Tire selection method and device for sanitation operation vehicles

Abstract

The present invention provides a tire selection method and device for a sanitation operation vehicle, and relates to the technical field of tire selection for sanitation operation vehicles, including obtaining historical tire wear characteristics and tire design parameters of a vehicle consistent with a current vehicle model during sanitation operation, and determining a multi-source data pair of tires of the current vehicle; processing the multi-source data pair of tires according to a principal component analysis algorithm and a supervised learning algorithm, determining at least one key feature affecting each target performance of the tire of the current vehicle, and then constructing a scoring function corresponding to each target performance; based on a random forest model, determining the importance of each target performance of the tire of the current vehicle to the implementation effect of the sanitation operation, and then setting weights, performing weighted summation on the scoring functions corresponding to each target performance, constructing a multi-objective function of the tire of the current vehicle, and determining the tire selection of the current vehicle; so as to alleviate the technical problem in the prior art that wheel selection is time-consuming and labor-intensive but has a high error rate.

Description

Technical Field

[0001] The invention relates to the technical field of tire selection for sanitation operation vehicles, and in particular to a tire selection method and device for sanitation operation vehicles. Background Art

[0002] The tire selection for sanitation vehicles has a significant impact on vehicle performance, operating efficiency and operating costs. Traditional selection methods mainly rely on experience and lack data support, making it difficult to meet the needs of different operating environments. In addition, there are many types of tires on the market with different performance parameters, and manual selection is time-consuming, labor-intensive and prone to errors. Summary of the invention

[0003] The purpose of the present invention is to provide a tire selection method and device for sanitation operation vehicles, so as to alleviate the technical problem in the prior art that wheel selection is time-consuming and labor-intensive but has a high error rate.

[0004] In a first aspect, an embodiment of the present invention provides a tire selection method for a sanitation vehicle, comprising:

[0005] Obtain historical tire wear characteristics of a vehicle consistent with the vehicle model currently to be selected during sanitation operations and tire design parameters collected by a vehicle factory interface, and determine a tire multi-source data pair for the vehicle currently to be selected;

[0006] Processing the tire multi-source data pair according to a principal component analysis algorithm and a supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected;

[0007] Using at least one key feature corresponding to each of the target performances, constructing a scoring function corresponding to each of the target performances, wherein the scoring function is used to describe a target performance of the tire of the current vehicle to be selected;

[0008] Determine the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model;

[0009] Setting a weight according to the importance of each target performance, performing weighted summation on the scoring function corresponding to each target performance, and constructing a multi-objective function of the tire of the current vehicle to be selected;

[0010] The multi-objective function is calculated to determine the tire type of the current vehicle to be selected.

[0011] Furthermore, the step of obtaining historical tire wear characteristics of a vehicle consistent with the vehicle model of the current vehicle to be selected during sanitation operations and tire design parameters collected by the vehicle factory interface to determine the tire multi-source data pair of the current vehicle to be selected includes:

[0012] Obtain the vehicle model of the vehicle to be selected, as well as the business type and target street environment in the sanitation work;

[0013] Based on big data, first historical tire wear characteristics identified by a deep learning algorithm for each sanitation operation vehicle corresponding to the vehicle model are obtained; wherein the historical tire wear characteristics are used to characterize the business preference problems existing in the vehicle model in the sanitation operation application;

[0014] The business preference question is screened according to the business type and the target street environment, and a second historical tire wear feature corresponding to the business preference question that is consistent with the current vehicle to be selected is determined;

[0015] Acquire tire design parameters corresponding to the vehicle model in real time based on the vehicle manufacturer interface;

[0016] Based on the second historical tire wear characteristics and the tire design parameters, a tire multi-source data pair of the current vehicle to be selected is constructed.

[0017] Furthermore, the step of processing the tire multi-source data pair according to the principal component analysis algorithm and the supervised learning algorithm to determine at least one key feature affecting each target performance of the tire of the current vehicle to be selected includes:

[0018] Constructing a covariance matrix between the tire multi-source data pairs;

[0019] Based on the eigenvalue sorting of the covariance matrix, reducing the principal components in the tire multi-source data pair into eigenvectors;

[0020] The feature vector is input into a supervised learning algorithm model constructed based on each target performance, and at least one key feature corresponding to each target performance is output.

[0021] Furthermore, the key features are enhanced by the following formula:

[0022]

[0023] in, is the original key feature of the output, is the enhanced feature vector, is the indicator function, when each feature in the tire multi-source data pair Belongs to the sanitation feature set The value is 1 when , otherwise it is 0. is the PCA projection matrix, The sanitation set includes operation frequency and garbage density characteristics, and α is the domain knowledge reinforcement coefficient.

[0024] Further, the target performance includes a first target performance, and the first target performance includes wear resistance, cut resistance, heat resistance and grip; using at least one key feature corresponding to each of the target performances, the step of constructing a scoring function corresponding to each of the target performances includes:

[0025] The average wear rate in the key feature is converted into a scoring value to determine a scoring function for evaluating the wear resistance; or, the theoretical wear rate is calculated by the average wear rate in the key feature, and the actual working parameters of the vehicle consistent with the vehicle model currently to be selected in the sanitation operation are collected to calculate the wear resistance scoring function as follows:

[0026]

[0027] in, , , The coefficient is calibrated through the sanitation operation log; is the ratio of the actual wear rate of the sanitation vehicle to the theoretical wear rate, T is the actual and theoretical operating temperature, and the actual maximum operating power of the sanitation vehicle and rated power ;

[0028] Converting the tread hardness and the frequency of cut damage in the key features into score values, and determining a score function for evaluating the cut resistance;

[0029] Converting the tire operating temperature and heat dissipation efficiency in the key features into scoring values, and determining a scoring function for evaluating the heat resistance;

[0030] The grip value and pattern design complexity in the key features are converted into scoring values, and a scoring function for evaluating the grip is determined.

[0031] Further, the target performance includes a second target performance, and the second target performance includes cost performance; and the step of constructing a scoring function corresponding to each of the target performance using at least one key feature corresponding to each of the target performances further includes:

[0032] Based on the price in the key feature and the score value of the first target performance, a score function for evaluating the cost performance is determined.

[0033] Further, based on the scoring function corresponding to each target performance and the pre-trained random forest model, the step of determining the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work includes:

[0034] Input the calculation results of the scoring function corresponding to each of the target performances into a pre-trained random forest model, and output a benchmark value of the implementation effect of the sanitation operation; wherein the pre-trained random forest model is used to predict the implementation effect of the tire of the current vehicle to be selected in the sanitation operation;

[0035] Repeat the following process until each of the target performances has been traversed: select the current feature sample from the target performance; disrupt the calculation results of the scoring function corresponding to the current feature sample at different times, while keeping the calculation results of the scoring function corresponding to each of the remaining target performances unchanged; input the calculation results of the scoring function corresponding to the current feature sample and the calculation results of the scoring function corresponding to each of the remaining target performances into the previously trained random forest model, and output the work implementation effect value of the sanitation operation corresponding to the current feature sample;

[0036] The work implementation effect value corresponding to each of the current feature samples and the work implementation effect benchmark value are calculated to determine the importance of each of the current feature samples; wherein the difference is proportional to the importance.

[0037] Furthermore, before the step of processing the tire multi-source data pair according to the principal component analysis algorithm and the supervised learning algorithm to determine at least one key feature affecting each target performance of the tire of the current vehicle to be selected, the method further includes:

[0038] Isolating abnormal data points according to a detection model for characterizing tire characteristics of the vehicle model in sanitation operations, and removing abnormal wear characteristics and abnormal design parameter data from the tire multi-source data pair;

[0039] Based on the parameter data and simulation data on both sides of the missing data and the eliminated data in the tire multi-source data pair, the tire multi-source data pair is multi-interpolated; wherein the simulation data is data obtained by simulating the tire characteristics of the vehicle of the vehicle model in sanitation operations;

[0040] The tire multi-source data after interpolation are standardized and mapped to the same measurement unit dimension.

[0041] In a second aspect, an embodiment of the present invention further provides a tire selection device for a sanitation vehicle, comprising:

[0042] The first determination module obtains the historical tire wear characteristics of a vehicle consistent with the vehicle model of the current vehicle to be selected during sanitation operations and the tire design parameters collected by the vehicle factory interface, and determines the tire multi-source data pair of the current vehicle to be selected;

[0043] A second determination module processes the tire multi-source data pair according to a principal component analysis algorithm and a supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected;

[0044] A first construction module constructs a scoring function corresponding to each target performance by using at least one key feature corresponding to each target performance, wherein the scoring function is used to describe a target performance of the tire of the current vehicle to be selected;

[0045] A third determination module determines the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model;

[0046] The second construction module sets a weight according to the importance of each target performance, performs weighted summation on the scoring function corresponding to each target performance, and constructs a multi-objective function of the tire of the current vehicle to be selected;

[0047] The selection module calculates the multi-objective function and determines the tire type of the current vehicle to be selected.

[0048] In a third aspect, an embodiment provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any of the aforementioned implementation methods are implemented.

[0049] The embodiment of the present invention provides a tire selection method and device for sanitation operation vehicles, which integrates the historical tire wear characteristics of the vehicle model currently performing sanitation operations and the tire design parameters collected from the vehicle factory interface to determine a tire multi-source data pair; uses a principal component analysis algorithm to reduce the dimension of the tire multi-source data pair, and then extracts each key feature that has a greater impact on each target performance of the tire from the tire multi-source data pair based on a supervised learning algorithm; constructs a scoring function for each target performance based on these key features to describe each target performance of the tire; then inputs the calculation result of the scoring function of each target performance into a pre-trained random forest model, outputs the importance of each target performance, and sets the corresponding weight of each target performance; performs a weighted sum calculation based on the scoring function of each target performance and its corresponding weight to form a multi-objective function of the tire of the current vehicle to be selected, calculates this multi-objective function, and obtains the tire selection of the current vehicle to be selected, which can realize the tire selection of sanitation operation vehicles from many influencing factors in a relatively simple and accurate manner.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A flow chart of a tire selection method for a sanitation vehicle provided in an embodiment of the present invention;

[0054] Figure 2 A schematic diagram of functional modules of a tire selection device for a sanitation vehicle provided by an embodiment of the present invention;

[0055] Figure 3 A schematic diagram of the hardware architecture of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Based on this, a tire selection method and device for a sanitation vehicle provided in an embodiment of the present invention can relatively easily realize accurate tire selection for a sanitation vehicle.

[0058] To facilitate understanding of this embodiment, a tire selection method for a sanitation vehicle disclosed in an embodiment of the present invention is first introduced in detail. This method can be applied to intelligent control devices such as a host computer, a server, and a controller.

[0059] Figure 1 A flow chart of a tire selection method for a sanitation vehicle provided in an embodiment of the present invention.

[0060] Reference Figure 1 The method can be implemented by the following steps, including:

[0061] Step S102, obtaining historical tire wear characteristics of vehicles consistent with the vehicle model currently to be selected during sanitation operations and tire design parameters collected by the vehicle factory interface, and determining a tire multi-source data pair for the current vehicle to be selected.

[0062] It is understandable that the tire models equipped on vehicles of the same model may be interoperable. Here, the historical tire wear characteristics of such vehicles are determined based on the current vehicle model, and the design parameters of the wheels configured for such vehicles are obtained from the vehicle manufacturer to jointly determine the multi-source tire data pair of the current vehicle to be selected; wherein, the historical tire wear characteristics can be obtained through wheel images captured by the camera of historical vehicles of the current vehicle model.

[0063] It should be noted that, as an optional embodiment, tire wear characteristics can also be collected through the following hardware structure, including:

[0064] In the sanitation operation scene, corrosion-resistant packaging is used to collect tire pressure / temperature sensors that are suitable for sewage and chemical corrosion environments; the three-dimensional wear scanner captures the uneven wear caused by garbage squeezing through high-frequency sampling (once every 10 minutes); the on-board CAN bus interface collects sanitation operation parameters in real time (number of lifts, compression device load); the edge computing layer has an embedded GPU processor and a lightweight PCA model to support real-time feature extraction on site (power consumption <15W); the cloud analysis layer sets up a distributed storage cluster to store all sanitation fleet data and establish a regional feature library (such as northern snow melting agent corrosion data).

[0065] Step S104 , processing the tire multi-source data pairs according to the principal component analysis algorithm and the supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected.

[0066] Here, the principal component analysis algorithm is an unsupervised learning algorithm. Combining it with the supervised learning algorithm can reduce the dimension of multi-source tire data pairs and extract the key features that have a greater impact on the target performance of the tire.

[0067] Step S106: construct a scoring function corresponding to each target performance using at least one key feature corresponding to each target performance.

[0068] Among them, the features that have a greater impact on tire performance are extracted from multi-source data, called key features, and then such key features are used to construct a scoring function so that the scoring function can be used to describe a target performance of the tire of the current vehicle to be selected; reasonable mapping rules are designed according to the characteristics of each dimension (target performance) to convert the tire multi-source data pairs into scoring values ​​between 0 and 100.

[0069] Step S108, based on the scoring function corresponding to each target performance and the pre-trained random forest model, determine the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work.

[0070] Here, the implementation effect of the sanitation work based on the trained random forest model is used as an indicator to determine the importance of each target performance, and this importance is used as the weight coefficient of each target performance in the subsequent multi-objective function.

[0071] Step S110, setting a weight according to the importance of each target performance, performing weighted summation on the scoring function corresponding to each target performance, and constructing a multi-objective function of the tire of the current vehicle to be selected.

[0072] like is the weight corresponding to each target performance, is the score value of each objective performance, then the multi-objective function is:

[0073]

[0074] Step S112, calculating the multi-objective function to determine the tire type of the current vehicle to be selected.

[0075] Here, the multi-objective function is set to ensure that each dimension of the target performance can fairly and reasonably reflect the actual performance of the tire, and can intuitively know what performance the tire selection of the current vehicle to be selected should have, thereby achieving accurate selection.

[0076] In a preferred embodiment of practical application, the historical tire wear characteristics of the vehicle model currently performing sanitation operations and the tire design parameters collected from the vehicle factory interface are integrated to determine a multi-source data pair of tires; the multi-source data pair of tires is reduced in dimension using a principal component analysis algorithm, and then each key feature that has a greater impact on each target performance of the tire is extracted from the multi-source data pair of tires based on a supervised learning algorithm; a scoring function for each target performance is constructed based on these key features to describe each target performance of the tire; the calculation result of the scoring function for each target performance is then input into a pre-trained random forest model, and the importance of each target performance is output, so as to set the corresponding weight of each target performance; a weighted sum calculation is performed based on the scoring function of each target performance and its corresponding weight to form a multi-objective function of the tire of the current vehicle to be selected, and this multi-objective function is calculated to obtain the tire selection of the current vehicle to be selected, so that the tire selection of the sanitation operation vehicle can be realized relatively simply and accurately from many influencing factors.

[0077] In some embodiments, the step of integrating the tire multi-source data pair of the current vehicle model in step S102 may include:

[0078] Step 1.1), obtain the vehicle model of the current vehicle to be selected, as well as the business type and target street environment in the sanitation work.

[0079] Here, based on this vehicle model, the tire wear characteristics of the same vehicle model are obtained from big data.

[0080] In step 1.2), based on big data, the first historical tire wear characteristics of each sanitation vehicle corresponding to the vehicle model are obtained based on the deep learning algorithm.

[0081] In actual applications, each sanitation vehicle is equipped with a high-definition camera to regularly collect tire surface images. The images are processed using deep learning algorithms (such as convolutional neural networks (CNN)) to identify wear characteristics such as tire tread depth and uneven wear. The collected wear characteristics are transmitted to the central database in real time using the Internet of Things technology, so that the first historical tire wear characteristics can be obtained from big data based on the same vehicle model and other conditions.

[0082] Among them, historical tire wear characteristics are used to characterize the business preference problems of vehicle models in sanitation operation applications. Uneven wear includes thick sides and thin middle: if the edges of the tire are thicker than the middle part, this may be caused by oversteering, insufficient tire inflation or suspension system problems. Generally speaking, if the thickness difference exceeds a certain proportion (such as 20%), it can be regarded as uneven wear. Uneven wear also includes the thickness of a specific pattern not meeting the requirements: each tire has its designed minimum tread depth. When the tread depth in a certain area is lower than the minimum value, it is considered that there is a wear problem in that area. For example, for some tires, if the tread depth is less than 1.6 mm, it is considered to be unsuitable for continued use. Uniform wear also includes other abnormal wear patterns: including but not limited to single-sided wear, center wear, edge wear, wavy wear or patchy wear, etc., all of which indicate that the tire is subjected to abnormal pressure distribution during operation or there is a problem with the vehicle.

[0083] Step 1.3), the business preference questions are screened according to the business type and the target street environment, and the second historical tire wear characteristics corresponding to the business preference questions that are consistent with the current vehicle to be selected are determined.

[0084] As an optional embodiment, the business type and target street environment in sanitation work can also be obtained, so as to obtain the second historical tire wear characteristics corresponding to the vehicles with the same vehicle model as the current selected vehicle, the same business type in sanitation work, and operating in the same target street environment based on big data, so as to determine a more accurate tire multi-source data pair.

[0085] Step 1.4), based on the car manufacturer interface, obtain the tire design parameters corresponding to the vehicle model in real time.

[0086] Establish a tire manufacturer data interface to regularly obtain the latest tire performance parameters, including but not limited to: tread formula, tread hardness, pattern design, load index, speed level, etc.

[0087] Step 1.5), based on the second historical tire wear characteristics and tire design parameters, a tire multi-source data pair for the current vehicle to be selected is constructed.

[0088] Here, it can be understood that the tire data from two sources are integrated into a tire multi-source data pair that constitutes the vehicle to be selected.

[0089] Furthermore, in order to ensure the processing accuracy of subsequent steps, the measurement units of multi-source data are unified before step S104, specifically including:

[0090] In step 2.1), the abnormal data points are isolated based on the detection model for characterizing the tire characteristics of the vehicle model in sanitation operations, and the abnormal wear characteristics and abnormal design parameter data in the tire multi-source data pair are eliminated.

[0091] Among them, outlier detection algorithms (such as IsolationForest) can be used to identify and process abnormal data points. By building a model that can effectively isolate outliers, these outliers can be marked for further review or exclusion. Outlier detection algorithms (such as IsolationForest) are used to identify various types of abnormal data; for tire wear data and performance parameters, abnormal data refers to data points that do not conform to normal ranges or trends. They may be caused by measurement errors, transmission errors, use under extreme conditions, or other atypical factors. The following are several types of data that may be considered abnormal:

[0092] 1. Wear anomalies: If the wear depth at one tire location is significantly higher or lower than at other locations and there is no reasonable explanation (such as uneven vehicle load distribution), then this data point can be considered an anomaly. For example, if most of the time the center of the tire is thinner than the edge, but a certain measurement shows the opposite result, this may be an anomaly.

[0093] 2. Selection anomaly: This refers to data anomalies found during the tire selection process, not data directly obtained from sensors. For example, based on historical data analysis, a certain type of tire should perform better in a specific environment, but if the same type of tire actually used shows poor performance, the relevant data may be considered abnormal.

[0094] 3. Parameter anomalies: For tire performance parameters, such as tread hardness, pattern design, etc., if the product parameters of a batch of products deviate significantly from the standard values ​​provided by the manufacturer, these data are also considered abnormal. For example, if the tread hardness of a batch of tires is generally high or low, beyond the reasonable fluctuation range, then these data are abnormal.

[0095] In step 2.2), multiple interpolation is performed on the tire multi-source data pair based on the parameter data and simulated data on both sides of the missing data and the eliminated data in the tire multi-source data pair.

[0096] Multiple imputation is used to handle missing data and removed data. This method not only takes into account the existing observations (parameter data on both sides of the removed data), but also reflects uncertainty by simulating multiple possible filling values ​​(simulated data), thereby improving the overall quality of the data set. Among them, the simulated data is the data obtained by simulating the tire characteristics of vehicles in sanitation operations based on vehicle models; the goal of the data cleaning stage is to ensure that the data used for subsequent analysis is as accurate and complete as possible so that valuable information can be extracted from it to support tire selection decisions.

[0097] Missing Data refers to data that should exist but is not actually recorded. For tire wear and performance parameters, missing data may include but is not limited to:

[0098] Missing wear data: The degree of wear at certain points in time or locations is not correctly recorded. For example, a camera malfunction prevents the capture of tire images within a specific time period, or image processing fails to identify tread depth information.

[0099] Missing performance parameters: The data interface from the tire manufacturer may occasionally lose updates, resulting in the latest performance parameters of some tire models not being entered into the system in a timely manner.

[0100] In step 2.3), the tire multi-source data after the interpolation operation is standardized and mapped to the same measurement unit dimension.

[0101] Due to these wear characteristics, the performance parameter data come from different systems and measurement tools, and their original values ​​may be on very different scales, so they need to be mapped into a unified interval through standardization for subsequent comparison and analysis.

[0102] Here, multi-source data from different sources and units are standardized and mapped to a unified interval using the Z-score or Min-Max method to facilitate subsequent comparative analysis.

[0103] Z-score standardization: For each feature, calculate its mean and standard deviation, and then convert the original data into a form with a multiple of the standard deviation relative to the mean value of the feature. This method is suitable for cases where the data is roughly normally distributed.

[0104] Min-Max normalization: linearly transform the original data to a fixed range (such as [0, 1]) to ensure that all features have the same value range. This helps to eliminate the magnitude differences between different features and is particularly suitable for data sets with known minimum and maximum values.

[0105] By standardizing data from these different sources and units, information from various channels can be evaluated fairly on the same platform, thereby improving the effectiveness of model training and the quality of final decision-making.

[0106] Based on the above embodiment, the above step S104 extracts at least one key feature of the tire of the current vehicle to be selected that has a greater impact on each target performance, which can be achieved through the following steps, specifically including:

[0107] Step 3.1), construct the covariance matrix between tire multi-source data pairs.

[0108] The covariance matrix is ​​an n×n square matrix (n is the number of tire features from each source in the tire multi-source data pair), which describes the linear relationship between each pair of features (historical tire wear characteristics and tire design parameters) in the original data set (tire multi-source data pair).

[0109] For two features, historical tire wear characteristics and tire design parameters X and Y, their covariance is defined as:

[0110]

[0111] Where E represents the expected value or mean. If feature X calculates covariance with itself, the variance of the feature is obtained, that is, .

[0112] The eigenvalues ​​and corresponding eigenvectors of the above covariance matrix can be completed by solving the following equations:

[0113]

[0114] Where C is the covariance matrix, λ is the unknown eigenvalue, and I is the identity matrix.

[0115] In step 3.2), based on the eigenvalue sorting of the covariance matrix, the principal components in the tire multi-source data pairs are reduced to eigenvectors.

[0116] Arrange all the eigenvalues ​​obtained from large to small, while keeping their one-to-one correspondence with their respective eigenvectors (each feature in the tire multi-source data pair). Select the most important principal components according to the size of the eigenvalues ​​(usually select the principal components with a cumulative contribution rate of more than 80%). These principal components can capture the maximum variance of the original data, that is, the changes in these principal components have a relatively large impact on the features.

[0117] In step 3.3), the feature vector is input into a supervised learning algorithm model built based on each target performance, and at least one key feature corresponding to each target performance is output.

[0118] Here, supervised learning algorithms include linear regression, logistic regression, support vector machine, etc.; the feature vector after dimension reduction is used as input, and the target performance of the tire is used as the target output;

[0119] If you are using a linear model (such as linear regression), you can understand the degree of influence of each principal component on the target performance by examining the coefficient size. For nonlinear models, you can use features such as feature importance scores, partial dependence plots, etc. to interpret the model.

[0120] As an optional embodiment, each key feature corresponding to each target performance can be enhanced in combination with the sanitation operation scenario to make the tire selection in the subsequent steps more accurate. Specifically, the enhancement can be performed through the following formula:

[0121]

[0122] in, is the original key feature output from step 3.3), is the enhanced feature vector, is the indicator function, when each feature in the tire multi-source data pair Belongs to the sanitation feature set The value is 1 when , otherwise it is 0. is the PCA projection matrix, It is a combination of sanitation characteristics including operation frequency and garbage density characteristics, and α is the domain knowledge reinforcement coefficient (empirical value 0.3-0.5).

[0123] In some embodiments, the target performance includes a first target performance and a second target performance; the first target performance includes wear resistance, cut resistance, heat resistance and grip, and the second target performance includes cost performance; the second target performance is a composite feature formed by combining at least one key feature and the first target performance scoring function; for step S106, the scoring functions of the first target performance and the second target performance can be calculated respectively, which not only simplifies the evaluation process, but also provides an intuitive method to compare the relative values ​​of different tires, specifically including:

[0124] In step 4.1), the average wear rate in the key feature is converted into a scoring value, and the scoring function for evaluating wear resistance is determined:

[0125]

[0126] in, is the average wear rate (e.g. mm / km), and are the minimum and maximum average wear rates of all tire samples, respectively. This formula assigns higher scores to tires with lower wear rates through linear transformation, because lower wear rates mean better wear resistance.

[0127] As an optional implementation, the theoretical wear rate can also be calculated by the average wear rate in the key features, and the actual working parameters of the sanitation vehicle are collected to calculate the wear resistance score function as follows:

[0128]

[0129] in, , , The coefficient is calibrated through the sanitation operation log; is the ratio of the actual wear rate of the sanitation vehicle to the theoretical wear rate, T is the actual and theoretical operating temperature, and the actual maximum operating power of the sanitation vehicle and rated power .

[0130] In step 4.2), the tread hardness and cut damage frequency among the key characteristics are converted into scoring values, and a scoring function for evaluating cut resistance is determined.

[0131]

[0132] Where H represents the tread hardness and h is the actual cutting damage frequency. and are the maximum and minimum values ​​of hardness; and are the maximum and minimum values ​​of the cut damage frequency. The first term in the formula is scored based on hardness, and the second term in the formula is based on the cut damage frequency, with the two components weighted differently to reflect their impact on overall cut resistance.

[0133] In step 4.3), the tire operating temperature and heat dissipation efficiency among the key characteristics are converted into scoring values, and a scoring function for evaluating heat resistance is determined.

[0134]

[0135] Among them, T is the operating temperature and E is the heat dissipation efficiency. and are the minimum and maximum values ​​within the operating temperature range; and are the minimum and maximum values ​​of heat dissipation efficiency. The first term in the formula takes into account the performance at low temperatures, while the second term in the formula focuses on efficient heat dissipation capabilities.

[0136] In step 4.4), the grip value and pattern design complexity in the key features are converted into scoring values, and the scoring function for evaluating grip is determined.

[0137]

[0138] Among them, G is the grip value, W is the complexity of the pattern design, and are the minimum and maximum values ​​of the grip score; and It is the minimum and maximum value of the pattern design complexity. The grip value directly reflects the performance of the tire under various road conditions, and the pattern design complexity affects the performance on slippery roads.

[0139] In step 4.5), a scoring function for evaluating cost performance is determined based on the price in the key feature and the scoring value of the first target performance.

[0140]

[0141] Among them, P is the price, and V is the comprehensive performance score, such as the weighted average of the calculation results (score values) of the above four dimensions (the first objective performance score function); and are the minimum and maximum values ​​within the price range; and It is the minimum and maximum value of the comprehensive performance score; the first price part in the formula is used for reverse scoring, that is, the lower the price, the higher the score; the second performance score in the formula is used for positive scoring, the better the performance, the higher the score.

[0142] In practical applications, step S108 can determine the impact of each target performance on the implementation effect of the work through the random forest model used to predict the implementation effect of the sanitation work, and then determine the importance of each target performance, so that the multi-objective function in the subsequent steps can be constructed more accurately, specifically including:

[0143] In step 5.1), the calculation results of the scoring function corresponding to each target performance are input into the pre-trained random forest model to output the benchmark value of the implementation effect of the sanitation operation.

[0144] Among them, the pre-trained random forest model is used to predict the working effect of the tires of the current selected vehicle in sanitation operations. The working effect benchmark value of the sanitation operation can be understood as the mean square error (MSE), R² score, etc. used to characterize the working effect of the sanitation operation; for example, when the scoring values ​​of the scoring function corresponding to each target performance are not disrupted, the outputted working effect benchmark value R² score of the sanitation operation is 0.85. Repeat the following steps 5.2) until each target performance is traversed:

[0145] In step 5.2), the current feature sample is selected from the target performance; the calculation results of the scoring function corresponding to the current feature sample at different times are disrupted, while the calculation results of the scoring function corresponding to each remaining target performance remain unchanged; the calculation results of the scoring function corresponding to the current feature sample and the calculation results of the scoring function corresponding to each remaining target performance are input into the previously trained random forest model, and the work implementation effect value of the sanitation operation corresponding to the current feature sample is output.

[0146] Here, if the target performance includes wear resistance, heat resistance, cut resistance, grip and cost performance, and wear resistance is the current feature sample, the tire wear resistance score values ​​at each time and under each working condition are shuffled, and then input into the pre-trained random forest model together with the original sorted score values ​​of heat resistance, cut resistance, grip and cost performance, and the output R² score is 0.65; according to the above process, the R² scores of the work implementation effect values ​​of sanitation operations corresponding to heat resistance, cut resistance, grip and cost performance are determined respectively.

[0147] In step 5.3), the difference between the work implementation effect value corresponding to each current feature sample and the work implementation effect benchmark value is calculated to determine the importance of each current feature sample.

[0148] Assume that the R² scores of the work implementation effect values ​​of sanitation operations corresponding to wear resistance, heat resistance, cut resistance, grip and cost-effectiveness are: 0.65, 0.75, 0.82, 0.84, and 0.78 respectively; compare each R² score with the R² score of the work implementation effect benchmark value of sanitation operations; if the difference is larger, the greater the impact on the work implementation effect of sanitation operations, the more important the target performance is, that is, the difference is directly proportional to the importance.

[0149] As an optional embodiment, step S108 can also calculate the contribution of each feature sample in each target performance through each node in each tree in the random forest model used to predict the implementation effect of sanitation work, and then average all trees to obtain the importance weight of each feature sample in each target performance to determine the impact of each target performance on the implementation effect of the work, as described in the following formula:

[0150]

[0151] in, is the importance weight of the jth feature sample in each target performance, is the number of trees in the random forest model, is the combination of nodes containing the jth feature sample in the tth tree, is the number of samples in node v, is the total sample size, The mean square error reduction of node v, The adjustment weight value corresponding to the job type of sanitation work.

[0152] In an embodiment of the present invention, on this basis, the corresponding weight of each target performance is set according to its importance, that is, the higher the importance, the greater the weight; as an optional embodiment, the weight coefficient corresponding to each target performance can be calculated according to the proportional relationship between the importance.

[0153] The embodiment of the present invention realizes accurate data analysis based on relatively simple steps, and can select appropriate tire models according to driving habits or business conditions of sanitation operations, slow down the wear rate of tires, and thus extend the service life of tires; by accurately identifying the key features that affect each target performance of the tire, it can not only ensure the safety and efficiency of vehicle operation, but also help enterprises make more wise tire selection decisions and maximize economic benefits.

[0154] In some embodiments, Figure 2 As shown, an embodiment of the present invention provides a tire selection device for a sanitation vehicle, comprising:

[0155] The first determination module obtains the historical tire wear characteristics of a vehicle consistent with the vehicle model of the current vehicle to be selected during sanitation operations and the tire design parameters collected by the vehicle factory interface, and determines the tire multi-source data pair of the current vehicle to be selected;

[0156] A second determination module processes the tire multi-source data pair according to a principal component analysis algorithm and a supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected;

[0157] A first construction module constructs a scoring function corresponding to each target performance by using at least one key feature corresponding to each target performance, wherein the scoring function is used to describe a target performance of the tire of the current vehicle to be selected;

[0158] A third determination module determines the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model;

[0159] The second construction module sets a weight according to the importance of each target performance, performs weighted summation on the scoring function corresponding to each target performance, and constructs a multi-objective function of the tire of the current vehicle to be selected;

[0160] The selection module calculates the multi-objective function and determines the tire type of the current vehicle to be selected.

[0161] Furthermore, the first determination module is used to obtain the vehicle model of the current vehicle to be selected, as well as the business type and target street environment in sanitation work; based on big data, obtain the first historical tire wear characteristics of each sanitation vehicle corresponding to the vehicle model based on the deep learning algorithm; wherein the historical tire wear characteristics are used to characterize the business preference problems existing in the vehicle model of the vehicle in the sanitation operation application; screen the business preference problems according to the business type and the target street environment, and determine the second historical tire wear characteristics corresponding to the business preference problems that are consistent with the current vehicle to be selected; obtain the tire design parameters corresponding to the vehicle model in real time based on the vehicle factory interface; based on the second historical tire wear characteristics and the tire design parameters, construct a tire multi-source data pair for the current vehicle to be selected.

[0162] Furthermore, the second determination module is used to construct a covariance matrix between the tire multi-source data pairs; based on the eigenvalue sorting of the covariance matrix, the principal components in the tire multi-source data pairs are reduced into eigenvectors; the eigenvectors are input into a supervised learning algorithm model constructed based on each target performance, and at least one key feature corresponding to each of the target performance is output.

[0163] Furthermore, the key features are enhanced by the following formula:

[0164]

[0165] in, is the original key feature of the output, is the enhanced feature vector, is the indicator function, when each feature in the tire multi-source data pair Belongs to the sanitation feature set The value is 1 when , otherwise it is 0. is the PCA projection matrix, The sanitation set includes operation frequency and garbage density characteristics, and α is the domain knowledge reinforcement coefficient.

[0166] Furthermore, the target performance includes a first target performance, which includes wear resistance, cut resistance, heat resistance and grip; a first building module is used to convert the average wear rate in the key feature into a scoring value, and determine a scoring function for evaluating the wear resistance; convert the tread hardness and cut damage frequency in the key feature into a scoring value, and determine a scoring function for evaluating the cut resistance; convert the tire operating temperature and heat dissipation efficiency in the key feature into a scoring value, and determine a scoring function for evaluating the heat resistance; convert the grip value and pattern design complexity in the key feature into a scoring value, and determine a scoring function for evaluating the grip.

[0167] Furthermore, the target performance includes a second target performance, and the second target performance includes cost performance; the first building module is also used to determine a scoring function for evaluating the cost performance based on the price in the key feature and the scoring value of the first target performance.

[0168] Furthermore, the third determination module inputs the calculation results of the scoring function corresponding to each of the target performances into a pre-trained random forest model, and outputs a benchmark value of the work implementation effect of the sanitation operation; wherein the pre-trained random forest model is used to predict the work implementation effect of the tires of the current vehicle to be selected in the sanitation operation; repeats the following process until each of the target performances is traversed: select a current feature sample from the target performance; scramble the calculation results of the scoring function corresponding to the current feature sample at different times, while keeping the calculation results of the scoring function corresponding to each of the remaining target performances unchanged; input the calculation results of the scoring function corresponding to the current feature sample and the calculation results of the scoring function corresponding to each of the remaining target performances into the pre-trained random forest model, and output the work implementation effect value of the sanitation operation corresponding to the current feature sample; perform difference calculation on the work implementation effect value corresponding to each of the current feature samples and the work implementation effect benchmark value to determine the importance of each of the current feature samples; wherein the difference is proportional to the importance.

[0169] Furthermore, before the step of processing the tire multi-source data pair according to the principal component analysis algorithm and the supervised learning algorithm in the second determination module to determine at least one key feature affecting each target performance of the tire of the current vehicle to be selected, the device is also used to isolate abnormal data points according to a detection model used to characterize the tire characteristics of the vehicle model in sanitation operations, and eliminate abnormal wear characteristics and abnormal design parameter data in the tire multi-source data pair; based on the missing data in the tire multi-source data pair and the parameter data and simulation data on both sides of the eliminated data, the tire multi-source data pair is multiple interpolated; wherein the simulation data is data obtained by simulating the tire characteristics of the vehicle model in sanitation operations; and the tire multi-source data pair after the interpolation operation is standardized and mapped to the same measurement unit dimension.

[0170] An embodiment of the present invention provides an electronic device for implementing an electronic device. In this embodiment, the electronic device may be, but is not limited to, a personal computer (PC), a notebook computer, a monitoring device, a server, or other computer device with analysis and processing capabilities.

[0171] As an exemplary embodiment, see Figure 3 The electronic device 110 includes a communication interface 111, a processor 112, a memory 113 and a bus 114. The processor 112, the communication interface 111 and the memory 113 are connected via the bus 114. The memory 113 is used to store a computer program that supports the processor 112 to execute the method. The processor 112 is configured to execute the program stored in the memory 113.

[0172] The machine-readable storage medium mentioned in this article can be any electronic, magnetic, optical or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium can be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), any type of storage disk (such as CD, DVD, etc.), or similar storage medium, or a combination thereof.

[0173] The non-volatile medium may be a non-volatile memory, a flash memory, a storage drive (such as a hard drive), any type of storage disk (such as a CD, DVD, etc.), or a similar non-volatile storage medium, or a combination thereof.

[0174] It can be understood that the specific operation methods of each functional module in this embodiment can refer to the detailed description of the corresponding steps in the above method embodiment, and will not be repeated here.

[0175] The computer-readable storage medium provided in the embodiment of the present invention stores a computer program. When the computer program code is executed, the method described in any of the above embodiments can be implemented. For specific implementation, please refer to the method embodiment, which will not be described in detail here.

[0176] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0177] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0178] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0179] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.

Claims

1. A tire selection method for a sanitation vehicle, characterized in that: include: Obtain historical tire wear characteristics of a vehicle consistent with the vehicle model currently to be selected during sanitation operations and tire design parameters collected by a vehicle factory interface, and determine a tire multi-source data pair for the vehicle currently to be selected; Processing the tire multi-source data pair according to a principal component analysis algorithm and a supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected; Using at least one key feature corresponding to each of the target performances, constructing a scoring function corresponding to each of the target performances, wherein the scoring function is used to describe a target performance of the tire of the current vehicle to be selected; Determine the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model; Setting a weight according to the importance of each target performance, performing weighted summation on the scoring function corresponding to each target performance, and constructing a multi-objective function of the tire of the current vehicle to be selected; The multi-objective function is calculated to determine the tire type of the current vehicle to be selected.

2. The method according to claim 1, characterized in that The step of obtaining historical tire wear characteristics of a vehicle consistent with the vehicle model of the current vehicle to be selected during sanitation operations and tire design parameters collected by the vehicle factory interface to determine the tire multi-source data pair of the current vehicle to be selected includes: Obtain the vehicle model of the vehicle to be selected, as well as the business type and target street environment in the sanitation work; Based on big data, first historical tire wear characteristics identified by a deep learning algorithm for each sanitation operation vehicle corresponding to the vehicle model are obtained; wherein the historical tire wear characteristics are used to characterize the business preference problems existing in the vehicle model in the sanitation operation application; The business preference question is screened according to the business type and the target street environment, and a second historical tire wear feature corresponding to the business preference question that is consistent with the current vehicle to be selected is determined; Acquire tire design parameters corresponding to the vehicle model in real time based on the vehicle manufacturer interface; Based on the second historical tire wear characteristics and the tire design parameters, a tire multi-source data pair of the current vehicle to be selected is constructed.

3. The method according to claim 1, characterized in that The step of processing the tire multi-source data pair according to the principal component analysis algorithm and the supervised learning algorithm to determine at least one key feature affecting each target performance of the tire of the current vehicle to be selected comprises: Constructing a covariance matrix between the tire multi-source data pairs; Based on the eigenvalue sorting of the covariance matrix, reducing the principal components in the tire multi-source data pair into eigenvectors; The feature vector is input into a supervised learning algorithm model constructed based on each target performance, and at least one key feature corresponding to each target performance is output.

4. The method according to claim 1, characterized in that: The target performance includes a first target performance, wherein the first target performance includes wear resistance, cut resistance, heat resistance and grip; using at least one key feature corresponding to each of the target performances, the step of constructing a scoring function corresponding to each of the target performances comprises: The average wear rate in the key feature is converted into a scoring value to determine a scoring function for evaluating the wear resistance; or, the theoretical wear rate is calculated by the average wear rate in the key feature, and the actual working parameters of the vehicle consistent with the vehicle model currently to be selected in the sanitation operation are collected to calculate the wear resistance scoring function as follows: in, , , The coefficient is calibrated through the sanitation operation log; is the ratio of the actual wear rate of the sanitation vehicle to the theoretical wear rate, T is the actual and theoretical operating temperature, and the actual maximum operating power of the sanitation vehicle and rated power ; Converting the tread hardness and the frequency of cut damage in the key features into score values, and determining a score function for evaluating the cut resistance; Converting the tire operating temperature and heat dissipation efficiency in the key features into scoring values, and determining a scoring function for evaluating the heat resistance; The grip value and pattern design complexity in the key features are converted into scoring values, and a scoring function for evaluating the grip is determined.

5. The method according to claim 4, characterized in that The target performance includes a second target performance, and the second target performance includes cost performance; using at least one key feature corresponding to each of the target performances, the step of constructing a scoring function corresponding to each of the target performances also includes: Based on the price in the key feature and the score value of the first target performance, a score function for evaluating the cost performance is determined.

6. The method according to claim 1, characterized in that The step of determining the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model includes: Input the calculation results of the scoring function corresponding to each of the target performances into a pre-trained random forest model, and output a benchmark value of the implementation effect of the sanitation operation; wherein the pre-trained random forest model is used to predict the implementation effect of the tire of the current vehicle to be selected in the sanitation operation; Repeat the following process until each of the target performances has been traversed: select the current feature sample from the target performance; disrupt the calculation results of the scoring function corresponding to the current feature sample at different times, while keeping the calculation results of the scoring function corresponding to each of the remaining target performances unchanged; input the calculation results of the scoring function corresponding to the current feature sample and the calculation results of the scoring function corresponding to each of the remaining target performances into the previously trained random forest model, and output the work implementation effect value of the sanitation operation corresponding to the current feature sample; The work implementation effect value corresponding to each of the current feature samples and the work implementation effect benchmark value are calculated to determine the importance of each of the current feature samples; wherein the difference is proportional to the importance.

7. The method according to claim 1, characterized in that Before the step of processing the tire multi-source data pair according to the principal component analysis algorithm and the supervised learning algorithm to determine at least one key feature affecting each target performance of the tire of the current vehicle to be selected, the method further includes: Isolating abnormal data points according to a detection model for characterizing tire characteristics of the vehicle model in sanitation operations, and removing abnormal wear characteristics and abnormal design parameter data from the tire multi-source data pair; Based on the parameter data and simulation data on both sides of the missing data and the eliminated data in the tire multi-source data pair, the tire multi-source data pair is multi-interpolated; wherein the simulation data is data obtained by simulating the tire characteristics of the vehicle of the vehicle model in sanitation operations; The tire multi-source data after interpolation are standardized and mapped to the same measurement unit dimension.

8. A tire selection device for a sanitation vehicle, characterized in that: include: The first determination module obtains the historical tire wear characteristics of a vehicle consistent with the vehicle model of the current vehicle to be selected during sanitation operations and the tire design parameters collected by the vehicle factory interface, and determines the tire multi-source data pair of the current vehicle to be selected; A second determination module processes the tire multi-source data pair according to a principal component analysis algorithm and a supervised learning algorithm to determine at least one key feature that affects each target performance of the tire of the current vehicle to be selected; A first construction module constructs a scoring function corresponding to each target performance by using at least one key feature corresponding to each target performance, wherein the scoring function is used to describe a target performance of the tire of the current vehicle to be selected; A third determination module determines the importance of each target performance of the tire of the current vehicle to be selected to the implementation effect of the sanitation work based on the scoring function corresponding to each target performance and the pre-trained random forest model; The second construction module sets a weight according to the importance of each target performance, performs weighted summation on the scoring function corresponding to each target performance, and constructs a multi-objective function of the tire of the current vehicle to be selected; The selection module calculates the multi-objective function and determines the tire type of the current vehicle to be selected.

9. An electronic device, characterized in that: The method comprises a memory, a processor, and a program stored in the memory and capable of being run on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

Images