A method and system for monitoring vehicle driving status based on big data

Abstract

This invention relates to the field of data processing technology, specifically to a method and system for monitoring vehicle driving status based on big data. The method includes: collecting pressure data from the bottom of the vehicle's fuel tank; selecting neighboring data segments for each pressure data point; determining anomaly factors for each pressure data point based on its prominence relative to its neighboring data segments and the dispersion of pressure data within those segments; calculating a correction coefficient for the threshold of each pressure data point by comprehensively considering the anomaly factors and the data change trends of neighboring data segments; correcting the threshold of the pressure data using the correction coefficient; and performing anomaly detection based on the corrected threshold. This method improves the accuracy of anomaly detection in pressure data and ensures the accuracy of fuel level monitoring during vehicle operation.

Description

Technical Field

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

[0002] In today's rapidly evolving world of intelligent vehicles, accurate vehicle status monitoring is crucial for ensuring driving safety, and fuel level detection is an indispensable key component. Traditional fuel level monitoring largely relies on float-type sensors, which are susceptible to the irregular shape of the fuel tank and frequent changes in vehicle posture, resulting in inaccurate and untimely fuel level readings. Therefore, current technology employs a hydrostatic method for fuel level monitoring. This method utilizes a pressure sensor installed at the bottom of the fuel tank to accurately calculate changes in fuel level by measuring the hydrostatic pressure of the liquid, thus achieving real-time fuel level monitoring.

[0003] Because vehicle driving environments are complex and varied, factors such as bumps and tilts can cause fluctuations in pressure data. These fluctuations may originate from normal driving conditions or may indicate an abnormality in the actual fuel level. Therefore, accurate anomaly detection in pressure data is crucial.

[0004] The CUSUM algorithm is commonly used for data anomaly detection. Its core principle is to perform cumulative sum calculations on the data to keenly detect changes and determine if anomalies have occurred. Under normal circumstances, data fluctuations are considered random changes within a reasonable range. Once the cumulative sum at a data point exceeds a pre-set threshold, the data is considered anomaly. For example, Chinese patent application CN118673012A discloses a method for cleaning hydropower unit monitoring data based on outlier detection. This method first collects monitoring data from the long-term operation of the hydropower unit, then uses the CUSUM algorithm to perform preliminary analysis on the collected data to obtain feature vectors. Next, the DPC algorithm is used to identify the discrete values ​​in the feature vectors, calculate the local density and relative distance between data points, and thus determine and label anomaly clusters. Finally, the CUSUM and DPC algorithms are integrated to accurately identify and remove outliers from the detected data.

[0005] However, in vehicle condition monitoring scenarios, the traditional CUSUM algorithm has revealed obvious limitations. This algorithm pre-sets a uniform 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 changes in pressure data at different times and the impact of possible noise interference on the actual distribution of data.

[0006] Specifically, when pressure data is affected by noise such as electromagnetic interference, fluctuations may occur that do not accurately reflect changes in oil level. Because fixed thresholds fail to account for the dynamic characteristics of the data, these instantaneous fluctuations within the normal range may accumulate and exceed the threshold, leading to misjudgments as anomalies. Furthermore, when the oil level rises or falls normally and slowly, it appears in the data as a series of continuous, small-amplitude pressure changes. Since fixed thresholds cannot be dynamically adjusted according to data trends, the accumulated sum may take a considerable amount of time to exceed the threshold. This results in a slow response to such genuine oil level changes, making it impossible to detect normal oil level changes promptly and accurately. These misjudgments and delays severely reduce the accuracy and reliability of anomaly detection, thus significantly impacting the accurate monitoring of vehicle status. Summary of the Invention

[0007] To address the issues that traditional CUSUM algorithms are susceptible to noise interference and unable to adapt to dynamic changes in pressure data when monitoring vehicle fuel levels, thus reducing the accuracy and reliability of anomaly detection and affecting the accurate monitoring of vehicle status, this invention proposes a vehicle driving status monitoring method based on big data, comprising:

[0008] Pressure data at the bottom of the vehicle's fuel tank is collected according to the preset collection duration and frequency. The algorithm uniformly sets a threshold for each pressure data point; it selects neighboring data segments for each pressure data point, and determines the anomaly factor of the pressure data point based on the prominence of the pressure data point in the neighboring data segment and the dispersion of the pressure data points contained in the neighboring data segment.

[0009] The correction factor for calculating the threshold of each pressure data point is:

[0010]

[0011] In the formula, It is the first The correction factor for the threshold of each pressure data point. It is the first Anomalies in the stress data and The first The number of times the change in pressure data is positive and the number of times the change is negative between any two adjacent pressure data segments. , For functions that take the maximum value and functions that take the minimum value, The total number of changes. , For the first The first data segment adjacent to the pressure data. The, the A variable, It is a natural exponential function. For normalization function, It is the absolute value symbol;

[0012] The pressure data threshold is corrected using a correction factor, and the CUSUM algorithm is used to detect anomalies based on the corrected threshold of all pressure data in order to monitor the oil level during vehicle operation.

[0013] The aforementioned technical solution acquires pressure data at regular time intervals by setting specific collection durations and frequencies, ensuring the systematic and complete nature of data collection and covering changes in the bottom pressure of the vehicle's fuel tank over different time periods. Furthermore, it acquires neighboring data segments based on the local characteristics of each pressure data point. The prominence reflects the deviation of the current pressure data from its neighboring segments, while the dispersion reflects the distribution of pressure data within those segments. The anomaly factor, combining these two factors, quantifies the likelihood of anomalies in individual pressure data points. When pressure data deviates significantly from neighboring segments and the dispersion of pressure data within those segments is large, the anomaly factor is high, indicating that the pressure data is more likely to be anomalous. This provides a targeted basis for subsequent threshold adjustments and improves the sensitivity to anomaly data identification. Furthermore, the threshold for each pressure data point is dynamically adjusted based on the degree of anomaly in the pressure data itself (reflected by anomaly factors) and the diversity and trend of pressure changes in adjacent data segments (comparison of the number of positive and negative changes and the sum of the differences between adjacent changes). If the anomaly factor is large and the pressure data changes in adjacent data segments are complex (large differences in the number of positive and negative changes and a large sum of the differences between adjacent changes), the correction coefficient is also large, and the preset threshold for that pressure data is significantly adjusted to better reflect the actual situation of the current pressure data. This effectively overcomes the problem that the uniform threshold set for each pressure data point in the traditional CUSUM algorithm cannot adapt to dynamic data changes, reduces false positives and false negatives caused by fixed thresholds, and improves the accuracy of anomaly detection. Furthermore, the corrected threshold for the pressure data allows the CUSUM algorithm to more accurately adapt to the actual situation of different pressure data during anomaly detection. The corrected threshold can reasonably distinguish between normal fluctuations and abnormal changes when there is noise interference in the data, avoiding false positives caused by noise. Even when the oil level changes slowly, the calculation of the cumulative sum can be adjusted in a timely manner, detecting the true change in oil level more quickly, thereby improving the accuracy and reliability of oil level detection during vehicle operation.

[0014] Preferably, the outliers in the pressure data satisfy the following relationship:

[0015] , It is the first Anomalies in the stress data For the first The prominence of a stress data point within its adjacent data segments For the first The degree of dispersion of pressure data contained in neighboring data segments of a pressure data set, and the degree of prominence. and the degree of dispersion The determination is based on the following formula:

[0016] , , For the first The average of each pressure data point and its adjacent data points. For the first The mean of the nearest data segments of each pressure data point. For hyperparameters, For the first The number of pressure data points corresponding to the mode of the nearest neighboring data segment for each pressure data point. For the first The total number of pressure data points contained in the neighboring data segments of each pressure data point. It is a natural exponential function. It is the absolute value symbol.

[0017] The aforementioned technical solution effectively quantifies the deviation of the current pressure data from the mean of its neighboring data segments by assessing the prominence of the current pressure data. This highlights potentially anomalous pressure data. The dispersion of pressure data within neighboring data segments effectively reflects the distribution of pressure data within those segments. Dispersion is closely related to anomaly detection; high dispersion indicates greater volatility and a higher likelihood of anomalies. Setting the anomaly factor as the product of prominence and dispersion allows for a comprehensive consideration of various characteristics of the pressure data within neighboring data segments, enabling a holistic assessment of the likelihood of each pressure data point being an anomaly from different perspectives.

[0018] Preferably, another method for determining the degree of dispersion of pressure data contained in neighboring data segments is to use the variance of pressure data contained in neighboring data segments as the degree of dispersion.

[0019] The above technical solution takes into account that changes in the dispersion of pressure data are an important clue in the anomaly detection of oil levels. When the dispersion suddenly increases, it may indicate abnormal fluctuations in the oil level, whether due to equipment failure, fluid leakage, or other abnormal conditions. By detecting changes in the dispersion, these potential anomalies can be detected more promptly, providing a more sensitive anomaly detection capability.

[0020] Preferably, the method for correcting the threshold of the pressure data using a correction factor is as follows:

[0021] The correction factor for the threshold of each pressure data point is multiplied by the threshold of that pressure data point, and the product is used as the corrected threshold for that pressure data point.

[0022] Preferably, the method for selecting the nearest data segment for each pressure data point is as follows:

[0023] Select before each pressure data acquisition time A series of consecutive adjacent pressure data points constitute the neighboring data segment of that pressure data. This is the preset quantity.

[0024] Preferably, the method for anomaly detection using the CUSUM algorithm based on a threshold corrected from all pressure data is as follows:

[0025] Calculate the cumulative sum at each pressure data point. If the cumulative sum at a pressure data point is greater than the corrected threshold for that pressure data, the pressure data is considered abnormal; if the cumulative sum at a pressure data point is not greater than the corrected threshold for that pressure data, the pressure data is considered normal.

[0026] The above technical solution takes into account that traditional fixed thresholds cannot accurately adapt to the characteristics of different pressure data and the complex changes in pressure data. By modifying the threshold, the judgment criteria can be adjusted in real time according to the changes in pressure data. Customized judgments can be made for the specific situation of each pressure data, thereby more accurately identifying real abnormal data, reducing the possibility of false positives and false negatives, and improving the accuracy and reliability of anomaly detection.

[0027] Preferably, the method of using the CUSUM algorithm to perform anomaly detection on all pressure data based on a threshold corrected from all pressure data further includes:

[0028] When the CUSUM algorithm performs anomaly detection on all pressure data, if an anomaly is detected in a certain pressure data, it determines that the oil level display corresponding to the time when the pressure data was collected is abnormal; if no anomaly is detected in a certain pressure data, it determines that the oil level display corresponding to the time when the pressure data was collected is normal.

[0029] 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 pressure data collection, analysis, and anomaly detection, it further realizes the effective handling of abnormal situations, so that the entire system forms a complete closed loop from data detection to problem feedback, improving the timeliness and reliability of detection.

[0030] The present invention also provides a vehicle driving status monitoring system based on big data. The vehicle driving status monitoring system includes a pressure sensor, an analog-to-digital converter, a data processing unit, and a communication unit. The pressure sensor is used to collect pressure data at the bottom of the vehicle's fuel tank, the analog-to-digital converter is used to convert the pressure data from analog to digital, the data processing unit is used to execute the vehicle driving status monitoring method, and the communication unit is used to notify the detection results.

[0031] The detection system in the aforementioned technical solution encompasses several key components, including pressure sensors, analog-to-digital converters, data processing units, and communication units. Each component has a clearly defined function and works collaboratively to achieve a complete process from collecting, converting, and processing pressure data from the bottom of the vehicle's fuel tank to finally notifying personnel of the test results. This completeness ensures the orderly operation of the entire monitoring system.

[0032] Preferably, the communication unit employs wireless communication technology.

[0033] Preferably, the data processing unit further includes a data storage module, which stores the collected pressure data, intermediate data during the calculation process, and anomaly detection results for a storage time of not less than [duration not specified]. For subsequent data tracing, The preset duration.

[0034] The data storage module of the aforementioned technical solution achieves complete retention of pressure data, intermediate data, and anomaly detection results, obtaining comprehensive data information on the oil level detection process. This provides a detailed data foundation for in-depth analysis of each stage of the detection process and accurate interpretation of dynamic oil level changes. Furthermore, the data traceability function effectively assists staff in exploring long-term oil level change patterns. In the event of anomalies, historical data can be efficiently traced back to investigate the root cause. Simultaneously, in specific event correlation analysis, relevant details of the oil level at that time can be accurately verified, greatly enhancing the data's usability and decision support capabilities.

[0035] The present invention has the following effects:

[0036] This invention calculates anomaly factors by integrating multiple features of neighboring data for each pressure data point, and then derives a correction coefficient for the threshold. This enables dynamic adjustment of the threshold for each pressure data point. Whether it is an instantaneous fluctuation caused by noise interference or a continuous small change caused by a slow and gradual change in oil level, the corrected threshold can more accurately match the actual data situation, effectively reducing misjudgments and omissions, and improving the accuracy and timeliness of remote oil level anomaly detection. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0038] Figure 2 This is a schematic diagram of the method flow of S22 of the present invention;

[0039] Figure 3 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0040] Reference Figure 1 The present invention provides a vehicle driving status monitoring method based on big data, including the following steps. -step :

[0041] S1: Collect pressure data at the bottom of the vehicle's fuel tank according to the preset collection duration and collection frequency.

[0042] The static pressure method takes into account that the oil in the tank will generate a pressure at the bottom of the tank due to gravity, and this pressure is proportional to the height of the oil level in the tank.

[0043] A high-precision pressure sensor is used to collect pressure data at the bottom of the vehicle's fuel tank in real time. The collection time is set to two hours and the collection frequency is once per second to ensure that a sufficiently detailed and representative data sample is obtained to accurately reflect the pressure status at the bottom of the vehicle's fuel tank.

[0044] The pressure value measured by the pressure sensor at the bottom of the oil tank is essentially determined by the static pressure generated by the oil level (i.e., the oil level). The pressure data directly reflects the oil level. As long as the pressure sensor is accurate and there are no other interfering factors, the oil level can be accurately calculated by accurately measuring the pressure and applying the static pressure formula.

[0045] However, during vehicle operation, pressure data will fluctuate due to factors such as vehicle bumps and tilting. These fluctuations are normal and unrelated to changes in the oil level itself. Accurate anomaly detection in the pressure data can distinguish between normal pressure fluctuations caused by vehicle operating conditions and pressure changes due to actual changes in the oil level. After eliminating normal fluctuations, the remaining pressure data anomalies are more likely caused by actual changes in the oil level, allowing for a more accurate assessment of the oil level based on the pressure data.

[0046] S2: Analyze the changing characteristics of the pressure data and adaptively adjust the threshold of each pressure data point to obtain the adjusted threshold for each pressure data point.

[0047] This invention uses the CUSUM algorithm to detect anomalies in the bottom pressure data of the fuel tank.

[0048] The CUSUM algorithm is a statistical method for detecting points of change in a data sequence. Its basic principle is to monitor the degree of deviation of the data by calculating the cumulative sum of the data sequence, and to 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 these differences.

[0049] In the current scenario, the traditional CUSUM algorithm sets a fixed, uniform threshold for all stress data, then calculates the cumulative sum at each stress data point. When the cumulative sum at a stress data point exceeds the preset threshold, that stress data is considered abnormal. However, this method of anomaly detection based on a preset fixed threshold is easily affected by noise and cannot adapt to dynamic changes in data, leading to inaccurate anomaly detection results.

[0050] Therefore, this step performs precise correction calculations on 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, anomaly detection is performed based on the corrected threshold for each pressure data, which can improve the accuracy of anomaly detection results.

[0051] Specifically, this step involves in-depth analysis of the numerical characteristics and trends of each pressure data point and its neighboring pressure data to ensure the accuracy and reliability of anomaly detection. For example... Figure 2 As shown, the steps include:

[0052] S21: Obtain the outlier factors for each stress data point.

[0053] Based on the overall scheme's logic, this step focuses on an in-depth analysis of the numerical presentation of each pressure data point and its neighboring pressure data to identify the outlier factors for each pressure data point. Specifically, if a pressure data point stands out significantly compared to its neighboring pressure data points, and the distribution of these neighboring pressure data points is highly dispersed, it indicates that the pressure data point deviates significantly from the norm in the local data environment, suggesting a large outlier factor and a correspondingly increased likelihood of anomalies.

[0054] To implement the above analysis logic, the following operations are performed:

[0055] First, select the nearest data segment for each pressure data point as follows:

[0056] For any pressure data point, trace back backward from the time the pressure data was collected and select... A series of consecutive adjacent pressure data, The preset quantity is set here. The value is 50, which is an empirical value. These 50 pressure data points constitute the adjacent data segment of this pressure data. Alternatively, if fewer than 50 pressure data points were collected before this pressure data point, to ensure the completeness and accuracy of the analysis, the data can be retrieved one minute prior to the time the pressure data was collected. This operation effectively avoids analytical bias caused by insufficient pressure data and ensures the scientific validity and reliability of the anomaly factor calculation.

[0057] Next, the prominence of each pressure data point within its neighboring data segments is evaluated, as well as the dispersion of the pressure data contained within those neighboring data segments.

[0058] Among them, the The formula for calculating the prominence of a pressure data point within its neighboring data segments is as follows:

[0059]

[0060] In the formula, For the first The prominence of a stress data point within its adjacent data segments For the first The average of each pressure data point and its adjacent data points. For the first The mean of the nearest data segments of each pressure data point. For hyperparameters, This value is an empirical value; this hyperparameter exists to prevent... and To ensure that the values ​​are equal, the formula avoids situations where a value of 0 fails to accurately reflect the prominence of pressure data, thus guaranteeing that the formula effectively measures the prominence of pressure data. It is the absolute value symbol.

[0061] In this formula, Taking into account the first The pressure data itself, as well as the surrounding data, are averaged to obtain a value that reflects the overall level of this local data. The emphasis is on the mean of only the nearest data segment, excluding the first segment. The pressure data itself, and The difference lies in whether or not the first The average value is calculated from the pressure data itself. It is the first A basic deviation measure of the pressure data from its neighboring data segments; if this deviation measure is large, it indicates that the pressure data is... The pressure data point deviates significantly from its neighboring data points, i.e., the first... The more prominent a stress data point is in its adjacent data segments, the more significant it becomes, and vice versa.

[0062] Among them, the The formula for calculating the dispersion of pressure data contained in the neighboring data segments of a pressure data set is:

[0063]

[0064] In the formula, For the first The degree of dispersion of pressure data contained in the neighboring data segments of a pressure data set. For the first The number of pressure data points corresponding to the mode of the nearest neighboring data segment for each pressure data point. For the first The total number of pressure data points contained in the neighboring data segments of each pressure data point. It is a natural exponential function.

[0065] In the formula An example of the method for obtaining it is: for example, the first There are 10 data points in the neighboring data segment of a stress data point: 2, 3, 3, 4, 5, 5, 5, 6, 7, 8. The number 5 appears most frequently, is the mode, and appears 3 times. Therefore, for this neighboring data segment... .

[0066] In this formula, Reflects the first The relative proportion of the pressure data corresponding to the mode in the neighboring data segment of a pressure data point to the pressure data in the neighboring segment is considered. A smaller relative proportion indicates a higher probability of a positive pressure reading in the first stress data point. Within a range of neighboring data points for a given stress data point, the fewer the number of stress data points corresponding to the mode relative to the total number of stress data points, the more dispersed the stress data distribution is within that range, according to data distribution. A smaller number of stress data points corresponding to the mode indicates that the stress data within that range is not concentrated at a specific value, but rather distributed relatively evenly across different values. Conversely, a larger relative proportion suggests that the stress data within that range is concentrated at a specific value. The higher the concentration of stress data within a range, the lower the dispersion. Therefore, this relative proportion and dispersion are negatively correlated. Thus, a negative correlation exponential function can be constructed to accurately represent this relationship between relative proportion and dispersion.

[0067] Finally, by combining the prominence of each pressure data point within its neighboring data segments and the dispersion of pressure data within those neighboring segments, the outlier factors of that pressure data are determined:

[0068]

[0069] In the formula, It is the first Anomalies in the stress data. These anomalies are positively correlated with both dispersion and prominence.

[0070] In real-world stress data detection scenarios, relying solely on a single indicator to determine whether stress data is abnormal is often inaccurate and incomplete. For example, considering only the degree of prominence might reveal a stress data point that is relatively prominent compared to surrounding data, even if the surrounding data is highly concentrated, meaning the data may not actually be abnormal. Conversely, considering only dispersion might mean that even if the stress data is widely distributed, a stress data point with a small difference from the mean of its surrounding data cannot be easily identified as abnormal. This formula, by combining prominence and dispersion to calculate the anomaly factor, avoids these one-sided judgments, more accurately identifying truly abnormal stress data and providing a more reliable basis for subsequent processing and analysis.

[0071] S22: Determine the correction factor for the threshold of each pressure data point.

[0072] Through the preceding series of analytical steps, the anomalous factors for each pressure data point were identified. However, it should be clearly pointed out that the generation mechanism of these anomalous factors primarily focuses on analyzing the numerical appearance of individual pressure data points and their immediate neighbors. While this analytical approach can capture the anomalous characteristics of individual pressure data points to some extent, it has certain limitations in terms of comprehensiveness and accuracy.

[0073] Specifically, because it only considers numerical performance, it cannot definitively determine whether a particular pressure data point truly represents the actual and significant pressure state inside the fuel tank. More importantly, it struggles to effectively identify and eliminate the adverse effects of noise data. In real-world fuel tank pressure data testing scenarios, the presence of noise data undoubtedly interferes with the accurate assessment of the true pressure condition, potentially leading to biases and errors in subsequent analysis and decision-making.

[0074] A deeper exploration of the physical processes inside the fuel tank reveals an inherent and close causal relationship between the dynamic changes in pressure data and fuel level changes. Under normal operating conditions, whether the fuel level rises or falls, it follows a gradual pattern, a process that inevitably requires a certain time span. In stark contrast, noise data, due to its inherent unreliability and randomness, often exhibits abrupt changes, contradicting the gradual logic of fuel level changes.

[0075] When a pressure data point and its adjacent data segments exhibit a highly gradual trend, this supports the authenticity of the pressure data. In other words, such pressure data is more likely to reflect the true pressure changes within the fuel tank, rather than being influenced by noise. Based on this judgment, any subtle deviations at this pressure data point should be taken seriously, as they may contain information about the true pressure changes within the tank. To ensure that these potentially valuable deviations are not overlooked or ignored during subsequent anomaly detection, the correction factor for the threshold at this pressure data point is intentionally set to a small value. This results in a smaller corrected threshold for the pressure data, making it easier to identify even relatively small deviations when using CUSUM for anomaly detection. This improves the accuracy, reliability, and sensitivity of the overall pressure data anomaly detection.

[0076] In light of the above, this step involves in-depth analysis of the data change trends of each pressure data point and its adjacent data segments, and based on this, a correction coefficient for the threshold of each pressure data point is calculated.

[0077]

[0078] In the formula, It is the first The correction factor for the threshold of each pressure data point. It is the first Anomalies in the stress data and The first The number of times the change in pressure data is positive and the number of times the change is negative between any two adjacent pressure data segments. , For functions that take the maximum value and functions that take the minimum value, The total number of changes. , For the first The first data segment adjacent to the pressure data. The, the A variable, It is a natural exponential function. For normalization function, It is the absolute value symbol.

[0079] In this formula, The larger the value, the more prominent the pressure data is in its adjacent data segments, and the more likely there is an anomaly. Deviations caused by such potentially abnormal pressure data should be given more attention. Logically, the correction coefficient of the pressure data threshold should be smaller in this case. In this way, even if the pressure data has a small deviation, it is easy to detect, so that potential anomalies can be discovered in time.

[0080] In this formula, This ratio can be quantified as follows: The greater the ratio of the variation in the changes of all adjacent pressure data in the neighboring data segment, the more likely the changes in the pressure data in that segment are either predominantly positive (the number of positive changes far exceeds the number of negative changes) or predominantly negative (the number of negative changes far exceeds the number of positive changes). In other words, the changes in the pressure data are mainly concentrated in one direction (positive or negative), lacking a balance between positive and negative changes. Therefore, the lower the diversity of the changes, the more gradual the trend of the pressure data in the neighboring data segment tends to be. This gradual trend indicates a higher degree of authenticity of the pressure data. Deviations caused by such pressure data should also be taken seriously, so the correction coefficient for its threshold should be smaller to ensure that the deviation can be easily detected as a significant change.

[0081] As can be seen from the above analysis, in this formula, The focus is on highlighting the prominence and likelihood of anomalies of stress data within adjacent data segments, while This focuses on describing the diversity of pressure data changes in adjacent data segments, or the trend characteristics of those changes. Therefore, the two are first multiplied together to obtain... This comprehensively considers the combined impact of these two important characteristics of pressure data on the threshold correction coefficient. Furthermore, considering... and All with Since the correlation is negative, we obtain the result by constructing a negative correlation function. This aligns perfectly with the expected adjustment of the correction factor, enabling the threshold after adjusting the pressure data based on the correction factor to more sensitively detect data deviations.

[0082] In this formula, Indicates the first The sum of the differences between all adjacent changes in pressure data within a neighboring data segment is the most reliable indicator of a gradual change in pressure data within that segment. This also means that the first... The higher the authenticity of a pressure data point, the smaller the correction coefficient for that pressure data threshold should be, so that deviations can be quickly detected as significant changes.

[0083] In summary, by comprehensively considering various factors, this formula derives a correction coefficient for the threshold of each pressure data point that more accurately reflects the characteristics of each pressure data point, thus laying a solid foundation for setting pressure data thresholds reasonably and sensitively in the future.

[0084] This step calculates a threshold correction coefficient for each pressure data point. This coefficient can dynamically adapt to changes in data characteristics, effectively improving the rationality and sensitivity of threshold settings.

[0085] S23: Correct the preset threshold using a correction coefficient.

[0086] When the correction factor for a certain pressure data threshold is small, it means that the deviation in that pressure data is more likely to originate from actual changes in the fuel tank itself. This is because actual pressure changes within the fuel tank often exhibit consistency and logic, rather than random abnormal fluctuations. In this case, the deviation in the pressure data is more valuable for analysis and should be considered a significant change. According to the operating logic of the CUSUM algorithm, when a deviation is deemed significant, the cumulative sum will correspondingly increase or decrease. This requires setting a smaller threshold for the pressure data. A smaller threshold lowers the standard for judging a deviation as a significant change, allowing even relatively small deviations to be captured and identified in a timely manner. This improves the sensitivity and response speed of the entire pressure data anomaly detection system to actual changes in fuel tank pressure, ensuring that the system can accurately and efficiently detect the fuel tank pressure status.

[0087] Therefore, the threshold of the pressure data is corrected using a correction coefficient for each pressure data point. The specific method is as follows:

[0088] The correction factor for the threshold of each pressure data point is multiplied by the threshold value of that pressure data point, and the product is used as the corrected threshold value for that pressure data point. This can be expressed by the formula:

[0089]

[0090] In the formula, Indicates the first The threshold after adjusting for each pressure data point. A fixed threshold is preset for this pressure data, and the preset threshold is the same for all pressure data. This value is an empirical value, and the unit is kilopascals. Indicates the first Correction coefficients for the threshold of each pressure data point.

[0091] This formula aims to rationally adjust the pressure data threshold based on the possibility that the deviation in the pressure data stems from actual changes in the fuel tank (reflected by the magnitude of the correction coefficient for the pressure data threshold), thereby improving the detection capability of actual changes in fuel tank pressure. The formula obtains the corrected threshold by multiplying a preset fixed threshold of the pressure data by its correction coefficient. This allows different pressure data to adjust the threshold according to their own correction coefficients, achieving dynamic threshold correction. This enables the algorithm to adapt to different driving scenarios and better capture the true anomalies in pressure data under these scenarios, rather than simply misjudging normal driving state changes as abnormal, thus achieving more accurate detection results.

[0092] In this formula, when the correction coefficient for a pressure data threshold is small (i.e., the deviation is more likely to originate from the actual changes in 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 explanation.

[0093] In summary, the logical explanation and calculation formula of this scheme maintain a high degree of consistency. The calculation formula accurately quantifies the idea of ​​adjusting the threshold based on the correction coefficient of the pressure data in the logical explanation, thereby effectively realizing the reasonable correction of the oil tank pressure data threshold according to the established logic.

[0094] S3: The CUSUM algorithm is used to detect anomalies in all pressure data based on the threshold corrected by all pressure data, so as to realize oil level monitoring in the vehicle driving state.

[0095] The method of using the CUSUM algorithm to perform anomaly detection on all pressure data based on a threshold corrected for all pressure data is as follows:

[0096] Calculate the cumulative sum at each pressure data point:

[0097]

[0098] In the formula, For the first The cumulative sum at each pressure data point, To find the maximum value function, For the first The cumulative sum at each pressure data point, For the first One pressure data point, This is the mean of all pressure data.

[0099] This formula, by calculating the cumulative sum at each pressure data point, can grasp the overall trend of pressure data changes. The concept of cumulative sum is the core of the CUSUM algorithm; it considers historical information of the data sequence, not just the current data point. This allows the algorithm to capture potential subtle changes in the data, helping to discover gradual anomalies in the data sequence, rather than just isolated outliers. For vehicle driving status monitoring, this holistic grasp of data trends can more accurately reflect the long-term changing trend of pressure data during vehicle operation, rather than focusing only on instantaneous pressure values, resulting in greater stability and reliability.

[0100] Furthermore, if the cumulative sum at a certain pressure data point is greater than the corrected threshold for that pressure data, the pressure data is considered abnormal; if the cumulative sum at a certain pressure data point is not greater than the corrected threshold for that pressure data, the pressure data is considered normal.

[0101] This invention, by considering the cumulative sum of pressure data and using a modified threshold for anomaly detection, can more accurately determine whether pressure data is abnormal compared to the traditional CUSUM algorithm that uses a fixed threshold. In practical applications, it can more accurately detect abnormal situations such as slow oil level changes (e.g., slow leaks in the fuel tank or during refueling) or abnormal pressure fluctuations in complex driving environments, reducing false positives and false negatives.

[0102] Furthermore, after performing anomaly detection on all pressure data, the following operations are also performed:

[0103] when During the process of anomaly detection of all pressure data, if an anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the time of acquisition of that 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 time of acquisition of that pressure data is normal. This anomaly handling mechanism improves the timeliness of the anomaly detection method.

[0104] When pressure data is abnormal, after ruling out interference factors such as pressure sensor malfunction, it means that the oil level has actually changed. For example, if a vehicle collision causes the fuel tank to rupture and leak, the pressure data measured by the pressure sensor will decrease due to the drop in oil level. This abnormal pressure data accurately reflects the true change in oil level. Alternatively, if the fuel pump malfunctions, causing abnormal pressure fluctuations in the fuel tank, this will also be reflected in the pressure data. Anomaly detection can capture these changes, thereby enabling oil level monitoring.

[0105] In another embodiment of the present invention, a vehicle driving status monitoring system based on big data is also provided, specifically as follows: Figure 3 As shown, it includes:

[0106] Pressure sensor S100: Used to collect pressure data at the bottom of the vehicle's fuel tank. As the source of data acquisition, it is installed at the center of the bottom of the vehicle's fuel tank or in the area where the pressure change is most representative, and it is necessary to ensure that the sensor has good contact with the bottom of the vehicle's fuel tank.

[0107] Analog-to-digital converter S200: Used to convert pressure data from analog to digital. It should have the characteristics of fast conversion and high resolution. Its conversion speed should meet the acquisition frequency requirements of the pressure sensor to ensure the real-time nature of the data and prevent data backlog or loss due to the conversion process. High resolution ensures that the converted digital signal can accurately reproduce the subtle changes in the pressure data, providing an accurate basis for subsequent data processing. In addition, the analog-to-digital converter should have a certain calibration function, which can be automatically calibrated periodically or manually calibrated according to instructions to maintain the stability of conversion accuracy and prevent conversion errors caused by equipment aging or environmental changes.

[0108] Data processing unit S300: This unit executes the vehicle driving status monitoring method to obtain the oil level monitoring results during vehicle operation. It is the core computing module of the entire system. To enable further data traceability, data processing unit S300 also includes a data storage module S301 for storing collected pressure data, intermediate data from the calculation process, and anomaly detection results. The storage time is no less than [duration missing]. For subsequent data tracing, The preset duration is set to 1 month based on experience. The data storage module S301 has data backup and recovery functions to prevent data loss due to storage device failure or other unexpected situations. For example, it uses redundant storage technology or regularly and automatically backs up data to external storage devices to ensure data security and integrity.

[0109] Communication Unit S400: Used to notify staff of oil level monitoring results, employing wireless communication technologies, including but not limited to Bluetooth, Wi-Fi, and any of 4G / 5G. Flexible configuration is possible based on the specific application scenario. In areas with good signal coverage and where data transmission speed requirements are not extremely high, Bluetooth or Wi-Fi can meet short-range data transmission needs, offering advantages such as low cost and low power consumption. 4G / 5G communication technology provides wider coverage and higher data transmission rates, ensuring timely and accurate notification of detection results. The Communication Unit S400 also features encrypted transmission capabilities, ensuring data security and confidentiality during transmission and preventing data theft or tampering. For example, it uses encryption protocols such as SSL / TLS to encrypt transmitted data, ensuring data integrity and privacy.

Claims

1. A method for monitoring vehicle driving status based on big data, characterized in that, include: Pressure data at the bottom of the vehicle's fuel tank is collected according to the preset collection duration and frequency. The algorithm uniformly sets a threshold for each pressure data point; Select the neighboring data segment for each pressure data point, and determine the outlier factor of the pressure data based on the prominence of the pressure data in the neighboring data segment and the dispersion of the pressure data contained in the neighboring data segment. The outliers in the stress data satisfy the following relationship: , It is the first Anomalies in the stress data For the first The prominence of a stress data point within its adjacent data segments For the first The degree of dispersion of pressure data contained in neighboring data segments of a pressure data set, and the degree of prominence. and the degree of dispersion The determination is based on the following formula: , , For the first The average of each pressure data point and its adjacent data points. For the first The mean of the nearest data segments of each pressure data point. For hyperparameters, For the first The number of pressure data points corresponding to the mode of the nearest neighboring data segment for each pressure data point. For the first The total number of pressure data points contained in the neighboring data segments of each pressure data point. It is a natural exponential function. It is the absolute value symbol; The correction factor for calculating the threshold of each pressure data point is: In the formula, It is the first The correction factor for the threshold of each pressure data point. It is the first Anomalies in the stress data and The first The number of times the change in pressure data between any two adjacent data points is positive and the number of times the change is negative. , For functions that take the maximum value and functions that take the minimum value, The total number of changes. , For the first The first data segment adjacent to the pressure data. The, the A variable, It is a natural exponential function. For normalization function, It is the absolute value symbol; The threshold of the pressure data is corrected using a correction factor. The algorithm performs anomaly detection based on a threshold corrected from all pressure data in order to monitor the oil level while the vehicle is in motion.

2. The vehicle driving status monitoring method based on big data according to claim 1, characterized in that, Another method to determine the degree of dispersion of pressure data contained in neighboring data segments is to use the variance of the pressure data contained in neighboring data segments as the degree of dispersion.

3. The vehicle driving status monitoring method based on big data according to claim 1, characterized in that, The method for correcting the threshold of pressure data using a correction factor is as follows: The correction factor for the threshold of each pressure data point is multiplied by the threshold of that pressure data point, and the product value is used as the corrected threshold of that pressure data point.

4. The vehicle driving status monitoring method based on big data according to claim 1, characterized in that, The method for selecting the nearest data segment for each pressure data point is as follows: Select before each pressure data acquisition time A series of consecutive adjacent pressure data points constitute the neighboring data segment of that pressure data. This is the preset quantity.

5. The vehicle driving status monitoring method based on big data according to claim 1, characterized in that, use The algorithm uses a threshold corrected from all pressure data to perform anomaly detection as follows: Calculate the cumulative sum at each pressure data point. If the cumulative sum at a pressure data point is greater than the corrected threshold for that pressure data, then that pressure data is considered abnormal. If the cumulative sum at a certain pressure data point is not greater than the corrected threshold for that pressure data, then that pressure data is not abnormal.

6. The vehicle driving status monitoring method based on big data according to claim 5, characterized in that, use The algorithm performs anomaly detection based on a threshold adjusted from all pressure data, and further includes: when During the process of anomaly detection of all pressure data, if an anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the time when the pressure data was collected is abnormal; if no anomaly is detected in a certain pressure data, it is determined that the oil level display corresponding to the time when the pressure data was collected is normal.

7. A vehicle driving status monitoring system based on big data, characterized in that, The vehicle driving status monitoring system includes a pressure sensor, an analog-to-digital converter, a data processing unit, and a communication unit; A pressure sensor is used to collect pressure data at the bottom of the vehicle's fuel tank; an analog-to-digital converter is used to convert the pressure data from analog to digital; a data processing unit is used to execute the condition monitoring method as described in any one of claims 1-6; and a communication unit is used to notify the detection results.

8. The vehicle driving status monitoring system based on big data according to claim 7, characterized in that, The communication unit employs wireless communication technology.

9. The vehicle driving status monitoring system based on big data according to claim 7, characterized in that, The data processing unit further includes a data storage module, which stores the collected pressure data, intermediate data during the calculation process, and anomaly detection results for a storage time of not less than [duration not specified]. For subsequent data tracing, The preset duration.

Images