Automobile tire pressure monitoring method and system based on multi-sensor fusion

Abstract

The invention provides an automobile tire pressure monitoring method and system based on multi-sensor fusion, and the method comprises the steps: obtaining the temperature, pressure and vibration parameters of a working environment of a sensor module, building a multiple regression analysis model with the temperature, pressure and vibration as independent variables and the sensor output precision as a dependent variable, and obtaining a sensor performance degradation mathematical model; for deformation and vibration interference in the high-speed running process of a tire, a self-adaptive noise elimination algorithm based on the minimum mean square error criterion is adopted, preprocessing such as filtering and smoothing is carried out on original data of a sensor, and the signal-to-noise ratio and stability of the data are improved; a general data analysis module is designed for data formats and protocol differences of different types of sensors, automatic identification and conversion of heterogeneous sensor data are achieved by configuring a sensor data format template, and the data are mapped to a unified data model.

Description

Technical Field

[0001] The present invention belongs to the technical field of tire pressure monitoring, and in particular relates to a method and system for monitoring automobile tire pressure based on multi-sensor fusion. Background Art

[0002] In the automobile tire pressure monitoring system, the sensor module is installed inside the tire or in the space between the tire and the wheel hub. It works in harsh environments for a long time and faces multiple tests such as high temperature, high pressure, vibration, and impact. These extreme conditions may cause the sensor's measurement accuracy to decrease, data transmission to be unstable, and even cause damage or failure of the sensor. At the same time, when the tire is driving at high speed, due to uneven road surface, sudden braking, etc., it will produce severe deformation and vibration, which will interfere with the sensor's data collection and transmission. In addition, the high temperature environment inside the tire will also affect the working performance and life of the sensor.

[0003] Therefore, a monitoring method is urgently needed to determine the status of the sensor. Summary of the invention

[0004] The present invention proposes a vehicle tire pressure monitoring method and system based on multi-sensor fusion to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a method for monitoring automobile tire pressure based on multi-sensor fusion, comprising the following steps:

[0006] Acquire working environment data of the sensor module, and construct a sensor performance degradation model based on the working environment data;

[0007] Preprocessing the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data;

[0008] Extracting data features of the preprocessed data, standardizing the data features and mapping them to a unified data model;

[0009] Dynamically select different data fusion algorithms according to the working status and data quality indicators, fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data;

[0010] The fused data is analyzed according to a preset tire pressure threshold range to obtain a sensor state.

[0011] Preferably, constructing a sensor performance degradation model based on the working environment data includes:

[0012] The temperature, pressure and vibration parameters of the sensor under different working environments are obtained as independent variables, and the output accuracy of the sensor under the corresponding environment is obtained as the dependent variable. A multiple linear regression algorithm is used to combine the independent variables and the dependent variables to establish a sensor performance degradation model.

[0013] Preferably, obtaining preprocessed data includes:

[0014] The sensor performance degradation model is used to identify and compensate for changes in sensor performance. A tire vibration noise model is established based on the vibration and deformation characteristics of the tire when driving at high speed. The original data collected by the sensor is decomposed in the time-frequency domain using the wavelet analysis method to extract signal characteristics at different scales. The scale and translation parameters of the wavelet basis function are adaptively adjusted according to the minimum mean square error criterion. The signal characteristics at different scales are recursively estimated and predicted using the Kalman filter algorithm to dynamically track the changing trend of the signal and improve the smoothness and stability of the data. According to the tire vibration noise model and the wavelet analysis results, the state transfer matrix and observation matrix of the Kalman filter are adaptively adjusted.

[0015] Preferably, extracting data features of the preprocessed data, standardizing the data features and mapping them to a unified data model comprises:

[0016] According to the pre-configured data format template, the data format and protocol type are judged. If the match is successful, data parsing is performed to extract key fields as data features. According to the field names and types defined in the unified data model, the extracted data features are converted and mapped to generate standardized data records.

[0017] Preferably, extracting key fields includes:

[0018] A machine learning algorithm is used to extract features from the preprocessed data to obtain data feature vectors. The feature vectors are classified through a clustering algorithm to identify different types of sensor data. For the different identified sensor types, the corresponding parsing rules are obtained from the pre-established protocol parsing library, the raw data is parsed, and complete data fields are extracted.

[0019] Preferably, the method also includes: in the data fusion process, introducing a data quality assessment mechanism, through real-time assessment of data integrity, accuracy and timeliness indicators, dynamically adjusting the weights and confidence levels of different sensor data, and eliminating or downgrading low-quality data.

[0020] The present invention also provides a vehicle tire pressure monitoring system based on multi-sensor fusion, comprising:

[0021] A sensor performance degradation model building module, used to obtain working environment data of the sensor module and build a sensor performance degradation model based on the working environment data;

[0022] A data preprocessing module, used to preprocess the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data;

[0023] A general data analysis module, used to extract data features of the preprocessed data, standardize the data features and map them to a unified data model;

[0024] The multi-model adaptive fusion module is used to dynamically select different data fusion algorithms according to the working status and data quality indicators, and fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data;

[0025] The fault diagnosis module is used to analyze the fused data according to a preset tire pressure threshold range to obtain a sensor status.

[0026] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0027] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method when executed by a processor.

[0028] The present invention also provides a computer program product, comprising a computer program, characterized in that the steps of the method are implemented when the computer program is executed by a processor.

[0029] Compared with the prior art, the present invention has the following advantages and technical effects:

[0030] The present invention discloses a method and system for monitoring automobile tire pressure based on multi-sensor fusion. Aiming at the problems of high-speed deformation and vibration interference of tires, an adaptive noise elimination algorithm is used to pre-process sensor data to improve data quality. By establishing a sensor performance degradation model, real-time evaluation of the working status of the sensor is achieved. A general data parsing module is designed to solve the identification and conversion of heterogeneous sensor data. A multi-model adaptive fusion strategy is adopted to dynamically select the optimal algorithm combination to achieve data fusion. A data quality evaluation mechanism is introduced to downgrade low-quality data. A fault diagnosis and fault-tolerant mechanism is established to start a backup or switching algorithm when a sensor fails. The present invention effectively solves the accuracy and reliability problems of tire pressure monitoring under high-speed driving conditions through the organic combination of multiple innovative technologies, improves the anti-interference ability and robustness of the system, and provides reliable technical support for ensuring driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0032] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0034] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0035] Embodiment 1

[0036] like Figure 1 As shown, this embodiment provides a method for monitoring automobile tire pressure based on multi-sensor fusion, comprising the following steps:

[0037] Acquire the working environment data of the sensor module and build a sensor performance degradation model based on the working environment data;

[0038] Preprocessing the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data;

[0039] Extract data features of preprocessed data, standardize the data features and map them to a unified data model;

[0040] Dynamically select different data fusion algorithms according to the working status and data quality indicators, fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data;

[0041] The fused data is analyzed according to the preset tire pressure threshold range to obtain the sensor status.

[0042] The specific steps include:

[0043] S101. Obtain the temperature, pressure, and vibration parameters of the working environment of the sensor module, establish a multivariate regression analysis model with temperature, pressure, and vibration as independent variables and sensor output accuracy as a dependent variable, and obtain a mathematical model of sensor performance degradation.

[0044] Specifically, the temperature, pressure and vibration parameters of the sensor under different working environments are obtained as independent variables, and the output accuracy of the sensor under the corresponding environment is obtained as the dependent variable to construct a data set. The obtained data set is preprocessed to remove abnormal data and normalize the data to make the numerical range of each feature consistent. According to the preprocessed data set, a regression analysis model is established using a multivariate linear regression algorithm to obtain the mathematical relationship between temperature, pressure and vibration parameters and the output accuracy of the sensor. The fitting effect and prediction accuracy of the regression model are evaluated by cross-validation. If the model performance does not meet the requirements, the hyperparameters of the model are adjusted or other regression algorithms are selected, such as polynomial regression, support vector regression, etc.

[0045] In this embodiment, taking the temperature sensor as an example, its temperature, pressure and vibration parameters, as well as the corresponding output accuracy data, can be collected under different environments. For example, the temperature change (20-150°C), pressure change (0.1-1.0MPa) and vibration frequency (10-100Hz) during driving can be recorded, and the output accuracy of the temperature sensor can be measured at the same time. Data preprocessing is a key step in model building. First, it is necessary to eliminate abnormal data, such as sudden temperature changes or sudden pressure drops caused by equipment failure. Then the data is normalized to unify the parameters of different dimensions into the range of 0-1. This can eliminate the dimensional effect and make the model more stable. For example, the minimum-maximum normalization method can be used to map the temperature, pressure and vibration parameters to a unified interval respectively. Multiple linear regression is a commonly used modeling method. Assume that the output accuracy y of the temperature sensor is linearly related to temperature x1, pressure x2 and vibration x3: y=β0+β1x1+β2x2+β3x3+ε. The least squares method can be used to estimate the regression coefficients β0, β1, β2, and β3, thereby establishing a mathematical model. This method is simple and intuitive, easy to understand and implement. Model evaluation is an important part of ensuring the reliability of the model. Cross-validation is a commonly used evaluation method that can effectively avoid overfitting problems. For example, using 10-fold cross-validation, the data set is randomly divided into 10 parts, 9 parts are used for training each time, and 1 part is used for testing, and repeated 10 times, and the average error is taken as the model performance indicator. If the root mean square error (RMSE) is too large, it means that the model fit is not good and needs to be adjusted. When the linear model cannot meet the requirements, a more complex regression algorithm can be considered. Polynomial regression can capture nonlinear relationships, such as y=β0+β1x1+β2x2^2+β3x1x2+ε. Support vector regression (SVR) can handle high-dimensional feature space and realize nonlinear mapping through kernel functions, which is suitable for complex sensor performance degradation models. Model application is the ultimate goal of the whole process.

[0046] S102. In view of the deformation and vibration interference of the tire during high-speed driving, an adaptive noise elimination algorithm based on the minimum mean square error criterion is used to perform pre-processing such as filtering and smoothing on the raw data of the sensor to improve the data signal-to-noise ratio and stability.

[0047] Specifically, according to the vibration and deformation characteristics of the tire when driving at high speed, a tire vibration noise model is established to describe the statistical characteristics and spectrum distribution law of the noise. The wavelet analysis method is used to decompose the raw data collected by the sensor in the time-frequency domain, extract the signal characteristics at different scales, and achieve the preliminary separation of noise and effective signals. According to the minimum mean square error criterion, the scale and translation parameters of the wavelet basis function are adaptively adjusted to match the statistical characteristics of the noise signal, so as to achieve accurate estimation and elimination of noise. The denoised signal is recursively estimated and predicted by the Kalman filter algorithm, and the change trend of the signal is dynamically tracked to further improve the smoothness and stability of the data. According to the tire vibration noise model and the wavelet analysis results, the state transfer matrix and observation matrix of the Kalman filter are adaptively adjusted to adapt to the dynamic characteristics of the signal and improve the filtering performance. The denoised and smoothed sensor data are fused with the tire deformation model, and the deformation parameters of the tire are estimated by the least squares method to achieve accurate measurement and compensation of tire deformation.

[0048] In this embodiment, when the tire is running at high speed, complex vibrations and deformations will be generated, which will cause noise and affect the measurement accuracy of the sensor. In order to accurately describe these noise characteristics, it is necessary to establish a tire vibration noise model. The model can be constructed by analyzing a large amount of measured data and extracting the statistical characteristics and spectrum distribution law of the noise. For example, it may be found that the tire noise has the most concentrated energy in the 100-200Hz frequency band and exhibits the characteristics of Gaussian distribution. Wavelet analysis is a powerful time-frequency analysis tool that can effectively decompose the raw data collected by the sensor. By selecting a suitable wavelet basis function, such as Daubechies wavelet, the signal can be decomposed into wavelet coefficients of different scales. Larger scale coefficients usually correspond to low-frequency signals, while smaller scale coefficients correspond to high-frequency noise. By setting a reasonable threshold, effective signals and noise components can be preliminarily distinguished. In order to further improve the accuracy of noise estimation, it is necessary to adaptively adjust the parameters of the wavelet basis function according to the minimum mean square error criterion. This can be achieved through an iterative optimization algorithm, such as a gradient descent method. By continuously adjusting the scale and translation parameters of the wavelet basis function, it is made to match the statistical characteristics of the noise signal as much as possible, thereby achieving more accurate noise estimation and elimination. Kalman filtering is a recursive estimation algorithm that is very suitable for processing signals in dynamic systems. In tire sensor data processing, the denoised signal can be used as an observation value to establish an appropriate state space model. For example, it can be assumed that the tire deformation parameters follow a first-order Markov process, and then these parameters can be estimated and predicted by the Kalman filter algorithm. This not only smoothes the data, but also predicts the future state of the tire. In order to improve the performance of the Kalman filter, it is necessary to adaptively adjust the filter parameters based on the tire vibration noise model and the wavelet analysis results. For example, if the wavelet analysis shows that the high-frequency components of the signal suddenly increase, the value of the observation noise covariance matrix can be increased accordingly to make the filter more sensitive to the new observations. This adaptive mechanism can make the filter better adapt to changes in the dynamic characteristics of the tire. Finally, the processed sensor data needs to be fused with the tire deformation model. The tire deformation model is usually based on finite element analysis or experimental data, which describes the deformation characteristics of the tire under different load and speed conditions. Through the least squares method, the actual tire deformation parameters, such as lateral deformation or ground pressure distribution, can be estimated. These estimation results can be used to compensate the sensor measurements to improve the overall measurement accuracy. Based on the tire deformation compensation results, the data acquisition strategy can be further optimized. For example, when a severe tire deformation is detected, the sampling frequency can be increased to capture rapidly changing signals; when the tire is in a stable state, the sampling frequency can be appropriately reduced to save computing resources. At the same time, the sensor range can be dynamically adjusted according to the actual degree of deformation to ensure that the measured value is always within the optimal accuracy range. This adaptive acquisition strategy can not only improve measurement accuracy, but also optimize system resource utilization and improve overall performance.

[0049] S103. Aiming at the differences in data formats and protocols of different types of sensors, a universal data parsing module is designed. By configuring the sensor data format template, the automatic recognition and conversion of heterogeneous sensor data is realized, and the data is mapped to a unified data model.

[0050] Specifically, sensor data is obtained, and the data format and protocol type are determined according to the pre-configured data format template. If the match is successful, data parsing is performed to extract key fields; according to the field name and type defined in the unified data model, the extracted key fields are converted and mapped to generate standardized data records; machine learning algorithms are used to extract features from standardized data to obtain data feature vectors, and the feature vectors are classified through clustering algorithms to identify different types of sensor data; for different identified sensor types, corresponding parsing rules are obtained from the pre-established protocol parsing library, the original data is parsed, and complete data fields are extracted; the parsed data fields are matched with the unified data model, and the data is converted into a standardized format through the mapping relationship between the field name and type, and saved in the database; for data that cannot match the data format template, natural language processing technology is used to perform semantic analysis on the data content, extract key information, and generate a standardized data format according to the definition of the unified data model; through data quality detection and verification, the converted standardized data is checked to determine the integrity, consistency and accuracy of the data, ensure that the data meets the requirements of the unified data model, and ensure the availability and reliability of the data.

[0051] In this embodiment, when acquiring sensor data, the communication protocol and data format of the sensor must first be determined. For example, a temperature sensor may output data in ASCII code format through a serial port, while a pressure sensor may output data in binary format through a CAN bus. The pre-configured data format template is similar to a rule base, which contains data format definitions for various common sensors. For example, a template may define the data format of a specific model of acceleration sensor as: "start symbol (1 byte) + timestamp (4 bytes) + X-axis acceleration (2 bytes) + Y-axis acceleration (2 bytes) + Z-axis acceleration (2 bytes) + checksum (1 byte) + end symbol (1 byte)". When a data stream is received, the system will try to match it with a known template. If the structure of the data stream is completely consistent with a template, the match is successful, and the value of each field can be extracted according to the field order and type defined in the template. For example, if a hexadecimal data is received: "AA01020304006400C8012C0D0A", assuming that it matches the data format template of the acceleration sensor mentioned above, the timestamp (01020304), X-axis acceleration (0064, decimal 100), Y-axis acceleration (00C8, decimal 200), Z-axis acceleration (012C, decimal 300) and other information can be extracted. The unified data model defines a set of standardized data field names and types, such as "temperature", "humidity", "pressure", etc., and specifies the data type of each field, such as "temperature" is a floating point number and "humidity" is a percentage. After extracting the key fields from the sensor data, type conversion and mapping need to be performed according to the unified data model. For example, if the name of the temperature field in a sensor data is "TEMP" and the unit is Celsius, while in the unified data model, the name of the temperature field is "temperature" and the unit is Kelvin, then you need to add 273.15 to the value of the "TEMP" field and map the field name to "temperature". Doing so ensures that the data is compatible with different types of sensor data after standardization, which facilitates data analysis and calculation. Machine learning algorithms, such as support vector machines (SVM) or random forests, can be used for feature extraction. For example, for time series data, you can calculate its statistical features such as mean, variance, kurtosis, skewness, and spectrum features after fast Fourier transform. These features can form a feature vector to characterize the characteristics of the original data. Clustering algorithms, such as K-means or DBSCAN, can classify feature vectors. For example, feature vectors of data from different types of sensors can be used as input and automatically divided into several clusters through clustering algorithms, each cluster representing a type of sensor. In this way, different types of sensor data can be automatically identified without manual intervention. For different types of sensors identified, corresponding parsing rules need to be obtained from the protocol parsing library.For example, for CAN bus data, the parsing rules may include the correspondence between the message ID and the data field, as well as the length and parsing method of each field. For example, a message with a CAN message ID of 0x100 may contain engine speed information, where the 3rd and 4th bytes represent the speed in rpm. According to these rules, the complete data field can be extracted from the original data. The parsed data field is matched with the unified data model, and the data is converted to a standardized format through the mapping relationship between the field name and type. For example, the engine speed field "EngineSpeed" parsed from the CAN bus is mapped to the "engine_speed" field in the unified data model, and the unit is converted from rpm to rps. For data that cannot match the data format template, natural language processing technology can be used for semantic analysis. For example, for a piece of text data "Current temperature: 25.6 degrees Celsius, humidity: 60%", the entities "temperature" and "humidity", as well as the corresponding values ​​and units, can be identified through the natural language processing model. Then, according to the definition of the unified data model, this information is converted into a standardized data format. For example, map "temperature" to the "temperature" field, map "humidity" to the "humidity" field, and convert the values ​​and units. Data quality detection and verification are key steps to ensure data reliability. For example, you can check the integrity of the data to determine whether there are missing values; check the consistency of the data to determine whether there are logical contradictions; and check the accuracy of the data to determine whether there are abnormal values. For example, you can set the reasonable range of temperature data to -50 degrees Celsius to 100 degrees Celsius. If a temperature data exceeds this range, it may be an abnormal value and needs further inspection and processing. Through these steps, you can ensure that the data meets the requirements of the unified data model, ensure the availability and reliability of the data, and lay a solid foundation for subsequent data analysis and application.

[0052] S104. According to the system working status and data quality indicators, a multi-model adaptive fusion strategy is adopted to dynamically select different fusion algorithms such as Kalman filtering, Bayesian estimation, neural network, etc., and data fusion performance optimization is achieved through adaptive combination of algorithms.

[0053] Specifically, the system working status and data quality indicators are obtained, and the system status and data quality are judged according to the preset threshold value to see whether they meet the requirements. If they meet the requirements, the next step is entered; if they do not meet the requirements, the system parameters are adjusted, and the status and quality indicators are re-acquired until they meet the requirements. For different system states and data qualities, several fusion algorithms with the highest matching degree are selected from the pre-established algorithm library, including but not limited to Kalman filtering, Bayesian estimation and neural network, and the best algorithm combination is determined according to the algorithm performance and computational complexity. The selected algorithm combination is adaptively optimized, and the algorithm parameters are dynamically adjusted to adapt to the current system state and data quality to obtain the optimal fusion strategy. The optimized multi-model adaptive fusion strategy is used to fuse multi-source heterogeneous data, and the noise suppression, feature extraction and information enhancement of the data are realized through the algorithm combination, so as to improve the accuracy and reliability of data fusion. The quality of the fused data is evaluated, and the key quality indicators are extracted. By comparing with the preset threshold, it is judged whether the fusion result meets the application requirements. If it meets the requirements, the fusion result is output; if not, it returns to step 3 and adjusts the fusion strategy. The fusion results are output to the subsequent application modules, and the fusion algorithms and strategies are dynamically optimized according to the feedback information to achieve continuous improvement and adaptive adjustment of fusion performance. A fusion algorithm performance evaluation mechanism is established to comprehensively evaluate the fusion accuracy, computational efficiency, robustness and other indicators of different algorithm combinations, and the algorithm library is updated according to the evaluation results to provide a basis for subsequent algorithm selection and achieve adaptive optimization of fusion strategies.

[0054] In this embodiment, the current working status information of the system is obtained, for example: the CPU utilization is 75%, the memory occupancy is 80%, the data processing delay is 200 milliseconds, and the data packet loss rate is 0.5%. At the same time, data quality indicators are obtained, for example: data integrity is 99.8%, data accuracy is 99.5%, and data consistency is 99.7%. The thresholds of the system working status and data quality are pre-set, for example: the CPU utilization is less than 90%, the memory occupancy is less than 85%, the data processing delay is less than 300 milliseconds, the data packet loss rate is less than 1%, the data integrity is higher than 99%, the data accuracy is higher than 98%, and the data consistency is higher than 99%. The obtained indicators are compared with the thresholds to determine whether the system status and data quality meet the requirements. If all indicators meet the threshold requirements, the subsequent data fusion steps are performed; if any indicator does not meet the threshold requirements, the system parameters need to be adjusted, for example: increase server resources, optimize data processing procedures, adjust data collection frequency, etc., and re-acquire status and quality indicators, and repeat this process until all indicators meet the requirements. Doing so can ensure that the system is in a good working state before data fusion, and the data quality meets the requirements, thereby ensuring the reliability and effectiveness of data fusion. For different system states, such as high load, low load, unstable network, etc., and different data qualities, such as high noise, missing data, data anomaly, etc., an algorithm library is established in advance. The algorithm library contains a variety of fusion algorithms, such as: Kalman filtering is suitable for processing data with linear relationships and Gaussian noise; Bayesian estimation is suitable for processing data with uncertainty and prior knowledge; neural networks are suitable for processing data with complex nonlinear relationships. According to the current system state and data quality indicators, select several algorithms with the highest matching degree from the algorithm library. For example: the current system is in a high-load state and the data noise is high. Kalman filtering and neural network algorithms can be selected because Kalman filtering can effectively suppress Gaussian noise, while neural networks can process complex data relationships under high load. Determining the best algorithm combination requires considering algorithm performance and computational complexity. For example: on the premise of ensuring fusion accuracy, give priority to algorithms with lower computational complexity to reduce the system burden. After selecting the two algorithms of Kalman filtering and neural network, the algorithm parameters need to be adaptively optimized. For example, the initial state covariance matrix and process noise covariance matrix of the Kalman filter, as well as the network structure and learning rate of the neural network, all need to be dynamically adjusted according to the current system state and data quality. In this way, the algorithm can better adapt to the current data characteristics and system environment, thereby obtaining the optimal fusion strategy. Doing so can improve the accuracy and efficiency of data fusion and make it better adaptable to actual application scenarios. The optimized Kalman filter and neural network combination strategy is used to fuse multi-source heterogeneous data.For example, for the three sensor data of temperature, humidity and pressure, the Kalman filter is first used to preprocess each sensor data, suppress noise and smooth the data; then the preprocessed data is input into the neural network model, and the nonlinear mapping ability of the neural network is used to extract data features and enhance information; finally, the fusion result output by the neural network is weighted averaged with the output of the Kalman filter to obtain the final fusion data. Through this algorithm combination, the advantages of different algorithms can be fully utilized to improve the accuracy and reliability of data fusion. The quality of the fused data is evaluated and key quality indicators are extracted, such as root mean square error, mean absolute error, correlation coefficient, etc. These indicators are compared with the preset thresholds to determine whether the fusion result meets the application requirements. For example, the preset root mean square error threshold is 0.1. If the root mean square error of the fusion result is 0.08, it meets the requirements and the fusion result is output; if the root mean square error of the fusion result is 0.12, it does not meet the requirements and needs to return to step 3, adjust the fusion strategy, and re-perform data fusion and quality evaluation. The fusion result is output to subsequent application modules, such as environmental monitoring system, equipment fault diagnosis system, etc. According to the feedback information of the application module, such as prediction accuracy, alarm accuracy, etc., the fusion algorithm and strategy are dynamically optimized. For example, if the prediction accuracy of the application module feedback is low, you can consider adjusting the network structure of the neural network or increasing the amount of training data; if the alarm accuracy of the feedback is low, you can consider adjusting the parameters of the Kalman filter or introducing other types of fusion algorithms. In this way, the continuous improvement and adaptive adjustment of the fusion performance can be achieved. Establish a fusion algorithm performance evaluation mechanism to comprehensively evaluate the fusion accuracy, computational efficiency, robustness and other indicators of different algorithm combinations. For example, test multiple algorithm combinations such as Kalman filter and neural network combination, Bayesian estimation and neural network combination, and record their fusion accuracy, computational time, sensitivity to abnormal data and other indicators on different data sets. Update the algorithm library according to the evaluation results, for example, remove the algorithm combination with poor performance from the algorithm library, or add the new algorithm combination with excellent performance to the algorithm library. Provide a basis for subsequent algorithm selection and realize adaptive optimization of fusion strategy.

[0055] S105. During the data fusion process, a data quality assessment mechanism is introduced to dynamically adjust the weights and confidence levels of different sensor data through real-time assessment of indicators such as data integrity, accuracy, and timeliness, and to eliminate or downgrade data of poor quality to improve the reliability of the fusion results.

[0056] Specifically, for the standardized data set, the completeness, accuracy and timeliness indicators of each data are calculated to obtain the data quality assessment results. According to the preset data quality threshold, it is judged whether the quality of each data meets the standard. If the data quality is lower than the threshold, it is marked as poor quality data. For poor quality data, data elimination or weight reduction processing methods are adopted. If the data is seriously missing or the error rate is high, the data is directly eliminated; if the data quality is slightly lower than the threshold, its weight and confidence are reduced. According to the data quality assessment results, the weights and confidences of different sensor data are dynamically adjusted. Higher quality data are given higher weights and confidences, and poor quality data are given lower weights and confidences. The sensor data after quality assessment and weight adjustment are fused, and the weighted average, Kalman filtering and other algorithms are used to obtain the fused result data. The reliability of the fusion result is evaluated and the confidence of the fused data is calculated. If the confidence is higher than the preset threshold, the fusion result is considered reliable; if the confidence is lower than the threshold, the data weight and fusion algorithm are readjusted until a reliable fusion result is obtained.

[0057] In this embodiment, for camera data, its quality can be evaluated by indicators such as image clarity and contrast; for radar data, it can be evaluated by indicators such as signal-to-noise ratio and measurement accuracy; for GPS data, its accuracy can be judged by indicators such as the number of satellites and geometric precision factor. Assuming that the image clarity threshold preset by the system is 0.8, if the clarity of a frame of image is 0.75, it will be marked as poor quality data. For data of poor quality, the system will adopt corresponding processing strategies. If a GPS receiver continues to output abnormal coordinates, it may be due to equipment failure or signal blocking, and the system will directly remove the data of the device. For data that slightly deviates from the threshold, such as radar ranging error slightly greater than expected, its weight in the fusion process can be reduced instead of being completely discarded. Dynamically adjusting data weights is an effective way to improve fusion accuracy. For example, camera data may be more reliable on sunny days and given a higher weight; while in rainy and foggy weather, the reliability of radar data may be higher, and the weight of radar data should be increased at this time. This dynamic adjustment can adapt to different environmental conditions and ensure the stability of the fusion results. The choice of data fusion algorithm depends on the specific application scenario and data characteristics. In the vehicle positioning task, the Kalman filter algorithm can be used to fuse the data of GPS and inertial navigation system, and the prediction-correction mechanism of Kalman filter can be used to effectively suppress the errors of each and improve the positioning accuracy. In the task of vehicle flow statistics, it may be more appropriate to use the weighted average method, which comprehensively considers the reliability of camera recognition results and ground sensor coil counts. The reliability evaluation of fusion results is an important means of system self-optimization. For example, by comparing the fused vehicle position with the actual road network data, the matching degree is calculated as a confidence index. If the road network matching degree of a fusion result is less than 90%, the system will consider the result to be unreliable, and then adjust the weight of each sensor data or try other fusion algorithms until a satisfactory result is obtained. Through this iterative optimization method, the multi-sensor fusion system can continuously improve the quality and reliability of data processing and provide more accurate information support for intelligent traffic decision-making. This method is not only applicable to the transportation field, but can also be extended to multiple application scenarios that require multi-source data fusion, such as environmental monitoring and industrial control, reflecting the universal importance of data quality management and adaptive fusion strategies.

[0058] S106. Establish a fault diagnosis and fault tolerance mechanism for the tire pressure monitoring system. By real-time monitoring the working status of each module of the system, start the backup sensor or switch the fusion algorithm when the sensor fails or the data is abnormal, so as to ensure the continuous and reliable operation of the system and improve the fault tolerance and robustness of the system.

[0059] Specifically, according to the preset tire pressure threshold range, the tire pressure data collected by each sensor of the tire pressure monitoring system is obtained in real time to determine whether the tire pressure data is within the normal range. If the tire pressure data exceeds the normal range, the fault diagnosis process is triggered to analyze the sensor working status and data transmission link to determine the fault type and location. If it is determined to be a sensor failure, the backup sensor of the corresponding tire is switched to obtain the tire pressure data of the backup sensor to ensure the continuous operation of the monitoring system. If the data is determined to be abnormal, the preset data correction algorithm is called to correct or filter the abnormal data to obtain the corrected tire pressure data. According to the historical tire pressure data and the current road conditions, the normal range threshold of the tire pressure is dynamically adjusted to improve the adaptive ability and robustness of the monitoring system. The Kalman filter algorithm is used to fuse the tire pressure data of multiple sensors to obtain a more accurate and stable tire pressure estimate, reducing the impact of a single sensor failure on the system. Through fault diagnosis and fault tolerance mechanisms, the high reliability and continuity of the tire pressure monitoring system are achieved to ensure driving safety.

[0060] In this embodiment, the tire pressure monitoring system is an important part of ensuring driving safety. The system obtains the tire pressure data of each tire sensor in real time and compares it with the preset normal range. For example, the normal tire pressure range of a certain model is 180-220kPa, and the system collects tire pressure data once a second. When abnormal tire pressure is detected, such as the pressure of the front left tire drops to 160kPa, the system will immediately trigger the fault diagnosis process. Fault diagnosis first analyzes the working status of the sensor. Assuming that the front left wheel sensor continues to output abnormally low pressure values, while other tire data are normal, the system may determine that the sensor is faulty. At this time, the system will automatically switch to the backup sensor of the tire. The backup sensor may use different measurement principles, such as direct measurement instead of indirect measurement, to improve system reliability. If it is judged that the data is abnormal rather than a sensor failure, the system will call the preset correction algorithm. For example, if the data suddenly jumps to 300kPa, which is obviously inconsistent with the actual situation, the system may use the moving average method to smooth the data, or interpolate and correct it according to the trend of historical data. An important feature of the system is that it can dynamically adjust the normal range of tire pressure according to the actual situation. For example, when driving at high speed, the increase in tire temperature will cause the tire pressure to increase naturally. At this time, the system may increase the upper limit of the normal range to 240kPa. For example, when driving on a rugged mountain road, in order to increase the contact area between the tire and the ground, the system may reduce the lower limit of the normal range to 160kPa. This adaptive ability greatly improves the practicality and reliability of the system. In order to obtain a more accurate tire pressure estimate, the system uses the Kalman filter algorithm to fuse multiple sensor data. Assuming that three sensors measure the tire pressure values ​​of 205kPa, 208kPa and 203kPa at a certain moment, considering the measurement errors and historical data of each sensor, the Kalman filter may obtain the optimal estimate of 206kPa. This method not only improves the measurement accuracy, but also enhances the system's anti-interference ability. Through the above mechanism, the tire pressure monitoring system can maintain high reliability and continuity in various complex situations. Even if a single sensor fails, the system can still maintain normal operation through backup sensors and data fusion technology. This multiple guarantee ensures real-time monitoring of tire status during driving, greatly reduces the probability of dangerous situations such as tire blowouts, and provides a strong guarantee for driving safety.

[0061] Furthermore, a Kalman filter algorithm is used to fuse tire pressure data from multiple sensors to obtain a more accurate and stable tire pressure estimate, reducing the impact of a single sensor failure on the system.

[0062] Furthermore, the tire pressure data collected by multiple tire pressure sensors in real time are obtained, and the tire pressure data are preprocessed to remove obviously abnormal data points, so as to obtain a set of tire pressure data after preliminary processing. According to the principle of the Kalman filter algorithm, the state equation and the observation equation are established. The state equation describes the dynamic change process of the tire pressure, and the observation equation describes the relationship between the sensor measurement value and the actual tire pressure. The initial parameters of the Kalman filter algorithm are determined, including the initial state estimate, the state covariance matrix, the process noise covariance matrix, and the measurement noise covariance matrix. For the tire pressure data at each moment, the prediction and update steps of the Kalman filter algorithm are recursively executed. The prediction step predicts the state estimate and state covariance matrix at the current moment based on the state estimate and state equation at the previous moment; the update step updates the state estimate and state covariance matrix based on the observation value and observation equation at the current moment. Through the recursive calculation of the Kalman filter algorithm, the fused tire pressure estimate is obtained. The estimate comprehensively considers the measurement information of multiple sensors and has higher accuracy and stability. Determine whether the fused tire pressure estimate is within the normal range. If it is beyond the normal range, the tire pressure abnormality alarm is triggered to prompt the driver to deal with it in time to ensure driving safety. Continuously monitor the change trend of the fused tire pressure estimate, and use the trend analysis algorithm to determine whether there are abnormal conditions such as slow leakage in the tire pressure. If an abnormal trend is detected, it will warn of potential tire pressure problems and prompt the driver to check and maintain regularly.

[0063] Furthermore, the tire pressure monitoring system collects tire pressure data in real time through multiple sensors to ensure driving safety. In the data preprocessing stage, the system will remove obviously abnormal data points, such as sudden zero values ​​or values ​​beyond the physically possible range. For example, if a sensor reports that the tire pressure drops from 35psi to 0psi in a short period of time, this may be a sensor failure rather than a real tire pressure change, and the system will remove it. The Kalman filter algorithm is a recursive estimation method used to fuse multiple sensor data. In tire pressure monitoring, the state equation can describe the change of tire pressure over time, such as a linear model considering the influence of temperature. The observation equation reflects the relationship between the sensor measurement value and the actual tire pressure, including measurement errors. The initial parameter setting is crucial. For example, the initial state estimation can use the first valid measurement value, and the state covariance matrix reflects the uncertainty of the initial estimate. During the execution of the algorithm, the prediction step uses the state at the previous moment and the dynamic model to predict the current state. For example, if the tire pressure was estimated to be 32psi at the previous moment, the prediction model may predict the current tire pressure to be 32.5psi considering that the temperature rise causes a slight increase in tire pressure. The update step corrects the prediction result according to the actual observation value. Assuming that the three sensors measure 32.8psi, 32.6psi, and 32.7psi respectively, the algorithm will combine this information and may come up with a final estimate of 32.7psi. The fused tire pressure estimate has higher accuracy and stability. For example, a single sensor may be disturbed by vibration and produce fluctuations, but the fusion algorithm can smooth these fluctuations. If a sensor fails completely, the data from other sensors can still ensure the normal operation of the system. The system compares the fusion result with the preset threshold. For example, the normal range is 28-35psi. If the estimated value is less than 28psi, the low pressure alarm is triggered. Trend analysis is crucial to predicting potential problems. The system may find that the tire pressure drops by 0.1psi every day. Although it is still within the normal range, this continuous downward trend may indicate that the tire is slowly deflated. With early warning, the driver can check and repair in time before the problem worsens, avoiding more serious safety hazards. This comprehensive approach greatly improves the reliability and predictive ability of the tire pressure monitoring system. By fusing multi-source data, eliminating outliers, and smoothing fluctuations, the system can more accurately reflect the actual tire pressure status. At the same time, the trend analysis function enables the system to shift from passive alarm to active warning, providing the driver with more comprehensive tire health information, effectively reducing safety risks such as tire blowouts, and improving overall driving safety.

[0064] This embodiment also provides a vehicle tire pressure monitoring system based on multi-sensor fusion, including:

[0065] A sensor performance degradation model building module, used to obtain working environment data of the sensor module and build a sensor performance degradation model based on the working environment data;

[0066] A data preprocessing module, used to preprocess the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data;

[0067] A general data analysis module, used to extract data features of the preprocessed data, standardize the data features and map them to a unified data model;

[0068] The multi-model adaptive fusion module is used to dynamically select different data fusion algorithms according to the working status and data quality indicators, and fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data;

[0069] The fault diagnosis module is used to analyze the fused data according to a preset tire pressure threshold range to obtain a sensor status.

[0070] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0071] This embodiment also provides a computer-readable storage medium on which a computer program is stored, characterized in that the steps of the method are implemented when the computer program is executed by a processor.

[0072] This embodiment also provides a computer program product, including a computer program, characterized in that the steps of the method are implemented when the computer program is executed by a processor.

[0073] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for monitoring automobile tire pressure based on multi-sensor fusion, characterized in that: The following steps are involved: Acquire working environment data of the sensor module, and construct a sensor performance degradation model based on the working environment data; Preprocessing the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data; Extracting data features of the preprocessed data, standardizing the data features and mapping them to a unified data model; Dynamically select different data fusion algorithms according to the working status and data quality indicators, fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data; The fused data is analyzed according to a preset tire pressure threshold range to obtain a sensor state.

2. The method according to claim 1, characterized in that Building a sensor performance degradation model based on the working environment data includes: The temperature, pressure and vibration parameters of the sensor under different working environments are obtained as independent variables, and the output accuracy of the sensor under the corresponding environment is obtained as the dependent variable. A multiple linear regression algorithm is used to combine the independent variables and the dependent variables to establish a sensor performance degradation model.

3. The method according to claim 1, characterized in that Obtaining preprocessed data includes: The sensor performance degradation model is used to identify and compensate for changes in sensor performance. A tire vibration noise model is established based on the vibration and deformation characteristics of the tire when driving at high speed. The original data collected by the sensor is decomposed in the time-frequency domain using the wavelet analysis method to extract signal characteristics at different scales. The scale and translation parameters of the wavelet basis function are adaptively adjusted according to the minimum mean square error criterion. The signal characteristics at different scales are recursively estimated and predicted using the Kalman filter algorithm to dynamically track the changing trend of the signal and improve the smoothness and stability of the data. According to the tire vibration noise model and the wavelet analysis results, the state transfer matrix and observation matrix of the Kalman filter are adaptively adjusted.

4. The method according to claim 1, characterized in that: Extracting data features of the preprocessed data, standardizing the data features and mapping them to a unified data model includes: According to the pre-configured data format template, the data format and protocol type are judged. If the match is successful, data parsing is performed to extract key fields as data features. According to the field names and types defined in the unified data model, the extracted data features are converted and mapped to generate standardized data records.

5. The method according to claim 4, characterized in that The key fields to be extracted include: A machine learning algorithm is used to extract features from the preprocessed data to obtain data feature vectors. The feature vectors are classified through a clustering algorithm to identify different types of sensor data. For the different identified sensor types, the corresponding parsing rules are obtained from the pre-established protocol parsing library, the raw data is parsed, and complete data fields are extracted.

6. The method according to claim 1, characterized in that The method also includes: in the data fusion process, introducing a data quality assessment mechanism, through real-time assessment of data integrity, accuracy and timeliness indicators, dynamically adjusting the weights and confidence levels of different sensor data, and eliminating or downgrading low-quality data.

7. An automobile tire pressure monitoring system based on multi-sensor fusion, characterized in that: include: A sensor performance degradation model building module, used to obtain working environment data of the sensor module and build a sensor performance degradation model based on the working environment data; A data preprocessing module, used to preprocess the raw data collected by the sensor according to the sensor performance degradation model to obtain preprocessed data; A general data analysis module, used to extract data features of the preprocessed data, standardize the data features and map them to a unified data model; The multi-model adaptive fusion module is used to dynamically select different data fusion algorithms according to the working status and data quality indicators, and fuse the data in the unified data model through the selected data fusion algorithm to obtain fused data; The fault diagnosis module is used to analyze the fused data according to a preset tire pressure threshold range to obtain a sensor status.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images