Automobile anomaly detection result interpretation method based on natural language generation

Abstract

The invention relates to the field of data processing, in particular to an automobile anomaly detection result interpretation method based on natural language generation, and the method comprises the steps: periodically carrying out the detection of all operation data of a target vehicle; determining whether to perform data exception judgment according to whether the numerical value of each operation data is within an exception data range; according to whether the target vehicle is in a data abnormal state, determining whether to perform abnormal analysis on each abnormal detection data, and according to the number of abnormal data items and the correlation coefficient, determining an abnormal analysis mode; and generating an anomaly detection result according to the anomaly reason of each anomaly detection data, and generating an anomaly detection result explanation according to the anomaly detection result, thereby improving the accuracy and content richness of the generated anomaly detection result explanation.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method for explaining automobile anomaly detection results based on natural language generation. Background Art

[0002] The results of automobile anomaly detection based on natural language processing technology can reduce the proportion of manual intervention in the analysis process, realize the automation of the generation process of anomaly detection result explanation, and reduce the cost of understanding the interpretation of anomaly detection results for non-professionals. However, the current existing methods for interpreting automobile anomaly detection results based on natural language generation can only judge the abnormal problems of the vehicle based on the anomaly detection data obtained at the current moment when obtaining anomaly detection results, and fail to trace the fault based on the anomaly detection data, resulting in low reliability of the generated anomaly detection results, and further resulting in low effectiveness of the generated anomaly detection result interpretation for non-professionals. Summary of the Invention

[0003] To this end, the present invention provides a membrane filtration sewage treatment system based on refined control to overcome the problem that the existing technology fails to trace faults for abnormal detection data, resulting in low reliability of the generated abnormal detection results, and further resulting in low effectiveness of the interpretation of the generated abnormal detection results for non-professionals.

[0004] To achieve the above objectives, the present invention provides a method for interpreting vehicle anomaly detection results based on natural language generation, comprising:

[0005] Periodically test various operating data of the target vehicle;

[0006] Determine whether to perform data anomaly judgment based on whether each operating data is within the abnormal data range to obtain abnormal detection data;

[0007] Determine whether to perform abnormality analysis on each abnormality detection data based on whether the target vehicle is in a data abnormality state, and determine the abnormality analysis method based on the number of abnormal data items and the correlation coefficient. The abnormality analysis method is to perform correlation analysis or abnormality tracing on each abnormality detection data;

[0008] Anomaly detection results are generated based on the anomaly causes of each anomaly detection data, and anomaly detection result explanations are generated based on the anomaly detection results.

[0009] Furthermore, various operating data of the target vehicle are periodically tested, wherein for any operating data, if the value of the operating data is not within the abnormal data range, a data abnormality judgment is performed on the operating data;

[0010] Each item of operating data has an abnormal data range corresponding to it.

[0011] Furthermore, the abnormality judgment of the operating data includes:

[0012] Detect the data fluctuation status of the operating data obtained from each test within the preset test period;

[0013] If the data fluctuation state of the operating data is in an abnormal fluctuation state, the operating data obtained from each detection within the preset detection period are recorded as abnormal detection data and stored in association with the corresponding abnormal fluctuation state;

[0014] If the data fluctuation state of the operating data is not in an abnormal fluctuation state, it is determined that the operating data is not abnormal detection data.

[0015] Furthermore, if it is under period triggering condition, it is determined whether to adjust for the preset detection period according to the fluctuation moment difference value;

[0016] If the fluctuation time difference value is greater than the preset fluctuation time difference value, the duration of the preset detection period is reduced according to the detection time difference value;

[0017] The reduction value of the duration of the preset detection period is positively correlated with the difference value of the detection time;

[0018] The periodic triggering condition is that the driving state change coefficient in the change detection period at the current moment is greater than the preset state change coefficient or the driving environment change coefficient is greater than the preset environment change coefficient;

[0019] The fluctuation time difference value is the interval length between the fluctuation start time and the cycle trigger time, and the detection time difference value is the interval length between the fluctuation start time and the cycle trigger time.

[0020] Furthermore, if the target vehicle is in a data abnormality state, an abnormality analysis method is determined based on the number of abnormal data items and the associated reference coefficient. The abnormality analysis method includes performing a correlation analysis on each abnormal detection data item and performing abnormality tracing on each abnormal detection data item.

[0021] The data abnormal state is that there is abnormal detection data in any item of operating data of the target vehicle.

[0022] Furthermore, the correlation reference coefficient is the average value of the correlation coefficients corresponding to each abnormality detection data;

[0023] For any one abnormality detection data, the correlation coefficient is determined according to the number of data items of the operating data that are correlated with the abnormality detection data and the degree of coincidence of abnormal periods of the operating data that are correlated.

[0024] Furthermore, if the number of abnormal data items is greater than the preset number of abnormal data items, or the correlation reference coefficient is greater than the preset correlation reference coefficient, correlation analysis is performed on each abnormal detection data item, and the abnormal detection data item is divided according to the correlation coefficient of each abnormal detection data item to obtain a plurality of data groups;

[0025] The correlation coefficient between the abnormal detection data in any data group is greater than the preset correlation coefficient;

[0026] For a single data group, the priority interpretation coefficient of the corresponding abnormal detection data is set according to the correlation coefficient of each abnormal detection data in the data group, and several abnormal detection data for abnormal cause analysis are determined according to the priority interpretation coefficient.

[0027] Furthermore, if the number of abnormal data items is less than or equal to the preset number of abnormal data items and the correlation coefficient is less than or equal to the preset correlation coefficient, abnormality tracing is performed for each abnormality detection data;

[0028] Determining a priority tracing coefficient based on the abnormal correlation degree of each operating data of the target vehicle, wherein the priority tracing coefficient is positively correlated with the abnormal correlation degree of each operating data;

[0029] For any item of operation data, the abnormality association degree is based on the number of data items of abnormality detection data that are associated with the operation data.

[0030] Furthermore, if the abnormal detection data has no correlation with other operating data, abnormal driving characteristics within the abnormal change range of the abnormal detection data that cannot be traced back to abnormality are detected, and the abnormal cause of the abnormal detection data is determined based on the abnormal driving characteristics.

[0031] Furthermore, if all abnormality detection data have completed abnormality analysis, an abnormality detection result is generated according to the abnormal cause of each abnormality detection data, and an abnormality detection result explanation is generated through the pre-trained model;

[0032] The abnormality detection results include abnormality detection database data, abnormal event description data, abnormality detection cause explanation, and vehicle body signal real-time data.

[0033] Compared with the prior art, the beneficial effect of the present invention lies in that the technical solution of the present invention determines whether to trace the abnormal cause of the abnormal detection data based on the number of abnormal data items obtained and the correlation coefficient, and analyzes the abnormal cause of the abnormal detection data in combination with the environmental factors and driving process during actual operation, so as to ensure that the abnormal cause given in the abnormal detection result is more accurate, and thus ensure the effectiveness of the final obtained abnormal detection result interpretation for non-professionals.

[0034] Furthermore, the present invention judges whether the operating data is abnormal detection data based on the operating data detection results at a single moment and the fluctuation state of the operating data within a preset detection period, avoiding making judgments on the abnormal detection data based only on the range of the operating data at a single moment, ensuring the accuracy of the judgment results on the abnormal detection data, and further ensuring the accuracy of subsequent abnormal detection results.

[0035] Furthermore, in the present invention, if the driving state or driving environment of the automobile changes, the preset detection period used for judging the fluctuation state of the operating data is adjusted. The fluctuation state of the operating data is detected according to the fixed preset detection period. Interference data is likely to exist within the preset detection period, which affects the judgment of the fluctuation state, resulting in low accuracy of the obtained fluctuation state judgment result. By adjusting the preset detection period according to the actual scenario, the accuracy of the fluctuation state judgment result of the operating data can be improved.

[0036] Furthermore, in the present invention, if the number of abnormal data items is less than or equal to the preset number of abnormal data items and the correlation coefficient between the abnormal detection data is less than or equal to the preset correlation coefficient, the abnormal cause of the abnormal detection data is traced to avoid omissions in the abnormal cause given only based on the current abnormal detection data, thereby ensuring the reliability of the abnormal detection results, and further ensuring the effectiveness of the final obtained abnormal detection result interpretation for non-professionals. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of a method for interpreting automobile anomaly detection results based on natural language generation according to the present invention;

[0038] Figure 2 This is a flow chart of the present invention for determining whether to perform data abnormality judgment based on whether the value of each operating data is within the abnormal data range;

[0039] Figure 3 This is a flow chart of the present invention for determining whether to adjust for a preset detection period based on the fluctuation moment difference value;

[0040] Figure 4 This is a flow chart of the present invention for determining an abnormality analysis method based on the number of abnormal data items and the associated reference coefficient. DETAILED DESCRIPTION

[0041] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0042] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0043] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0044] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0045] See also Figures 1 to 4 As shown, the present invention provides a method for interpreting automobile anomaly detection results based on natural language generation, comprising:

[0046] Periodically test various operating data of the target vehicle;

[0047] Determine whether to perform data anomaly judgment based on whether each operating data is within the abnormal data range to obtain abnormal detection data;

[0048] Determine whether to perform abnormality analysis on each abnormality detection data based on whether the target vehicle is in a data abnormality state, and determine the abnormality analysis method based on the number of abnormal data items and the correlation coefficient. The abnormality analysis method is to perform correlation analysis or abnormality tracing on each abnormality detection data;

[0049] Anomaly detection results are generated based on the anomaly causes of each anomaly detection data, and anomaly detection result explanations are generated based on the anomaly detection results.

[0050] Specifically, various operating data of the target vehicle are periodically tested, wherein for any operating data, if the value of the operating data is not within the abnormal data range, a data abnormality judgment is performed on the operating data;

[0051] Each item of operating data has an abnormal data range corresponding to it.

[0052] Among them, the target vehicle is a vehicle for which an abnormality detection result needs to be obtained. The various operating data of the target vehicle in the present invention include but are not limited to engine speed, vehicle speed, coolant temperature, oil pressure, external temperature and battery voltage. A cyclic data detection period is applied in the present invention. The duration of the data detection period can be determined by the user. The higher the user's real-time requirements for the abnormality detection results, the shorter the duration of the data detection period. A data detection period duration is provided, and the data detection period is 5 minutes. At the end of each data detection period, the various operating data of the target vehicle are detected, and it is determined whether the value of each operating data is within the abnormal data range; the abnormal data range set for each operating data can be set by the user according to actual needs. How to set the abnormal data range for each operating data is easy to understand for technicians in this field and will not be elaborated here; if the value of the operating data is within the abnormal data range, only the operating data detected at the current moment is recorded as abnormal detection data.

[0053] Specifically, the judgment of data anomalies in operating data includes:

[0054] Detect the data fluctuation status of the operating data obtained from each test within the preset test period;

[0055] If the data fluctuation state of the operating data is in an abnormal fluctuation state, the operating data obtained from each detection within the preset detection period are recorded as abnormal detection data and stored in association with the corresponding abnormal fluctuation state;

[0056] If the data fluctuation state of the operating data is not in an abnormal fluctuation state, it is determined that the operating data is not abnormal detection data.

[0057] Among them, for any one item of operation data, the data fluctuation state is determined according to the operation data fluctuation value within the preset detection period. If the operation data fluctuation value is greater than the preset data fluctuation value, it is determined that the operation data is in an abnormal fluctuation state within the preset detection period. If the operation data fluctuation value is less than or equal to the preset data fluctuation value, it is determined that the operation data is not in an abnormal fluctuation state within the preset detection period. The operation data fluctuation value is the ratio of the sum of the absolute values ​​of the differences obtained by subtracting the average detection value from the detection value of each operation data within the preset detection period to the number of data detection periods within the preset detection period. The average detection value is the average value of the detection values ​​of each operation data. For any one item of operation data, the preset number According to the value of the fluctuation value, the user can set it according to actual needs and historical records. The higher the user's requirements for the timeliness of the abnormality detection results, the smaller the value of the preset data fluctuation value. A method for determining the preset data fluctuation value is provided. The operating data obtained from a single moment of detection is recorded as the historical record of abnormality detection data and recorded as a reference record. The average value of the fluctuation values ​​of each operating data in the reference record that meets the user's requirements for the timeliness of abnormality detection results is recorded as the preset data fluctuation value; the operating data obtained from each detection within the preset detection period are all recorded as abnormal detection data and stored in association with the corresponding abnormal fluctuation state, that is, when storing abnormal detection data, the abnormal fluctuation state of the abnormal detection data is stored together.

[0058] Specifically, if it is under periodic triggering conditions, it is determined whether to adjust for the preset detection period according to the difference value of the fluctuation moment;

[0059] If the fluctuation time difference value is greater than the preset fluctuation time difference value, the duration of the preset detection period is reduced according to the detection time difference value;

[0060] The reduction value of the duration of the preset detection period is positively correlated with the difference value of the detection time;

[0061] The periodic triggering condition is that the driving state change coefficient in the change detection period at the current moment is greater than the preset state change coefficient or the driving environment change coefficient is greater than the preset environment change coefficient;

[0062] The fluctuation time difference value is the interval length between the fluctuation start time and the cycle trigger time, and the detection time difference value is the interval length between the fluctuation start time and the cycle trigger time.

[0063] Among them, if the fluctuation moment difference value is less than or equal to the preset fluctuation moment difference value, the duration of the preset detection period is not adjusted; the value of the preset fluctuation moment difference value can be set by the user according to actual needs and historical records. The higher the user's reliability requirements for the abnormality detection results, the smaller the value of the preset fluctuation moment difference value. A method for determining the preset fluctuation moment difference value is provided, and the historical records adjusted for the preset detection period are recorded as period reference records. The average value of each fluctuation moment difference value in the period reference records that meet the user's requirements for the reliability of the abnormality detection results is recorded as the preset fluctuation moment difference value; the value of the duration of the preset detection period before adjustment can be set by the user according to actual needs. The higher the user's reliability requirements for the abnormality detection results, the longer the duration of the preset detection period. A value for the duration of the preset detection period is provided, and the value of the duration of the preset detection period is 60 minutes;

[0064] The present invention applies a cyclic change detection cycle. The duration of the change detection cycle can be determined by the user. The higher the user's requirements for the accuracy of the abnormality detection results, the shorter the change detection cycle. A change detection cycle is provided, which is 30 minutes. At the end of each change detection cycle, the driving state change coefficient and the driving environment change coefficient are detected.

[0065] The driving state change coefficient is determined according to the actual load weight change value, the average speed change value and the sudden speed change number change value of the target vehicle. The driving state change coefficient is the sum of the products of the actual load weight change value, the average speed change value and the sudden speed change number change value and the corresponding influence coefficient. The actual load weight change value is the absolute value of the difference between the actual load weight of the target vehicle at the current detection moment and the previous detection moment, the average speed change value is the absolute value of the difference between the average speed of the target vehicle in the current change detection cycle and the previous change detection cycle, and the sudden speed change number change value is the absolute value of the difference between the average speed of the target vehicle in the current change detection cycle and the previous change detection cycle. The absolute value of the difference in the number of sudden speed changes completed by the target vehicle. If the target vehicle performs a sudden brake or a sudden acceleration, it is recorded as completing a sudden speed change. The user can set the values ​​of the influence coefficients corresponding to the actual load weight change value, the average vehicle speed change value, and the sudden speed change number change value according to actual needs. The values ​​of the influence coefficients corresponding to the actual load weight change value, the average vehicle speed change value, and the sudden speed change number change value are provided. The value of the influence coefficient corresponding to the actual load weight change value is 0.4, the value of the influence coefficient corresponding to the average vehicle speed change value is 0.3, and the value of the influence coefficient corresponding to the sudden speed change number change value is 0.3;

[0066] The driving environment variation coefficient is determined according to the terrain complexity difference and the weather severity difference. The driving environment variation coefficient is the natural logarithm of the product of the terrain complexity difference and the weather severity difference. The terrain complexity difference is the absolute value of the difference between the terrain complexity in the current change detection period and the previous change detection period. The weather severity difference is the absolute value of the difference between the weather severity in the current change detection period and the previous change detection period. The terrain complexity is the number of complex road types passed by the target vehicle. The complex road types in the present invention include but are not limited to rural roads and mountain roads. The weather severity is the sum of the products of wind speed, precipitation and visibility and the corresponding influence coefficients. How to obtain the wind speed, precipitation and visibility during driving is easy to understand for those skilled in the art and will not be elaborated here. The user can provide a value of the influence coefficient corresponding to wind speed, precipitation and visibility according to actual needs. The value of the influence coefficient corresponding to wind speed is 0.3, the value of the influence coefficient corresponding to precipitation is 0.4, and the value of the influence coefficient corresponding to visibility is 0.3;

[0067] The values ​​of the preset state change coefficient and the preset environment change coefficient can be set by the user according to actual needs and historical records. The higher the user's requirement for the reliability of the abnormal detection result, the smaller the value of the preset state change coefficient, and the smaller the value of the preset environment change coefficient. A method for determining the value of the preset state change coefficient is provided, and the minimum value of each driving state change coefficient in the periodic reference record that meets the user's requirement for the reliability of the abnormal detection result is recorded as the preset state change coefficient. A method for determining the value of the preset environment change coefficient is provided, and the minimum value of each driving environment change coefficient in the periodic reference record that meets the user's requirement for the reliability of the abnormal detection result is recorded as the preset environment change coefficient.

[0068] Specifically, if the target vehicle is in a data abnormality state, the abnormality analysis method is determined based on the number of abnormal data items and the associated reference coefficient. The abnormality analysis method includes performing correlation analysis on each abnormal detection data and abnormality tracing on each abnormal detection data.

[0069] The data abnormal state is that there is abnormal detection data in any item of operating data of the target vehicle.

[0070] The number of abnormal data items is the number of items included in the abnormality detection data of the target vehicle.

[0071] Specifically, the correlation reference coefficient is the average value of the correlation coefficients corresponding to each abnormality detection data;

[0072] For any one abnormality detection data, the correlation coefficient is determined according to the number of data items of the operating data that are correlated with the abnormality detection data and the degree of coincidence of abnormal periods of the operating data that are correlated.

[0073] Among them, the correlation coefficient is the natural logarithm of the product of the number of data items of the operating data with a correlation relationship and the average abnormal cycle overlap. The average abnormal cycle overlap is the average value of the abnormal cycle overlaps of each operating data with a correlation relationship. For any item of operating data that has a correlation relationship with the abnormal detection data, the abnormal cycle overlap is negatively correlated with the abnormal interval time. The abnormal interval time is the interval between the start time of the abnormality and the sudden change time of the operating data within the preset detection period.

[0074] Specifically, if the number of abnormal data items is greater than the preset number of abnormal data items, or the correlation reference coefficient is greater than the preset correlation reference coefficient, correlation analysis is performed on each abnormal detection data, and the abnormal detection data is divided according to the correlation coefficient of each abnormal detection data to obtain several data groups;

[0075] The correlation coefficient between the abnormal detection data in any data group is greater than the preset correlation coefficient;

[0076] For a single data group, the priority interpretation coefficient of the corresponding abnormal detection data is set according to the correlation coefficient of each abnormal detection data in the data group, and several abnormal detection data for abnormal cause analysis are determined according to the priority interpretation coefficient.

[0077] Among them, the values ​​of the preset number of abnormal data items and the preset correlation reference coefficient can be set by the user according to actual needs and historical records. The higher the user's reliability requirements for the abnormality detection results, the smaller the value of the preset number of abnormal data items and the smaller the value of the preset correlation coefficient. A method for determining the value of the preset number of abnormal data items is provided, and the historical records without abnormality tracing are recorded as analysis reference records. The minimum value of the number of abnormal data items in the analysis reference records that meet the user's requirements for the reliability of the abnormality detection results is recorded as the preset number of abnormal data items. A method for determining the value of the preset correlation reference coefficient is provided, and the minimum value of the correlation reference coefficient in the analysis reference records that meet the user's requirements for the reliability of the abnormality detection results is recorded as the preset correlation reference coefficient.

[0078] For any two anomaly detection data, the correlation coefficient is the sum of the product of the performance similarity and the device similarity and the corresponding weight coefficient. If the two anomaly detection data both represent the same performance of the same device, the performance similarity of the two anomaly detection data is recorded as 1. If the two anomaly detection data both represent different performances of the same device or different devices, the performance similarity of the two anomaly detection data is recorded as 0. If the two anomaly detection data are both operating data of the same device, the device similarity of the two anomaly detection data is recorded as 1. If the two anomaly detection data are respectively operating data of different devices, the device similarity of the two anomaly detection data is recorded as 0. The user can provide a value of the performance similarity and the weight coefficient corresponding to the device similarity according to actual needs. The value of the weight coefficient corresponding to the performance similarity is 0.5, and the value of the weight coefficient corresponding to the device similarity is 0.5.

[0079] The user can set the preset correlation coefficient according to actual needs and historical records. The higher the user's reliability requirements for the anomaly detection results, the larger the value of the preset correlation coefficient. A method for determining the value of the preset correlation coefficient is provided. The average value of the correlation coefficient between the anomaly detection data in each data group in the analysis reference record that meets the user's reliability requirements for the anomaly detection results is recorded as the preset correlation coefficient.

[0080] The priority interpretation coefficient is positively correlated with the correlation coefficient of the corresponding anomaly detection data. The higher the priority interpretation coefficient, the more priority it has in becoming the anomaly detection data for anomaly cause analysis. The number of data items of the selected anomaly detection data is the preferred number of data items. The user can set the preferred number of data items according to actual needs. The higher the user's reliability requirements for the anomaly detection results, the larger the number of preferred data items. A value for the preferred number of data items is provided, and the preferred number of data items is 5.

[0081] Specifically, if the number of abnormal data items is less than or equal to the preset number of abnormal data items and the correlation coefficient is less than or equal to the preset correlation coefficient, anomaly tracing is performed for each abnormal detection data;

[0082] Determining a priority tracing coefficient based on the abnormal correlation degree of each operating data of the target vehicle, wherein the priority tracing coefficient is positively correlated with the abnormal correlation degree of each operating data;

[0083] For any item of operation data, the abnormality association degree is based on the number of data items of abnormality detection data that are associated with the operation data.

[0084] Among them, the operating data with a larger priority tracing coefficient has a higher priority as the basis for analyzing the abnormal cause of the abnormality detection data. For any item of operating data, the abnormality correlation degree is the number of data items of the abnormality detection data that are associated with the operating data.

[0085] Specifically, if there is no correlation between the abnormal detection data and other operating data, abnormal driving characteristics within the abnormal change range of the abnormal detection data that cannot be traced back to the abnormality are detected, and the abnormal cause of the abnormal detection data is determined based on the abnormal driving characteristics.

[0086] Among them, if the abnormal detection data is the operating data of a single moment in time and is within the warning data range, the abnormal change range is the time range that meets the preset feature detection time and takes the abnormal detection data as the cutoff time; if the abnormal detection data is in an abnormal fluctuation state, the abnormal change range is the preset detection period corresponding to the abnormal detection data. The abnormal driving characteristics include but are not limited to sudden acceleration, sudden change of direction, vehicle overload and vehicle bumps;

[0087] The user can determine whether the abnormal driving characteristics within the abnormal change range are the abnormal cause corresponding to the abnormal detection data based on the historical records. This is easy for technicians in this field to understand and will not be elaborated here.

[0088] Specifically, if all anomaly detection data have completed anomaly analysis, anomaly detection results are generated based on the anomaly causes of each anomaly detection data, and an explanation of the anomaly detection results is generated through the pre-trained model;

[0089] The abnormality detection results include abnormality detection database data, abnormal event description data, abnormality detection cause explanation, and vehicle body signal real-time data.

[0090] Among them, the anomaly detection database data refers to the abnormal events that have been detected, including the anomaly detection data of the car in various abnormal scenarios and the timestamp of the anomaly occurrence; the abnormal event description data describes each abnormal event, including the specific situation of the abnormality, the operating data related to the anomaly detection data, and any possible external factors, so that the large language model can have a deeper understanding of the background and context of the abnormal event; the anomaly detection cause explanation is a specific explanation of the cause of each abnormal event. Each anomaly detection result is associated with one or more reasons, explaining why the event is considered abnormal; the real-time data of the vehicle body signal comes from various sensors and systems of the vehicle, and records the real-time information when the abnormal event occurs, including but not limited to vehicle speed, speed, fuel level, temperature and other data. The generation of anomaly detection database data, abnormal event description data, anomaly detection cause explanation, and real-time data of the vehicle body signal can be obtained through statistical data and machine learning models, which will not be elaborated here.

[0091] For example, the anomaly detection database data is:

[0092] Timestamp: 2020.7.10, 12:15:32;

[0093] Abnormal detection data: Oil temperature sensor temperature reading: 120℃;

[0094] Normal temperature range: 70℃-110℃.

[0095] Abnormal event description data can be:

[0096] Event description: The engine oil overheat alarm lasts for 3 minutes;

[0097] Impact of the incident: The engine's lubrication effect is weakened, exacerbating mechanical wear.

[0098] The reasons for anomaly detection can be explained as follows:

[0099] Possible reasons: The oil circulation system is faulty and the oil cannot be cooled effectively;

[0100] Secondary reasons: The engine runs at high load for a long time, and the oil temperature continues to rise;

[0101] Recommended measures: Reduce the engine load immediately, check the oil circulation system and oil filter, replace the oil and repair related parts if necessary.

[0102] The real-time data of vehicle body signals can be:

[0103] Timestamp: July 10, 2020, 12:15:32 PM

[0104] Engine speed: 2000 rpm;

[0105] Speed: 110km / h;

[0106] Coolant temperature: 90°C;

[0107] Oil pressure: 3.5 bar;

[0108] Fuel capacity: 70%;

[0109] External temperature: 35°C

[0110] Battery voltage is 12.3V.

[0111] The data contained in the anomaly detection results is used as the input data of the pre-training model to obtain the explanation of the anomaly detection results and convert the anomaly detection results into content that can be understood by non-professionals. The pre-training model is a model obtained by incremental pre-training on the general knowledge question-answering reasoning ability of the large language model. The pre-training model's ability to judge and analyze the anomaly detection results is regulated by the pre-training stage. After the training is completed, the anomaly detection results can be interpreted. How to pre-train the language model based on the anomaly detection database data, abnormal event description data, anomaly detection cause explanation data, and vehicle body signal data is easy for technicians in this field to understand and will not be elaborated here.

[0112] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0113] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for explaining automobile anomaly detection results based on natural language generation, characterized in that: include: Periodically test various operating data of the target vehicle; Determine whether to perform data anomaly judgment based on whether each operating data is within the abnormal data range to obtain abnormal detection data; Determine whether to perform abnormal analysis on each abnormal detection data according to whether the target vehicle is in a data abnormal state, and determine the abnormal analysis method according to the number of abnormal data items and the correlation coefficient, wherein the abnormal analysis method is to perform correlation analysis or abnormal tracing on each abnormal detection data; An anomaly detection result is generated according to the anomaly cause of each anomaly detection data, and an anomaly detection result explanation is generated according to the anomaly detection result.

2. The method for explaining automobile anomaly detection results based on natural language generation according to claim 1, characterized in that: Periodically testing various operating data of the target vehicle, wherein for any operating data, if the value of the operating data is not within the abnormal data range, then performing data abnormality judgment on the operating data; Each item of operating data has an abnormal data range corresponding to it.

3. The method for explaining automobile anomaly detection results based on natural language generation according to claim 2, characterized in that: Data anomaly judgment for operating data, including: Detect the data fluctuation status of the operating data obtained from each test within a preset test period; If the data fluctuation state of the operating data is in an abnormal fluctuation state, the operating data obtained from each detection within the preset detection period are recorded as abnormal detection data and stored in association with the corresponding abnormal fluctuation state; If the data fluctuation state of the operating data is not in an abnormal fluctuation state, it is determined that the operating data is not abnormal detection data.

4. The method for explaining automobile anomaly detection results based on natural language generation according to claim 3, characterized in that: If it is under periodic triggering conditions, determine whether to adjust the preset detection period according to the difference value of the fluctuation time; If the fluctuation moment difference value is greater than the preset fluctuation moment difference value, the duration of the preset detection cycle is reduced and adjusted according to the detection moment difference value; The reduction value of the duration of the preset detection cycle is positively correlated with the difference value of the detection time; The periodic trigger condition is that the driving state change coefficient in the change detection period at the current moment is greater than the preset state change coefficient or the driving environment change coefficient is greater than the preset environment change coefficient; The fluctuation time difference value is the interval between the fluctuation start time and the cycle trigger time, and the detection time difference value is the interval between the fluctuation start time and the cycle trigger time.

5. The method for explaining automobile anomaly detection results based on natural language generation according to claim 4, characterized in that: If the target vehicle is in a data abnormality state, determine the abnormality analysis method according to the number of abnormal data items and the associated reference coefficient. The abnormality analysis method includes performing correlation analysis on each abnormal detection data, and performing abnormality tracing on each abnormal detection data; The data abnormal state is that there is abnormal detection data in any item of operating data of the target vehicle.

6. The method for explaining automobile anomaly detection results based on natural language generation according to claim 5, characterized in that: The correlation reference coefficient is the average value of the correlation coefficients corresponding to each abnormal detection data; For any one abnormality detection data, the correlation coefficient is determined according to the number of data items of the operating data that are associated with the abnormality detection data and the degree of coincidence of abnormal periods of the operating data that are associated with the abnormality detection data.

7. The method for explaining automobile anomaly detection results based on natural language generation according to claim 6, characterized in that: If the number of abnormal data items is greater than the preset number of abnormal data items, or the correlation reference coefficient is greater than the preset correlation reference coefficient, correlation analysis is performed on each abnormal detection data, and the abnormal detection data is divided according to the correlation coefficient of each abnormal detection data to obtain a number of data groups; The correlation coefficient between the abnormal detection data in any data group is greater than the preset correlation coefficient; For a single data group, the priority interpretation coefficient of the corresponding abnormal detection data is set according to the correlation coefficient of each abnormal detection data in the data group, and several abnormal detection data for abnormal cause analysis are determined according to the priority interpretation coefficient.

8. The method for explaining automobile anomaly detection results based on natural language generation according to claim 7, characterized in that: If the number of abnormal data items is less than or equal to the preset number of abnormal data items and the correlation coefficient is less than or equal to the preset correlation coefficient, anomaly tracing is performed for each abnormal detection data; Determine a priority tracing coefficient according to the abnormal correlation of each operating data of the target vehicle, wherein the priority tracing coefficient is positively correlated with the abnormal correlation of each operating data; For any item of operation data, the abnormality association degree is based on the number of data items of abnormality detection data that are associated with the item of operation data.

9. The method for explaining automobile anomaly detection results based on natural language generation according to claim 8, characterized in that: If the abnormal detection data has no correlation with other operating data, abnormal driving characteristics within the abnormal change range of the abnormal detection data that cannot be traced back to the abnormality are detected, and the abnormal cause of the abnormal detection data is determined based on the abnormal driving characteristics.

10. The method for explaining automobile anomaly detection results based on natural language generation according to claim 9, characterized in that: If all abnormal detection data have completed abnormal analysis, an abnormal detection result is generated according to the abnormal cause of each abnormal detection data, and an abnormal detection result explanation is generated through the pre-trained model; The abnormality detection result includes abnormality detection database data, abnormal event description data, abnormality detection cause explanation, and vehicle body signal real-time data.

Images