Tire full life cycle supply chain management method and system

Abstract

The invention relates to a tire full life cycle supply chain management method and system, and the method comprises the steps of intelligent sensor deployment, data processing, parallel calculation, data standardization, traceability graph construction, storage and visualization. Data processing and feature extraction are performed by using edge computing and distributed computing clusters, data consistency and exception handling are ensured, problem node positioning is realized by constructing a data traceability graph, the processed data are stored in a distributed database, a tire three-dimensional model is generated and rasterization processing is performed, and a tire three-dimensional image is obtained. Visual monitoring of the tire state is achieved, and supply chain management is optimized.

Description

Technical Field

[0001] The present application relates to the field of information technology, and in particular to a tire full life cycle supply chain management method and system. Background Art

[0002] In digital data processing and tire life cycle supply chain management systems, there are technical challenges in real-time processing and accuracy assurance of massive tire life cycle data. A large amount of data is generated in various links such as tire production, transportation, storage, use, maintenance, and scrapping. These data are time-sensitive, have diverse formats, and are sourced in a dispersed manner. Efficient collection, transmission, and processing of these data is a huge challenge. At the same time, due to the harsh environment in which tires are used, noise, packet loss, distortion, and other problems are prone to occur during data collection, resulting in inaccurate data and affecting subsequent analysis and decision-making.

[0003] In addition, tire life cycle data has obvious time series characteristics. How to use the time sequence and correlation of the data to build a reasonable data model and algorithm to achieve real-time verification and error correction of the data, thereby improving data quality and system reliability, is also a technical problem that needs to be solved urgently.

[0004] The solution proposed by the present invention for the shortcomings of the above content is: intelligent sensors collect tire data, edge computing nodes perform data preprocessing, distributed computing clusters perform parallel computing, and locate problem nodes through data tracing diagrams, and finally store and render tire data to generate a three-dimensional model. The present invention includes the steps of data collection, filtering, preprocessing, parallel computing, data unification, problem location, storage and visualization to ensure the efficiency and accuracy of data processing. Summary of the invention

[0005] In order to solve the problems existing in the above-mentioned prior art, the purpose of this application is to provide a tire full life cycle supply chain management method and system.

[0006] The present application discloses a tire full life cycle supply chain management method, comprising the following steps:

[0007] S1. Deploy smart sensors with multi-core heterogeneous parallel processing capabilities, which collect raw data of tire pressure in real time and obtain tire internal structure data;

[0008] S2, the edge computing node filters the original data of the tire pressure in step S1 to remove redundant pressure information, pre-processes the tire internal structure data in step S1, and generates a tire internal structure feature map;

[0009] S3, uploading the original data of tire pressure and the tire internal structure characteristic map in step S2 to the distributed computing cluster, and the distributed computing framework distributes the original data of tire pressure and the tire internal structure characteristic map to different computing nodes for parallel computing;

[0010] S4. If the hardware multithreading of the computing node is abnormal, the dynamic branch prediction mechanism is activated to distribute the data processed by the node to other normal nodes, and the cache consistency protocol is used to ensure data consistency;

[0011] S5. After each computing node completes the calculation of the original data of the tire pressure and the tire internal structure characteristic diagram, the formats of the original data of the tire pressure and the tire internal structure characteristic diagram are unified through a unified data standard;

[0012] S6. Based on the unified original data of tire pressure and the characteristic diagram of tire internal structure, a tire data traceability diagram is constructed to record the entire process from data collection to data processing. If an abnormality occurs in the data processing link, the problem node is located according to the traceability diagram;

[0013] S7, storing the processed original data of tire pressure and the characteristic diagram of tire internal structure in a distributed database, rendering the original data of tire pressure and the characteristic diagram of tire internal structure to obtain a visualization result;

[0014] S8, the visualization server generates a three-dimensional tire model according to the original data of tire pressure and the tire internal structure characteristic map, and rasterizes the three-dimensional model to obtain a three-dimensional tire image;

[0015] S9, mapping the original data of tire pressure and the characteristic map of the internal structure of the tire to the three-dimensional model of the tire, rendering the mapped model, and obtaining a tire visualization result.

[0016] Preferably, in step S1, an intelligent sensor integrating a multi-core heterogeneous processor, pressure and image sensors is deployed to collect tire pressure data and internal structure images in real time, and upload them to a data processing center. The data processing center monitors pressure anomalies and issues alarms, processes images to extract texture features, performs spectrum analysis for early warning, and fuses pressure and image data.

[0017] Preferably, in step S2, the edge node receives and groups the tire pressure data uploaded by the vehicle-mounted sensor, filters and fuses the valid data to remove noise, analyzes the trend of data changes in the tire, processes features through a convolutional neural network, performs pooling and classification to determine the internal structural characteristics of the tire, and uses a recurrent neural network for time series analysis to build a structural feature change model and determine the state. Finally, the processed data and feature graphs are stored and visualized.

[0018] Preferably, in step S3, the edge node receives and groups the tire pressure data uploaded by the vehicle-mounted sensor, filters and denoises to obtain valid data, processes data features through a convolutional neural network, performs pooling and classification to identify the internal structural features of the tire, uploads the data to a distributed computing cluster for parallel computing, identifies the degree of deformation in the feature map based on the relationship model between the tire pressure data and the internal structure, and uses a regression model to correct the pressure data.

[0019] Preferably, in step S4, the operation information of the computing nodes is monitored, the thread status is recorded and the exceptions are handled, the normal node set is determined, and a data allocation plan is formulated and data migration is triggered based on the node load and data volume to achieve data synchronization and cache consistency maintenance between nodes, collect tire pressure and structural feature data, assign unique identifiers and record information.

[0020] Preferably, in step S5, the original data of tire pressure and the format of tire internal structure characteristic map uploaded by each computing node are collected and standardized. If the data format does not meet the standard, the format is converted, the structure characteristic map is analyzed using an image recognition algorithm, the pressure data is corrected in combination with the regression model, the corrected pressure data and characteristic map are converted into a standard format, the feature vector is extracted, and the data is packaged into a data packet with a unified data standard according to preset rules for subsequent processing.

[0021] Preferably, in step S6, the original data of tire pressure and the internal structure characteristic diagram of the tire are collected, a unique identifier is assigned and a timestamp is recorded, the data preprocessing module cleans and calibrates the data, extracts key features to form a feature vector, stores normal data in a database, and generates a data traceability diagram. Abnormal data triggers an alarm, and the problem node is traced back and located through the identifier. The abnormal pressure data is further analyzed in the structure characteristic diagram to determine the degree of deformation, and the traceability diagram and abnormal nodes are displayed through a graphical interface.

[0022] Preferably, in the step S7, the tire pressure data and the internal structure characteristic diagram completed by the computing node are collected. If the pressure data is abnormal, the deformation in the structure diagram is analyzed by an image recognition algorithm, and the pressure data is corrected. The corrected pressure data is standardized, and the characteristic vector of the structure diagram is extracted. The standardized pressure data and the structure diagram are combined into a data packet, which is segmented and verified according to the database protocol and then stored. The pressure data is visualized into a curve graph through a rendering engine, and the structure characteristic diagram is rendered.

[0023] Preferably, in step S8, the processor obtains the tire model from the model library, uses the graphics library and combines the internal structure feature map information to build a three-dimensional tire model. If the pressure is abnormal, the model is adjusted to reflect the deformation. The model is rasterized to obtain an initial image. If the resolution is insufficient, it is enhanced through a super-resolution algorithm. The processed image is used to eliminate aliasing, denoise and enhance color to obtain a three-dimensional tire image, which is output by a visualization server.

[0024] Preferably, in step S9, tire pressure data and internal structure feature maps are obtained, and the data format is determined. These data are used to construct a mapping relationship and initial model data of the tire three-dimensional model, thereby forming a geometric model of the tire. After the geometric model is verified, attribute mapping is performed to generate an attribute model, and a lighting model is created to achieve a lighting effect. If the lighting effect meets the visual characteristics, rasterization processing is performed to generate a two-dimensional image, and image rendering is completed to achieve visualization of the tire three-dimensional model. Finally, the pressure data, structure feature map and visualization results are stored in a distributed database, and the storage structure is determined.

[0025] A tire full life cycle supply chain management system described in the present application includes: an intelligent sensor module, an edge computing node module, a distributed computing cluster module, a data standardization module, a data traceability construction module, a distributed database storage module, a tire three-dimensional model generation module, and a tire three-dimensional model optimization module;

[0026] The smart sensor module is connected to the edge computing node module to collect tire pressure data and internal structure images in real time, and upload them to the data processing center using the communication module in the smart sensor;

[0027] The edge computing node module is connected to the smart sensor module and the distributed computing cluster module, and is used to receive tire pressure data from the smart sensor, process filtering, denoising, and feature extraction, perform pooling and classification through a convolutional neural network, and determine the internal structure characteristics of the tire, wherein the processed tire pressure data and the internal structure feature map are stored through a distributed database, the tire pressure data and the internal structure feature map of the tire are uploaded to the distributed computing cluster, and the distributed computing framework distributes the pressure data and the internal structure data to different computing nodes;

[0028] The distributed computing cluster module is connected to the edge computing node module and the data standardization module, and is used to receive data uploaded by the edge node, distribute it to different computing nodes for parallel computing, and perform exception processing. If the computing node is abnormal, the dynamic branch prediction mechanism is started to reallocate the data;

[0029] The data standardization module is connected with the distributed computing cluster module and the data traceability construction module to convert the data into data with unified data standards;

[0030] The data traceability construction module is connected with the data standardization module and the distributed database storage module, and is used to construct a tire data traceability diagram based on the standardized data. When data processing is abnormal, the problem node is located through the traceability diagram, wherein the original data is transmitted to the data preprocessing module, and the data integrity is judged according to the rule base; normal data is stored, and the data is associated with the corresponding identifier and timestamp and stored in the distributed database;

[0031] The distributed database storage module is connected to the data tracing construction module and the tire three-dimensional model generation module, and is used to store the processed data in the distributed database, visualize the pressure data through the rendering engine, and render the structural feature map, wherein the standard data packet is divided into multiple data blocks according to the data storage protocol of the distributed database;

[0032] The tire 3D model generation module is connected to the distributed database storage module and the tire 3D model optimization module, and is used to construct a tire 3D model using the graphics library and the internal structure feature map information, and the tire 3D image after rasterization is output by the visualization server, wherein the processor accesses the model library to obtain the tire model file, and uses the graphics library to modify the tire 3D model to obtain the deformed tire 3D model;

[0033] The tire three-dimensional model optimization module is connected to the tire three-dimensional model generation module, and is used to construct a mapping relationship and initial model data of the tire three-dimensional model, create a lighting model, achieve a lighting effect, and perform rasterization processing to generate a two-dimensional image, wherein pressure data, internal structure feature maps and visualization result images are stored through a distributed database.

[0034] The tire full life cycle supply chain management method and system described in the present application have the advantages that, in order to solve the problem of multi-source heterogeneous tire data processing and the reliability and traceability requirements in the data processing process, the present invention deploys intelligent sensors with multi-core heterogeneous parallel processing capabilities to collect tire pressure and internal structure data in real time, uses edge computing nodes to preprocess data, and distributed computing clusters to perform parallel computing, and introduces a dynamic branch prediction mechanism and a cache consistency protocol to ensure the reliability of the computing process. At the same time, a tire data traceability diagram is constructed to achieve traceability of the data processing process. Finally, by rendering the processed data, a three-dimensional tire model is generated and visualized, so as to achieve an intuitive presentation of the tire status.

[0035] The present invention effectively solves the problems of data collection, processing, reliability assurance, traceability and visualization throughout the tire life cycle, and improves the efficiency and accuracy of tire status monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1This is the process of a tire full life cycle supply chain management method described in this application Figure 1 ;

[0037] Figure 2 This is the process of a tire full life cycle supply chain management method described in this application Figure 2 ;

[0038] Figure 3 It is a structural block diagram of a tire full life cycle supply chain management system described in this application. DETAILED DESCRIPTION

[0039] like Figure 1-Figure 2 As shown, a tire full life cycle supply chain management method described in this application includes the following steps:

[0040] S1. Deploy smart sensors with multi-core heterogeneous parallel processing capabilities, which collect raw data of tire pressure in real time and obtain tire internal structure data;

[0041] S2, the edge computing node filters the original data of the tire pressure in step S1 to remove redundant pressure information, pre-processes the tire internal structure data in step S1, and generates a tire internal structure feature map;

[0042] S3, uploading the original data of tire pressure and the tire internal structure characteristic map in step S2 to the distributed computing cluster, and the distributed computing framework distributes the original data of tire pressure and the tire internal structure characteristic map to different computing nodes for parallel computing;

[0043] S4. If the hardware multithreading of the computing node is abnormal, the dynamic branch prediction mechanism is activated to distribute the data processed by the node to other normal nodes, and the cache consistency protocol is used to ensure data consistency;

[0044] S5. After each computing node completes the calculation of the original data of the tire pressure and the tire internal structure characteristic diagram, the formats of the original data of the tire pressure and the tire internal structure characteristic diagram are unified through a unified data standard;

[0045] S6. Based on the unified original data of tire pressure and the characteristic diagram of tire internal structure, a tire data traceability diagram is constructed to record the entire process from data collection to data processing. If an abnormality occurs in the data processing link, the problem node is located according to the traceability diagram;

[0046] S7, storing the processed original data of tire pressure and the characteristic diagram of tire internal structure in a distributed database, rendering the original data of tire pressure and the characteristic diagram of tire internal structure to obtain a visualization result;

[0047] S8, the visualization server generates a three-dimensional tire model according to the original data of tire pressure and the tire internal structure characteristic map, and rasterizes the three-dimensional model to obtain a three-dimensional tire image;

[0048] S9, mapping the original data of tire pressure and the characteristic map of the internal structure of the tire to the three-dimensional model of the tire, rendering the mapped model, and obtaining a tire visualization result.

[0049] like Figure 1-Figure 2 As shown, in step S1, a smart sensor is deployed, wherein the smart sensor integrates a multi-core heterogeneous processor, a pressure sensor, and an image sensor;

[0050] The pressure sensor in the intelligent sensor collects tire pressure data in real time to obtain raw data of tire pressure;

[0051] The image sensor in the smart sensor collects the internal structure image of the tire to obtain the original image of the structure;

[0052] The integrated communication module in the smart sensor uploads the raw data of tire pressure to the data processing center. If the raw data of tire pressure is lower than the preset pressure threshold, the alarm mechanism is triggered;

[0053] The image processor in the smart sensor segments the original image and determines multiple sub-images;

[0054] The data processing center receives the original data of tire pressure. If the deviation between the original data of tire pressure and historical data is greater than the set value, the data is marked as abnormal;

[0055] The data processing center adopts a multi-core heterogeneous parallel processing architecture to obtain texture feature information of each sub-image;

[0056] The data processing center performs spectrum analysis based on the texture feature information, and triggers an early warning mechanism if the difference between the spectrum analysis result and the standard spectrum is greater than a set range;

[0057] The data processing center fuses the original pressure data and the original structure image to obtain fused data.

[0058] Specifically, in step S1, a smart sensor is deployed, which integrates a multi-core heterogeneous processor, a pressure sensor, and an image sensor, including using an R-Car series chip from Renesas Electronics, which integrates multiple ARM Cortex-A57 and Cortex-A53 cores, and a PowerVR GX6650 GPU from Imagination Technologies;

[0059] The pressure sensor in the smart sensor uses a piezoresistive pressure sensor, and the sampling frequency is set to 100Hz to collect tire pressure data in real time and obtain the original data of tire pressure, including 2.5bar;

[0060] The image sensor in the smart sensor uses a CMOS image sensor with a resolution of 1280x720 pixels to collect the internal structure image of the tire and obtain the original image of the structure;

[0061] The 4G communication module integrated in the smart sensor uploads the raw pressure data to the data processing center. If the raw pressure data is lower than the preset pressure threshold of 2.0 bar, the alarm mechanism is triggered;

[0062] The image processor in the smart sensor uses an edge detection-based image segmentation algorithm, including the Canny algorithm, to segment the original image of the structure into multiple sub-images and determine 9 sub-images;

[0063] The data processing center receives the original pressure data. If the deviation between the original pressure data and the historical data is greater than the set value of 0.2 bar, the data will be marked as abnormal;

[0064] The data processing center adopts a multi-core heterogeneous parallel processing architecture, uses the OpenCL framework, calls GPU resources, and uses the gray-level co-occurrence matrix algorithm to obtain texture feature information such as contrast, energy, and entropy for each sub-image;

[0065] The data processing center performs fast Fourier transform spectrum analysis based on texture feature information. If the difference between the high-frequency component ratio in the spectrum analysis result and the high-frequency component ratio in the standard spectrum is greater than the set range of 10%, the early warning mechanism is triggered;

[0066] The data processing center uses an image fusion algorithm based on wavelet transform to fuse the original pressure data and the original structure image to obtain fused data.

[0067] like Figure 1-Figure 2 As shown, in step S2, the edge node receives the original data of tire pressure uploaded by the vehicle-mounted sensor, groups the data according to the time series, and obtains a set of tire pressure data in different time periods;

[0068] The tire pressure data set in each time period is analyzed through data filtering. If the fluctuation range of the pressure data value is less than the tire pressure threshold, the pressure data set in this time period is judged to be redundant information, and the redundant information is discarded to obtain valid pressure data. The effective pressure data is fused using the preprocessing layer to obtain the tire data after removing the noise signal.

[0069] According to the change trend of the intra-tire data, the change characteristics of the intra-tire data are extracted. If the change trend of the intra-tire data shows that the fluctuation range of the tire pressure data value is greater than the tire pressure threshold, the pressure data set in this time period is determined to be valid tire pressure data;

[0070] Performing convolution calculation on the intra-fetal data change feature through a first convolutional neural network to obtain a plurality of first convolutional feature maps, and performing a first pooling process according to the first convolutional feature maps to obtain a first pooling feature map;

[0071] Performing convolution calculation on the first pooled feature map through a second convolutional neural network to obtain multiple second convolutional feature maps, and performing a second pooling process on the second convolutional feature map to obtain a second pooled feature map;

[0072] The second pooled feature map is classified through a fully connected layer to obtain a third feature map, and the internal structural features of the tire are determined according to the third feature map;

[0073] The structural features are analyzed in time series by using a recurrent neural network, a time series change model of the internal structural features of the tire is constructed, and the internal structural state of the tire is determined based on the time series change model.

[0074] The processed tire pressure data and internal structure characteristic map are stored in a distributed database, and the tire pressure data and internal structure characteristic map are processed to obtain visualization results.

[0075] Specifically, in step S2, the edge node receives the original tire pressure data uploaded by the vehicle-mounted sensor in real time, including receiving 100 pressure data points per second, and grouping these data points into time periods of 10 seconds according to the timestamp, to obtain a tire pressure data set of 10 time periods;

[0076] Analyze the tire pressure data set in each time period, including calculating the standard deviation of the tire pressure data in each time period. If the standard deviation in a certain time period is less than the preset threshold value of 0.1PSI, the pressure data set in this time period is determined to be redundant information;

[0077] Discard these redundant information and only keep the valid pressure data with a standard deviation greater than 0.1PSI. Use the Kalman filter algorithm to fuse the valid pressure data to obtain the in-fetal data after removing the noise signal.

[0078] According to the change trend of tire data, including the tire pressure value rising by more than 2PSI within 1 minute, the change characteristics of tire data are extracted. If the change trend of tire data shows that the fluctuation range of tire pressure data value is greater than the tire pressure threshold, the pressure data set within this time period is judged to be valid tire pressure data;

[0079] The first convolutional neural network including 64 3x3 convolution kernels is used to perform convolution calculation on the intra-fetal data change characteristics to obtain 64 first convolutional feature maps, and the first convolutional feature map is subjected to maximum pooling processing with a step size of 2 to obtain the first pooling feature map;

[0080] Perform convolution calculation on the first pooled feature map through a second convolutional neural network including 128 3x3 convolution kernels to obtain 128 second convolutional feature maps, and perform average pooling processing with a step size of 2 on the second convolutional feature map to obtain a second pooled feature map;

[0081] The second pooled feature map is classified by a fully connected layer including 1024 neurons to obtain a third feature map, and the internal structural features of the tire, including the deformation mode of the crown or the sidewall, are determined according to the third feature map;

[0082] The structural features are analyzed in time series through the long short-term memory network (LSTM), and a time series change model of the internal structural features of the tire is constructed. According to the time series change model, the internal structural state of the tire is determined, including whether there is slow air leakage.

[0083] The processed tire pressure data and internal structure feature maps are stored in a distributed database based on Hadoop, and the tire pressure data and internal structure feature maps are rendered based on WebGL to obtain visualization results.

[0084] like Figure 1-Figure 2 As shown, in step S3, the edge node receives the original data of tire pressure uploaded by the vehicle-mounted sensor, groups the data according to the time series, and obtains a set of tire pressure data in different time periods;

[0085] The tire pressure data set in each time period is analyzed through data filtering. If the pressure data value fluctuation range is less than the tire pressure threshold, the pressure data set in the time period is judged to be redundant information, and the redundant information is discarded to obtain valid pressure data;

[0086] The effective pressure data is fused by the preprocessing layer to remove the noise signal and obtain the denoised in-fetal data. The change characteristics of the in-fetal data are extracted according to the change trend of the in-fetal data.

[0087] Performing convolution calculation on the intra-fetal data change feature through a first convolutional neural network to obtain a plurality of first convolutional feature maps, and performing a first pooling process according to the first convolutional feature maps to obtain a first pooling feature map;

[0088] Performing convolution calculation on the first pooled feature map through a second convolutional neural network to obtain multiple second convolutional feature maps, and performing a second pooling process on the second convolutional feature map to obtain a second pooled feature map;

[0089] The second pooled feature map is classified through a fully connected layer to obtain a third feature map, and the internal structural features of the tire are determined according to the third feature map;

[0090] The tire pressure data and tire internal structure feature map are uploaded to the distributed computing cluster. The distributed computing framework distributes the pressure data and internal structure data to different computing nodes for parallel computing to obtain the tire pressure data calculated by each computing node.

[0091] According to the pre-established tire pressure data and internal structure relationship model, the tire pressure data range is judged. If the tire pressure data range exceeds the preset threshold range, the tire internal structure characteristic diagram at the corresponding time of the calculation node is obtained to obtain the internal structure characteristic diagram of the calculation node;

[0092] The preset marked area in the feature map is identified by a pre-trained image recognition algorithm, and the deformation degree of the marked area is determined. According to the deformation degree of the marked area, a regression model based on the relationship between the deformation degree and the pressure data is adopted to obtain the corrected pressure data.

[0093] Specifically, in step S3, the edge node receives the original data of tire pressure uploaded by the vehicle-mounted sensor, including sampling 100 times per second, and groups the data according to the time series, including every 5 minutes as a group, to obtain tire pressure data sets in different time periods;

[0094] The tire pressure data set in each time period is analyzed through data filtering. If the pressure data value fluctuation range is less than the preset 0.5PSI tire pressure threshold, the pressure data set in this time period is judged to be redundant information, and the redundant information is discarded to obtain valid pressure data;

[0095] The effective pressure data is fused by the preprocessing layer, and the noise signal is removed by applying the Kalman filter algorithm to obtain the denoised fetal data. According to the change trend of the fetal data, the change characteristics of the fetal data, including the pressure rise rate or drop rate, are extracted;

[0096] The first convolutional neural network is used to perform convolution calculation on the intra-fetal data change characteristics, using a 3x3 convolution kernel with a step size of 1 to obtain multiple first convolution feature maps, and the first pooling process is performed according to the first convolution feature map, using a 2x2 maximum pooling to obtain a first pooling feature map;

[0097] The first pooled feature map is convolved by the second convolutional neural network, and a 3x3 convolution kernel is also used with a step size of 1 to obtain multiple second convolutional feature maps. A second pooling process is performed according to the second convolutional feature map, and 2x2 average pooling is adopted to obtain the second pooled feature map.

[0098] Classifying the second pooled feature map through a fully connected layer, including using a Softmax classifier, to obtain a third feature map, and determining internal structural features of the tire based on the third feature map;

[0099] The tire pressure data and tire internal structure feature map are uploaded to a distributed computing cluster, including a Hadoop cluster. The distributed computing framework distributes the pressure data and internal structure data to different computing nodes, including 10 computing nodes, for parallel computing. Using the MapReduce framework, the tire pressure data calculated by each computing node is obtained.

[0100] According to the pre-established tire pressure data and internal structure relationship model, including the support vector machine model, the tire pressure data range is determined. If the tire pressure data range exceeds the preset 2.0-2.5PSI threshold range, a tire internal structure characteristic diagram at the corresponding time of the calculation node is obtained to obtain the internal structure characteristic diagram of the calculation node;

[0101] Through pre-trained image recognition algorithms, including Faster R-CNN algorithms, preset marked areas in the feature map are identified, and the degree of deformation of the marked areas is determined, including a deformation degree of 10%. According to the degree of deformation of the marked areas, a regression model based on the relationship between the degree of deformation and pressure data, including a linear regression model, is used to obtain corrected pressure data.

[0102] like Figure 1-Figure 2 As shown, in step S4, the computing node runtime information is obtained, and the register in the processor records the thread state. If the thread state is abnormal, the interrupt handler is triggered to obtain the abnormal thread information;

[0103] According to the abnormal thread information, the interrupt handler reads the pre-established node status list. If the node status information in the list indicates that there are normal nodes, the set of normal nodes participating in the calculation is determined;

[0104] According to the set of normal nodes participating in the calculation, the execution unit counts the amount of data to be processed by each thread, and obtains the average amount of data according to the amount of data to be processed and the number of normal nodes participating in the calculation;

[0105] The average data volume and the current load of each normal node are used to determine the data allocation plan, which includes the mapping relationship between data blocks and target nodes, and obtains the data block redistribution list;

[0106] According to the data block reallocation list, the fault detection module triggers the data migration instruction, which includes the data block identifier and the target node address, and obtains the set of data blocks to be migrated;

[0107] According to the set of data blocks to be migrated, after the data migration instruction is triggered, the source node sends the data block to the target node. After receiving the data block, the target node sends a cache update request through the internal bus to obtain the data synchronization instruction between nodes;

[0108] According to the inter-node data synchronization instruction, the target node cache data is updated. The synchronization module compares the cache information of each node. If the cache information is different, the cache consistency protocol is triggered to obtain the cache consistency maintenance instruction.

[0109] Obtain the original data of tire pressure and the characteristic diagram of tire internal structure, assign a unique identifier to each data, the identifier set is S, record the data collection time and data source information according to the identifier and timestamp, and obtain the original tire data;

[0110] According to the cache consistency maintenance instructions, the cache data of each node is updated to ensure data consistency. The branch prediction mechanism is dynamically adjusted according to the data allocation plan to determine the data processing flow.

[0111] Specifically, in step S4, the computing node continuously monitors the thread status through the internal sensor, including the register status value of 0x00FF indicating thread abnormality, triggering the interrupt handler with interrupt vector number 15, and obtaining abnormal thread information;

[0112] The interrupt handler accesses the node status list, including the list showing that the node A is in a normal state, the node B is in an abnormal state, and the node C is in a normal state, and then determines that nodes A and C participate in the calculation;

[0113] The execution unit counts the amount of data to be processed by each thread, including 10MB of data to be processed by node A, 15MB of data to be processed by node B, and 5MB of data to be processed by node C. The average data amount is (10+15+5) / 2=15MB;

[0114] The current load of node A is 60%, and the current load of node C is 40%. According to the load balancing algorithm, including the weighted polling algorithm, it is determined that the 15MB data of node B is allocated to node C for processing;

[0115] The fault detection module generates a data migration instruction, including an instruction format of "MOVE DATA_BLOCK_ID=001TARGET_NODE=C", and obtains an instruction for migrating data block 001 to node C;

[0116] Data block 001 is migrated from node B to node C. After receiving the data block, node C sends a cache update request, including a request to update the data at the cache line address 0x1000.

[0117] The synchronization module compares the cache information of nodes A and C and finds that the data of cache line address 0x1000 of node C is inconsistent, triggering the MESI protocol;

[0118] Get tire pressure data, including tire pressure value of 2.5 bar, internal structure feature map resolution of 1024x1024 pixels, assigned identifier Tire_001, timestamp of 2023-10-27 10:00:00, and get original tire data;

[0119] According to the MESI protocol, the cache line address 0x1000 data of node C is updated to be consistent with node A. According to the data allocation plan, the dynamic branch prediction mechanism reallocates the task originally assigned to node B to nodes A and C to determine the new data processing flow.

[0120] like Figure 1-Figure 2 As shown, in step S5, the original data of the tire pressure calculated by each computing node and the tire internal structure characteristic diagram are obtained, and the standard format of the original data of the tire pressure and the tire internal structure characteristic diagram is determined according to the pre-established data format specification. The format of the pressure data uploaded by each computing node is identified through the preset data format template. If the format of the pressure data is inconsistent with the standard format, the format conversion program is started to obtain the pressure data in the standard format;

[0121] According to the preset image format standard of the internal structure feature map, the format of the internal structure feature map uploaded by each computing node is identified. If the format of the internal structure feature map is inconsistent with the standard format, the image format conversion program is started to obtain the internal structure feature map in a unified format;

[0122] Obtain the pressure data after format conversion, and determine the tire pressure data range according to the pre-established tire pressure data and internal structure relationship model;

[0123] If the tire pressure data range exceeds the preset threshold range, the tire internal structure feature map at the corresponding time of the calculation node is obtained, and the preset marked area in the feature map is identified by a pre-trained image recognition algorithm to determine the deformation degree of the marked area;

[0124] According to the deformation degree of the marked area, a regression model based on the relationship between the deformation degree and the pressure data is used to obtain the corrected pressure data and update the original pressure data;

[0125] The corrected pressure data of each computing node is converted into first pressure data in a standard format through a preset data format template, and the image feature extraction algorithm based on a convolutional neural network is used to extract the feature vector of the internal structure feature map uploaded by each computing node to determine the feature vector set;

[0126] According to the preset image format standard of the internal structure feature map, the internal structure feature map uploaded by each computing node is converted into a second internal structure feature map in a unified format, the first pressure data and the second internal structure feature map are obtained, and the next step of data processing is determined;

[0127] According to the preset data packaging rules, the first pressure data and the second internal structure characteristic map are combined into a standard data packet to obtain the output result of the final unified data standard.

[0128] Specifically, in step S5, firstly, the tire pressure data calculated by each calculation node is obtained, including the pressure data of node A being 3.5MPa and node B being 2.8MPa, and their corresponding internal structure characteristic graphs. The standard format of these data should be determined according to the preset data format specification, including the uniform use of MPa unit for pressure data and the uniform use of JPEG format for internal structure characteristic graphs;

[0129] The pressure data format uploaded by each node is identified through a preset data format template, including an XML template. If the pressure data format of node A conforms to the XML template, and the pressure data of node B is in CSV format, a format conversion program is started to convert the CSV format of node B into XML format to obtain pressure data in a standardized format.

[0130] According to the preset image format standard, including requiring all images to be in JPEG format of 1024x768 pixels, identify the format of the internal structure feature map uploaded by each node. If the feature map of node A is in PNG format and the feature map of node B meets the JPEG standard, start the image format conversion program to convert the PNG format of node A into JPEG format to obtain the internal structure feature map in a unified format.

[0131] Obtain the pressure data after format conversion, including 3.5MPa of node A and 2.8MPa of node B. According to the relationship model between tire pressure data and internal structure, when the pressure is greater than 3.0MPa, it is considered to be a high pressure state. It is determined that the pressure data range of node A is high pressure and node B is normal.

[0132] If the pressure data range of node A exceeds the preset threshold range (including 3.0MPa), the internal structure feature map of the tire at the corresponding moment of node A is obtained, and the preset marked area in the feature map, including the marked points at specific positions, is identified through a pre-trained image recognition algorithm, including an image classification model based on ResNet, and the deformation degree of the marked area is determined, including a 10% change in the spacing between the marked points;

[0133] According to the deformation degree of the marked area of ​​10%, a regression model based on the relationship between the deformation degree and the pressure data is adopted, including a linear regression model y=kx+b, where y is the corrected pressure, x is the deformation degree, k and b are model parameters, to obtain the corrected pressure data, including 3.6MPa, and update the original pressure data;

[0134] The corrected pressure data of each node, including 3.6MPa of node A and 2.8MPa of node B, are converted into first pressure data in a standard format through a preset data format template, and a convolutional neural network-based image feature extraction algorithm, including a VGG-16 network, is used to extract feature vectors of the internal structure feature graph uploaded by each node, including obtaining a 1x4096-dimensional feature vector and determining a feature vector set;

[0135] According to the preset image format standard of the internal structure feature map, the internal structure feature map uploaded by each node is converted into a second internal structure feature map in a unified format, the first pressure data and the second internal structure feature map are obtained, and the next step of data processing is determined to be data packaging;

[0136] According to the preset data packaging rules, including packaging the pressure data and feature map data in JSON format, combining the first pressure data and the second internal structure feature map into a standard data packet, and obtaining the final output result of the unified data standard, including generating a JSON file containing the pressure data and feature map data.

[0137] like Figure 1-Figure 2 As shown, in step S6, the original data of tire pressure and the internal structure characteristic map of the tire are obtained, each data is assigned a unique identifier, the identifier set is S, the number of internal structure characteristic maps is N, the i-th characteristic map is represented by Xi, and the set is represented by X={X1, X2, ...XN}. The data collection time and data source information are recorded according to the identifier and timestamp to obtain the original data;

[0138] The original data is transmitted to the data preprocessing module for data cleaning, removal of duplicate and erroneous data, and judgment of data integrity based on the rule base. If the data is missing, it is marked to obtain intermediate data;

[0139] The intermediate data is formatted, heterogeneous data is converted into a unified format, and the data is calibrated according to the calibration rules to obtain standard data;

[0140] Using feature extraction algorithm, key features are extracted from standard data to form feature vectors. The feature vector is represented by V = {v1, v2, ...vn}, where n represents the feature dimension. According to the rules, it is determined whether there are abnormal values ​​in the feature vector. If there are abnormal values, abnormal alarm information is generated.

[0141] Store normal data, associate the data with the corresponding identifier and timestamp and store them in a distributed database, generate the node information of the data traceability graph based on the data storage record, and obtain the node information set;

[0142] Obtain abnormal alarm information, extract abnormal data identifiers, trace back abnormal data in the database according to the identifiers, and obtain the abnormal data processing flow;

[0143] If the abnormal data identifier exists in the tire pressure data, then the tire pressure data calculated by each computing node is obtained, and the tire pressure data range is determined according to the pre-established tire pressure data and internal structure relationship model;

[0144] If the tire pressure data range exceeds the preset threshold range, the tire internal structure feature map at the corresponding time of the calculation node is obtained, and the preset marked area in the feature map is identified by a pre-trained image recognition algorithm to determine the deformation degree of the marked area;

[0145] The data traceability diagram is displayed through a graphical interface. The abnormal nodes are highlighted in the traceability diagram according to the processing flow of the abnormal data. Based on the abnormal node information, the corresponding data processing link information is obtained to obtain the problem node location result.

[0146] Specifically, in step S6, tire pressure data and internal structure characteristic graphs are obtained, a unique identifier is assigned to each data, including an identifier set S = {S1, S2, ... Sn}, the number of internal structure characteristic graphs is 100, the 50th characteristic graph is represented by X50, the set is represented by X = {X1, X2, ... X100}, and the data collection time and data source information are recorded according to the identifier S1 and the timestamp 2023-10-26, 10:00:00, to obtain the original data containing specific data information;

[0147] The original data is transmitted to the data preprocessing module, and the data cleaning algorithm is used to remove duplicate and erroneous data, including removing duplicate values ​​in the pressure data and erroneous values ​​that are obviously beyond the normal range, including negative pressure or pressure exceeding 10MPa. The data integrity is judged according to the rule base. If the data is missing by more than 20%, it is marked to obtain the marked intermediate data;

[0148] The intermediate data is converted into JSON format by format conversion algorithm, and the sensor data is calibrated by standard pressure value according to the calibration rules to obtain standard data. The principal component analysis (PCA) algorithm is used to extract key features from the standard data to form a feature vector V = {v1, v2, ... v10}, including reducing the 100 feature dimensions of the original data to 10, and judging whether there are abnormal values ​​in the feature vector according to the quartile range rule. If the value of v3 is greater than the upper quartile plus 1.5 times the interquartile range, an abnormal alarm message is generated;

[0149] The normal data is associated with the corresponding identifier S1 and the timestamp 2023-10-26, 10:00:00 and stored in the database, and the node information of the data traceability graph is generated according to the data storage record to obtain a node information set;

[0150] Obtain abnormal alarm information, extract abnormal data identifier S3, trace back abnormal data in the database according to identifier S3, and obtain the processing flow of abnormal data from collection to processing;

[0151] If the abnormal data identifier S3 exists in the tire pressure data, then the tire pressure data calculated by each computing node is obtained, including the pressure data of node A being 3.5 MPa, and according to the pre-established tire pressure data and internal structure relationship model, it is determined whether the pressure data is between 2.0 MPa and 3.0 MPa;

[0152] If the tire pressure data range of 3.5MPa exceeds the preset threshold range of 2.0MPa to 3.0MPa, then obtain the tire internal structure feature map X30 at the corresponding time of the computing node A, identify the preset marked area in the feature map X30 through the pre-trained ResNet-50 image recognition algorithm, and determine the deformation degree of the marked area, including the deformation coefficient of 0.8;

[0153] The data traceability diagram is displayed through a graphical interface. The abnormal nodes are highlighted in the traceability diagram according to the processing flow of the abnormal data S3. According to the abnormal node information, the corresponding data processing link is obtained as the data preprocessing link, and the problem node location result is that an abnormality occurs in the data preprocessing link.

[0154] like Figure 1-Figure 2 As shown, in step S7, the tire pressure data calculated by each computing node and the tire internal structure characteristic diagram at the corresponding moment are obtained to obtain the original tire pressure data and the internal structure diagram;

[0155] If the tire pressure data range exceeds the preset threshold range, the preset marked area in the tire internal structure feature map at the time corresponding to the calculation node is identified by a pre-trained image recognition algorithm to determine the deformation degree of the marked area;

[0156] According to the deformation degree of the marked area, a regression model based on the relationship between the deformation degree and the pressure data is used to obtain the corrected pressure data and update the original pressure data;

[0157] The corrected pressure data of each computing node is converted into first pressure data in a standard format through a preset data format template to obtain normalized pressure data;

[0158] An image feature extraction algorithm based on a convolutional neural network is used to extract the feature vectors of the internal structure feature graph uploaded by each computing node and determine the feature vector set;

[0159] According to a preset image format standard of the internal structure feature map, the internal structure feature map uploaded by each computing node is converted into a second internal structure feature map in a unified format to obtain a normalized internal structure feature map;

[0160] Acquire the first pressure data and the second internal structure characteristic map, and combine the first pressure data and the second internal structure characteristic map into a standard data packet according to a preset data packaging rule to obtain a combined data packet;

[0161] According to the data storage protocol of the distributed database, the standard data packet is divided into multiple data blocks, and verification information is generated through a data verification algorithm to obtain the data blocks to be stored and the verification information;

[0162] The first pressure data is mapped into a pressure curve graph through a rendering engine, and the second internal structure feature graph is rendered to obtain a visualization result.

[0163] Specifically, in step S7, the tire pressure data calculated by each computing node is obtained, including the pressure value uploaded by node A being 60PSI, the pressure value uploaded by node B being 65PSI, and the tire internal structure characteristic diagrams of these nodes at the corresponding moments;

[0164] If the tire pressure data 60PSI of node A exceeds the preset threshold range of 55PSI to 63PSI, a pre-trained image recognition algorithm, including a ResNet-50 model, is used to identify the preset marked area in the tire internal structure feature map at the corresponding moment of node A, and the deformation degree of the marked area is determined to be 5%;

[0165] According to the deformation degree of the marked area of ​​5%, a regression model based on the relationship between the deformation degree and the pressure data is adopted, including a linear regression model y=ax+b, where y represents the corrected pressure data, and x represents the deformation degree. The corrected pressure data is calculated to be 62PSI, and the original pressure data is updated by 60PSI;

[0166] Through the preset data format template, the corrected pressure data of each computing node, including 62PSI of node A and 64PSI of node B, is converted into the first pressure data in the standard format. In this example, the JSON format {"node":"A","pressure":62} is used;

[0167] An image feature extraction algorithm based on a convolutional neural network, including the VGG-16 model, is used to extract feature vectors of the internal structure feature graph uploaded by each computing node and determine a feature vector set;

[0168] According to a preset image format standard of the internal structure feature map, including unified conversion into PNG format and adjusting the resolution to 512x512 pixels, the internal structure feature map uploaded by each computing node is converted into a second internal structure feature map in a unified format;

[0169] Acquire the first pressure data and the second internal structure characteristic graph, and combine the first pressure data and the second internal structure characteristic graph into a standard data packet according to a preset data packaging rule, including sorting and combining the data according to timestamps and node IDs;

[0170] According to the data storage protocol of the distributed database, including HDFS of Hadoop, the standard data packet is divided into a plurality of data blocks, including each data block having a size of 128MB, and verification information is generated through a data verification algorithm, including MD5;

[0171] Through a rendering engine, including OpenGL, the first pressure data is mapped into a pressure curve graph, with the horizontal axis being time and the vertical axis being the pressure value, and the second internal structure feature graph is image rendered to obtain a visualization result.

[0172] like Figure 1-Figure 2 As shown, in step S8, according to the tire pressure data and the internal structure characteristic diagram, the processor accesses the model library to obtain the corresponding tire model file and obtains a preliminary tire data model;

[0173] Obtain a preliminary tire data model, use the graphics library and combine the internal structure feature map information to build a tire 3D model, and determine the basic shape of the tire 3D model;

[0174] Obtain the constructed tire three-dimensional model. If the pressure value exceeds the preset range, determine the tire deformation parameter according to the pressure value, and use the graphics library to modify the tire three-dimensional model to obtain a deformed tire three-dimensional model.

[0175] Acquire the deformed tire three-dimensional model, and use a renderer to perform rasterization processing to obtain a first rasterized image of the tire three-dimensional model;

[0176] Acquire a first rasterized image of the tire three-dimensional model, and if the resolution of the first rasterized image is lower than a preset resolution threshold, use a super-resolution reconstruction algorithm to process it to obtain a second rasterized image of the tire three-dimensional model with a high resolution;

[0177] A second rasterized image of a high-resolution tire three-dimensional model is obtained. If there are aliasing on the edge of the image, an anti-aliasing algorithm is used to process the image to obtain a third rasterized image of the tire three-dimensional model with smooth edges.

[0178] A third rasterized image of the smooth-edge tire 3D model is obtained. If there is noise in the image, an image denoising algorithm is used to process the image to obtain a fourth rasterized image of a clear tire 3D model.

[0179] Acquire a fourth rasterized image of a clear tire 3D model, and if the color saturation of the fourth rasterized image of the clear tire 3D model is lower than a preset value, perform color enhancement processing to obtain a fifth rasterized image of the tire 3D model with rich colors;

[0180] The fifth rasterized image of the colorful tire three-dimensional model is obtained, and the visualization server outputs the final tire three-dimensional image to obtain the tire visualization result.

[0181] Specifically, in step S8, the processor accesses the model library according to the tire pressure data (including 1.5 Bar) and the internal structure characteristic diagram (including the tread pattern depth of 8 mm and the number of sidewall steel wire layers of 2 layers), matches the corresponding tire model file (including the Michelin Pilot Sport4 tire model) through the retrieval algorithm, and obtains a preliminary tire data model;

[0182] Obtain the tire data model, use the OpenGL graphics library, and combine the internal structure feature map information (including the angle of each pattern block is 30 degrees) to construct a three-dimensional model with a tread pattern depth of 8 mm and two layers of sidewall steel wires, and determine the basic shape of the tire three-dimensional model;

[0183] Obtain the constructed tire 3D model. If the pressure value is 2.5 Bar, which exceeds the preset range of 1.8 Bar to 2.2 Bar, then determine the tire deformation parameters (including the radial deformation coefficient of 0.05) according to the pressure value of 2.5 Bar, and use the OpenGL graphics library to modify the tire 3D model to obtain a deformed tire 3D model that meets the pressure of 2.5 Bar.

[0184] The deformed tire three-dimensional model is obtained, and rasterized using a ray tracing renderer, with a sampling rate set to 16 samples per pixel, to obtain a first rasterized image of the tire three-dimensional model with a resolution of 1024x768;

[0185] Obtain a first rasterized image of the tire 3D model. If its resolution 1024x768 is lower than a preset resolution threshold 1920x1080, use a super-resolution reconstruction algorithm based on deep learning (including an SRCNN algorithm, including three convolution layers, with convolution kernel sizes of 9x9, 1x1, and 5x5, respectively) to process the image and obtain a second rasterized image of the high-resolution tire 3D model with a resolution of 1920x1080.

[0186] The high-resolution image is obtained. If jagged edges are detected by an edge detection algorithm (including a Sobel operator), a fast approximate anti-aliasing algorithm (FXAA) is used to process the image to obtain a third rasterized image of the tire 3D model with smooth edges.

[0187] The smooth edge image is obtained. If the peak signal-to-noise ratio (PSNR) of the image is found to be lower than 30 dB by calculating the peak signal-to-noise ratio (PSNR), a non-local mean-based image denoising algorithm is used for processing. The search window size is set to 21x21 and the similarity window size is set to 7x7 to obtain a fourth rasterized image of a clear tire 3D model.

[0188] The clear image is obtained. If the color saturation of the image is lower than the preset value of 0.8, a color enhancement process is performed to increase the saturation to 0.9, thereby obtaining a fifth rasterized image of the tire three-dimensional model with rich colors.

[0189] The colorful image is obtained, and the visualization server outputs the final tire three-dimensional image to obtain the tire visualization result.

[0190] like Figure 1-Figure 2 As shown, in step S9, tire pressure data and internal structure characteristic diagram are obtained, and data formats of the tire pressure data and internal structure characteristic diagram are determined;

[0191] According to the data format of the tire pressure data and the internal structure characteristic map, a mapping relationship of the tire three-dimensional model is constructed to obtain mapping relationship parameters of the tire three-dimensional model;

[0192] Using the mapping relationship parameters of the tire three-dimensional model, data conversion is performed on the tire pressure data and the internal structure characteristic map to obtain initial model data of the tire three-dimensional model;

[0193] The geometric shape of the tire three-dimensional model is constructed by using the initial model data of the tire three-dimensional model to obtain the geometric model of the tire three-dimensional model;

[0194] If the geometric model of the tire three-dimensional model conforms to the mapping relationship between the tire pressure data and the internal structure characteristic map, then attribute mapping is performed on the geometric model of the tire three-dimensional model to obtain an attribute model of the tire three-dimensional model;

[0195] Generate a lighting model of the three-dimensional tire model according to the attribute model of the three-dimensional tire model to obtain the lighting effect of the three-dimensional tire model;

[0196] If the lighting effect of the three-dimensional tire model meets the visual characteristics of the tire pressure data and the internal structure feature map, the three-dimensional tire model is rasterized to obtain a two-dimensional image of the three-dimensional tire model;

[0197] Performing image rendering according to the two-dimensional image of the three-dimensional tire model to obtain a visualization result of the three-dimensional tire model;

[0198] The tire pressure data, the internal structure characteristic diagram and the visualization result of the tire three-dimensional model are stored through the distributed database, and the storage structure of the tire visualization result is determined.

[0199] Specifically, in step S9, tire pressure data is obtained, including a pressure value of 3.5 bar recorded by each tire pressure sensor, and an internal structure feature map is obtained, including an internal structure image of the tire obtained by X-ray scanning, with a resolution of 1024x768 pixels, and the pressure data is determined to be in a floating point format, and the internal structure feature map is in a grayscale image format;

[0200] According to these data formats, a mapping relationship is constructed, including mapping the pressure value to the color change of the tire model, 3.5 bar corresponds to the RGB color value (255,0,0), and mapping the grayscale value of the internal structure feature map to the texture details of the model surface. The grayscale value range of 0-255 corresponds to the texture depth of 0.1-1.0 mm, and a mapping relationship parameter file is obtained. This mapping relationship parameter file is used to convert the data, including using a Python script to read the pressure data and image data, and converting the pressure value and grayscale value into the vertex color and texture coordinates of the tire model through a linear interpolation algorithm to obtain an initial model data file;

[0201] By reading the initial model data file, using OpenGL and other graphics libraries to build geometric shapes, including building a triangular patch model through vertex coordinates and patch indices, a geometric model containing 10,000 triangular patches is obtained;

[0202] If the vertex color and texture coordinates of each triangle in the geometric model are consistent with the definition in the mapping relationship parameter file, attribute mapping is performed, including setting the color attribute of each triangle to the corresponding RGB value and mapping the texture coordinates to the corresponding texture image to obtain a model with color and texture attributes;

[0203] Generate a lighting model based on the attribute model, including using the Phong lighting model, setting the coefficients of ambient light, diffuse light, and specular light to 0.2, 0.5, and 0.3 respectively, to obtain a model with lighting effects;

[0204] If the high-pressure area of ​​the tire model is highlighted in red under the lighting effect, and the internal structure texture is clearly visible, which meets the visual characteristics, rasterization processing is performed, including using the Z-buffer algorithm for depth testing, projecting the 3D model onto a 2D plane, and obtaining a 1920x1080 pixel 2D image;

[0205] According to the two-dimensional image, use the rendering engine including Blender to render the image, set the rendering parameters including the sampling number as 100 and the output format as PNG, and obtain the final visualization result image;

[0206] The pressure data, internal structure feature graphs and visualization result images are stored through a distributed database including MongoDB, and the storage structure is determined to include a pressure value field, an image field and a result image field for each tire record.

[0207] like Figure 3 As shown, a tire full life cycle supply chain management system described in the present application includes: an intelligent sensor module, an edge computing node module, a distributed computing cluster module, a data standardization module, a data traceability construction module, a distributed database storage module, a tire three-dimensional model generation module, and a tire three-dimensional model optimization module;

[0208] The smart sensor module is connected to the edge computing node module to collect tire pressure data and internal structure images in real time, and upload them to the data processing center using the communication module in the smart sensor;

[0209] The edge computing node module is connected to the smart sensor module and the distributed computing cluster module, and is used to receive tire pressure data from the smart sensor, process filtering, denoising, and feature extraction, perform pooling and classification through a convolutional neural network, and determine the internal structure characteristics of the tire, wherein the processed tire pressure data and the internal structure feature map are stored through a distributed database, the tire pressure data and the internal structure feature map of the tire are uploaded to the distributed computing cluster, and the distributed computing framework distributes the pressure data and the internal structure data to different computing nodes;

[0210] The distributed computing cluster module is connected to the edge computing node module and the data standardization module, and is used to receive data uploaded by the edge node, distribute it to different computing nodes for parallel computing, and perform exception processing. If the computing node is abnormal, the dynamic branch prediction mechanism is started to reallocate the data;

[0211] The data standardization module is connected with the distributed computing cluster module and the data traceability construction module to convert the data into data with unified data standards;

[0212] The data traceability construction module is connected with the data standardization module and the distributed database storage module, and is used to construct a tire data traceability diagram based on the standardized data. When data processing is abnormal, the problem node is located through the traceability diagram, wherein the original data is transmitted to the data preprocessing module, and the data integrity is judged according to the rule base; normal data is stored, and the data is associated with the corresponding identifier and timestamp and stored in the distributed database;

[0213] The distributed database storage module is connected to the data tracing construction module and the tire three-dimensional model generation module, and is used to store the processed data in the distributed database, visualize the pressure data through the rendering engine, and render the structural feature map, wherein the standard data packet is divided into multiple data blocks according to the data storage protocol of the distributed database;

[0214] The tire 3D model generation module is connected to the distributed database storage module and the tire 3D model optimization module, and is used to construct a tire 3D model using the graphics library and the internal structure feature map information, and the tire 3D image after rasterization is output by the visualization server, wherein the processor accesses the model library to obtain the tire model file, and uses the graphics library to modify the tire 3D model to obtain the deformed tire 3D model;

[0215] The tire three-dimensional model optimization module is connected to the tire three-dimensional model generation module, and is used to construct a mapping relationship and initial model data of the tire three-dimensional model, create a lighting model, achieve a lighting effect, and perform rasterization processing to generate a two-dimensional image, wherein pressure data, internal structure feature maps and visualization result images are stored through a distributed database.

[0216] In the description of this specification, the description with reference to the terms "some embodiments", "other embodiments", "ideal embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example.

[0217] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0218] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A tire full life cycle supply chain management method, characterized in that: The following steps are involved: S1. Deploy a multi-core heterogeneous parallel processing intelligent sensor, which collects raw data of tire pressure in real time and obtains tire internal structure data; S2, the edge computing node filters the original data of the tire pressure in step S1 to remove redundant pressure information, pre-processes the tire internal structure data in step S1, and generates a tire internal structure feature map; S3, uploading the original data of tire pressure and the tire internal structure characteristic map in step S2 to the distributed computing cluster, and the distributed computing framework distributes the original data of tire pressure and the tire internal structure characteristic map to different computing nodes for parallel computing; S4. If the hardware multithreading of the computing node is abnormal, the dynamic branch prediction mechanism is activated to distribute the data processed by the node to other normal nodes, and the cache consistency protocol is used to ensure data consistency; S5. After each computing node completes the calculation of the original data of the tire pressure and the tire internal structure characteristic diagram, the formats of the original data of the tire pressure and the tire internal structure characteristic diagram are unified through a unified data standard; S6. Based on the unified original data of tire pressure and the characteristic diagram of tire internal structure, a tire data traceability diagram is constructed to record the entire process from data collection to data processing. If an abnormality occurs in the data processing link, the problem node is located according to the traceability diagram; S7, storing the processed original data of tire pressure and the characteristic diagram of tire internal structure in a distributed database, rendering the original data of tire pressure and the characteristic diagram of tire internal structure to obtain a visualization result; S8, the visualization server generates a three-dimensional tire model according to the original data of tire pressure and the tire internal structure characteristic map, and rasterizes the three-dimensional model to obtain a three-dimensional tire image; S9, mapping the original data of tire pressure and the characteristic map of the internal structure of the tire to the three-dimensional model of the tire, rendering the mapped model, and obtaining a tire visualization result.

2. A tire full life cycle supply chain management method according to claim 1, characterized in that: In step S1, the original data of tire pressure and tire internal structure data are collected by intelligent sensors, including: Deploy intelligent sensor nodes to obtain raw data of tire pressure, and integrate communication modules in the intelligent sensor nodes to upload the raw data of tire pressure to a data processing center; The data processing center receives the raw data of the tire pressure, and if the raw data of the tire pressure is lower than a preset pressure threshold, an alarm mechanism is triggered, and the data processing center stores the raw data of the tire pressure; The intelligent sensor node collects the original image of the internal structure of the tire, and the image processor integrated in the intelligent sensor node segments the original image of the internal structure of the tire to determine sub-images; The data processing center adopts a multi-core heterogeneous parallel processing architecture to obtain texture feature information of each sub-image.

3. The tire full life cycle supply chain management method according to claim 1, characterized in that: The edge computing node filters the original data of the tire pressure in step S1 to remove redundant pressure information, preprocesses the tire internal structure data in step S1, and generates a tire internal structure feature map, including: The edge node receives the original data of tire pressure uploaded by the vehicle-mounted sensor, groups the original data of tire pressure according to the time series, and obtains tire pressure data sets of different time periods; Analyze the tire pressure data set in each time period through data filtering. If the pressure data value fluctuation range is less than the tire pressure threshold, determine that the tire pressure data set in the time period is redundant information, discard the redundant information, and obtain valid pressure data; The effective pressure data is fused by using a preprocessing layer to remove noise signals therein, thereby obtaining denoised in-fetal data, and according to the change trend of the in-fetal data, the change characteristics of the in-fetal data are extracted.

4. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S3, the original data of tire pressure and the tire internal structure characteristic map in step S2 are uploaded to the distributed computing cluster, and the distributed computing framework distributes the original data of tire pressure and the tire internal structure characteristic map to different computing nodes for parallel computing, including: Obtain the tire ID associated timestamp to form an initial data pair, and according to the tire ID, combine the original data of the tire pressure and the tire internal structure characteristic map into a data packet, and upload it to the distributed computing cluster; The distributed computing cluster receives the data packet and distributes the data packet to an idle node in the computing cluster according to a pre-established distribution table. If the node is processing data, the data packet is cached in the queue of the node. After receiving the data packet, the nodes in the computing group use a fast Fourier transform algorithm to extract the periodic variation characteristics in the pressure value and obtain the pressure fluctuation frequency.

5. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S4, if the hardware multithreading of the computing node is abnormal, the dynamic branch prediction mechanism is started to distribute the data processed by the node to other normal nodes for the cache consistency protocol to ensure data consistency, including: Obtaining computing node runtime information and thread status recorded in processor registers, and triggering an interrupt handler if the thread status is abnormal; The interrupt processing program reads a pre-established node status list and determines a set of normal nodes participating in the calculation according to the node status information in the list; The execution unit counts the amount of data to be processed of each thread, and obtains the average amount of data according to the amount of data to be processed and the number of normal nodes participating in the calculation; A data allocation scheme is determined according to the average data volume and the current load of each normal node, wherein the data allocation scheme includes a mapping relationship between data blocks and target nodes.

6. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S5, after each computing node completes the calculation of the original data of tire pressure and the tire internal structure characteristic diagram, the format of the original data of tire pressure and the tire internal structure characteristic diagram is unified through a unified data standard, including: Obtain tire pressure data calculated by each computing node, and determine the tire pressure data range based on a pre-established tire pressure data and internal structure relationship model; If the tire pressure data range exceeds a preset threshold range, a tire internal structure characteristic map at a corresponding time of the calculation node is obtained, a preset marked area in the tire internal structure characteristic map is identified by a pre-trained image recognition algorithm, and a deformation degree of the marked area is determined; According to the deformation degree of the marked area, a regression model based on the relationship between the deformation degree and the pressure data is adopted to obtain the corrected pressure data.

7. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S6, a tire data traceability diagram is constructed based on the unified original tire pressure data and the tire internal structure characteristic diagram to record the entire process from data collection to data processing. If an abnormality occurs in the data processing link, the problem node is located according to the traceability diagram, including: Obtain tire pressure data and tire internal structure feature maps, assign a unique identifier to each data, the identifier set is S, the number of internal structure feature maps is N, the i-th feature map is represented by Xi, and the set is represented by X = {X1, X2, ... XN}. Record the data collection time and data source information according to the identifier and timestamp to obtain the original data; The raw data is transmitted to the data preprocessing module for data cleaning to remove duplicate and erroneous data.

8. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S7, the processed original data of tire pressure and the tire internal structure characteristic map are stored in a distributed database, and the original data of tire pressure and the tire internal structure characteristic map are rendered to obtain a visualization result, including: According to a preset image format standard of the internal structure feature map, the internal structure feature map uploaded by each computing node is converted into a second internal structure feature map in a unified format to obtain a normalized internal structure feature map; Acquire the first pressure data and the second internal structure characteristic map, and combine the first pressure data and the second internal structure characteristic map into a standard data packet according to a preset data packaging rule to obtain a combined data packet; According to the data storage protocol of the distributed database, the standard data packet is divided into multiple data blocks, and verification information is generated through a data verification algorithm to obtain the data blocks to be stored and the verification information; The first pressure data is mapped into a pressure curve graph through a rendering engine, and the second internal structure feature graph is rendered to obtain a visualization result.

9. The tire full life cycle supply chain management method according to claim 1, characterized in that: In step S8, the visualization server generates a tire three-dimensional model according to the original data of tire pressure and the tire internal structure characteristic map, and rasterizes the three-dimensional model to obtain a tire three-dimensional image, including: According to the tire pressure data and the internal structure characteristic diagram, the processor accesses the model library to obtain the tire model file and obtains a preliminary tire data model; Obtain a preliminary tire data model, use a graphics library and combine the internal structure feature map information to build a tire three-dimensional model to determine the basic shape of the tire three-dimensional model; The constructed tire 3D model is obtained. If the pressure value exceeds the preset range, the tire deformation parameter is determined according to the pressure value, and the tire 3D model is modified using the graphics library to obtain a deformed tire 3D model.

10. A tire full life cycle supply chain management system, characterized in that: include: The smart sensor module is connected to the edge computing node module to collect tire pressure data and internal structure images in real time; The edge computing node module is connected to the smart sensor module and the distributed computing cluster module to receive tire pressure data from the smart sensor, process filtering, denoising, and feature extraction, perform pooling and classification through a convolutional neural network, and determine the internal structural characteristics of the tire; The distributed computing cluster module is connected to the edge computing node module and the data standardization module to receive data uploaded by the edge nodes, distribute them to different computing nodes for parallel computing, and handle exceptions. If a computing node is abnormal, the dynamic branch prediction mechanism is activated to reallocate data. The data standardization module is connected with the distributed computing cluster module and the data traceability construction module to convert the data into data with unified data standards; The data traceability construction module is connected with the data standardization module and the distributed database storage module to construct a tire data traceability diagram based on the standardized data. When data processing is abnormal, the problem node is located through the traceability diagram; The distributed database storage module is connected with the data tracing construction module and the tire three-dimensional model generation module to store the processed data in the distributed database, visualize the pressure data through the rendering engine, and render the structural feature map; The tire 3D model generation module is connected with the distributed database storage module and the tire 3D model optimization module, and is used to construct the tire 3D model using the graphics library and the internal structure feature map information. The tire 3D image after rasterization processing is output by the visualization server; The tire three-dimensional model optimization module is connected to the tire three-dimensional model generation module, and is used to construct the mapping relationship and initial model data of the tire three-dimensional model, create a lighting model, and perform rasterization processing to generate a two-dimensional image.

Images