Vehicle driving state monitoring method and system based on big data

Abstract

The invention relates to the technical field of data processing, in particular to a vehicle driving state monitoring method and system based on big data, and the method comprises the steps: collecting pressure data of the bottom of a vehicle oil tank, and then selecting adjacent data segments of the pressure data, determining an abnormal factor of each piece of pressure data according to the prominence degree of each piece of pressure data relative to the adjacent data segment and the dispersion degree of the pressure data contained in the adjacent data segment, and calculating a correction coefficient of a threshold value of each piece of pressure data by comprehensively considering the abnormal factor and the data change trend of the adjacent data segment; and correcting the threshold value of the pressure data by using the correction coefficient, and carrying out anomaly detection based on the corrected threshold value. According to the method, the accuracy of abnormal detection of the pressure data is improved, and the accuracy of oil level monitoring in the vehicle running state is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing. Specifically, it relates to a method and system for monitoring vehicle driving status based on big data. Background Art

[0002] In the current rapid development of vehicle intelligence, accurately monitoring the vehicle status is of great significance for ensuring driving safety, and oil level detection is an essential key link among them. Most traditional vehicle oil levels are based on the principle of float sensors. This method is easily affected by the irregular shape of the fuel tank and the frequent changes in the vehicle driving posture, resulting in the inability to timely and accurately feedback the true information of the oil volume. Therefore, the prior art uses the hydrostatic pressure method for oil volume monitoring. This method relies on a pressure sensor installed at the bottom of the fuel tank to accurately calculate the change in oil volume by measuring the liquid hydrostatic pressure, thereby realizing real-time monitoring of the oil volume.

[0003] Due to the complex and changeable driving environment of vehicles, factors such as bumps and tilts will cause fluctuations in pressure data. These fluctuations may stem from normal driving conditions or may be manifestations of abnormal real oil volumes. Therefore, accurate anomaly detection of pressure data is crucial.

[0004] The CUSUM (Cumulative Sum) algorithm is commonly used for data anomaly detection. Its core principle is to perform cumulative sum operations on the data to keenly capture changes in the data and then determine whether an anomaly occurs. Under normal circumstances, the fluctuations in the data are regarded as random changes within a certain reasonable range. Once the cumulative sum at a certain data point exceeds a pre-set threshold, it is determined that the data is abnormal. For example, in the Chinese patent application document with the publication number CN118673012A, a method for cleaning detection data of a hydropower unit based on outlier detection is disclosed. This method first collects the detection data of the long-term operation of the hydropower unit, and then uses the CUSUM algorithm to preliminarily analyze the collected data to obtain feature vectors. Then, the DPC algorithm is used to identify the discrete values in the feature vectors, calculate the local density and relative distance between data points, thereby determining and marking abnormal clusters. Finally, the CUSUM algorithm and the DPC algorithm are integrated to accurately identify and remove outliers in the detection data.

[0005] However, in the scenario of vehicle status monitoring, the traditional CUSUM algorithm exposes obvious limitations. This algorithm pre-sets a unified fixed threshold for all collected pressure data. Throughout the detection process, it always uses this fixed standard to measure whether each pressure data deviates from the normal range to an abnormal degree, completely ignoring the dynamic change characteristics of pressure data at different times and the influence of factors such as possible noise interference on the actual distribution of the data.

[0006] Specifically, when the pressure data is affected by noise such as electromagnetic interference, there will be some fluctuations that do not truly reflect the change in the oil level. Since the fixed threshold fails to consider the dynamic characteristics of the data, these instantaneous fluctuations within the normal range may cause the cumulative sum to exceed the threshold and thus be misjudged as abnormal. In addition, when the oil level is rising or falling slowly and normally, it is manifested as a series of continuous small-amplitude pressure changes in the data. Due to the fixed threshold being unable to be dynamically adjusted according to the data change trend, the cumulative sum may take a long time to exceed the threshold, resulting in a slow response to this real oil level change and being unable to detect the normal oil level change situation in a timely and accurate manner. This misjudgment and delay seriously reduce the accuracy and reliability of anomaly detection, and thus greatly affect the accurate monitoring of the vehicle status. Summary of the Invention

[0007] To solve the problem that in the monitoring of the oil level of a vehicle, the traditional CUSUM algorithm is vulnerable to noise interference and unable to adapt to the dynamic changes of pressure data, reducing the accuracy and reliability of anomaly detection and affecting the accurate monitoring of the vehicle status, the present invention proposes a method for monitoring the driving status of a vehicle based on big data, including: Collecting the pressure data at the bottom of the vehicle fuel tank according to a preset acquisition duration and acquisition frequency, and through the algorithm to uniformly set the threshold for each pressure data; selecting the adjacent data segments of each pressure data, and determining the anomaly factor of the pressure data according to the prominence of the pressure data in the adjacent data segments and the dispersion degree of the pressure data included in the adjacent data segments; Calculating the correction coefficient of the threshold for each pressure data as:

[0008] In the formula, is the correction coefficient of the threshold of the th pressure data, is the anomaly factor of the th pressure data, and are respectively the number of times the change amount of two adjacent pressure data in the adjacent data segment of the th pressure data is positive and the number of times the change amount is negative, and are the maximum value function and the minimum value function, is the total number of change amounts, and are the th and th change amounts in the adjacent data segment of the th pressure data, is the natural exponential function, is the normalization function, is the absolute value symbol; The threshold of the pressure data is corrected by using a correction coefficient, and the CUSUM algorithm is used to perform anomaly detection based on the corrected threshold of all pressure data to monitor the oil level in the vehicle driving state.

[0009] Through the above technical solution, by setting a specific acquisition duration and acquisition frequency, pressure data is obtained at regular time intervals, ensuring the systematicness and integrity of data acquisition, and being able to cover the changes in the bottom pressure of the vehicle fuel tank in different time periods. Furthermore, the adjacent data segments of the pressure data are obtained. Starting from the local characteristics of each pressure data, the prominence can reflect the deviation of the current pressure data from its adjacent data segments, and the dispersion degree reflects the distribution dispersion of the pressure data included in the adjacent data segments. The anomaly factor combines these two factors and can quantify the anomaly possibility of a single pressure data. When the pressure data deviates significantly in the adjacent data segments and the dispersion degree of the pressure data included in the adjacent data segments is large, the anomaly factor of the pressure data will be large, indicating that this pressure data is more likely to be abnormal data, providing a targeted basis for the subsequent correction of the threshold and improving the sensitivity of identifying abnormal data. Further, according to the anomaly degree of the pressure data itself (reflected by the anomaly factor) and the diversity and trend of the pressure changes in the adjacent data segments (the comparison of the number of positive and negative change amounts and the sum of the differences between adjacent change amounts), the threshold of each pressure data is dynamically adjusted. If the anomaly factor is large and the pressure data changes in the adjacent data segments are relatively complex (the difference in the number of positive and negative change amounts is large and the sum of the differences between adjacent change amounts is large), the correction coefficient is also large, and the preset threshold of this pressure data will be adjusted greatly, making the threshold more suitable for the actual situation of the current pressure data, effectively overcoming the problem that the unified threshold set for each pressure data in the traditional CUSUM algorithm cannot adapt to the dynamic changes of the data, reducing misjudgments and missed judgments caused by the fixed threshold, and improving the accuracy of anomaly detection. Further, through the corrected threshold of the pressure data, the CUSUM algorithm can more accurately adapt to the actual situation of different pressure data during the anomaly detection process. The corrected threshold can reasonably distinguish normal fluctuations and abnormal changes when there is noise interference in the data, avoiding misjudgments caused by noise. When the oil level changes slowly, it can also adjust the calculation of the cumulative sum in time and detect the real change of the oil level faster, thereby improving the accuracy and reliability of the oil level detection in the vehicle driving state.

[0010] Preferably, the anomaly factor of the pressure data satisfies the following relational expression: , is the anomaly factor of the th pressure data, is the prominence of the th pressure data in its adjacent data segments, For the degree of dispersion of the pressure data included in the adjacent data segment of the th pressure data, and the prominence and the degree of dispersion , , are determined based on the following formula: is the mean value of the th pressure data and its adjacent data segment, is the mean value of the adjacent data segment of the th pressure data, is a hyperparameter, is the number of pressure data corresponding to the mode in the adjacent data segment of the th pressure data, is the total number of pressure data included in the adjacent data segment of the th pressure data, is the natural exponential function,

[0011] is the absolute value symbol.

[0012] The above technical solution effectively quantifies the deviation degree of the current pressure data relative to the mean value of its adjacent data segment through the prominence of the current pressure data, thereby highlighting those pressure data that may be abnormal. Through the degree of dispersion of the pressure data included in the adjacent data segment of the current pressure data, the distribution of the pressure data within the adjacent data segment can be well reflected. The degree of dispersion is closely related to the abnormal judgment. A high degree of dispersion of the pressure data means that the pressure data has a large volatility and is more likely to have abnormal data. Setting the abnormal factor as the product relationship between the prominence and the degree of dispersion can comprehensively consider various characteristics of the pressure data in the adjacent data segment and comprehensively evaluate the possibility of each pressure data becoming an abnormal data from different angles.

[0013] Preferably, another method for determining the degree of dispersion of the pressure data included in the adjacent data segment of the pressure data is: using the variance value of the pressure data included in the adjacent data segment of the pressure data as the degree of dispersion.

[0014] Preferably, the method for correcting the threshold of the pressure data using a correction coefficient is: Multiply the correction coefficient of the threshold value of each pressure data by the threshold value of this pressure data, and use the obtained product value as the corrected threshold value of this pressure data.

[0015] Preferably, the method for selecting the adjacent data segment of each pressure data is as follows: Select before the acquisition moment of each pressure data consecutive adjacent pressure data to form the adjacent data segment of this pressure data, is a preset quantity.

[0016] Preferably, the method for anomaly detection using the CUSUM algorithm based on the corrected threshold values of all pressure data is as follows: Calculate the cumulative sum at each pressure data. If the cumulative sum at a certain pressure data is greater than the corrected threshold value of this pressure data, this pressure data is abnormal; if the cumulative sum at a certain pressure data is not greater than the corrected threshold value of this pressure data, this pressure data is not abnormal.

[0017] The above technical solution takes into account that the traditional fixed threshold cannot accurately adapt to the characteristics of different pressure data and the complex changes of pressure data. The corrected threshold can adjust the judgment criterion in real time according to the changes of pressure data, and can perform customized judgment according to the specific situation of each pressure data, so as to more accurately identify the true abnormal data, reduce the possibility of misjudgment and missed judgment, and improve the accuracy and reliability of anomaly detection.

[0018] Preferably, using the CUSUM algorithm to perform anomaly detection on all pressure data based on the corrected threshold values of all pressure data further includes: During the process of using the CUSUM algorithm to perform anomaly detection on all pressure data, if an anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the acquisition moment of this pressure data is abnormal; if no anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the acquisition moment of this pressure data is normal.

[0019] The anomaly handling mechanism designed in the above technical solution is an important part of the entire monitoring method. After completing a series of operations such as acquisition, analysis, and anomaly detection of pressure data, it further realizes the effective handling of abnormal situations, making the entire system form a complete closed loop from data detection to problem feedback, and improving the timeliness and reliability of detection.

[0020] The present invention also provides a vehicle driving state monitoring system based on big data. The vehicle driving state monitoring system includes a pressure sensor, an analog-to-digital conversion device, a data processing unit, and a communication unit. The pressure sensor is used to collect pressure data at the bottom of the vehicle fuel tank. The analog-to-digital conversion device is used to perform analog-to-digital conversion on the pressure data. The data processing unit is used to execute the vehicle driving state monitoring method. The communication unit is used to notify the detection result.

[0021] The detection system in the above technical solution covers multiple key components such as a pressure sensor, an analog-to-digital conversion device, a data processing unit, and a communication unit. Each component has a clear division of labor and works in coordination to jointly achieve a complete process from the collection, conversion, and processing of pressure data at the bottom of the vehicle fuel tank to finally notifying the detection result to the staff. This integrity ensures the orderly operation of the entire monitoring system.

[0022] Preferably, the communication unit uses wireless communication technology.

[0023] Preferably, the data processing unit further includes a data storage module. The data storage module is used to store the collected pressure data, intermediate data during the calculation process, and abnormal detection results, and the storage duration is not less than for subsequent data traceability, which is a preset duration.

[0024] The data storage module in the above technical solution realizes the complete retention of pressure data, intermediate data, and abnormal detection results, obtaining comprehensive data information on the oil level detection process, providing a detailed data basis for in-depth analysis of each link in the detection process and accurate interpretation of the dynamic changes in the oil level. And the data traceability function effectively assists the staff in deeply exploring the long-term change law of the oil level, being able to efficiently trace historical data to investigate the root cause when an abnormal situation occurs, and accurately verifying the relevant details of the oil level at that time in the correlation analysis of specific events, greatly improving the data utilization value and decision-making support ability.

[0025] The present invention has the following effects: The present invention calculates the abnormal factor by comprehensively considering various characteristics of the adjacent data of each pressure data, and then obtains the correction coefficient of the threshold, realizing the dynamic adjustment of the threshold for each pressure data. Whether it is an instantaneous fluctuation caused by noise interference or a continuous small change caused by the slow gradual change of the oil level, the corrected threshold can more accurately fit the actual situation of the data, effectively reducing misjudgment and missed judgment, and improving the accuracy and timeliness of abnormal detection of the remote oil level. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic flowchart of the method of the present invention; Figure 2It is a schematic flowchart of the method of S22 of the present invention; Figure 3 It is a schematic structural diagram of the system of the present invention. Specific embodiments

[0027] Referring to Figure 1 , a vehicle driving state monitoring method based on big data provided by the present invention includes the steps - Step : S1: Collect the pressure data at the bottom of the vehicle fuel tank according to the preset collection duration and collection frequency.

[0028] The static pressure method takes into account that the oil in the fuel tank will generate a pressure at the bottom of the fuel tank due to the action of gravity, and this pressure is proportional to the height of the oil level in the fuel tank.

[0029] A high-precision pressure sensor is used to collect the pressure data at the bottom of the interior of the vehicle fuel tank in real time. The collection duration is set to two hours, and the collection frequency is once per second to ensure obtaining sufficiently fine and representative data samples to accurately reflect the pressure condition at the bottom of the vehicle fuel tank.

[0030] The pressure value measured by the pressure sensor at the bottom of the fuel tank is essentially determined by the static pressure generated by the oil height (i.e., the oil level). The pressure data directly reflects the height of the oil level. As long as the pressure sensor measures accurately and there are no other interference factors, the oil level can be accurately calculated by accurately measuring the pressure and according to the static pressure formula.

[0031] However, during the vehicle driving process, the pressure data will fluctuate due to factors such as vehicle bumping and tilting, but these are normal fluctuations and have nothing to do with the change of the oil level itself. By accurately detecting the abnormality of the pressure data, the normal fluctuations of the pressure caused by factors such as the vehicle driving state can be distinguished from the pressure change caused by the real change of the oil level. After excluding the normal fluctuation factors, the remaining abnormal pressure data is more likely to be caused by the real change of the oil level, so that the oil level situation can be more accurately judged according to the pressure data.

[0032] S2: Analyze the change characteristics of the pressure data and adaptively correct the threshold of each pressure data to obtain the corrected threshold of each pressure data.

[0033] The present invention uses the CUSUM algorithm to perform anomaly detection on the pressure data at the bottom of the fuel tank.

[0034] The CUSUM algorithm is a statistical method used to detect change points in a data sequence. Its basic principle is to monitor the deviation degree of the data by calculating the cumulative sum of the data sequence, and trigger an event when the cumulative sum exceeds a preset threshold. The CUSUM algorithm detects abnormal changes in the data sequence by calculating the difference between each data point and the expected value and accumulating the difference values.

[0035] In the current scenario, the traditional CUSUM algorithm sets a fixed and unified threshold for all pressure data, and then calculates the cumulative sum at each pressure data. When the cumulative sum at a certain pressure data exceeds the preset threshold, it is considered that the pressure data has an abnormality. However, since this method of detecting abnormalities by presetting a fixed threshold is easily affected by noise and cannot adapt to the dynamic changes of the data, the abnormality detection results are inaccurate.

[0036] Therefore, in this step, precise correction calculations are carried out for the unified threshold set by the CUSUM algorithm for each pressure data to obtain the corrected threshold for each pressure data. The corrected threshold can more accurately reflect the dynamic change trend of the pressure data. Furthermore, based on the corrected threshold of each pressure data for abnormality detection, the accuracy of the abnormality detection results can be improved.

[0037] Specifically, this step mainly deeply analyzes the numerical characteristics and change trends of each pressure data and its neighboring pressure data to ensure the accuracy and reliability of abnormality detection. As Figure 2 shown, it includes the following steps: S21: Obtain the abnormality factor of each pressure data.

[0038] According to the logic of the overall scheme, this link focuses on deeply analyzing the numerical presentation of each pressure data and its neighboring pressure data to determine the abnormality factor of each pressure data. Specifically, if a certain pressure data is extremely prominent compared with its neighboring pressure data, and the distribution of the neighboring pressure data of this pressure data shows a highly discrete state, it indicates that the degree of deviation of this pressure data from the normal state in the local data environment is relatively high, indicating that the abnormality factor of this pressure data is relatively large, and the possibility of abnormality also increases accordingly.

[0039] To implement the above analysis logic, the following specific operations are carried out: First, select the neighboring data segments of each pressure data according to the following method: For any pressure data, trace back in time from the acquisition moment of this pressure data and select consecutively adjacent pressure data, is the preset quantity, and here it is set It is 50. This value is an empirical value. These 50 pressure data constitute the adjacent data segment of this pressure data. Additionally, if the number of pressure data collected before this pressure data is less than 50, to ensure the integrity and accuracy of the analysis, one minute can be traced back from the collection moment of this pressure data to obtain the pressure data within this minute. This operation can effectively avoid analysis deviations caused by insufficient pressure data quantity and ensure the scientificity and reliability of abnormal factor calculation.

[0040] Next, evaluate the prominence of each pressure data in its adjacent data segment and the dispersion degree of the pressure data included in the adjacent data segment of this pressure data.

[0041] Among them, the formula for calculating the prominence of the

[0042] In the formula, is the prominence of the th pressure data in its adjacent data segment, is the mean value of the th pressure data and its adjacent data segment, is the mean value of the adjacent data segment of the th pressure data, is a hyperparameter, , this value is an empirical value. The existence of this hyperparameter is to prevent and from being equal, avoiding the situation where it is difficult to accurately reflect the prominence of pressure data because the value of the formula is 0, and ensuring that the formula can effectively measure the prominence of pressure data, is the absolute value symbol.

[0043] In this formula, comprehensively considers the th pressure data itself and the situation of its surrounding adjacent data, and obtains a value that can reflect the overall level of this local data by calculating the mean value. It is emphasized that only the mean value of the adjacent data segment is considered, not including the th pressure data itself. The difference between and lies in whether the th pressure data itself is included in the mean value calculation. is a basic deviation measure of the th pressure data from its adjacent data segment. If this deviation measure is large, it indicates that the The more prominent a pressure data is in its adjacent data segment, and vice versa.

[0044] Among them, for the adjacent data segment of the

[0045] pressure data, the calculation formula for the dispersion degree of the pressure data it contains is: In the formula, is the dispersion degree of the pressure data contained in the adjacent data segment of the th pressure data, is the number of pressure data corresponding to the mode in the adjacent data segment of the th pressure data,

[0046] is the total number of pressure data contained in the adjacent data segment of the th pressure data, and is the natural exponential function.

[0047] In this formula, reflects the relative proportion in terms of quantity between the pressure data corresponding to the mode in the adjacent data segment of the th pressure data and the pressure data in the adjacent data segment. When this relative proportion is smaller, it means that within the adjacent data segment of the

[0048] th pressure data, the number of pressure data corresponding to the mode is less relative to the total number of pressure data. From the perspective of data distribution, this indicates that the pressure data in this adjacent data segment is more discrete, because the small number of pressure data corresponding to the mode means that the pressure data within the adjacent data segment is not significantly concentrated on a specific value but is more dispersed among different values. Conversely, if this relative proportion is larger, it means that the pressure data within the adjacent data segment is concentrated on a specific value, and the higher the concentration degree of the pressure data within the data segment, the smaller the dispersion degree. It can be seen that this relative proportion and the dispersion degree are in a negative correlation relationship. Therefore, a negative correlation exponential function is constructed to accurately represent the relationship between this relative proportion and the dispersion degree.

[0049] In the formula, is the anomaly factor of the th pressure data. The anomaly factor has a positive correlation with both the degree of dispersion and the degree of prominence.

[0050] In the actual pressure data detection scenario, relying solely on a single indicator to judge whether the pressure data is abnormal is often not accurate and comprehensive enough. For example, just looking at the degree of prominence, there may be a situation where although a certain pressure data is relatively prominent compared to the surrounding pressure data, the distribution of the surrounding pressure data is very concentrated, and actually this pressure data may not be a truly abnormal pressure data; or only considering the degree of dispersion, even if the pressure data is distributed very dispersedly, but the difference between a certain pressure data and the mean value of the surrounding pressure data is not large, it cannot be easily determined as an abnormal pressure data. And this formula calculates the anomaly factor by comprehensively considering the degree of prominence and the degree of dispersion, which can avoid these one-sided judgment problems and can more accurately screen out the truly abnormal pressure data, providing a more reliable basis for subsequent processing and analysis.

[0051] S22: Determine the correction coefficient of the threshold for each pressure data.

[0052] Through a series of previous analysis steps, the anomaly factor of each pressure data is obtained. However, it should be clearly pointed out that the generation mechanism of these anomaly factors mainly focuses on analyzing the numerical appearance of a single pressure data and the pressure data immediately adjacent to it. Although this analysis method can capture the abnormal characteristics of a single pressure data to a certain extent, from the perspectives of comprehensiveness and accuracy, it has certain limitations.

[0053] Specifically, because it only considers the numerical performance, it is impossible to definitely judge whether a certain pressure data actually represents a real and prominent pressure state inside the fuel tank. More importantly, it is difficult to effectively identify and eliminate the adverse effects brought by noise data. In the actual fuel tank pressure data detection scenario, the existence of noise data will undoubtedly interfere with the accurate judgment of the real pressure situation, and may further lead to deviations and mistakes in a series of subsequent analysis and decisions.

[0054] By deeply exploring the physical process inside the fuel tank, it can be known that there is an inherent and close causal relationship between the dynamic change of the pressure data in the fuel tank and the change of the oil level. Under normal operating conditions, whether the oil level rises or falls, it follows a gradual change law, and this process necessarily requires a certain time span to complete. In sharp contrast, due to the inherent unreality and randomness of noise data, it often shows a sudden change characteristic, which is incompatible with the gradual change logic followed by the oil level change.

[0055] When the data change trend of a certain pressure data and its adjacent data segments shows a high degree of gradualness, this provides support for the authenticity of the pressure data. That is to say, such pressure data can better reflect the real pressure change inside the fuel tank rather than being interfered by noise data. Based on this judgment, any slight deviation at the position of this pressure data should be taken seriously because these deviations may very likely contain information about the real pressure change inside the fuel tank. In order to ensure that these potentially valuable deviations will not be overlooked or ignored during the subsequent anomaly detection process, the correction coefficient of the threshold at the position of this pressure data is specifically set to a relatively small value. In this way, the corrected threshold of this pressure data will also be relatively small. When using CUSUM for anomaly detection later, even if only a relatively small deviation occurs at the position of this pressure data, it can be more easily identified, thereby improving the accuracy, reliability, and sensitivity of the entire pressure data anomaly detection.

[0056] In view of the above, in this step, by deeply mining the data change trend information of each pressure data and its adjacent data segments, the correction coefficient of the threshold of each pressure data is calculated based on this:

[0057] In the formula, is the correction coefficient of the threshold of the th pressure data, is the anomaly factor of the th pressure data, and are respectively the number of times the change amount of two adjacent pressure data in the adjacent data segment of the th pressure data is positive and the number of times the change amount is negative. and are the maximum value function and the minimum value function, is the total number of change amounts, and are the th and the th, the th change amounts in the adjacent data segment of the th pressure data, is the natural exponential function, is the normalization function, is the absolute value symbol.

[0058] In this formula, The larger it is, the more prominent the pressure data is in its adjacent data segment, and the more likely it is that there are abnormal situations. For the deviation caused by such potentially abnormal pressure data, more attention should be paid. Logically, the correction coefficient of the threshold of this pressure data should be smaller at this time. In this way, even if a small deviation occurs in this pressure data, it is easy to be detected, so that potential abnormal situations can be discovered in time.

[0059] In this formula, This ratio can quantify the diversity of the change amounts of all adjacent pressure data in the adjacent data segment of the

[0060] In this formula, as can be seen from the above analysis, focuses on reflecting the prominence degree and abnormal possibility of the pressure data in the adjacent data segment, while focuses on describing the diversity of the change amounts of the pressure data in the adjacent data segment, or rather, the trend characteristics of the change. Therefore, multiplying the two first gives which comprehensively considers the combined influence of these two important characteristics of the pressure data on the threshold correction coefficient. Further, considering that and are both negatively correlated with Therefore, by constructing a negative correlation function, we get which exactly meets the expectation of adjusting the correction coefficient, making the threshold after correcting the pressure data based on the correction coefficient able to detect data deviation more sensitively.

[0061] In this formula, represents the sum of the difference amounts between the change amounts of all adjacent (pressure data) in the adjacent data segment of the pressure data. The smaller this value is, the greater the credibility that the change trend of the pressure data in this adjacent data segment tends to be a gradual change. This also means that the pressure data has a higher authenticity. Then, the correction coefficient of the threshold of this pressure data should also be smaller so that the deviation can be quickly detected as a significant change.

[0062] In summary, by comprehensively considering various aspects of information, the formula can obtain the correction coefficient of the threshold for each pressure data, which can more accurately reflect the characteristics of each pressure data, thus laying a solid foundation for reasonably and sensitively setting the threshold of pressure data subsequently.

[0063] The threshold correction coefficient of each pressure data calculated in this step can dynamically adapt to the characteristics of data changes and effectively improve the rationality and sensitivity of threshold setting.

[0064] S23: Use the correction coefficient to correct the preset threshold.

[0065] When the correction coefficient of the threshold of a certain pressure data is small, it means that the deviation of this pressure data is more likely to stem from the true changes of the fuel tank itself. This is because the true changes in the pressure inside the fuel tank usually have coherence and logic, rather than random abnormal fluctuations. In this case, the deviation generated by this pressure data is more valuable for analysis and should be regarded as a significant change. According to the operating logic of the CUSUM algorithm, when the generated deviation is determined to be significant, the cumulative sum will correspondingly continue to increase or decrease. This requires setting a smaller threshold for this pressure data. A smaller threshold can lower the standard for determining a deviation as a significant change, enabling even relatively small deviations to be captured and identified in a timely manner, thereby enhancing the sensitivity and response speed of the entire pressure data anomaly detection system to the true pressure changes in the fuel tank and ensuring that the system can accurately and efficiently detect the pressure condition of the fuel tank.

[0066] Therefore, use the correction coefficient of the threshold of each pressure data to correct the threshold of this pressure data. The specific method is as follows: Multiply the correction coefficient of the threshold of each pressure data by the threshold of this pressure data, and use the obtained product value as the corrected threshold of this pressure data. It is expressed by the formula:

[0067] In the formula, represents the corrected threshold of the th pressure data, is the preset fixed threshold of this pressure data, and the preset thresholds of all pressure data are the same, , this value is an empirical value with the unit of kilopascal, represents the correction coefficient of the threshold of the th pressure data.

[0068] This formula is to reasonably adjust the threshold of pressure data according to the possibility that the deviation of pressure data stems from the real change of the fuel tank (reflected by the magnitude of the correction coefficient of the threshold of pressure data), so as to improve the detection ability of the real pressure change of the fuel tank. This formula multiplies the preset fixed threshold of pressure data by its threshold correction coefficient to obtain the corrected threshold, enabling different pressure data to adjust the threshold according to their own correction coefficients, realizing dynamic threshold correction, making the algorithm adaptable to different driving scenarios, and being able to better capture the real anomalies of pressure data in these scenarios, rather than simply misjudging the normal driving state changes as anomalies, thus achieving more accurate detection results.

[0069] In this formula, when the correction coefficient of the threshold of a pressure data is small (that is, the deviation is more likely to stem from the real change of the fuel tank), the corrected threshold will also be smaller, which is consistent with the direction of adjusting the threshold according to the magnitude of the correction coefficient in the above logical elaboration.

[0070] To sum up, the logical elaboration and calculation formula of this scheme maintain a high degree of consistency. The calculation formula accurately quantifies the idea of adjusting the threshold according to the correction coefficient of pressure data in the logical elaboration, so as to be able to effectively realize the reasonable correction of the threshold of fuel tank pressure data according to the established logic.

[0071] S3: Use the CUSUM algorithm to perform anomaly detection on all pressure data based on the corrected threshold of all pressure data to realize the fuel level monitoring in the vehicle driving state.

[0072] The method of using the CUSUM algorithm to perform anomaly detection on all pressure data based on the corrected threshold of all pressure data is as follows: Calculate the cumulative sum at each pressure data:

[0073] In the formula, is the cumulative sum at the th pressure data, is the maximum value function, is the cumulative sum at the th pressure data, is the th pressure data, is the mean value of all pressure data.

[0074] By calculating the cumulative sum at each pressure data point, this formula can comprehensively capture the changing trend of pressure data. The concept of cumulative sum is the core of the CUSUM algorithm, which takes into account the historical information of the data sequence and is not limited to the current data point. This enables the algorithm to detect potential subtle changes in the data, helping to identify gradual abnormal changes in the data sequence rather than just isolated abnormal points. For vehicle driving state monitoring, this comprehensive grasp of the data changing trend can more accurately reflect the long-term changing trend of pressure data during vehicle driving, rather than just focusing on instantaneous pressure values, and has stronger stability and reliability.

[0075] Furthermore, if the cumulative sum at a certain pressure data point is greater than the corrected threshold of this pressure data, this pressure data is considered abnormal; if the cumulative sum at a certain pressure data point is not greater than the corrected threshold of this pressure data, this pressure data is not considered abnormal.

[0076] By considering the cumulative sum of pressure data and using the corrected threshold for anomaly detection, the present invention can more accurately determine whether the pressure data is abnormal compared with the traditional CUSUM algorithm using a fixed threshold. In practical applications, for those slow fuel level changes (such as slow fuel tank leakage or refueling process) or abnormal pressure fluctuations in complex driving environments, it can more accurately detect abnormal situations and reduce false positives and false negatives.

[0077] Furthermore, after performing anomaly detection on all pressure data, the following operations are also carried out: When During the process of the algorithm performing anomaly detection on all pressure data, if an abnormal pressure data is detected, it is determined that the fuel level display corresponding to the acquisition moment of this pressure data is abnormal; if no abnormal pressure data is detected, it is determined that the fuel level display corresponding to the acquisition moment of this pressure data is normal. This anomaly handling mechanism improves the timeliness of the anomaly detection method.

[0078] When the pressure data is abnormal, after excluding interference factors such as the malfunction of the pressure sensor itself, it means that the fuel level has actually changed. For example, when the vehicle collides and the fuel tank is damaged and leaks, the pressure data measured by the pressure sensor will decrease due to the fuel level drop, and this abnormal pressure data accurately reflects the real change of the fuel level. Or when the fuel pump works abnormally and causes abnormal pressure fluctuations in the fuel tank, it will also be reflected in the pressure data, and these changes can be captured through anomaly detection, thereby realizing the monitoring of the fuel level.

[0079] In another embodiment of the present invention, a vehicle driving state monitoring system based on big data is also provided, specifically as Figure 3 shown, including: Pressure sensor S100: It is used to collect the pressure data at the bottom of the vehicle fuel tank. As the source of data collection, it is installed at the center of the bottom of the vehicle fuel tank or in the area where the pressure change is most representative, and it is necessary to ensure good contact between the sensor and the bottom of the vehicle fuel tank.

[0080] Analog-to-digital conversion device S200: It is used to perform analog-to-digital conversion on the pressure data. It should have the characteristics of fast conversion and high resolution. Its conversion speed should be able to meet the acquisition frequency requirements of the pressure sensor to ensure the real-time nature of the data, and there will be no data backlog or loss due to the conversion process. High resolution ensures that the converted digital signal can accurately restore the subtle changes in the pressure data, providing an accurate basis for subsequent data processing. In addition, the analog-to-digital conversion device should have a certain calibration function, which can be automatically calibrated regularly or manually calibrated according to instructions to maintain the stability of the conversion accuracy and prevent conversion errors caused by equipment aging or environmental changes.

[0081] Data processing unit S300: It is used to execute the vehicle driving state monitoring method to obtain the oil level monitoring result in the vehicle driving state. It belongs to the core operation module of the entire system. In order to further trace the data, the data processing unit S300 is also provided with a data storage module S301, which is used to store the collected pressure data, intermediate data during the calculation process, and abnormal detection results. The storage duration is not less than , for subsequent data tracing. is a preset duration, and the empirical value is set to 1 month. And the data storage module S301 has the functions of data backup and recovery to prevent data loss due to storage device failures or other unexpected situations. For example, redundant storage technology or the method of regularly automatically backing up data to an external storage device is adopted to ensure the security and integrity of the data.

[0082] Communication unit S400: It is used to notify the staff of the oil level monitoring result. It adopts wireless communication technology, including but not limited to any one of Bluetooth, Wi-Fi, 4G / 5G. It is flexibly configured according to the actual application scenario. When the signal is good in the area where the fuel tank is located and the requirement for data transmission speed is not extremely high, Bluetooth or Wi-Fi can meet the short-distance data transmission needs and has the advantages of low cost and low power consumption. 4G / 5G communication technology can provide a larger coverage area and high-speed data transmission rate to ensure that the detection results can be notified in a timely and accurate manner. The communication unit S400 also has an encrypted transmission function to ensure the security and confidentiality of the data during the transmission process and prevent the data from being stolen or tampered with. For example, encryption protocols such as SSL / TLS are used to encrypt the transmitted data to ensure the integrity and privacy of the data.

Claims

1. A method for monitoring the driving state of a vehicle based on big data, characterized in that Including: Collect the pressure data at the bottom of the vehicle fuel tank according to the preset acquisition duration and acquisition frequency, and uniformly set the threshold value of each pressure data through the algorithm; Select adjacent data segments of each pressure data, and determine the anomaly factor of the pressure data according to the prominence of the pressure data in the adjacent data segments and the dispersion degree of the pressure data included in the adjacent data segments; Calculate the correction coefficient of the threshold value of each pressure data as follows: ； Wherein, is the correction coefficient of the threshold value of the -th pressure data, is the anomaly factor of the -th pressure data, and are respectively the number of times the change amount of every two adjacent pressure data in the adjacent data segment of the -th pressure data is positive and the number of times the change amount is negative, , are the maximum value function and the minimum value function, is the total number of change amounts, , are the -th and -th change amounts in the adjacent data segment of the -th pressure data, is the natural exponential function, is the normalization function, is the absolute value symbol; The threshold of the pressure data is corrected using a correction coefficient, and the algorithm performs anomaly detection based on the corrected threshold of all pressure data to monitor the oil level in the vehicle driving state.

2. The vehicle driving state monitoring method based on big data according to claim 1, characterized in that, The anomaly factor of the pressure data satisfies the following relational expression: , is the anomaly factor of the nth pressure data, is the prominence of the nth pressure data in its adjacent data segment, is the dispersion degree of the pressure data included in the adjacent data segment of the nth pressure data, and the said prominence and the said dispersion degree are determined based on the following formula: , , is the mean of the th pressure data and its adjacent data segments, is the mean of the adjacent data segments of the th pressure data, is a hyperparameter, is the number of pressure data corresponding to the mode in the adjacent data segments of the th pressure data, is the total number of pressure data included in the adjacent data segments of the th pressure data, is the natural exponential function, is the absolute value symbol.

3. The vehicle driving state monitoring method based on big data according to claim 2, characterized in that Another determination method for the dispersion degree of the pressure data included in the adjacent data segments of the pressure data is: taking the variance value of the pressure data included in the adjacent data segments of the pressure data as the dispersion degree.

4. The vehicle driving state monitoring method based on big data according to claim 1, characterized in that The method for correcting the threshold value of the pressure data by using the correction coefficient is: Multiply the correction coefficient of the threshold value of each pressure data by the threshold value of the pressure data, and take the obtained product value as the corrected threshold value of the pressure data.

5. The vehicle driving state monitoring method based on big data according to claim 1, characterized in that The method for selecting adjacent data segments of each pressure data is: Select consecutive adjacent pressure data before each acquisition time of the pressure data to form the adjacent data segment of the pressure data, where is a preset quantity. ​ 6. The vehicle driving state monitoring method based on big data according to claim 1, characterized in that Utilize The method for anomaly detection using the algorithm based on the threshold value after correcting all pressure data is as follows: Calculate the cumulative sum at each pressure data. If the cumulative sum at a certain pressure data is greater than the corrected threshold value of the pressure data, the pressure data is abnormal; If the cumulative sum at a certain pressure data is not greater than the corrected threshold value of the pressure data, the pressure data is not abnormal.

7. The vehicle driving state monitoring method based on big data according to claim 6, characterized in that, Utilize The algorithm performs anomaly detection based on the threshold values after correcting all pressure data, and further includes: When During the process of the algorithm performing anomaly detection on all pressure data, if an anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the acquisition time of this pressure data is abnormal; if no anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the acquisition time of this pressure data is normal.

8. A vehicle driving state monitoring system based on big data, characterized in that, The vehicle driving state monitoring system includes a pressure sensor, an analog-to-digital conversion device, a data processing unit, and a communication unit; The pressure sensor is used to collect the pressure data at the bottom of the vehicle fuel tank. The analog-to-digital conversion device is used to perform analog-to-digital conversion on the pressure data. The data processing unit is used to execute the state monitoring method described in any one of claims 1-7, and the communication unit is used to notify the detection result.

9. The vehicle driving state monitoring system based on big data according to claim 8, characterized in that, The communication unit adopts wireless communication technology.

10. The vehicle driving state monitoring system based on big data according to claim 8, wherein The data processing unit further includes a data storage module, which is used to store the collected pressure data, intermediate data during the calculation process, and anomaly detection results, and the storage duration is not less than , for subsequent data traceability, which is a preset duration.

Images