A vehicle tire pressure monitoring platform and method based on the Internet of Things

Abstract

The present invention belongs to the field of vehicle-mounted tire pressure monitoring, and discloses a vehicle-mounted tire pressure monitoring platform and method based on the Internet of Things; the platform and method include: collecting comprehensive status data of each tire, and constructing a joint structure diagram of the tires; calculating the difference data between the joint structure diagram at a current time point and a previous time point to form a difference package, and transmitting the difference package using an improved difference technology; building and training a monitoring and early warning model, importing the difference package into the trained monitoring and early warning model, and obtaining early warning information; displaying the early warning information through an on-board interface, and adjusting the vehicle, so as to realize intelligent, efficient and high-quality monitoring of vehicle-mounted tire pressure based on the Internet of Things.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-mounted tire pressure monitoring, and more specifically, to a vehicle-mounted tire pressure monitoring platform and method based on the Internet of Things. Background Art

[0002] For cars, tire pressure is an important factor in the use of the car. Too high or too low tire pressure will cause uneven force on the tire and increase the risk of tire blowout. Especially when driving at high speeds, abnormal tire pressure is one of the main causes of tire blowout. Incorrect tire pressure will affect the contact area between the tire and the ground, resulting in longer braking distance, reduced steering ability, and even dangerous situations such as skidding.

[0003] In the process of monitoring tire pressure on vehicles based on the Internet of Things (IoT), although the technology provides convenience for driving safety and vehicle management, it also faces some problems and challenges. The main problem is that a car has more than one wheel, and as a whole, each wheel cannot work independently. As the road conditions or the working state of the car change, the wheels will affect each other. However, due to the complexity of the environment and the dynamic changes in the degree of mutual influence between the wheels, the specific impact is difficult to analyze, which increases the difficulty of tire pressure judgment. In addition, the tire pressure monitoring system usually uses low-power wireless technology (such as RF, BLE, LoRa, etc.) to transmit data to the vehicle controller or cloud platform. However, in some complex environments (such as tunnels, mountainous areas or urban high-rise areas), there may be delays or packet loss during data transmission, and the signal may be interfered or lost. Especially in remote areas, data upload may be limited due to network coverage, resulting in reduced real-time performance and affecting the monitoring effect.

[0004] In view of this, the present invention proposes a vehicle-mounted tire pressure monitoring method based on the Internet of Things to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a vehicle-mounted tire pressure monitoring method based on the Internet of Things, comprising: step S1: collecting comprehensive status data of each tire and constructing a joint structure diagram of the tire;

[0006] Step S2: Calculate the difference data between the joint structure graph at the current time point and the previous time point to form a difference packet, and use the improved difference technology to transmit the difference packet;

[0007] Step S3: construct and train a monitoring and early warning model, import the difference package into the trained monitoring and early warning model, and obtain early warning information;

[0008] Step S4: Display the warning information through the vehicle interface and adjust the vehicle.

[0009] Preferably, the comprehensive status data includes real-time tire pressure data, temperature data, acceleration data, vehicle status data and basic vehicle information;

[0010] The method for obtaining real-time tire pressure data and temperature data is to divide the tire into S_Y areas, and the area of ​​the area , each area outputs a real-time tire pressure data and temperature data;

[0011] in, is the acquisition frequency, is the surface area of ​​the tire;

[0012] Adaptive frequency Collect acceleration data, and ,in, Indicates the acquisition frequency at the previous time point, Indicates the sampling frequency at the previous time point. Indicates the collection frequency at the current time point;

[0013] Vehicle status data includes driving speed, driving direction, steering angle and braking status;

[0014] The basic vehicle information includes the vehicle's appearance data and the number, size and material of its tires.

[0015] Preferably, the method for constructing a joint structure graph comprises:

[0016] For each tire, the comprehensive status data is fused to form a complete tire portrait;

[0017] Define the correlation between each tire, including pressure correlation, temperature correlation, acceleration correlation and stress state correlation, and extract the correlation between each tire based on the tire image;

[0018] Represent the extracted associations as an undirected graph ,in, is a set of nodes, each node is used to represent a tire, It is a set of directed edges. The weight of each edge is represented by the value of the association relationship between tires. The direction of the directed edge is from the core entity to the associated entity.

[0019] Through undirected graph Potential new associations are mined and added to the original undirected graph, and the resulting complete undirected graph is recorded as a joint structure graph.

[0020] Preferably, the method of fusing the comprehensive status data to form a complete tire image includes:

[0021] Define the core entities, related entities and associated entities of tire portraits. The core entities represent each tire, and the related entities represent their own characteristics, associations and vehicle status. The own characteristics include the real-time tire pressure data, temperature data and acceleration data corresponding to each core entity. The association is used to describe the relationship between any core entity and the associated entity. The associated entity represents other core entities except the selected core entity. The vehicle status includes vehicle status data and basic vehicle information.

[0022] For each core entity, the data of related entities are extracted from the comprehensive status data, the related entities are labeled and associated with the core entity, and then fused to form a complete tire portrait.

[0023] Preferably, the undirected graph Mining potential new relationships. Specific methods include:

[0024] For each node in an undirected graph, calculate its critical centrality ,in, , and Respectively represent The degree centrality, weighted degree centrality and structural similarity centrality of each node, , and represents the weight coefficient, Indicates the index of the tire;

[0025] ,in, Indicates The degree of the node, Indicates the total number of nodes;

[0026] ,in, Indicates except The set of all nodes outside the node, Indicates The node and The weights between nodes, represents the maximum weight of all directed edges;

[0027] ,in, Indicates The node and The similarity score between nodes, and , , , and represents the adjustment factor, Indicates The node and The attribute similarity between nodes, Indicates The node and The frequency of interaction between nodes, represents the time decay factor, Indicates The node and The time interval between the last interaction of the nodes. Indicates The node and The distance between nodes;

[0028] Extract the key centrality of all nodes and sort them in descending order, select the first e_u nodes as the core node set, and the remaining nodes are recorded as non-core nodes;

[0029] For each non-core node, calculate the path distance between it and all nodes in the core node set, and take the minimum value of the path distance as the benchmark distance from the non-core node to the core node set;

[0030] Set J_L levels, each level defines a partition range, and the partition range is continuous. Non-core nodes are allocated to each level according to the benchmark distance until all non-core nodes are allocated.

[0031] For nodes at the same level, by calculating the similarity scores between each data category, if the similarity score is greater than the judgment threshold, it means that there is an association relationship and a directed edge is arranged; for nodes across levels, the number of interaction behaviors of each data category in the time period a_u is calculated. If the number of interaction behaviors is greater than the preset number, it means that there is an association relationship and a directed edge is arranged.

[0032] Preferably, the method for arranging directed edges comprises:

[0033] Select a machine learning model, set the input of the machine learning model as a sample set, set the output as a tire performance index, train the machine learning model using historical data, and obtain a trained machine learning model;

[0034] Set a standard data set. For data categories whose similarity scores are greater than the judgment threshold or whose number of interactive behaviors is greater than the preset number, randomly traverse and select two data categories. Replace the corresponding data category values ​​in the standard data set with one and two data categories respectively to form a new data set, which is recorded as a disturbed data set. Then, three disturbed data sets are obtained.

[0035] Import the perturbed data set into the trained machine learning model to obtain the performance index, which is recorded as the permutation performance value;

[0036] Calculate the weight of a directed edge between two data categories ,in, Indicates Data categories and The weights between data categories, Indicates The permutation performance value of the perturbed data set output corresponding to each data category, Indicates The permutation performance value of the perturbed data set output corresponding to each data category, Indicates and The permutation performance values ​​of the perturbed data set outputs corresponding to each data category.

[0037] Preferably, the improved differential technology is used to transmit the difference packet, and the specific method includes:

[0038] Copy and extract the nodes in the joint structure graph to form a new node set. For each directed edge, calculate the weight difference between the current time point and the previous time point to form a difference edge set. , then the difference package ;

[0039] Set P_O continuous value ranges and arrange them in descending order, divide the weight differences into corresponding value ranges, record the weight differences in the same value range as difference blocks, and collect P_O difference blocks;

[0040] Set the priority of P_O difference blocks according to the order of the value range, arrange the priorities in descending order, add a checksum to each difference packet, and compress the difference block using a lightweight compression algorithm;

[0041] Transmit P_O difference blocks of the difference packet according to the priority;

[0042] The difference blocks received after transmission are merged piece by piece to form a complete difference packet.

[0043] Preferably, the method of constructing and training a monitoring and early warning model, importing the difference package into the trained monitoring and early warning model, and obtaining early warning information includes:

[0044] The framework of the monitoring and early warning model is based on neural network learning and LSTM technology, including input layer, hidden layer and output layer. The input of the input layer is set as multiple sets of data sets, including difference packages and basic vehicle information. The output layer sets three output heads in parallel, namely anomaly detection head, anomaly classification head and adjustment suggestion head, and the output is early warning information, which includes whether there is an abnormality in the tire, the cause of the abnormality and the abnormal adjustment suggestion.

[0045] The anomaly detection head is used to detect the tire pressure status of each tire. The activation function is Sigmoid, and the loss function is set to binary cross entropy. The anomaly classification head is used to detect the abnormal cause category of each tire. The abnormal cause category is defined as excessive tire pressure, low tire pressure, air leakage, sensor failure or other abnormalities. The activation function is Softmax, and the loss function is set to multi-class cross entropy. The adjustment suggestion head is used to generate abnormal adjustment suggestions for each tire. The task types include regression tasks and classification tasks. The activation functions are linear activation functions and ReLU, and the loss function is the accumulation of mean square error and multi-class cross entropy.

[0046] Set the total loss function of the model to the sum of the loss functions of the three output heads, and the optimization goal is to minimize the total loss function, using Adam or SGD optimizer;

[0047] The monitoring and early warning model is trained using the training set until the optimization target is reached, the trained monitoring and early warning model is obtained, and multiple sets of data sets at the current time point are input into the trained monitoring and early warning model to obtain early warning information.

[0048] Preferably, the training set includes T_P groups of historical samples, each group of samples includes a difference package of the vehicle and basic information of the vehicle and actual corresponding warning information;

[0049] Extract the timestamp of each sample, sort them according to the time series, divide them into batches, and import them into the monitoring and early warning model for training in sequence.

[0050] An on-vehicle tire pressure monitoring platform based on the Internet of Things, comprising:

[0051] Data collection and integration module: used to collect comprehensive status data of tires and build a joint structure diagram list;

[0052] Efficient transmission module: used to calculate differential data and transmit differential packets through improved differential technology;

[0053] Model prediction module: used to build and train monitoring and early warning models, import difference packages into the trained monitoring and early warning models, and obtain early warning information;

[0054] Visualization module: used to arrange warning information on the vehicle interface and interact with the user.

[0055] The technical effects and advantages of the vehicle tire pressure monitoring method based on the Internet of Things of the present invention are as follows:

[0056] 1. Joint monitoring of each tire strengthens the focus on the possible correlation between tires. By building a joint tire structure diagram, the correlation between tires is considered, which improves the accuracy of monitoring and the reliability of early warning. Better understand the overall status of the tire and reduce the inaccuracy and one-sidedness caused by isolated analysis of each tire. The adaptive acquisition frequency of acceleration data balances data accuracy and transmission efficiency. Multi-dimensional data collection and correlation analysis between tires improve the accuracy and reliability of tire pressure monitoring.

[0057] 2. Improved differential technology and lightweight compression algorithm reduce bandwidth occupancy. By only transmitting the changed part of the data, the data transmission volume is effectively reduced, the data transmission efficiency is significantly improved, and the network bandwidth consumption and communication costs are reduced.

[0058] 3. The monitoring and early warning model based on neural network learning and LSTM technology can improve the intelligence and autonomy of the monitoring system. It can not only detect anomalies, but also classify the causes of anomalies and provide corresponding adjustment suggestions, providing users with more comprehensive information. So that preventive measures can be taken in advance. By predicting potential risks, predictive maintenance can be achieved and maintenance costs can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a structural schematic diagram of an on-vehicle tire pressure monitoring platform based on the Internet of Things of the present invention;

[0060] Figure 2 The present invention is a schematic diagram of the steps of a vehicle tire pressure monitoring method based on the Internet of Things. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 and Figure 2 As shown, this embodiment provides a vehicle tire pressure monitoring method based on the Internet of Things, including:

[0064] The purpose of on-board tire pressure monitoring (TPMS, tire pressure monitoring) is to improve driving safety, extend tire life and optimize vehicle performance.

[0065] For cars, tire pressure is an important factor in the use of the car. Too high or too low tire pressure will cause uneven force on the tire and increase the risk of tire blowout. Especially when driving at high speeds, abnormal tire pressure is one of the main causes of tire blowout. Incorrect tire pressure will affect the contact area between the tire and the ground, resulting in longer braking distance, reduced steering ability, and even dangerous situations such as skidding.

[0066] To elaborate, too low tire pressure will cause excessive stress on the tire sidewall, increase wear, and increase the contact area between the tire and the ground, increase rolling resistance, and increase fuel consumption. It is estimated that for every 10% decrease in tire pressure, fuel consumption may increase by about 1%-2%. Low tire pressure will also cause the tire's shock absorption function to decrease, increase the burden on the suspension system, and accelerate its aging and damage; too high tire pressure will cause premature wear in the central part of the tread, and maintaining normal tire pressure can reduce the extra burden on the tire during driving and reduce the probability of damage caused by abnormal tire pressure. Abnormal tire pressure will increase the bumpy feeling when the vehicle is driving, affecting driving comfort.

[0067] In the process of monitoring tire pressure on board based on the Internet of Things (IoT), although the technology provides convenience for driving safety and vehicle management, it also faces some problems and challenges. The main one is that a car has more than one wheel, and as a whole, each wheel of the car cannot work independently. As the road conditions or the working state of the car change, the wheels will affect each other, but based on the complexity of the environment and the dynamic changes in the degree of mutual influence between the wheels, the specific impact is difficult to analyze, which increases the difficulty of tire pressure judgment; on-board tire pressure monitoring relies on sensors for data collection, but the sensors may cause inaccurate data or failure due to the external environment (such as high temperature, humidity, vibration), that is, they may be damaged by external forces (such as collisions, vibrations caused by bad road conditions), or affected by electromagnetic interference (such as other wireless devices, strong magnetic fields), resulting in abnormal data collection, and inaccurate or abnormal data makes it difficult to provide data support for analysis and judgment during the monitoring process.

[0068] In addition, tire pressure monitoring systems usually use low-power wireless technologies (such as RF, BLE, LoRa, etc.) to transmit data to the vehicle controller or cloud platform. However, in some complex environments (such as tunnels, mountainous areas or urban high-rise areas), there may be delays or packet loss during data transmission, and the signal may be interfered or lost. Especially in remote areas, data upload may be limited by network coverage, resulting in reduced real-time performance and affecting monitoring results.

[0069] Therefore, the tire pressure monitoring platform on the vehicle can detect tire pressure changes in real time, promptly remind the driver to correct abnormal tire pressure, avoid accidents, optimize fuel efficiency, thereby reducing vehicle exhaust emissions and achieving environmental protection effects. In general, tire pressure monitoring is an important means to ensure driving safety, reduce vehicle maintenance costs and improve driving experience. By monitoring tire pressure in real time, car owners can find problems in time and take corresponding measures to prevent accidents and extend the service life of tires.

[0070] Step S1: Collecting comprehensive status data of each tire and constructing a joint structure diagram of the tire;

[0071] Data collection and integration module: used to collect comprehensive status data of tires and build a joint structure diagram list;

[0072] Comprehensive status data includes real-time tire pressure data, temperature data, acceleration data, vehicle status data and basic vehicle information;

[0073] The method for obtaining real-time tire pressure data and temperature data is to divide the tire into S_Y areas, and the area of ​​the area , each area outputs a real-time tire pressure data and temperature data;

[0074] in, is the acquisition frequency, The surface area of ​​the tire is specifically the product of the width and circumference of the tire. In each area, a pressure sensor (such as a MEMS pressure sensor) and a temperature sensor (such as a thermistor or a thermocouple) are installed inside or outside the tire to directly measure the tire pressure and temperature.

[0075] Adaptive frequency Collect acceleration data, and ,in, Indicates the acquisition frequency at the previous time point, Indicates the sampling frequency at the previous time point. Indicates the collection frequency at the current time point;

[0076] The accelerometer can be mounted on the vehicle's suspension system, as close to the tire as possible.

[0077] Vehicle status data includes driving speed, driving direction, steering angle and braking status; the driving speed is collected by the wheel speed sensor installed on the wheel bearing or brake disc; the driving direction is usually directly collected and processed by the on-board ECU; the braking status can be obtained by the brake pedal sensor or brake pressure sensor; the steering angle can be obtained by the steering angle sensor.

[0078] The basic vehicle information includes the vehicle's appearance data and the number, size and material of its tires;

[0079] The basic information of the vehicle is obtained from the manufacturer's data or production and processing data. The appearance data includes the length, width and height of the vehicle, the horizontal and vertical axis spacing of the wheels, and the size of the wheel hub.

[0080] Methods for constructing joint structure graphs include:

[0081] For each tire, the comprehensive status data is fused to form a complete tire portrait;

[0082] Define the correlation between each tire, including pressure correlation, temperature correlation, acceleration correlation and stress state correlation, and extract the correlation between each tire based on the tire image;

[0083] Pressure correlation mainly refers to tire pressure difference, that is, the difference in tire pressure with other tires, which is used to determine whether there is an abnormal tire pressure distribution (such as the overall low tire pressure on the right side). Temperature correlation refers to the temperature difference between coaxial tires or diagonal tires (such as caused by sharp turns or long-term high speed). Acceleration correlation refers to the difference in lateral or longitudinal acceleration (the stress state of different tires when turning or braking). Stress state correlation refers to the difference in stress of each tire, which can be obtained by calculating the tire load through steering angle, braking state and driving speed.

[0084] Represent the extracted associations as an undirected graph ,in, is a set of nodes, each node is used to represent a tire (assuming there are four tires, four nodes will be formed), It is a set of directed edges. The weight of each edge is represented by the numerical value of the association between tires (such as pressure association, temperature association, acceleration association, and force state association). The edge weight can represent the strength of the association (for example, the greater the tire pressure difference, the higher the edge weight). The direction of the directed edge is from the core entity to the associated entity.

[0085] Through undirected graph Potential new associations are mined and added to the original undirected graph, and the resulting complete undirected graph is recorded as a joint structure graph.

[0086] Through new associations, it is possible to analyze and infer whether the abnormality of a certain tire will have an impact on other tires. For example, a tire with abnormal tire pressure may cause an increase in the load on other tires, thereby affecting their temperature or tire pressure changes. When the vehicle turns, accelerates or brakes, the state changes of different tires may show certain regularities, which can be further modeled by mining new associations. For example, the increase in temperature and pressure of the front outer tire when turning may be associated with the acceleration change of the rear inner tire. Coaxial tires (such as front wheel FL and front wheel FR) or diagonal tires (such as front wheel FL and rear wheel RR) may show potential symmetry or asymmetry. Mining new associations can help discover whether these symmetries are broken and why.

[0087] Methods for fusing comprehensive status data to form a complete tire image include:

[0088] Define the core entities, related entities and associated entities of tire portraits. The core entities represent each tire, and the related entities represent their own characteristics, associations and vehicle status. The own characteristics include the real-time tire pressure data, temperature data and acceleration data corresponding to each core entity. The association is used to describe the relationship between any core entity and the associated entity. The associated entity represents other core entities except the selected core entity. The vehicle status includes vehicle status data and basic vehicle information.

[0089] For each core entity, the data of related entities are extracted from the comprehensive status data, the related entities are labeled and associated with the core entity, and then fused to form a complete tire portrait.

[0090] Through undirected graph Mining potential new relationships. Specific methods include:

[0091] For each node in an undirected graph, calculate its critical centrality , which is used to measure the importance of a tire in the joint structure diagram, where , and Respectively represent The degree centrality, weighted degree centrality and structural similarity centrality of each node (tire), , and represents the weight coefficient, Indicates the index of the tire;

[0092] ,in, Indicates The degree of the node (tire) The number of edges connecting the nodes), Represents the total number of nodes (i.e. the total number of tires, usually 4);

[0093] Degree centrality is used to measure the number of connections (i.e., the number of edges) between a tire and other tires. In the tire scenario, degree centrality can indicate whether a tire frequently has abnormal connections with other tires, such as whether tire pressure fluctuations affect other tires.

[0094] ,in, Indicates except The set of all nodes outside the node, Indicates The node and The weights between nodes (such as the difference in tire pressure or temperature), represents the maximum weight of all directed edges;

[0095] Weighted degree centrality is used to measure the strength of the association between a tire and other tires (such as tire pressure difference, temperature difference, acceleration difference), and is calculated weightedly according to the weight of the edge (such as tire pressure difference or temperature difference).

[0096] ,in, Indicates The node and The similarity score between nodes, and , , , and represents the adjustment factor, Indicates The node and The attribute similarity between nodes (such as the cosine similarity of tire pressure and temperature, etc.), Indicates The node and The frequency of interaction between nodes (such as the number of times anomalies occur simultaneously), represents the time decay factor, Indicates The node and The time interval between the last interaction of the nodes. Indicates The node and The distance between nodes;

[0097] The calculation formula of the similarity score can integrate multiple dimensions. The structural similarity centrality is used to measure the attribute similarity between a tire and its neighbor tires, that is, whether it shows consistency with certain tires in certain characteristics (such as tire pressure, temperature, acceleration).

[0098] Extract the key centrality of all nodes and sort them in descending order, select the first e_u nodes as the core node set, these nodes represent the most important tires in the joint structure graph (may be the tires with frequent abnormalities or the strongest correlation with other tires). The remaining nodes are recorded as non-core nodes;

[0099] For each non-core node, calculate the path distance between it and all nodes in the core node set (the path distance can be the Euclidean distance), and take the minimum value of the path distance as the benchmark distance from the non-core node to the core node set;

[0100] Set J_L levels, each level defines a partition range, and the partition range is continuous. Non-core nodes are allocated to each level according to the benchmark distance until all non-core nodes are allocated.

[0101] According to the shortest path distance, non-core nodes are divided into different levels. For example, the first level: non-core nodes whose shortest path to the core node set is 1; the second level: non-core nodes whose shortest path to the core node set is 2, and so on, until all non-core nodes are assigned to a certain level.

[0102] For nodes at the same level, by calculating the similarity scores between each data category, if the similarity score is greater than the judgment threshold, it means that there is an association relationship and a directed edge is arranged; for nodes across levels, the number of interaction behaviors of each data category in the time period a_u is calculated. If the number of interaction behaviors is greater than the preset number, it means that there is an association relationship and a directed edge is arranged.

[0103] The data category refers to the various different classifications of data included in the comprehensive status data. For example, tire pressure data is a category of data, temperature data is a category of data, and driving direction can also be a category of data.

[0104] The determination threshold and the preset number of times can be set based on experimental data analysis or historical data analysis, or can be set based on experience.

[0105] Methods for arranging directed edges include:

[0106] Select a machine learning model (can be a neural network or deep learning), set the input of the machine learning model as the sample set, and the output as the tire performance index (the performance index can be the accuracy of the tire pressure warning, or the tire state deviation, which refers to the difference between the actual tire pressure and the model output tire pressure), use historical data to train the machine learning model, and obtain the trained machine learning model;

[0107] Set a standard data set. For data categories whose similarity scores are greater than the judgment threshold or whose number of interactive behaviors is greater than the preset number, randomly traverse and select two data categories. Replace the corresponding data category values ​​in the standard data set with one and two data categories respectively to form a new data set, which is recorded as a disturbed data set. Then, three disturbed data sets are obtained.

[0108] Import the perturbed data set into the trained machine learning model to obtain the performance index, which is recorded as the permutation performance value;

[0109] Calculate the weight of a directed edge between two data categories ,in, Indicates Data categories and The weights between data categories, Indicates The permutation performance value of the perturbed data set output corresponding to each data category, Indicates The permutation performance value of the perturbed data set output corresponding to each data category, Indicates and The permutation performance values ​​of the perturbed data set outputs corresponding to each data category.

[0110] The standard data set refers to data values ​​formed when the data values ​​of each data category are at standard values, and the standard value refers to the value required for each data category when the wheel is in a normal state.

[0111] Step S2: Calculate the difference data between the joint structure graph at the current time point and the previous time point to form a difference packet, and use the improved difference technology to transmit the difference packet;

[0112] Efficient transmission module: used to calculate differential data and transmit differential packets through improved differential technology;

[0113] In distributed systems or real-time data transmission scenarios, dynamic updates of joint structure graphs require efficient transmission. Directly transmitting the entire structure graph is expensive, especially when updates are frequent and network bandwidth is limited. Therefore, we design an improved difference technique that only captures and transmits the "difference packet" of the graph between two time points to reduce the amount of data and improve transmission efficiency.

[0114] The improved differential technology is used to transmit the differential packets. The specific methods include:

[0115] Copy and extract the nodes in the joint structure graph to form a new node set. For each directed edge, calculate the weight difference between the current time point and the previous time point to form a difference edge set. , then the difference package ;

[0116] When the size of the difference packet is large, it is transmitted in pieces, P_O continuous value ranges are set, and they are arranged in descending order. The weight differences are divided into corresponding value ranges, and the weight differences in the same value range are recorded as difference blocks. P_O difference blocks are collected;

[0117] Set the priority of P_O difference blocks according to the order of the value range, and arrange the priorities in descending order. Add a checksum (such as MD5 or SHA256) to each difference packet for the receiving end to verify the data integrity. Use a lightweight compression algorithm (such as Gzip or LZ4) to compress the difference block; reduce the transmission volume, and for data packets with small changes, directly use binary encoding (such as Protocol Buffers, MessagePack) instead of JSON to further reduce overhead.

[0118] Transmit P_O difference blocks of the difference packet according to the priority;

[0119] The difference blocks received after transmission are merged piece by piece to form a complete difference packet.

[0120] The higher the priority, the more important it is, the less likely it is to be lost, and the more it needs to be checked for packet loss or resent or sent first.

[0121] Step S3: construct and train a monitoring and early warning model, import the difference package into the trained monitoring and early warning model, and obtain early warning information;

[0122] Model prediction module: used to build and train monitoring and early warning models, import difference packages into the trained monitoring and early warning models, and obtain early warning information;

[0123] The model function is used to determine whether the current tire pressure of each tire is abnormal based on the input data; identify the specific reasons for abnormal tire pressure, such as excessive tire pressure, low tire pressure, air leakage, sensor failure, etc.; and provide adjustment suggestions for different abnormal reasons (such as increasing tire pressure, reducing tire pressure, checking sensors, etc.).

[0124] Construct and train a monitoring and early warning model, import the difference package into the trained monitoring and early warning model, and obtain early warning information by:

[0125] The framework of the monitoring and early warning model is based on neural network learning and LSTM technology, including input layer, hidden layer and output layer. The input of the input layer is set as multiple sets of data sets, including difference packages and basic vehicle information. The output layer sets three output heads in parallel, namely anomaly detection head, anomaly classification head and adjustment suggestion head, and the output is early warning information, which includes whether there is an abnormality in the tire, the cause of the abnormality and the abnormal adjustment suggestion.

[0126] The hidden layers include:

[0127] Time series feature extraction module: Use LSTM network to extract dynamic features in time series data. LSTM can capture the time dependency and change trend of data such as tire pressure and temperature. The input is the difference package, and the output is the time series dynamic feature representation, which serves as the input of subsequent tasks.

[0128] Static feature extraction module: Use a fully connected neural network (MLP) to extract high-level expressions of static features (such as vehicle speed, ambient temperature, load, etc.).

[0129] Feature fusion module: fuses time series features (LSTM output) and static features (MLP output) to form an overall feature vector through splicing operations. The fused feature vector is input into the task-specific network to complete anomaly detection, anomaly classification, and adjustment suggestion prediction.

[0130] The anomaly detection head is used to detect the tire pressure status (normal / abnormal) of each tire. The activation function is Sigmoid, and the loss function is set to binary cross entropy. The anomaly classification head is used to detect the abnormal cause category of each tire. The abnormal cause category is defined as excessive tire pressure (tire pressure exceeds the set upper limit), low tire pressure (tire pressure is lower than the set lower limit), air leakage (rapid drop in tire pressure), sensor failure (abnormal or lost tire pressure or temperature data) or other abnormalities (unclassifiable abnormalities). The activation function is Softmax, and the loss function is set to multi-class cross entropy. The adjustment suggestion head is used to generate abnormal adjustment suggestions for each tire. The task types include regression tasks (if the suggestion is numerical, such as adjusting the tire pressure to a certain target value) and classification tasks (if the suggestion is label type, such as "check for air leakage"), the activation functions are linear activation function (used for numerical output) and ReLU (used for label output), and the loss function is the accumulation of mean square error and multi-class cross entropy.

[0131] Set the total loss function of the model to the sum of the loss functions of the three output heads, and the optimization goal is to minimize the total loss function, using Adam or SGD optimizer;

[0132] The monitoring and early warning model is trained using the training set until the optimization target is reached, the trained monitoring and early warning model is obtained, and multiple sets of data sets at the current time point are input into the trained monitoring and early warning model to obtain early warning information.

[0133] The trained model is deployed in the vehicle's embedded devices (such as the tire pressure monitoring system controller), or deployed on a cloud server, and the tire pressure data is uploaded to the cloud for analysis via the Internet of Things (IoT).

[0134] The training set includes T_P groups of historical samples, each group of samples includes the vehicle difference package and basic vehicle information as well as the actual corresponding warning information;

[0135] Extract the timestamp of each sample, sort them according to the time series, divide them into batches, and import them into the monitoring and early warning model for training in sequence.

[0136] Step S4: displaying the warning information through the vehicle interface and adjusting the vehicle;

[0137] Visualization module: used to arrange warning information on the vehicle interface and interact with the user;

[0138] Visual interaction means that the warning information is displayed on the vehicle interface to facilitate observation and operation by personnel.

[0139] Example 2

[0140] See also Figure 1 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a vehicle tire pressure monitoring platform based on the Internet of Things, including:

[0141] Data collection and integration module: used to collect comprehensive status data of tires and build a joint structure diagram list;

[0142] Efficient transmission module: used to calculate differential data and transmit differential packets through improved differential technology;

[0143] Model prediction module: used to build and train monitoring and early warning models, import difference packages into the trained monitoring and early warning models, and obtain early warning information;

[0144] Visualization module: used to arrange warning information on the vehicle interface and interact with the user.

[0145] Example 3

[0146] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned vehicle-mounted tire pressure monitoring method based on the Internet of Things is implemented.

[0147] Since the electronic device introduced in this embodiment is an electronic device used to implement a vehicle-mounted tire pressure monitoring method based on the Internet of Things in the embodiment of this application, based on the vehicle-mounted tire pressure monitoring method based on the Internet of Things introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the vehicle-mounted tire pressure monitoring method based on the Internet of Things in the embodiment of this application, it belongs to the scope of protection of this application.

[0148] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0149] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A vehicle tire pressure monitoring method based on the Internet of Things, characterized in that: include: Step S1: Collecting comprehensive status data of each tire and constructing a joint structure diagram of the tire; The method for constructing a joint structure graph comprises: For each tire, the comprehensive status data is fused to form a complete tire portrait; Define the correlation between each tire, including pressure correlation, temperature correlation, acceleration correlation and stress state correlation, and extract the correlation between each tire based on the tire image; The extracted association relationship is represented as an undirected graph G = (V, E), where V is a set of nodes, each node is used to represent a tire, E is a set of directed edges, the weight of each edge is represented by the value of the association relationship between tires, and the direction of the directed edge is from the core entity to the associated entity; Through the undirected graph G, potential new associations are mined and added to the original undirected graph. The obtained complete undirected graph is recorded as a joint structure graph; Step S2: Calculate the difference data between the joint structure graph at the current time point and the previous time point to form a difference packet, and use the improved difference technology to transmit the difference packet; The improved differential technology is used to transmit the difference packet, and the specific method includes: Copy and extract the nodes in the joint structure graph to form a new node set. For each directed edge, calculate the weight difference between the current time point and the previous time point to form a difference edge set ΔE. The difference package C G =(V, ΔE); Set P_O continuous value ranges and arrange them in descending order, divide the weight differences into corresponding value ranges, record the weight differences in the same value range as difference blocks, and collect P_O difference blocks; Set the priority of P_O difference blocks according to the order of the value range, arrange the priorities in descending order, add a checksum to each difference packet, and compress the difference block using a lightweight compression algorithm; Transmit P_O difference blocks of the difference packet according to the priority; Merge the difference blocks received after transmission piece by piece to form a complete difference packet; Step S3: construct and train a monitoring and early warning model, import the difference package into the trained monitoring and early warning model, and obtain early warning information; Step S4: Display the warning information through the vehicle interface and adjust the vehicle.

2. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 1, characterized in that: The comprehensive status data includes real-time tire pressure data, temperature data, acceleration data, vehicle status data and basic vehicle information; The method for obtaining real-time tire pressure data and temperature data is to divide the tire into S_Y areas, and the area of ​​the area Each area outputs a real-time tire pressure data and temperature data; Among them, O_P is the acquisition frequency, L_p is the surface area of ​​the tire; The acceleration data is collected at an adaptive frequency Zs, and Wherein, Zs' represents the acquisition frequency at the previous time point, Zs" represents the acquisition frequency at the time point before that, and Zs represents the acquisition frequency at the current time point; Vehicle status data includes driving speed, driving direction, steering angle and braking status; The basic vehicle information includes the vehicle's appearance data and the number, size and material of its tires.

3. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 2 is characterized in that: The method of fusing the comprehensive status data to form a complete tire image includes: Define the core entities, related entities and associated entities of tire portraits. The core entities represent each tire, and the related entities represent their own characteristics, associations and vehicle status. The own characteristics include the real-time tire pressure data, temperature data and acceleration data corresponding to each core entity. The association is used to describe the relationship between any core entity and the associated entity. The associated entity represents other core entities except the selected core entity. The vehicle status includes vehicle status data and basic vehicle information. For each core entity, the data of related entities are extracted from the comprehensive status data, the related entities are labeled and associated with the core entity, and then fused to form a complete tire portrait.

4. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 3 is characterized in that: The specific method of mining potential new association relationships through the undirected graph G includes: For each node in the undirected graph, calculate its key centrality Gz(i) = α1×C d (i)+α2×C ω (i)+α3×C s (i), where C d (i) C ω (i) and C s (i) represents the degree centrality, weighted degree centrality and structural similarity centrality of the i-th node, respectively. α1, α2 and α3 represent weight coefficients, and i represents the index of the tire. Among them, k i represents the degree of the i-th node, and N represents the total number of nodes; Where N(i) represents the set of all nodes except the i-th node, ω(i, j) represents the weight between the i-th node and the j-th node, and Δω represents the maximum weight of all directed edges; Among them, S(i, j) represents the relationship between the i-th node and the j-th node. Similarity score, and β1, β2, β3 and β4 represent adjustment coefficients, sim(i, j) represents the attribute similarity between the i-th node and the j-th node, f(i, j) represents the frequency of interaction between the i-th node and the j-th node, λ represents the time decay factor, t(i, j) represents the time interval between the i-th node and the j-th node’s most recent interaction, and d(i, j) represents the distance between the i-th node and the j-th node; Extract the key centrality of all nodes and sort them in descending order, select the first e_u nodes as the core node set, and the remaining nodes are recorded as non-core nodes; For each non-core node, calculate the path distance between it and all nodes in the core node set, and take the minimum value of the path distance as the benchmark distance from the non-core node to the core node set; Set J_L levels, each level defines a partition range, and the partition range is continuous. Non-core nodes are allocated to each level according to the benchmark distance until all non-core nodes are allocated. For nodes at the same level, by calculating the similarity scores between each data category, if the similarity score is greater than the judgment threshold, it means that there is an association relationship and a directed edge is arranged; for nodes across levels, the number of interaction behaviors of each data category in the time period a_u is calculated. If the number of interaction behaviors is greater than the preset number, it means that there is an association relationship and a directed edge is arranged.

5. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 4 is characterized in that: The method for arranging directed edges comprises: Select a machine learning model, set the input of the machine learning model as a sample set, set the output as a tire performance index, train the machine learning model using historical data, and obtain a trained machine learning model; Set a standard data set. For data categories whose similarity scores are greater than the judgment threshold or whose number of interactive behaviors is greater than the preset number, randomly traverse and select two data categories. Replace the corresponding data category values ​​in the standard data set with one and two data categories respectively to form a new data set, which is recorded as a disturbed data set. Then, three disturbed data sets are obtained. Import the perturbed data set into the trained machine learning model to obtain the performance index, which is recorded as the permutation performance value; Calculate the weight of a directed edge between two data categories Among them, ω(c, v) represents the weight between the c-th data category and the v-th data category, Ao(c) represents the permutation performance value of the c-th data category corresponding to the perturbed data set output, Bo(c) represents the permutation performance value of the v-th data category corresponding to the perturbed data set output, and BA(c, v) represents the permutation performance value of the c-th and v-th data categories corresponding to the perturbed data set output.

6. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 5, characterized in that: The method of constructing and training a monitoring and early warning model, importing the difference package into the trained monitoring and early warning model, and obtaining early warning information includes: The framework of the monitoring and early warning model is based on neural network learning and LSTM technology, including input layer, hidden layer and output layer. The input of the input layer is set as multiple sets of data sets, including difference packages and basic vehicle information. The output layer sets three output heads in parallel, namely anomaly detection head, anomaly classification head and adjustment suggestion head, and the output is early warning information, which includes whether there is an abnormality in the tire, the cause of the abnormality and the abnormal adjustment suggestion. The anomaly detection head is used to detect the tire pressure status of each tire. The activation function is Sigmoid, and the loss function is set to binary cross entropy. The anomaly classification head is used to detect the abnormal cause category of each tire. The abnormal cause category is defined as excessive tire pressure, low tire pressure, air leakage, sensor failure or other abnormalities. The activation function is Softmax, and the loss function is set to multi-class cross entropy. The adjustment suggestion head is used to generate abnormal adjustment suggestions for each tire. The task types include regression tasks and classification tasks. The activation functions are linear activation functions and ReLU, and the loss function is the accumulation of mean square error and multi-class cross entropy. Set the total loss function of the model to the sum of the loss functions of the three output heads, and the optimization goal is to minimize the total loss function, using Adam or SGD optimizer; The monitoring and early warning model is trained using the training set until the optimization target is reached, the trained monitoring and early warning model is obtained, and multiple sets of data sets at the current time point are input into the trained monitoring and early warning model to obtain early warning information.

7. The vehicle-mounted tire pressure monitoring method based on the Internet of Things according to claim 6, characterized in that: The training set includes T_P groups of historical samples, each group of samples includes a vehicle difference package and basic vehicle information and actual corresponding warning information; Extract the timestamp of each sample, sort them according to the time series, divide them into batches, and import them into the monitoring and early warning model for training in sequence.

8. An on-vehicle tire pressure monitoring platform based on the Internet of Things, characterized in that: The method for monitoring tire pressure on a vehicle based on the Internet of Things according to any one of claims 1 to 7 comprises: Data collection and integration module: used to collect comprehensive status data of tires and build a joint structure diagram list; Efficient transmission module: used to calculate differential data and transmit differential packets through improved differential technology; Model prediction module: used to build and train monitoring and early warning models, import difference packages into the trained monitoring and early warning models, and obtain early warning information; Visualization module: used to arrange warning information on the vehicle interface and interact with the user.

Images