vehicle is not consistent, the time length of each start and stop time interval is not the same, and according to the different time length of the first preset time period and the actual start and stop of the vehicle, the number of start and stop time intervals is also different, which is not limited in the embodiments of the application.

[0094] It can be understood that, compared with the related art, the start and stop time intervals of the vehicle are divided according to the start and stop time of the vehicle, or the start and stop time intervals of the vehicle are divided by a fixed time threshold, which can cause the problem of inaccurate division of the state data of the vehicle. The method provided in the application can divide the start and stop time intervals according to the time node at which the battery current changes abruptly, which can make the division of the state data of the vehicle more accurate, and thus improve the accuracy of fault warning.

[0095] S103, dividing the state data in the first preset time period according to the M start and stop time intervals to obtain M state data sets.

[0096] Among them, one state data set is the state data of the vehicle in one start and stop time interval.

[0097] In some embodiments, as shown in Figure 5 The above step S103 can be specifically implemented as steps S1031-S1033.

[0098] S1031, dividing the state data in the first preset time period according to the M start and stop time intervals.

[0099] Specifically, the M start and stop time intervals are arranged in time sequence, and the state data in the first preset time period is divided into different start and stop intervals according to time. Among them, the category of the state data contained in each start and stop interval is consistent.

[0100] S1032, extracting the features of the state data in each start and stop interval, and normalizing the features of the state data.

[0101] In some embodiments, the above step S1032 can be specifically implemented as the following steps:

[0102] Step a1, extracting the features of each type of state data in each start and stop interval in time domain and frequency domain.

[0103] For example, the significant characteristics of the state data can be analyzed by analyzing the features of the state data in each start and stop time interval that change with time.

[0104] Optionally, if the characteristics of the state data do not change significantly in the time domain or the extraction effect is not good, the characteristics of the state data can be extracted in the frequency domain through Fourier transform.

[0105] For example, taking the battery voltage in the state data as an example, the characteristics of the battery voltage in the time domain include: maximum value, minimum value, mean value, variance, kurtosis, skewness, and waveform factor; and the characteristics of the battery voltage in the frequency domain include: maximum value, minimum value, mean value, center frequency, average frequency, and root mean square frequency.

[0106] For example, the characteristics of each type of state data extracted in the time domain and the frequency domain are 126 in total.

[0107] It can be understood that, although the length of each start-stop time interval is different, the type of state data in each start-stop time interval, and the characteristics and the number of characteristics extracted from each type of state data are the same.

[0108] Step a2, screening the characteristics of the state data that need to be normalized, and performing normalization processing on the characteristics.

[0109] In some embodiments, since the characteristics of the state data are divided into dimensional characteristics and dimensionless characteristics, normalization processing is needed. The normalization processing methods include: linear function normalization method, maximum and minimum standardization method, and neural network normalization method.

[0110] Optionally, in the embodiments of the present application, the normalization processing method can be the maximum and minimum standardization method.

[0111] S1033, according to the M start-stop time intervals, the normalized state data in the first preset time period is divided to obtain M state data sets.

[0112] S104, aggregating the state data in the M state data sets to obtain a target state data set.

[0113] In some embodiments, as shown in Figure 6 The above step S104 can be specifically implemented as steps S1041-S1042.

[0114] S1041, respectively aggregating each type of state data in the M state data sets to obtain the target state data of each type of state data in the M state data sets.

[0115] In some embodiments, each characteristic of each type of state data in the M state data sets is aggregated to obtain the target state data of each type of state data in the M state data sets.

[0116] The mean value can be used to aggregate each feature of each type of state data.

[0117] For example, the maximum value of the battery voltage in the M state data sets is aggregated by using the mean value to obtain the target state feature of the maximum value of the battery voltage in the M state data sets. The peak value of the battery voltage in the M state data sets is aggregated by using the mean value to obtain the target state feature of the peak value of the battery voltage in the M state data sets. Similarly, the target state features of all features of the battery voltage are obtained as the target state data of the battery voltage.

[0118] For example, the minimum value of the battery temperature in the M state data sets is aggregated by using the mean value to obtain the target state feature of the minimum value of the battery temperature in the M state data sets. The peak value of the battery temperature in the M state data sets is aggregated by using the mean value to obtain the target state feature of the peak value of the battery temperature in the M state data sets. Similarly, the target state features of all features of the battery temperature are obtained as the target state data of the battery temperature.

[0119] S1042, obtaining a target state data set according to the target state data of each type of state data in the M state data sets.

[0120] In some embodiments, the target state data of each type of state data obtained in the above step S1041 is collected as the target state data set.

[0121] It can be understood that the fault state of the parts of the vehicle depends not only on the state data in the current start-stop time interval of the vehicle, but also on the state data in the historical start-stop time interval of the vehicle. If the state data in a single start-stop time interval is input into the fault warning model to perform fault warning on the vehicle, the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle is ignored, which may cause the problem of low accuracy of fault warning. Therefore, the method provided in the present application considers the influence of the state data in the historical start-stop time interval on the fault state of the parts of the vehicle. By aggregating the state data in the multiple state data sets, the fault features of the obtained target state data are more obvious, and the current state of the vehicle is better reflected. By inputting the target state data into the fault warning model, the input state data is more representative, and the accuracy of fault warning is improved.

[0122] S105, determining a fault warning result according to the target state data set and the fault warning model.

[0123] In some embodiments, the target state data set is input into the fault early warning model to obtain a fault early warning result of the vehicle. For example, the method provided in the present application is described by taking the battery as an example. Therefore, after the target state data set in the embodiment of the present application is input into the fault early warning model, the fault early warning result of the battery can be obtained. The fault early warning result includes: fault and non-fault.

[0124] Compared with the related art, using a fixed time threshold to determine the start-stop time interval of the vehicle, and then dividing the state data of the vehicle according to the start-stop time interval, there is a problem that the size of the fixed time threshold is difficult to determine or the state data of the vehicle in one start-stop time interval is incomplete. The method provided in the present application determines the start-stop time interval of the vehicle by the change trend of the battery current of the vehicle and divides the state data of the vehicle, which can avoid dividing the state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the state data, and further improve the accuracy of fault early warning and the user experience.

[0125] Meanwhile, compared with the related art, the state data in a single start-stop time interval is input into the fault early warning model to perform fault early warning on the vehicle, without considering the influence of the state data in the historical start-stop time interval of the vehicle on the fault early warning of the vehicle. The method provided in the present application aggregates the state data in the plurality of state data sets to obtain a target state data set, and inputs the target state data set into the fault early warning model to obtain a fault early warning result, which can make the fault early warning result of the vehicle more accurate and reliable.

[0126] Figure 7 For the model training method provided in the embodiment of the present application, the present application will take the battery in the vehicle component as an example to describe the specific implementation of the model training method. The method includes the following steps: S201-S205.

[0127] S201, obtaining state data of the vehicle in a second preset time period.

[0128] The second preset time period includes a period of time before and after the vehicle is repaired; the state data includes the change trend of the battery current of the vehicle.

[0129] In some embodiments, according to the vehicle after-sales repair record card, the repair time of the vehicle with battery failure and the vehicle frame number of the vehicle are determined, and then the state data of the vehicle in a period of time before and after the vehicle is repaired is extracted from the cloud server according to the vehicle frame number.

[0130] For example, in this embodiment of the application, status data of 500 vehicles with battery failures are extracted. The status data includes status data from three months prior to repair and two months after repair.

[0131] In some embodiments, the specific implementation of obtaining the vehicle's status data within a second preset time period can be referred to steps S1011-S1012 above.

[0132] S202. Based on the changing trend of the battery current, determine the N start-stop time intervals of the vehicle within the second preset time period.

[0133] Where N is an integer greater than or equal to 1.

[0134] It is understandable that when a vehicle starts, the battery provides current to the starter motor, and the current provided by the battery is very high at the moment the vehicle starts. After the vehicle starts, the current provided by the battery tends to stabilize. Therefore, the start-stop time interval of the vehicle can be divided according to the trend of the battery current change.

[0135] In some embodiments, such as Figure 8 As shown, the above step S202 can be specifically implemented as: steps S2021-S2022.

[0136] S2021. Based on the changing trend of the battery current, determine the N+1 time nodes in the second preset time period where the battery current changes abruptly.

[0137] In some embodiments, when the vehicle is started, the battery provides a high current to the starter motor, and the battery current changes abruptly. After the starter motor is started, the current provided by the battery to the starter motor tends to stabilize, and the battery current is smaller and its change trend is more stable. Therefore, based on the change trend of the battery current, N+1 time points in the second preset time period where the battery current changes abruptly can be determined.

[0138] S2022. Based on the N+1 time points where the battery current changes abruptly, the second preset time period is divided into N start-stop time intervals.

[0139] In some embodiments, the time point at which the battery current changes abruptly is taken as the starting point of a start-stop time interval; and the next time point at which the battery current changes abruptly is taken as the ending point of the start-stop time interval.

[0140] Optionally, when the vehicle is started, the battery typically provides 150A-200A of current to the starter motor. Therefore, the current magnitude at which the battery current changes abruptly can be set to 150A; when the battery current reaches 150A, it is considered that the battery current has changed abruptly.

[0141] For example, the time nodes at which the battery current changes abruptly are arranged in chronological order, the first time node at which the battery current reaches 150 A is taken as the start point of the first start-stop time interval, the second time node at which the battery current reaches 150 A is taken as the end point of the first start-stop time interval, and the second time node is also the start point of the second start-stop time interval, and so on. In this way, the second preset time period can be divided into N start-stop time intervals.

[0142] It can be understood that, in actual division of the second preset time period, the time interval of each start-stop of the vehicle is not consistent, and thus the time length of each start-stop time interval is not the same. According to the different time length of the second preset time period and the actual start-stop of the vehicle, the number of start-stop time intervals is also different, which is not limited in the embodiments of the application.

[0143] It can be understood that, compared with the related art, the historical start-stop time interval of the vehicle is divided according to the time of starting and extinguishing the vehicle, or the historical start-stop time interval of the vehicle is divided by a fixed time threshold, which can cause the problem of inaccurate division of the historical state data of the vehicle. The method provided in the application determines the time nodes at which the battery current changes abruptly according to the change trend of the battery current, and divides the historical start-stop interval according to the time nodes at which the battery current changes abruptly, which can make the division of the historical state data more accurate, and thus make the training of the fault early warning model more accurate and reliable.

[0144] S203, dividing the state data in the second preset time period according to the N start-stop time intervals to obtain N state data sets.

[0145] In some embodiments, the specific implementation of the step S203 can refer to the steps S1031-S1033.

[0146] In some embodiments, the specific implementation of the step S203 can refer to the steps S1031-S1033.

[0147] In some embodiments, the step S203 further includes labeling the N state data sets according to the maintenance time.

[0148] For example, according to the maintenance time, the N state data sets are divided into a state data set before maintenance and a state data set after maintenance. The state data before the battery failure is more likely to be abnormal compared with the normal fault-free data. For example, the state data before the failure is lower or higher than the normal range. Therefore, taking the maintenance time in the after-sales maintenance record card as the benchmark, the state data set 15 days before maintenance is marked as 1, that is, the state data set is identified as a data set in the fault state; the remaining state data set is marked as 0, that is, the state data set is a data set in the non-fault state.

[0149] S204, constructing a training sample according to the N state data sets.

[0150] In some embodiments, as shown in the figure, Figure 9 The above step S204 can be specifically implemented as S2041-S2043.

[0151] S2041, using a sliding window algorithm, dividing the N state data sets into a plurality of sample data according to a preset window length.

[0152] Among them, the state data set in one preset window length is aggregated as a sample data; the preset window length is the length of K start-stop time intervals.

[0153] In some embodiments, the number M of start-stop time intervals of the vehicle in the first preset time period in the above step S102 and the number K of start-stop time intervals in the preset window length have the same value.

[0154] In some embodiments, when the N state data sets are divided by using the sliding window algorithm, the state data set before maintenance and the state data set after maintenance can be divided respectively according to the maintenance time, to obtain the sample data before maintenance and the sample data after maintenance.

[0155] For example, taking the state data set before maintenance as an example, among the N state data sets of the vehicle, t data sets before maintenance are sorted in time sequence, which is specifically represented as: D=D1, D2, D3, …, Dt, D t}, wherein, wherein D i represents the start-stop of the i th state data set of the vehicle, and D represents all the state data extracted by the vehicle.

[0156] Then, the step length of each sliding is set to 1, and the start pointer and the end pointer are respectively start and end, which point to the boundary of the preset window. Initially, start points to D1, and end points to D k D1 to D kaggregating K state data sets of D1 to D k+1 , to obtain the first sample data. The second time, start and end are slid back by 1 step respectively, at this time, start points to D2, and end points to D k+1 , to obtain the second sample data; and so on until end points to the last state data set D t , to obtain the last sample data.

[0157] Wherein, the state data set can be aggregated by averaging or maximum value.

[0158] For example, taking the battery voltage as an example, aggregating K state data sets of D1 to D k includes: aggregating the maximum value of the battery voltage in K state data sets by averaging to obtain the average of the maximum value of the battery voltage in K state data sets; aggregating the peak value of the battery voltage in K state data sets by averaging to obtain the maximum value of the peak value of the battery voltage in K state data sets; and so on, the average of all features of the battery voltage obtained is taken as the sample data of the battery voltage in the first sample data.

[0159] For example, taking the battery temperature as an example, aggregating K state data sets of D1 to D k includes: aggregating the minimum value of the battery temperature in K state data sets by averaging to obtain the target state feature of the minimum value of the battery temperature in N state data sets; aggregating the peak value of the battery temperature in N state data sets by averaging to obtain the target state feature of the peak value of the battery temperature in N state data sets; and so on, the target state feature of all features of the battery temperature obtained is taken as the sample data of the battery temperature in the first sample data.

[0160] Optionally, in the embodiment of the application, the preset window length can be the length of 7 start-stop time intervals.

[0161] S2042, according to the fault condition of the vehicle in each preset window length, determine the label of each sample data.

[0162] Wherein, the label includes: fault and no fault.

[0163] In some embodiments, for sample data before the repair time, the annotation result of K state data sets is determined when aggregating the state data in K state data sets each time. The number of state data sets annotated as 1 in K state data sets is counted, and according to the relationship between count and K, a preset annotation rule is used to annotate each sample data.

[0164] For example, the preset labeling rule is that when count≥K / 2, the label of the sample data is 1, that is, if half of the K state data sets are labeled as 1, the label of the aggregated sample data is 1. The label of the sample data that does not meet the preset labeling rule is 0.

[0165] In some embodiments, for sample data after the maintenance time, the label of each sample data is 0.

[0166] Wherein, 1 represents failure, and 0 represents no failure.

[0167] S2043, constructing training samples according to the plurality of sample data and the label of each sample data.

[0168] It can be understood that the failure state of the parts of the vehicle depends not only on the state data in a single start-stop time interval, but also on the state data in the historical start-stop time interval of the vehicle. Compared with the related art, which takes the state data and its label in a start-stop time interval as a sample data and ignores the influence of the state data in the historical start-stop time interval on the failure state of the parts of the vehicle, the method provided in the present application considers the influence of the state data in the historical start-stop time interval on the failure state of the parts of the vehicle. By using a sliding window algorithm, the state data in N state data sets is aggregated each time, and the aggregated state data and its label are taken as a sample data, which can make the failure features of each sample data more obvious, better reflect the failure state of the vehicle at that time, improve the representativeness of the sample data, make the training sample more accurate and reliable, and thus improve the accuracy of the failure warning model.

[0169] S205, training a failure warning model according to the training samples, to obtain a trained failure warning model.

[0170] Wherein, the failure warning model is used to issue a failure warning result to a user.

[0171] In some embodiments, as shown in Figure 10 The above step S205 can be specifically implemented as steps S2051-S2052.

[0172] S2051, dividing the training samples into a training set and a test set.

[0173] Wherein, the training set is used for training and parameter optimization of the failure warning model, and the test set is used for evaluating whether the performance of the failure warning model reaches the expected target.

[0174] Optionally, the ratio of the number of sample data of the training set and the test set can be 9:1.

[0175] S2052, select a machine learning model to train and optimize the fault early warning model, and obtain a trained fault early warning model.

[0176] Optionally, in the embodiments of the present application, the machine learning model can be an extreme gradient boosting (XGBoost) model.

[0177] It can be understood that, unlike in the related art, when training the fault early warning model, a fixed time threshold is usually used to determine the historical start-stop time interval of the vehicle, and the historical state data of the vehicle is divided according to the historical start-stop time interval of the vehicle. There is a problem that it is difficult to determine the fixed time threshold, or the historical state data of the vehicle in one start-stop time interval is incomplete. The method provided by the present application determines the historical start-stop interval of the vehicle by the change trend of the current of the vehicle battery, and divides the historical state data of the vehicle. It can avoid dividing the historical state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the historical state data, make the state data in each training sample more accurate and representative, and further make the fault early warning model trained according to the training sample more reliable, and improve the accuracy of the fault early warning result.

[0178] The above is an embodiment of the fault early warning method and model training method provided by the present application. In order to facilitate understanding, the above fault early warning method and model training method will be further described in the form of examples.

[0179] For example, as shown in the flowchart of the fault early warning method and model training method provided by the present application. Specifically, Figure 11

[0180] Step b1, extracting data before and after maintenance of the automobile battery from the cloud server.

[0181] Step b2, data cleaning is performed on the above extracted state data, and the state data is divided into different start-stop time intervals according to the current of the automobile battery.

[0182] Step b3, respectively extracting the features of the state data in each start-stop time interval in the time domain and the frequency domain, and normalizing the features.

[0183] Step b4, dividing the processed state data into before and after maintenance according to the maintenance time, and labeling the processed state data.

[0184] ​Step b5, using a sliding window algorithm, the state data in multiple start-stop time intervals is aggregated into one sample, and is re-labeled to reconstruct new training samples.

[0185] Step b6, selecting an appropriate machine learning method to establish a battery failure warning model.

[0186] Step b7, through the state data collected by the sensors of a vehicle and the historical state data of the vehicle, the current battery failure result of the vehicle can be predicted.

[0187] The above mainly introduces the scheme provided by the embodiments of the application from the method. In order to realize the above functions, the positioning device or electronic device of the vehicle includes the hardware structure and / or software module corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0188] The embodiments of the application can divide the functional modules of the positioning device or electronic device of the vehicle according to the above method, for example, the positioning device or electronic device of the vehicle can include functional modules corresponding to each functional division, or two or more functions can be integrated in one processing module. The integrated module can be realized in the form of hardware or software functional module. It should be noted that the division of modules in the embodiments of the application is illustrative, and is only a logical functional division. Actual implementation can have another division method.

[0189] Figure 12 is a block diagram of a failure warning device according to an exemplary embodiment, applied to a vehicle terminal; the vehicle terminal includes a failure warning model. The failure warning device 500 includes an acquisition module 501, a determination module 502, a division module 503 and an aggregation module 504.

[0190] The acquisition module 501 is configured to acquire state data of the vehicle in a first preset time period; the state data includes a change trend of the battery current of the vehicle.

[0191] The determination module 502 is configured to determine M start-stop time intervals of the vehicle in the first preset time period according to the change trend of the battery current; M is an integer greater than or equal to 1.

[0192] The dividing module 503 is configured to divide the state data in the first preset time period according to M start-stop time intervals to obtain M state data sets; one state data set is state data of the vehicle in one start-stop time interval.

[0193] The aggregating module 504 is configured to aggregate the state data in the M state data sets to obtain a target state data set.

[0194] The determining module 502 is further configured to determine a fault warning result according to the target state data set and a fault warning model.

[0195] In a possible implementation, the determining module 502 is specifically configured to determine M+1 time nodes at which the battery current mutates in the first preset time period according to a change trend of the battery current; and divide the first preset time period into M start-stop time intervals according to the M+1 time nodes at which the battery current mutates.

[0196] In a possible implementation, the aggregating module 504 is specifically configured to aggregate each type of state data in the M state data sets respectively to obtain target state data of each type of state data in the M state data sets; and obtain the target state data set according to the target state data of each type of state data in the M state data sets.

[0197] It can be understood that, compared with using a fixed time threshold to determine the start-stop time interval of the vehicle in the related art, and then dividing the state data of the vehicle according to the start-stop time interval, there is a problem that the size of the fixed time threshold is difficult to determine or the state data of the vehicle in one start-stop time interval is incomplete, the method provided in the present application determines the start-stop time interval of the vehicle and divides the state data of the vehicle according to the change trend of the battery current of the vehicle, which can avoid dividing the state data originally in one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the state data, and then improve the accuracy of fault warning and improve the user experience.

[0198] Meanwhile, compared with inputting the state data in a single start-stop time interval into a fault warning model to perform fault warning on the vehicle in the related art, the method provided in the present application aggregates the state data in the multiple state data sets to obtain a target state data set, and inputs the target state data set into the fault warning model to obtain a fault warning result, which can make the result of the fault warning on the vehicle more accurate and reliable.

[0199] Figure 13FIG. 6 is a block diagram of a model training apparatus 600 according to an example embodiment. The model training apparatus 600 includes an obtaining module 601, a determining module 602, a constructing module 603, and a training module 604.

[0200] The obtaining module 601 is configured to obtain state data of the vehicle in a second preset time period. The second preset time period includes a period of time before and after vehicle maintenance. The state data includes a change trend of a battery current of the vehicle.

[0201] The determining module 602 is configured to determine N start-stop time intervals of the vehicle in the second preset time period according to the change trend of the battery current. N is an integer greater than or equal to 1. The dividing module is configured to divide the state data in the second preset time period according to the N start-stop time intervals to obtain N state data sets. One state data set is state data of the vehicle in one start-stop time interval.

[0202] The constructing module 603 is configured to construct training samples according to the N state data sets.

[0203] The training module 604 is configured to train a fault warning model according to the training samples to obtain a trained fault warning model. The fault warning model is configured to issue a fault warning result to a user.

[0204] In a possible implementation, the determining module 602 is specifically configured to determine N+1 time nodes at which the battery current in the second preset time period has a sudden change according to the change trend of the battery current. The second preset time period is divided into N start-stop time intervals according to the N+1 time nodes at which the battery current has a sudden change.

[0205] In a possible implementation, the constructing module 603 is specifically configured to divide the N state data sets into a plurality of sample data by using a sliding window algorithm according to a preset window length. One state data set in one preset window length is aggregated into one sample data. The preset window length is the length of K start-stop time intervals. The label of each sample data is determined according to the fault condition of the vehicle in each preset window length. The label includes fault and no fault. The training samples are constructed according to the plurality of sample data and the label of each sample data.

[0206] It can be understood that, different from the related art, when the fault early warning model is trained, a fixed time threshold is usually used to determine the historical start-stop time interval of the vehicle, and the historical state data of the vehicle is divided according to the historical start-stop time interval of the vehicle, and there is a problem that the fixed time threshold is difficult to determine, or the historical state data of the vehicle in one start-stop time interval is incomplete. The method provided in the application determines the historical start-stop interval of the vehicle and divides the historical state data of the vehicle by the change trend of the current of the battery of the vehicle, which can avoid dividing the historical state data originally belonging to one start-stop time interval into different start-stop time intervals, improve the accuracy of dividing the historical state data, make the state data in each training sample more accurate and representative, and further make the fault early warning model trained according to the training sample more reliable and improve the accuracy of the fault early warning result.

[0207] As to the apparatus in the above-mentioned embodiments, the specific manners in which various modules perform operations have been described in detail in the embodiments of the method, and thus will not be described in detail here.

[0208] Figure 14 is a block diagram of a vehicle according to an example embodiment. As shown in Figure 14 , the vehicle 700 includes, but is not limited to, a processor 701 and a memory 702.

[0209] The memory 702 described above is configured to store executable instructions of the processor 701. It can be understood that the processor 701 is configured to execute the instructions to implement the fault early warning method and the model training method in the above-mentioned embodiments.

[0210] It should be noted that those skilled in the art can understand, Figure 14 that the vehicle structure shown in the above-mentioned embodiments does not constitute a limitation on the vehicle, and the vehicle can include more or fewer components than Figure 14 shown, or combine certain components, or different component arrangements.

[0211] The processor 701 is the control center of the vehicle, connects various parts of the vehicle through various interfaces and lines, and performs various functions of the vehicle and processes data by running or executing software programs and / or modules stored in the memory 702 and calling data stored in the memory 702, thereby monitoring the vehicle as a whole. The processor 701 can include one or more processing units. Optionally, the processor 701 can integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application programs, and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor 701.

[0212] The memory 702 can be used to store software programs and various data. The memory 702 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs (such as a determination unit, a processing unit, etc.) required by at least one function module, and the like. In addition, the memory 702 can include a high-speed random access memory, and can also include a non-volatile memory, for example, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device.

[0213] In an example embodiment, a computer readable storage medium including instructions, for example, the memory 702 including instructions, is also provided, which can be executed by the processor 701 of the vehicle 700 to implement the fault early warning method and the model training method in the above embodiments.

[0214] In actual implementation, Figure 12 the functions of the acquisition module 501, the determination module 502, the division module 503, and the aggregation module 504 in the above embodiment can be implemented by Figure 13 the functions of the acquisition module 601, the determination module 602, the construction module 603, and the training module 604 in the above embodiment can be implemented by Figure 14 the processor 701 calling a computer program stored in the memory 702. The specific execution process can refer to the description of the fault early warning method and the model training method in the above embodiments, which will not be repeated here.

[0215] Optionally, the computer readable storage medium can be a non-transitory computer readable storage medium, for example, the non-transitory computer readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a CB-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0216] In an example embodiment, the embodiments of the present application also provide a computer program product including one or more instructions, which can be executed by the processor 701 of the vehicle to complete the fault early warning method and the model training method in the above embodiments.

[0217] It should be noted that the instructions in the above computer readable storage medium or the one or more instructions in the computer program product are executed by the processor of the vehicle to implement each process of the above fault early warning method and model training method embodiments, and can achieve the same technical effects as the above fault early warning method and model training method. To avoid repetition, it will not be repeated here.

[0218] Through the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of functional modules is taken as an example, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete the full classification or part of the functions described above.

[0219] In several embodiments provided in the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0220] The units described as separate components can or can not be physically separated, and the components displayed as units can be one physical unit or multiple physical units, that is, can be located in one place or can be distributed to multiple different places. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0221] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0222] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical scheme of the embodiment of the present application essentially or the part that contributes to the prior art or the full classification or part of the technical scheme can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for making a device (which can be a single-chip microcomputer, a chip, etc.) or a processor (procbssor) execute all or part of the steps of the method of each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, and various storage program codes.

[0223] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any change or replacement within the technical scope disclosed by the present application should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A fault early warning method, characterized in that, The method is applied to a vehicle terminal; the vehicle terminal includes a fault warning model; the method includes: Acquire vehicle status data within a first preset time period; the status data includes the trend of changes in the vehicle's battery current; Based on the changing trend of the battery current, determine M+1 time points in the first preset time period where the battery current changes abruptly. Based on the M+1 time points at which the battery current changes abruptly, the first preset time period is divided into M start-stop time intervals, where M is an integer greater than or equal to 1. Based on the M start-stop time intervals, the status data within the first preset time period is divided to obtain M status datasets; wherein, one status dataset is the status data of the vehicle within one start-stop time interval; The state data within the M state datasets are aggregated to obtain the target state dataset; Based on the target state dataset and the fault warning model, the fault warning result is determined.

2. The method according to claim 1, characterized in that, The step of aggregating the state data within the M state datasets to obtain the target state dataset includes: Each type of state data in the M state datasets is aggregated to obtain the target state data for each type of state data in the M state datasets; The target state dataset is obtained based on the target state data of each type of state data in the M state datasets.

3. A model training method, characterized in that, The method includes: Acquire vehicle status data within a second preset time period; the second preset time period includes a period before and after vehicle maintenance; the status data includes the trend of changes in the vehicle's battery current; Based on the changing trend of the battery current, determine N+1 time points in the second preset time period where the battery current changes abruptly. Based on the N+1 time points at which the battery current changes abruptly, the second preset time period is divided into N start-stop time intervals, where N is an integer greater than or equal to 1. Based on the N start-stop time intervals, the status data within the second preset time period is divided to obtain N status datasets; wherein, one status dataset is the status data of the vehicle within one start-stop time interval; Based on the N state datasets, construct training samples; Based on the training samples, a fault warning model is trained to obtain a trained fault warning model; the fault warning model is used to issue fault warning results to users.

4. The method according to claim 3, characterized in that, The step of constructing training samples based on the N state datasets includes: A sliding window algorithm is used to divide the N state datasets into multiple sample data according to a preset window length; wherein, the state datasets within a preset window length are aggregated into one sample data; the preset window length is the length of K start and stop time intervals. Based on the fault status of the vehicle within each preset window length, a label is determined for each sample data entry; the label includes: faulty and not faulty; The training samples are constructed based on the multiple sample data and the label of each sample data.

5. A fault early warning device, characterized in that, Applied to a vehicle terminal; the vehicle terminal includes a fault warning model; the fault warning device includes: The acquisition module is used to acquire the vehicle's status data within a first preset time period; the status data includes the trend of the vehicle's battery current. The determining module is used to determine M start-stop time intervals of the vehicle within the first preset time period based on the changing trend of the battery current; where M is an integer greater than or equal to 1. The segmentation module is used to segment the status data within the first preset time period according to the M start-stop time intervals to obtain M status datasets; wherein, one status dataset is the status data of the vehicle within one start-stop time interval; The aggregation module is used to aggregate the state data within the M state datasets to obtain the target state dataset; The determining module is further configured to determine the fault warning result based on the target state dataset and the fault warning model; The determining module is specifically used to determine M+1 time nodes in the first preset time period where the battery current changes abruptly, based on the changing trend of the battery current; and to divide the first preset time period into M start-stop time intervals based on the M+1 time nodes where the battery current changes abruptly.

6. A model training device, characterized in that, include: The acquisition module is used to acquire the vehicle's status data within a second preset time period; The second preset time period includes a period of time before and after vehicle repair; The status data includes the trend of change in the vehicle's battery current; The determination module is used to determine N start-stop time intervals of the vehicle within the second preset time period based on the changing trend of the battery current. N is an integer greater than or equal to 1; The segmentation module is used to segment the status data within the second preset time period according to the N start-stop time intervals to obtain N status datasets; wherein, one status dataset is the status data of the vehicle within one start-stop time interval; A construction module is used to construct training samples based on the N state datasets; The training module is used to train the fault warning model based on the training samples to obtain the trained fault warning model; the fault warning model is used to issue fault warning results to the user. The determining module is specifically used to determine N+1 time nodes in the second preset time period where the battery current changes abruptly, based on the changing trend of the battery current; and to divide the second preset time period into N start-stop time intervals based on the N+1 time nodes where the battery current changes abruptly.

7. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is capable of performing the method as described in any one of claims 1 to 4.

Images