CN-121835442-B - Tire service life prediction method and system based on big data

CN121835442BCN 121835442 BCN121835442 BCN 121835442BCN-121835442-B

Abstract

The invention relates to the technical field of tire life prediction, in particular to a tire life prediction method and a system based on big data, comprising the steps of acquiring multisource running data and vehicle state information of a target tire, and a reference tire life database and sample characteristics corresponding to the target tire; and cleaning and extracting the characteristics of the multi-source operation data to obtain a wear characteristic sequence and a load characteristic sequence of the target tire, and constructing a wear trend sequence and a load evaluation sequence of the target tire according to the reference tire life database, the wear characteristic sequence and the load characteristic sequence. The invention can accurately predict the residual life of the target tire by collecting and fusing multi-source data and combining a attention mechanism, and is helpful for providing more accurate and personalized life prediction by analyzing the wear deviation and the load overrun condition of different time nodes.

Inventors

WANG CHUANZHU
GUO YONGFANG
MA QINGJUN
XU YAN
ZHANG YUNCHUAN
ZHOU WENMIN
WANG HONGLEI

Assignees

青岛泰凯英专用轮胎股份有限公司
泰凯英(青岛)专用轮胎技术研究开发有限公司

Dates

Publication Date: 20260508
Application Date: 20260313

Claims (10)

1. A tire life prediction method based on big data, comprising: Acquiring multisource operation data and vehicle state information of a target tire, and a reference tire life database and sample characteristics corresponding to the target tire; cleaning and extracting characteristics of the multi-source operation data to obtain a wear characteristic sequence and a load characteristic sequence of the target tire, and constructing a wear trend sequence and a load evaluation sequence of the target tire according to a reference tire life database, the wear characteristic sequence and the load characteristic sequence; according to the time sequence position of the time stamp in the reference tire life database, comparing the abrasion trend sequence with the load evaluation sequence to obtain a fusion characteristic sequence, and acquiring the attention weight characteristic of each associated monitoring point and the associated characteristic corresponding to the target tire through an attention mechanism based on the fusion characteristic sequence and the sample characteristic; carrying out global average pooling on a second historical monitoring data sequence in the plurality of historical monitoring data sequences to obtain pooling characteristics, wherein the second historical monitoring data sequence is a historical monitoring data sequence except the first historical monitoring data sequence; the pooled feature, the attention weight feature, the associated feature and the fusion feature sequence are subjected to splicing treatment to obtain a spliced feature, and the spliced feature is input into a mask network of a life prediction model to obtain a mask feature; And extracting running state characteristics of the vehicle state information based on the mask characteristics to obtain multi-dimensional state characteristics, carrying out regression processing on the multi-dimensional state characteristics, and outputting target residual life information of the target tire, wherein the target residual life information comprises predicted residual safety mileage, predicted abrasion depth and predicted tire burst probability.
2. The big data based tire life prediction method of claim 1, wherein the multi-source operational data includes wear monitoring data, mileage data, road condition data, and load history data; Acquiring multi-source operating data and vehicle state information of a target tire, comprising: Acquiring initial sensing data of a target tire, wherein the initial sensing data comprises tread depths, tire temperatures, running speeds and ground reaction forces of a plurality of monitoring periods; Performing outlier rejection and missing value filling processing on the initial sensing data to obtain multi-source operation data; Acquiring driving behavior information, environmental climate information and maintenance record information of a vehicle to which a target tire belongs; according to a preset driving behavior classification rule, carrying out labeling treatment on driving behavior information to obtain driving style characteristic information of the vehicle; Calculating a temperature and humidity index of the environmental climate information to obtain temperature and humidity characteristic information of the environmental climate information, and summarizing the maintenance record information and the temperature and humidity characteristic information to obtain environmental maintenance characteristic information of the vehicle; And obtaining the vehicle state information based on the driving style characteristic information and the environment maintenance characteristic information.
3. The method for predicting the service life of a tire based on big data according to claim 2, wherein the wear characteristic sequence comprises a wear depth value, a driving range value and a road surface roughness value, the load characteristic sequence comprises a vertical load value, a side load value and a tire pressure fluctuation value, the wear trend sequence comprises a reference wear curve and an actual wear curve, and the load evaluation sequence comprises a working condition load sequence and a structural stress sequence; Constructing a wear trend sequence and a load assessment sequence of the target tire according to the reference tire life database, the wear feature sequence and the load feature sequence, wherein the method comprises the following steps of: according to the time sequence position of the time stamp in the reference tire life database, sequencing and combining the load value under each working condition in the actual load curve and the impact load value of each abnormal high-load working condition to obtain a working condition load sequence; Sequencing and combining the structural stress estimated value under each working condition in the actual load curve and the structural stress peak value of each abnormal high-load working condition to obtain a structural stress sequence; Performing differential calculation on the reference abrasion curve and the actual abrasion curve, determining the abrasion deviation of each time node in the fusion characteristic sequence, generating an abrasion deviation vector of each time node based on the abrasion deviation of each time node, and combining the abrasion deviation vectors of each time node to obtain an abrasion deviation matrix of the fusion characteristic sequence; Calculating the ratio through the reference load curve and the actual load curve, determining the load overrun condition of each working condition in the fusion characteristic sequence, generating the load coefficient vector of each working condition based on the load overrun condition of each working condition, and combining the load coefficient vectors of each working condition to obtain a load coefficient matrix of the fusion characteristic sequence; And splicing the abrasion deviation matrix, the load coefficient matrix and the structural stress sequence to obtain an abrasion characteristic matrix, and determining the abrasion characteristic matrix as an abrasion load characteristic.
4. A tire life prediction method based on big data as in claim 3, wherein generating a wear deviation vector for each time node based on the wear deviation for each time node comprises: For any time node, constructing an initial abrasion vector of the time node, wherein the initial abrasion vector comprises normal abrasion zone bits, abnormal abrasion zone bits, a plurality of grades of mild abrasion identification bits and a plurality of grades of severe abrasion identification bits; if the abrasion deviation of the time node is larger than a preset threshold value, setting the numerical value of the abnormal abrasion zone position in the initial abrasion vector and the numerical value of the heavy abrasion mark position matched with the abrasion deviation grade as a first preset value, setting the numerical values of other mark positions in the initial abrasion vector as a second preset value, and generating an abrasion deviation vector of the time node; If the abrasion deviation of the time node is smaller than or equal to a preset threshold value, setting the numerical value of the normal abrasion zone bit in the initial abrasion vector and the numerical value of the light abrasion mark bit matched with the abrasion deviation grade as a first preset value, setting the numerical values of other mark bits in the initial abrasion vector as a second preset value, and generating the abrasion deviation vector of the time node.
5. The method of claim 4, wherein generating the load factor vector for each condition based on the overrun condition of the load for each condition comprises: for any working condition, constructing an initial load vector of the working condition, wherein the initial load vector comprises a plurality of candidate load grade identification bits; And determining a target load grade identification bit corresponding to the actual load of the working condition from the plurality of candidate load grade identification bits, setting the numerical value of the target load grade identification bit as a first preset value, and setting the numerical values of other load grade identification bits in the initial load vector as a second preset value to obtain a load coefficient vector of the working condition.
6. The method for predicting the service life of a tire based on big data according to claim 5, wherein the acquiring the attention weight feature of each associated monitoring point and the associated feature corresponding to the target tire through the attention mechanism based on the fusion feature sequence and the sample feature comprises: acquiring a first historical monitoring data sequence from the plurality of historical monitoring data sequences, wherein the type of monitoring data included in the first historical monitoring data sequence is a preset type; Performing target attention mechanism processing on the tire identifiers of the target tires and the monitoring data of the associated monitoring points contained in the first historical monitoring data sequence respectively to obtain attention weight characteristics of the target tires corresponding to different associated monitoring points; acquiring monitoring data of a current associated monitoring point and monitoring data of other associated monitoring points from a first historical monitoring data sequence, wherein the current associated monitoring point is any associated monitoring point, and the other associated monitoring points are associated monitoring points except the current associated monitoring point; Constructing a similarity matrix according to cosine distances between the monitoring data of the current associated monitoring point and the monitoring data of all the other associated monitoring points; and obtaining the corresponding association characteristic of the target tire according to the maximum matrix element in the similarity matrix and the pre-constructed normal distribution.
7. The method for predicting the life of a tire based on big data as set forth in claim 6, wherein the life prediction model includes a feature encoding layer, a feature fusion layer, and a life decoding layer; Extracting running state features of the vehicle state information based on the mask features to obtain multi-dimensional state features, performing regression processing on the multi-dimensional state features, and outputting target remaining life information of the target tire, wherein the method comprises the following steps: Extracting wear load characteristics of the fusion characteristic sequence based on the characteristic coding layer, and extracting running state characteristics of vehicle state information; Splicing the wear load characteristics and the running state characteristics based on the characteristic fusion layer, and performing nonlinear transformation processing on the spliced characteristics to obtain multidimensional state characteristics; And carrying out regression processing on the multidimensional state characteristics based on the life decoding layer, and outputting target residual life information of the target tire.
8. The method for predicting the service life of a tire based on big data as recited in claim 7, wherein the feature code layer comprises a sequence wear code unit, and the fusion feature sequence comprises a reference wear curve, an actual wear curve, a reference load curve and an actual load curve; extracting wear load features of the fused feature sequence based on the feature coding layer, comprising: performing differential calculation on the reference wear curve and the actual wear curve through a sequence wear coding unit, determining the wear deviation of each time node in the fusion feature sequence, generating a wear deviation vector of each time node based on the wear deviation of each time node, and combining the wear deviation vectors of each time node to obtain a wear deviation matrix of the fusion feature sequence; the method comprises the steps of performing ratio calculation on a reference load curve and an actual load curve through a sequence abrasion coding unit, determining the load overrun condition of each working condition in a fusion characteristic sequence, generating load coefficient vectors of each working condition based on the load overrun condition of each working condition, and combining the load coefficient vectors of each working condition to obtain a load coefficient matrix of the fusion characteristic sequence; And splicing the abrasion deviation matrix, the load coefficient matrix and the structural stress sequence to obtain an abrasion characteristic matrix, and determining the abrasion characteristic matrix as an abrasion load characteristic.
9. The tire life prediction method based on big data according to claim 8, wherein the feature encoding layer includes a text encoding unit and a numerical encoding unit; Extracting an operating state characteristic of vehicle state information, comprising: extracting driving style characteristics in the vehicle state information based on the text coding unit, and extracting environment temperature and humidity characteristics in the vehicle state information based on the numerical coding unit; determining running state characteristics according to driving style characteristics and environment temperature and humidity characteristics; the service life decoding layer comprises a full-connection layer and an output layer; regression processing is carried out on the multidimensional state characteristics based on the life decoding layer, and target residual life information of the target tire is output, wherein the regression processing comprises the following steps: based on the full-connection layer, carrying out dimension compression on the multidimensional state characteristics to obtain a one-dimensional life characteristic vector of the target tire; And based on the output layer, performing activation function mapping on the one-dimensional life characteristic vector, predicting the residual life probability distribution of the target tire at each time point in the future, and determining the time point corresponding to the maximum value of the residual life probability as the target residual life information of the target tire.
10. A big data based tire life prediction system adapted to a big data based tire life prediction method according to any one of claims 1 to 9, comprising: The data acquisition module is used for acquiring multisource operation data and vehicle state information of the target tire, and a reference tire service life database and sample characteristics corresponding to the target tire; The characteristic extraction module is used for cleaning and extracting the characteristics of the multi-source operation data to obtain a wear characteristic sequence and a load characteristic sequence of the target tire, and constructing a wear trend sequence and a load evaluation sequence of the target tire according to the reference tire life database, the wear characteristic sequence and the load characteristic sequence; The feature fusion module is used for comparing the abrasion trend sequence with the load evaluation sequence according to the time sequence position of the time stamp in the reference tire life database to obtain a fusion feature sequence, and acquiring the attention weight feature of each associated monitoring point and the associated feature corresponding to the target tire through an attention mechanism based on the fusion feature sequence and the sample feature; The device comprises an average pooling module, a pooling module and a storage module, wherein the average pooling module is used for carrying out global average pooling on a second historical monitoring data sequence in a plurality of historical monitoring data sequences to obtain pooling characteristics, wherein the second historical monitoring data sequence is a historical monitoring data sequence except for a first historical monitoring data sequence; the feature splicing module is used for carrying out splicing treatment on the pooled features, the attention weight features, the associated features and the fusion feature sequences to obtain spliced features, and inputting the spliced features into a mask network of the life prediction model to obtain mask features; The life prediction module is used for extracting running state characteristics of the vehicle state information based on the mask characteristics to obtain multi-dimensional state characteristics, carrying out regression processing on the multi-dimensional state characteristics, and outputting target residual life information of the target tire, wherein the target residual life information comprises predicted residual safety mileage, predicted abrasion depth and predicted tire burst probability.

Description

Tire service life prediction method and system based on big data Technical Field The invention relates to the technical field of tire life prediction, in particular to a tire life prediction method and system based on big data. Background At present, the traditional methods are usually based on static models to predict, and lack of adaptability to real-time changes, which means that the traditional methods are difficult to effectively influence different working conditions, environmental changes and load fluctuation, prediction results may be inaccurate, actual use states of tires cannot be reflected in real time, the traditional methods only depend on a single data source to predict service lives, other potential influence factors are ignored, and therefore prediction accuracy of the traditional methods is relatively low, and comprehensive states of the tires cannot be comprehensively reflected. In addition, the conventional method generally adopts a unified life prediction standard, and individual prediction cannot be provided for each tire according to specific use conditions, so that the prediction result is relatively general, specific requirements of different users and different vehicles cannot be met, potential safety hazards of the tires, such as excessive wear or overrun of bearing capacity, cannot be found in time, problems are found when the tires reach a critical point by the conventional method, and the risk of safety accidents is increased. Disclosure of Invention In order to achieve the above purpose, the invention provides a tire life prediction method based on big data, comprising the following steps: Acquiring multisource operation data and vehicle state information of a target tire, and a reference tire life database and sample characteristics corresponding to the target tire; cleaning and extracting characteristics of the multi-source operation data to obtain a wear characteristic sequence and a load characteristic sequence of the target tire, and constructing a wear trend sequence and a load evaluation sequence of the target tire according to a reference tire life database, the wear characteristic sequence and the load characteristic sequence; according to the time sequence position of the time stamp in the reference tire life database, comparing the abrasion trend sequence with the load evaluation sequence to obtain a fusion characteristic sequence, and acquiring the attention weight characteristic of each associated monitoring point and the associated characteristic corresponding to the target tire through an attention mechanism based on the fusion characteristic sequence and the sample characteristic; carrying out global average pooling on a second historical monitoring data sequence in the plurality of historical monitoring data sequences to obtain pooling characteristics, wherein the second historical monitoring data sequence is a historical monitoring data sequence except the first historical monitoring data sequence; the pooled feature, the attention weight feature, the associated feature and the fusion feature sequence are subjected to splicing treatment to obtain a spliced feature, and the spliced feature is input into a mask network of a life prediction model to obtain a mask feature; And extracting running state characteristics of the vehicle state information based on the mask characteristics to obtain multi-dimensional state characteristics, carrying out regression processing on the multi-dimensional state characteristics, and outputting target residual life information of the target tire, wherein the target residual life information comprises predicted residual safety mileage, predicted abrasion depth and predicted tire burst probability. Preferably, the multi-source operating data includes wear monitoring data, mileage data, road condition data, and load history data; Acquiring multi-source operating data and vehicle state information of a target tire, comprising: Acquiring initial sensing data of a target tire, wherein the initial sensing data comprises tread depths, tire temperatures, running speeds and ground reaction forces of a plurality of monitoring periods; Performing outlier rejection and missing value filling processing on the initial sensing data to obtain multi-source operation data; Acquiring driving behavior information, environmental climate information and maintenance record information of a vehicle to which a target tire belongs; according to a preset driving behavior classification rule, carrying out labeling treatment on driving behavior information to obtain driving style characteristic information of the vehicle; Calculating a temperature and humidity index of the environmental climate information to obtain temperature and humidity characteristic information of the environmental climate information, and summarizing the maintenance record information and the temperature and humidity characteristic information to obtain environmental maintenance charac