Vehicle operation information evaluation management system and method based on big data

Abstract

The invention discloses a vehicle operation information evaluation management system and method based on big data, and relates to the technical field of fault monitoring, and the method comprises the steps: extracting key abnormal features causing vehicle operation abnormality, and classifying the extracted key abnormal features; obtaining a potential association relationship between different key abnormal features; using a machine learning algorithm to construct a fault prediction model, and training the fault prediction model; generating a data prediction model by using the change conditions of different key abnormal characteristics and the potential association relationship when the vehicle is abnormal in history; in the running process of the vehicle, key abnormal feature data are collected in real time and input into the trained fault prediction model, and whether the vehicle has an abnormal risk or not is judged; and when the vehicle is judged to have the abnormal risk, calculating the time required for plan making, and inputting the required time into the data prediction model to predict future key abnormal data as a basis to make an emergency plan.

Description

Technical Field

[0001] The present invention relates to the technical field of fault monitoring, and specifically to a vehicle operation information evaluation and management system and method based on big data. Background Art

[0002] With the continuous improvement of the electronic and intelligent levels of automobiles, as well as the wide application of Internet of Things technology in the vehicle field, a large amount of data is generated during the operation of vehicles, such as vehicle speed, fuel consumption, fault codes, geographical locations, etc. These data provide rich materials for the evaluation and management of vehicle operation information based on big data. The construction of intelligent transportation systems requires a comprehensive and accurate evaluation and management of vehicle operation information to achieve goals such as traffic flow optimization, traffic safety improvement, and travel efficiency enhancement. Systems and methods based on big data can better process and analyze massive vehicle data, providing strong support for intelligent transportation. With the development of algorithm technology, more intelligent vehicle operation evaluation models and management systems have been developed based on advanced technologies such as deep learning, which can automatically identify complex vehicle operation patterns and abnormal situations, and achieve automated decision-making and early warning.

[0003] However, in the current judgment of vehicle operation information, the monitoring of faults usually only performs threshold judgment on a single data. However, in the actual complex vehicle operation, there are cases where a single data does not exceed the threshold, but due to the mutual influence between data, a fault anomaly occurs without any data exceeding the original threshold, resulting in untimely and inaccurate judgment of vehicle operation information; and due to the relatively fast change of operation information during vehicle operation, when adjusting and optimizing vehicle operation based on the evaluation results, due to the time difference in links such as information transmission and decision execution, the adjustment measures may not be able to keep up with the change of vehicle operation status in a timely manner. Summary of the Invention

[0004] The purpose of the present invention is to provide a vehicle operation information evaluation and management system and method based on big data to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A vehicle operation information evaluation and management method based on big data, the method includes the following steps:

[0007] S100. Collect the records of abnormal vehicle operation in the vehicle history, extract the vehicle operation information in the records, extract the key abnormal features that cause the abnormal vehicle operation, and classify the extracted key abnormal features;

[0008] Further, the specific steps for classifying the extracted key abnormal features are:

[0009] S101. Collect the records of abnormal vehicle operations in the vehicle's history through the cloud platform in the vehicle, extract the vehicle operation information in the records, and preprocess the extracted vehicle operation information, including missing value processing, denoising, and standardization. For missing value processing, collect the information intervals of all vehicle operation information, calculate the average of the information intervals, determine that the information with an information interval greater than the average is missing, calculate the average of the information on both adjacent sides of the missing part as the filling value, and fill the information at the missing part with the filling value.

[0010] For denoising, use the moving average filtering method to remove the noise in the vehicle operation information. Calculate the average and standard deviation of all operation information, and standardize the operation information. The formula is:

[0011]

[0012] In the formula, x’ represents the standardized operation information, x represents the original operation information, x aver represents the average of the operation information, and x stand represents the standard deviation of the operation information.

[0013] S102. Segment the vehicle operation information during abnormal occurrence, extract the data window of N seconds before and after the abnormal occurrence, and perform time-domain and frequency-domain analysis on the operation information within the extracted data window. Through time-domain analysis, obtain the data value characteristics of the operation information, including the average value, standard deviation, and peak value. During frequency-domain analysis, use the Fourier transform to extract the main frequency components of the operation information. Input the operation information characteristics after double-domain analysis into the correlation model to calculate the correlation value between each operation information and vehicle abnormality. The formula is:

[0014]

[0015] In the formula, r represents the correlation value between each operation information and vehicle abnormality, X represents the characteristics of each operation information, Y represents the vehicle abnormality label, and σX and σY represent the standard deviations of X and Y respectively. Sort the absolute values of the correlation values of all types of operation information calculated from small to large, calculate the difference between adjacent correlation values, compare and judge to obtain the maximum difference, use the maximum difference as the boundary node, and use the operation information greater than the boundary node as the key abnormal characteristics of vehicle abnormality.

[0016] S103. Collect abnormal samples and divide them into a training set and a test set. Construct a logistic regression model as h(x), use the training set to train the logistic regression model, input the extracted key abnormal characteristics into the logistic regression model to classify the key abnormal characteristics, and obtain different types of key abnormal characteristics.

[0017] By extracting and classifying key abnormal features, it is possible to quickly and accurately determine the system or component where the vehicle fault lies, helping maintenance personnel to conduct targeted inspections and repairs, improving the maintenance efficiency, and reducing the maintenance time and cost. Timely detection of abnormal features during vehicle operation can provide early warnings for problems that may affect driving safety, such as abnormalities in the braking system and failures in the steering system, enabling the driver to take corresponding measures to ensure driving safety.

[0018] S200. Conduct correlation analysis on different types of key abnormal features to obtain the potential correlation relationships between different key abnormal features;

[0019] Furthermore, the specific steps to obtain the potential correlation relationships between different key abnormal features are as follows:

[0020] S201. Collect the key abnormal features when the vehicle has abnormalities in history, discretely extract the collected key abnormal feature data, and transform it into discrete key abnormal feature data; Extract two different key abnormal features as a combination to draw a scatter plot, and use polynomial regression to fit the points in the scatter plot to obtain a polynomial G i =β0 + β1×G j +β2×G j 2 +…β n ×G j n , In the polynomial, G i and G j respectively represent two different key abnormal features extracted, β0 represents the constant term, and β1 to β n represent the coefficients of the first-order independent variable to the nth-order independent variable in the polynomial;

[0021] After obtaining the fitted polynomial, substitute the abscissa of each point in the scatter plot into the polynomial to calculate the fitted value of each point, corresponding to collecting the true value of each point in the scatter plot, and calculate the mean square error in the scatter plot. The formula is:

[0022]

[0023] In the formula, MSE represents the calculated mean square error, m represents the number of points in the scatter plot, G u represents the fitted value of the u-th point in the scatter plot, and G’ u represents the true value of the u-th point in the scatter plot;

[0024] Perform pairwise calculations on all key abnormal features to obtain the polynomial and the mean square error. Set the error threshold to MY; When the calculated mean square error MSE ≤ MY, it is determined that there is a correlation between the corresponding two key abnormal features. When MSE > MY, it is determined that there is no correlation between the corresponding two key abnormal features;

[0025] S202. After determining that there is an association between two key abnormal features, combine the associated key abnormal features to form an association group, integrate and record all association groups, and use the calculated polynomial in the association group as the potential association relationship F(G).

[0026] Different key abnormal features may be associated with each other, jointly leading to the occurrence of vehicle failures. Through association analysis, these potential association relationships can be discovered, and the complex causes of failures can be deeply understood, rather than being limited to the influence of a single feature. For example, finding an association between an abnormal increase in engine oil temperature and an abnormal decrease in the rotational speed of the cooling system fan helps to more comprehensively understand the cause of engine overheating failure. In actual fault diagnosis, the joint analysis of multiple abnormal features can more accurately determine the type and location of the fault than the judgment of a single feature;

[0027] Through association analysis, it is possible to avoid misjudgments and missed judgments that may occur when diagnosing based on a single feature alone. The mutual verification of multiple related features can improve the reliability of the diagnostic results and ensure that vehicle failures are handled promptly and accurately.

[0028] S300. For the potential association relationships between different key abnormal features mined, use machine learning algorithms to construct a fault prediction model, collect the key abnormal features during abnormal operation in the vehicle history as the training set, and train the fault prediction model;

[0029] Furthermore, the specific steps for training the fault prediction model are as follows:

[0030] S301. Calculate the average value and standard deviation of the data values of each key abnormal feature when the historical vehicle has an abnormality, and use the average value minus the standard deviation to obtain the fault threshold YG_d for each key abnormal feature.

[0031] In the association group, substitute the fault thresholds of the two key abnormal features in the association group into the potential association relationship respectively, calculate the fitted values of the two key abnormal features respectively, and use the fitted values as the associated fault threshold YG_g of the association group in the association group.

[0032] S302. Use deep learning algorithms to construct a fault prediction model, and collect the key abnormal features during abnormal operation in the vehicle history as the training set to train the fault prediction model.

[0033] Based on the real-time monitoring of key abnormal features and the prediction of the model, it is possible to issue an early warning in a timely manner before a vehicle failure occurs, enabling the driver or maintenance personnel to have enough time to take measures, such as arranging repairs, replacing parts, etc., to reduce the losses and safety risks brought by the failure.

[0034] An accurate fault prediction model can help formulate a more reasonable maintenance plan, shifting from traditional regular maintenance to condition-based maintenance. According to the prediction results of the model, only the vehicles about to have faults are maintained, avoiding unnecessary maintenance operations and saving maintenance resources and costs.

[0035] S400. Generate a data prediction model using the change situations and potential correlation relationships of different key abnormal features when the vehicle is abnormal in history;

[0036] Furthermore, the specific steps for generating a data prediction model using the change situations and potential correlation relationships of different key abnormal features when the vehicle is abnormal in history are as follows:

[0037] S401. Construct a curve of each key abnormal feature changing with time series using the change situations of different key abnormal features when the vehicle is abnormal in history, and use linear regression to interpret the curve to obtain the time prediction function Gt = f(t) of each key abnormal feature;

[0038] Extract the potential correlation relationship within the correlation group to calculate the fitted value after the mutual influence of two key abnormal features, and use the potential correlation relationship as the correlation prediction function Gn = F(G);

[0039] S402. Combine the two prediction functions to obtain a data prediction model, specifically: Ga = Max(Gt, Gn).

[0040] S500. During the vehicle operation, collect key abnormal feature data in real time and input it into the trained fault prediction model to judge whether the vehicle has abnormal risks;

[0041] Furthermore, the specific steps for judging whether the vehicle has abnormal risks are as follows:

[0042] S501. During the vehicle operation, collect key abnormal feature data in real time and input it into the trained fault prediction model. Let the real-time key abnormal feature data be Gs. The specific judgment rule is:

[0043] Individually judge each key abnormal feature. When Gs ≥ YG_d, judge that there are abnormal risks during the vehicle operation;

[0044] When Gs < YG_d, enter the correlation judgment mode, extract all correlation groups in the vehicle operation data. For each correlation group, let the real-time data values of the two key abnormal features within the correlation group be (Gs i , Gs j ). When (Gs i ≥YG_g i )|(Gs j ≥YG_g j) When it is determined that there is an abnormal risk during the vehicle operation.

[0045] S600. When it is determined that the vehicle has an abnormal risk, calculate the time required for formulating a plan, input the required time into the data prediction model to predict future key abnormal data, and use it as the basis to formulate an emergency plan.

[0046] Furthermore, the specific steps for inputting the required time into the data prediction model to predict future key abnormal data and using it as the basis to formulate an emergency plan are as follows:

[0047] S601. Collect the analysis and transmission time of all operation information during the vehicle operation in history, calculate the average value of all analysis and transmission times, and use the average value as the vehicle information analysis standard time. When it is determined that the vehicle has an abnormal risk, input the analysis standard time into the data prediction model to calculate the time prediction data value. Then, input the real-time data values of two key abnormal feature types into the data prediction model within the associated group to calculate the associated prediction data value. Select the maximum value between the time prediction data value and the associated prediction data value as the future data value for output;

[0048] S602. After it is determined that there is an abnormal risk during the vehicle operation, use the future data value to replace the real-time collected key abnormal feature data value as the basis to formulate an emergency plan.

[0049] When it is determined that the vehicle has an abnormal risk and an emergency plan is needed, predict the analysis data, predict in advance the data value of the vehicle after the analysis data time, and then formulate a plan, which solves the problem of plan lag. As time goes by and new data is generated, continuously update the prediction results, dynamically adjust and optimize the emergency plan to ensure that the plan always matches the actual situation and future development of the vehicle.

[0050] A vehicle operation information evaluation and management system based on big data. The vehicle operation information evaluation and management system includes a data collection module, a key abnormal feature search module, an association analysis module, a model construction module, a fault judgment module, and a data prediction module;

[0051] The data collection module is used to collect operation information during the historical operation of the vehicle;

[0052] The key abnormal feature search module is used to collect records of abnormal vehicle operations in history, extract the vehicle operation information in the records, extract the key abnormal features that cause abnormal vehicle operations, and classify the extracted key abnormal features;

[0053] The association analysis module is used to perform association analysis on different types of key abnormal features to obtain the potential association relationships between different key abnormal features;

[0054] The model construction module is used to analyze the key abnormal feature data when the vehicle has an abnormality in history, and construct a fault prediction model and a data prediction model respectively;

[0055] The fault judgment module is used to collect key abnormal feature data in real time during the vehicle operation and input it into the trained fault prediction model to judge whether there is an abnormal risk for the vehicle;

[0056] The data prediction module is used to input the information analysis standard time and the real-time key abnormal feature data value into the data prediction model to obtain the future data value.

[0057] The association analysis module includes an association judgment unit and a potential association relationship calculation unit;

[0058] The association judgment unit is used to calculate the mean square error in the scatter plot to judge whether there is an association between two key abnormal features;

[0059] The potential association relationship calculation unit is used to extract two different key abnormal features as a combination to draw a scatter plot, and use polynomial regression to fit the points in the scatter plot to obtain a polynomial as the potential association relationship.

[0060] The model construction module includes a fault prediction model unit and a data prediction model unit;

[0061] The fault prediction model unit uses a deep learning algorithm to construct a fault prediction model, and collects the key abnormal features during the abnormal operation of the vehicle history as a training set to train the fault prediction model;

[0062] The data prediction model unit is used to calculate a time prediction function and an association prediction function respectively, and combine the two functions to obtain a data prediction model.

[0063] Compared with the prior art, the beneficial effects of the present invention are:

[0064] 1. Through association analysis, the present invention can avoid misjudgment and missed judgment that may occur when diagnosing based on a single feature only. The mutual verification of multiple related features can improve the reliability of the diagnosis result and ensure that vehicle faults are processed in a timely and accurate manner.

[0065] 2. When the present invention judges that there is an abnormal risk for the vehicle and needs to know the emergency plan, it predicts the analysis data, predicts in advance the data value of the vehicle after the analysis data time, and formulates a plan to solve the problem of plan lag. As time goes by and new data is generated, the prediction result is continuously updated, and the emergency plan is dynamically adjusted and optimized to ensure that the plan always matches the actual situation and future development of the vehicle. Brief Description of the Drawings

[0066] Figure 1 This is the module distribution diagram of a vehicle operation information evaluation and management system based on big data according to the present invention;

[0067] Figure 2 This is the step schematic diagram of a vehicle operation information evaluation and management method based on big data according to the present invention. Specific embodiments

[0068] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0069] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution,

[0070] A vehicle operation information evaluation and management method based on big data, the method includes the following steps:

[0071] S100. Collect the records of abnormal vehicle operations in the vehicle history, extract the vehicle operation information in the records, extract the key abnormal features that cause the abnormal vehicle operation, and classify the extracted key abnormal features;

[0072] The specific steps for classifying the extracted key abnormal features are as follows:

[0073] S101. Collect the records of abnormal vehicle operations in the vehicle history through the cloud platform in the vehicle, extract the vehicle operation information in the records, and preprocess the extracted vehicle operation information, including missing value processing, denoising, and standardization; for missing value processing, collect the information intervals of all vehicle operation information, calculate the average value of the information intervals, determine that the information is missing if the information interval is greater than the average value, calculate the average value of the information on both sides adjacent to the missing part as the filling value, and fill the information at the missing part with the filling value;

[0074] For denoising, use the moving average filtering method to remove the noise in the vehicle operation information; calculate the average value and standard deviation of all operation information, and standardize the operation information. The formula is:

[0075]

[0076] In the formula, x' represents the standardized operation information, x represents the original operation information, x aver represents the average value of the operation information, and x stand represents the standard deviation of the operation information;

[0077] S102. Segment the vehicle operation information when an anomaly occurs, extract the data window of N seconds before and after the anomaly occurs, and perform time-domain and frequency-domain analyses on the operation information within the extracted data window. Obtain the data value characteristics of the operation information through time-domain analysis, including the mean value, standard deviation, and peak value. When performing frequency-domain analysis, use the Fourier transform to extract the main frequency components of the operation information. Input the operation information characteristics after the dual-domain analysis into the correlation model to calculate the correlation value between each type of operation information and the vehicle anomaly. The formula is as follows:

[0078]

[0079] In the formula, r represents the correlation value between each type of operation information and the vehicle anomaly, X represents the characteristics of each type of operation information, Y represents the vehicle anomaly label, and σX and σY represent the standard deviations of X and Y respectively. Sort the absolute values of the correlation values of all types of operation information calculated from small to large, calculate the difference between adjacent correlation values, compare and judge to obtain the maximum difference, use the maximum difference as the boundary node, and use the operation information greater than the boundary node as the key anomaly characteristics of the vehicle anomaly;

[0080] S103. Collect anomaly samples and divide them into a training set and a test set. Construct a logistic regression model as h(x), use the training set to train the logistic regression model, and input the extracted key anomaly characteristics into the logistic regression model to classify the key anomaly characteristics to obtain different types of key anomaly characteristics.

[0081] By extracting and classifying the key anomaly characteristics, it is possible to quickly and accurately determine the system or component where the vehicle fault is located, help maintenance personnel conduct targeted inspections and repairs, improve the maintenance efficiency, and reduce the maintenance time and cost. Timely discovery of the anomaly characteristics during vehicle operation can provide early warnings for problems that may affect driving safety, such as abnormal braking systems and steering system failures, enabling drivers to take corresponding measures to ensure driving safety.

[0082] S200. Conduct correlation analysis on different types of key anomaly characteristics to obtain the potential correlation relationships between different key anomaly characteristics;

[0083] The specific steps to obtain the potential correlation relationships between different key anomaly characteristics are as follows:

[0084] S201. Collect the key anomaly characteristics when the vehicle has an anomaly in history, discretely extract the collected key anomaly characteristic data, and convert it into discrete key anomaly characteristic data. Extract two different key anomaly characteristics as a combination to draw a scatter plot, and use polynomial regression to fit the points in the scatter plot to obtain the polynomial as G i =β0 + β1×G j +β2×G j2 + … β n × G j n , in the polynomial, G i and G j respectively represent two different key abnormal features extracted, β0 represents the constant term, and β1 to β n represent the coefficients of the first-order independent variable to the nth-order independent variable in the polynomial;

[0085] After obtaining the fitted polynomial, substitute the abscissa of each point in the scatter plot into the polynomial to calculate the fitted value of each point, collect the true value of each point in the scatter plot correspondingly, and calculate the mean square error in the scatter plot. The formula is:

[0086]

[0087] In the formula, MSE represents the calculated mean square error, m represents the number of points in the scatter plot, G u represents the fitted value of the u-th point in the scatter plot, and G’ u represents the true value of the u-th point in the scatter plot;

[0088] Calculate the polynomial and the mean square error pairwise for all key abnormal features. Set the error threshold to MY; when the calculated mean square error MSE ≤ MY, it is judged that there is an association between the corresponding two key abnormal features, and when MSE > MY, it is judged that there is no association between the corresponding two key abnormal features;

[0089] S202. After judging that there is an association between two key abnormal features, combine the associated key abnormal features to form an association group, integrate and record all association groups, and use the polynomial calculated in the association group as the potential association relationship F(G).

[0090] Different key abnormal features may be correlated with each other and jointly cause vehicle failures. Through association analysis, these potential association relationships can be discovered, and the complex causes of failures can be understood in depth, rather than being limited to the influence of a single feature. For example, finding an association between abnormal increase in engine oil temperature and abnormal decrease in the rotational speed of the cooling system fan helps to more comprehensively understand the cause of engine overheating failure. In actual fault diagnosis, the joint analysis of multiple abnormal features can more accurately determine the fault type and location than the judgment of a single feature;

[0091] Through association analysis, misjudgment and missed judgment situations that may occur when diagnosing based on a single feature can be avoided. The mutual verification of multiple related features can improve the reliability of the diagnosis result and ensure that vehicle failures are handled promptly and accurately.

[0092] S300. For the potential correlation relationships between different key abnormal features mined, use machine learning algorithms to construct a fault prediction model. Collect the key abnormal features during abnormal operation in the vehicle history as the training set, and train the fault prediction model;

[0093] The specific steps for training the fault prediction model are as follows:

[0094] S301. Calculate the average value and standard deviation of each key abnormal feature data value when the historical vehicle has an abnormality. Use the average value minus the standard deviation to obtain the fault threshold YG_d for each key abnormal feature;

[0095] In the association group, substitute the fault thresholds of the two key abnormal features in the association group into the potential correlation relationship respectively, and calculate the fitted values of the two key abnormal features respectively. In the association group, use the fitted values as the associated fault threshold YG_g of the association group;

[0096] S302. Use deep learning algorithms to construct a fault prediction model. Collect the key abnormal features during abnormal operation in the vehicle history as the training set to train the fault prediction model.

[0097] Based on the real-time monitoring of key abnormal features and the prediction of the model, it is possible to issue a warning in time before the vehicle fault occurs, enabling the driver or maintenance personnel to have enough time to take measures, such as arranging repairs, replacing parts, etc., to reduce the losses and safety risks brought by the fault.

[0098] An accurate fault prediction model can help formulate a more reasonable maintenance plan, changing from traditional regular maintenance to condition-based maintenance. According to the prediction results of the model, only maintain the vehicles that are about to have faults, avoiding unnecessary maintenance operations and saving maintenance resources and costs.

[0099] S400. Generate a data prediction model using the change situations and potential correlation relationships of different key abnormal features during vehicle abnormalities in history;

[0100] The specific steps for generating a data prediction model using the change situations and potential correlation relationships of different key abnormal features during vehicle abnormalities in history are as follows:

[0101] S401. Use the change situations of different key abnormal features during vehicle abnormalities in history to construct a curve of each key abnormal feature changing with time series. Use linear regression to interpret the curve to obtain the time prediction function Gt = f(t) of each key abnormal feature;

[0102] Extract the potential correlation relationship in the association group to calculate the fitted value after the mutual influence of the two key abnormal features, and use the potential correlation relationship as the association prediction function Gn = F(G);

[0103] S402. Combine the two prediction functions to obtain a data prediction model, specifically: Ga = Max(Gt, Gn).

[0104] S500. During the vehicle operation, collect key abnormal feature data in real time and input it into the trained fault prediction model to determine whether there is an abnormal risk for the vehicle.

[0105] The specific steps for determining whether there is an abnormal risk for the vehicle are as follows:

[0106] S501. During the vehicle operation, collect key abnormal feature data in real time and input it into the trained fault prediction model. Let the real-time key abnormal feature data be Gs. The specific judgment rule is as follows:

[0107] Individually judge each key abnormal feature. When Gs ≥ YG_d, it is judged that there is an abnormal risk during the vehicle operation.

[0108] When Gs < YG_d, enter the associated judgment mode, extract all associated groups in the vehicle operation data. For each associated group, let the real-time data values of the two key abnormal features in the associated group be (Gs i , Gs j ). When (Gs i ≥ YG_g i ) | (Gs j ≥ YG_g j ), it is judged that there is an abnormal risk during the vehicle operation.

[0109] S600. When it is judged that there is an abnormal risk for the vehicle, calculate the time required for formulating a plan, input the required time into the data prediction model to predict future key abnormal data, and formulate an emergency plan based on this.

[0110] The specific steps for inputting the required time into the data prediction model to predict future key abnormal data and formulating an emergency plan based on this are as follows:

[0111] S601. Collect the analysis and transmission time of all operation information during the historical vehicle operation, calculate the average value of all analysis and transmission times, and use the average value as the vehicle information analysis standard time. When it is judged that there is an abnormal risk for the vehicle, input the analysis standard time into the data prediction model to calculate the time prediction data value. Then, input the real-time data values of the two key abnormal features in the associated group into the data prediction model to calculate the associated prediction data value. Select the maximum value between the time prediction data value and the associated prediction data value as the future data value for output.

[0112] S602. After determining that there are abnormal risks during vehicle operation, replace the real-time collected key abnormal feature data values with future data values as the basis to formulate an emergency plan.

[0113] When it is determined that the vehicle has abnormal risks and an emergency plan is needed, predict the analysis data, predict in advance the data values of the vehicle after the analysis data time, and then formulate a plan, which solves the problem of plan lag. As time goes by and new data is generated, continuously update the prediction results, dynamically adjust and optimize the emergency plan to ensure that the plan always matches the actual situation and future development of the vehicle.

[0114] A vehicle operation information evaluation and management system based on big data. The vehicle operation information evaluation and management system includes a data collection module, a key abnormal feature search module, an association analysis module, a model construction module, a fault judgment module, and a data prediction module;

[0115] The data collection module is used to collect the operation information during the historical operation of the vehicle;

[0116] The key abnormal feature search module is used to collect the records of abnormal vehicle operations in history, extract the vehicle operation information in the records, extract the key abnormal features that cause abnormal vehicle operations, and classify the extracted key abnormal features;

[0117] The association analysis module is used to perform association analysis on different types of key abnormal features to obtain the potential association relationships between different key abnormal features;

[0118] The model construction module is used to analyze the key abnormal feature data when the vehicle has an abnormality in history, and construct a fault prediction model and a data prediction model respectively;

[0119] The fault judgment module is used to collect key abnormal feature data in real time during vehicle operation and input it into the trained fault prediction model to judge whether the vehicle has abnormal risks;

[0120] The data prediction module is used to input the information analysis standard time and the real-time key abnormal feature data values into the data prediction model to obtain future data values.

[0121] The association analysis module includes an association judgment unit and a potential association relationship calculation unit;

[0122] The association judgment unit is used to calculate the mean square error in the scatter plot to judge whether there is an association between two key abnormal features;

[0123] The potential correlation calculation unit is used to extract two different key abnormal features as a combination to draw a scatter plot, and use polynomial regression to fit the points in the scatter plot to obtain a polynomial as the potential correlation.

[0124] The model construction module includes a fault prediction model unit and a data prediction model unit;

[0125] The fault prediction model unit uses a deep learning algorithm to construct a fault prediction model, and collects the key abnormal features during abnormal operation in the vehicle history as a training set to train the fault prediction model;

[0126] The data prediction model unit is used to calculate a time prediction function and a correlation prediction function respectively, and combine the two functions to obtain a data prediction model.

[0127] Embodiment: Real-time evaluate the operation information of the vehicle, and extract the key abnormal features affecting vehicle abnormalities as temperature, vibration, and rotational speed; Through correlation analysis, it is obtained that there is a correlation between temperature and rotational speed, and a correlation group is constructed;

[0128] Construct a fault prediction model and a data prediction model respectively according to historical data;

[0129] Real-time collect the data values of the three key abnormal features of temperature, vibration, and rotational speed as 50, 20, and 3000; The individual fault thresholds of the three data are 70, 30, and 5000 respectively; When judged individually, there is no abnormal risk for the vehicle, and the correlation fault thresholds are calculated as 40 and 4000 in the correlation group;

[0130] It is judged in the correlation group that due to temperature reasons, the vehicle has an abnormal risk;

[0131] After formulating an emergency plan, set the standard time as 1s, input the standard time, real-time temperature, and rotational speed into the data prediction model, and obtain the time prediction temperature as 52 and the correlation prediction temperature as 60; When formulating the plan, it is necessary to formulate an emergency plan based on 60.

[0132] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. A vehicle operation information evaluation and management method based on big data, characterized in that: The method comprises the following steps: S100, collecting records of abnormal operation in the vehicle history, extracting vehicle operation information in the records, extracting key abnormal features that cause abnormal vehicle operation, and classifying the extracted key abnormal features; S200, performing correlation analysis on different types of key abnormal features to obtain potential correlation relationships between different key abnormal features; S300, using a machine learning algorithm to build a fault prediction model based on the potential correlation between the mined key abnormal features, collecting key abnormal features of abnormal operation in the history of the vehicle as a training set, and training the fault prediction model; S400, generating a data prediction model using the changes in different key abnormal characteristics and potential correlations when the vehicle is abnormal in history; S500, during the operation of the vehicle, collect key abnormal feature data in real time and input it into the trained fault prediction model to determine whether the vehicle has abnormal risks; S600: When it is determined that the vehicle has an abnormal risk, the time required for formulating a plan is calculated, and the required time is input into a data prediction model to predict future key abnormal data as a basis for formulating an emergency plan.

2. The vehicle operation information evaluation and management method based on big data according to claim 1 is characterized by: The specific steps of classifying the extracted key abnormal features in S100 are: S101, collecting records of abnormal operation in the history of the vehicle through a cloud platform in the vehicle, extracting vehicle operation information in the records, and preprocessing the extracted vehicle operation information, including missing value processing, denoising, and standardization; For missing value processing, collect the information intervals of all vehicle operation information, calculate the average value of the information intervals, judge that the information with an information interval greater than the average value is missing, calculate the average value of the information on both sides of the missing part as the filling value, and fill in the missing information with the filling value; For denoising, the sliding average filter method is used to remove the noise in the vehicle operation information; Calculate the mean and standard deviation of all running information and standardize the running information. The formula is: In the formula, x' represents the normalized operation information, x represents the original operation information, and x aver represents the average value of the running information, x stand Indicates the standard deviation of the running information; S102, segmenting the vehicle operation information when the abnormality occurs, extracting the data window of N seconds before and after the abnormality occurs, performing time domain and frequency domain analysis on the operation information in the extracted data window, and obtaining the data value characteristics of the operation information through time domain analysis, including the average value, standard deviation and peak value; in the frequency domain analysis, using Fourier transform to extract the main frequency component of the operation information; inputting the operation information characteristics after the dual-domain analysis into the correlation model to calculate the correlation value between each type of operation information and the vehicle abnormality, the formula is: In the formula, r represents the correlation value between each type of operation information and vehicle abnormality, X represents each type of operation information feature, Y represents the vehicle abnormality label, σX and σY represent the standard deviations of X and Y respectively; the absolute values ​​of the correlation values ​​of all types of operation information calculated are taken and sorted from small to large, the difference between adjacent correlation values ​​is calculated, and the maximum difference is obtained by comparison and judgment. The maximum difference is taken as the demarcation node, and the operation information greater than the demarcation node is taken as the key abnormality feature of the vehicle abnormality; S103, collect abnormal samples and divide them into training set and test set, build a logistic regression model h(x), use the training set to train the logistic regression model, input the extracted key abnormal features into the logistic regression model to classify the key abnormal features, and obtain different types of key abnormal features.

3. The vehicle operation information evaluation and management method based on big data according to claim 2 is characterized by: The specific steps of obtaining the potential correlation between different key abnormal features in S200 are: S201, collecting key abnormal features when the vehicle abnormalities occurred in history, performing discrete extraction on the collected key abnormal feature data, and converting them into discrete key abnormal feature data; extracting two different key abnormal features as a combination to draw a scatter plot, and using polynomial regression to fit the points in the scatter plot to obtain a polynomial G i =β0+β1×G i +β2×G j 2 +…β n ×G j n , in the polynomial, G i and G j They represent the extraction of two different key abnormal features, β0 represents a constant term, β1 to β n Represents the coefficients of the first-order independent variable to the n-order independent variable in the polynomial; After obtaining the fitted polynomial, substitute the horizontal coordinate of each point in the scatter plot into the polynomial to calculate the fitted value of each point. Correspondingly, collect the true value of each point in the scatter plot and calculate the mean square error in the scatter plot. The formula is: In the formula, MSE represents the calculated mean square error, m represents the number of points in the scatter plot, and G u Represents the fitted value of the u-th point in the scatter plot, G' u Represents the true value of the u-th point in the scatter plot; All key abnormal features are calculated pairwise to obtain polynomials and mean square errors, and the error threshold is set to MY; when the calculated mean square error MSE ≤ MY, it is judged that the corresponding two key abnormal features are associated; when MSE > MY, it is judged that the corresponding two key abnormal features are not associated; S202: When it is determined that two key abnormal features are associated, the associated key abnormal features are combined to form an associated group, all the associated groups are integrated and recorded, and the polynomials calculated in the associated groups are used as potential associated relationships F(G).

4. The vehicle operation information evaluation and management method based on big data according to claim 3 is characterized by: The specific steps of training the fault prediction model in S300 are: S301, calculating the average value and standard deviation of each key abnormal feature data value when the vehicle abnormality occurred in the past, and subtracting the standard deviation from the average value to obtain the fault threshold value of each key abnormal feature as YG_d; In the association group, the fault thresholds of the two key abnormal characteristics in the association group are respectively substituted into the potential association relationship, and the fitting values ​​of the two key abnormal characteristics are respectively calculated. The fitting values ​​are used as the association fault threshold YG_g of the association group in the association group; S302: Build a fault prediction model using a deep learning algorithm, and collect key abnormal features of abnormal operation in the vehicle history as a training set to train the fault prediction model.

5. The vehicle operation information evaluation and management method based on big data according to claim 3 is characterized by: The specific steps of generating a data prediction model by using the changes in different key abnormal characteristics and potential correlations when the vehicle is abnormal in history in S400 are: S401, constructing a curve of each key abnormality feature changing with time series by using the changes of different key abnormality features when the vehicle is abnormal in history, interpreting the curve by linear regression, and obtaining a time prediction function Gt=f(t) of each key abnormality feature; Extract potential correlations within the correlation group and calculate the fitting value after the two key abnormal features influence each other, and use the potential correlations as the correlation prediction function Gn=F(G); S402: Combining the two prediction functions to obtain a data prediction model, specifically: Ga=Max(Gt, Gn).

6. The vehicle operation information evaluation and management method based on big data according to claim 4 is characterized by: The specific steps of determining whether the vehicle has abnormal risk in S500 are: S501. During the operation of the vehicle, key abnormal feature data is collected in real time and input into the trained fault prediction model. Assuming that the real-time key abnormal feature data is Gs, the specific judgment rules are: Each key abnormal feature is judged separately. When Gs≥YG_d, it is judged that there is an abnormal risk in the vehicle operation process; When Gs<YG_d, enter the association judgment mode, extract all the association groups in the vehicle operation data, and for each association group, set the real-time data values ​​of the two key abnormal characteristics in the association group to be (Gs i , Gs j ), when (Gs i ≥YG_g i )|(Gs j ≥YG_g j ), it is judged that there is an abnormal risk in the vehicle operation process.

7. The vehicle operation information evaluation and management method based on big data according to claim 5 is characterized by: The specific steps of inputting the required time into the data prediction model in S600 to predict future key abnormal data as a basis for formulating an emergency plan are: S601. Collect the analysis transmission time of all operation information in the vehicle operation process in history, calculate the average value of all analysis transmission times, and use the average value as the vehicle information analysis standard time. When it is determined that the vehicle has an abnormal risk, input the analysis standard time into the data prediction model to calculate the time prediction data value, then input the two key abnormal feature real-time data values ​​into the data prediction model in the association group, calculate the association prediction data value, and select the maximum value of the time prediction data value and the association prediction data value as the future data value for output; S602: After determining that there is an abnormal risk during vehicle operation, replace the key abnormal characteristic data values ​​collected in real time with future data values ​​as a basis for formulating an emergency plan.

8. A vehicle operation information evaluation and management system based on big data, characterized by: The vehicle operation information evaluation and management system includes a data collection module, a key abnormal feature search module, a correlation analysis module, a model building module, a fault judgment module and a data prediction module; The data collection module is used to collect operation information of the vehicle during historical operation; The key abnormal feature search module is used to collect records of abnormal operation in the vehicle history, extract vehicle operation information in the records, extract key abnormal features that cause abnormal vehicle operation, and classify the extracted key abnormal features; The association analysis module is used to perform association analysis on different types of key abnormal features to obtain potential association relationships between different key abnormal features; The model building module is used to analyze key abnormal feature data when abnormalities occurred in the vehicle in the past, and to build a fault prediction model and a data prediction model respectively; The fault judgment module is used to collect key abnormal feature data in real time during vehicle operation and input it into the trained fault prediction model to determine whether the vehicle has abnormal risk; The data prediction module is used to input the information analysis standard time and real-time key abnormal characteristic data values ​​into the data prediction model to obtain future data values.

9. The vehicle operation information evaluation and management system based on big data according to claim 8 is characterized by: The association analysis module includes an association judgment unit and a potential association relationship calculation unit; The association judgment unit is used to calculate the mean square error in the scatter plot to determine whether there is an association between two key abnormal features; The potential correlation relationship calculation unit is used to extract two different key abnormal features as a combination to draw a scatter plot, and use polynomial regression to fit the points in the scatter plot to obtain a polynomial as a potential correlation relationship.

10. The vehicle operation information evaluation and management system based on big data according to claim 8, characterized in that: The model building module includes a fault prediction model unit and a data prediction model unit; The fault prediction model unit uses a deep learning algorithm to build a fault prediction model, and collects key abnormal features of abnormal operation in the history of the vehicle as a training set to train the fault prediction model; The data prediction model unit is used to calculate the time prediction function and the association prediction function respectively, and combine the two functions to obtain the data prediction model.

Images