Commercial vehicle gear recognition method based on working condition driving data

Abstract

The application discloses a commercial vehicle gear recognition method based on working condition driving data, which comprises the following steps: a vehicle-mounted intelligent terminal collects working condition data of vehicle driving and sends the working condition data to a big data management platform; a clustering algorithm of multiple modes is used to find out each cluster center calculated by all clustering algorithms; a ratio of center point coordinates of multiple cluster centers is regarded as a typical value capable of reflecting a transmission ratio level, and n-1 comparison values can be obtained by ratio calculation in sequence; a mode is inferred as a transmission ratio ratio according to a rule that a transmission ratio of a gearbox is designed as an equal ratio series; a speed / vehicle speed ratio value in a high-speed driving state is read to infer a speed / vehicle speed ratio value in the state; a speed / vehicle speed ratio value of a highest gear is used to recursively deduce speed / vehicle speed ratio values of the rest gears; and gear values under the working condition point can be obtained by combining the deduced speed / vehicle speed ratio values of each gear. Therefore, the problem of one-sidedness of a single algorithm calculation result can be avoided, and the gear recognition precision is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile and engine design and manufacturing, and in particular, to a commercial vehicle gear recognition method based on driving data of working conditions. BACKGROUND

[0002] As an important part of intelligent networked vehicle services, the commercial vehicle gear recognition technology based on driving data of working conditions not only greatly affects the convenience and accuracy of vehicle intelligent networked services, but also plays a crucial role in driving behavior mining and user portrait drawing.

[0003] The traditional method of directly reading the current gear of the vehicle through the on-board diagnostic port relies on the intelligent networked terminal monitoring variable collection protocol. When the protocol does not support this data, the gear data tag will be missing. Moreover, when the actual configuration of the transmission of the commercial vehicle is inconsistent with the preset model, it will cause errors in gear reading. The existing gear recognition schemes mainly include auxiliary recognition by external devices and data classification algorithm recognition.

[0004] In the existing technical solutions, the following shortcomings exist:

[0005] 1. The auxiliary recognition scheme by external devices increases the user's use cost, and the reliability of the device needs to be verified, which is not conducive to commercial promotion.

[0006] 2. The data classification algorithm recognition scheme lacks auxiliary markers of transmission design rules, consumes a large amount of computing resources, and has recognition accuracy errors, making it difficult to be put into practical application.

[0007] The information disclosed in this BACKGROUND section is only intended to increase an understanding of the general context in which the present application can be practiced. It is not intended to be in any way a surrender of prior art, nor is it intended to represent the applications described herein as being prior art to the application described herein. SUMMARY

[0008] The present application aims to provide a commercial vehicle gear recognition method based on driving data of working conditions, which uses a variety of modal machine learning clustering algorithms to speculate the gear speed / vehicle speed ratio in each gear of the working condition data points, avoiding the problem of one-sidedness of single algorithm calculation results, and improving the gear recognition accuracy.

[0009] To achieve the above object, the application provides a commercial vehicle gear recognition method based on working condition driving data, which comprises the following steps: a vehicle-mounted intelligent terminal collects working condition data of vehicle driving and sends the working condition data to a big data management platform; the big data management platform reads historical driving data files of vehicle driving and compares the working condition data; point cluster clustering is performed on the rotating speed / vehicle speed data of the working condition data by using a plurality of modal clustering algorithms, and each cluster center calculated by all the clustering algorithms is found out; the ratio of the center point coordinates of the plurality of cluster centers is regarded as a typical value capable of reflecting the transmission ratio level, a plurality of Y / X coordinate values are generated from the plurality of typical values, the plurality of Y / X coordinate values are arranged from small to large, and n-1 comparison values are obtained by sequentially calculating the ratio, wherein the rotating speed is the Y coordinate, the vehicle speed is the X coordinate, and n is the plurality of center point coordinates; according to the law that the transmission ratio of the gearbox is designed as an equal ratio series, the mode of the n-1 comparison values of Y / X of each cluster center is found out, and the mode is inferred as the transmission ratio ratio; the rotating speed / vehicle speed ratio in high-speed driving is read, and the rotating speed / vehicle speed ratio in this state is inferred as the rotating speed / vehicle speed ratio of the highest gear; the rotating speed / vehicle speed ratio of the remaining gears is recursively deduced from the rotating speed / vehicle speed ratio of the highest gear according to the total number of gears and the transmission ratio ratio; and the working condition data of any point is matched with the nearest neighbor in combination with the rotating speed / vehicle speed ratio of each gear deduced by recursion, so that the gear value under the working condition point can be obtained.

[0010] In a preferred embodiment, the total number of gears is the total number of gears of the current vehicle input by the user.

[0011] In a preferred embodiment, the vehicle speed in high-speed driving is between 110-120 km / h.

[0012] In a preferred embodiment, the plurality of modal clustering algorithms comprises K-Means, mean shift clustering, DBSCAN, GMM EM clustering, agglomerative hierarchical clustering or graph community detection.

[0013] In a preferred embodiment, the DBSCAN comprises the following steps: loading data; inputting the known total number of gears by the user; only keeping the rows with the Epm_nEng(rpm) column values between 1200-1500; converting the nested list into a two-dimensional NumPy array; calculating the mean of each column; calculating the standard deviation of each column; performing standardization processing on each column of data; using DBSCAN for clustering; obtaining the clustering label and core sample index; outputting the core point data; calculating the possible transmission ratio ratio; outputting the mode as the inferred transmission ratio ratio; extracting the rows with the vehicle speed between a certain value, calculating the minimum transmission ratio; converting the nested list into a two-dimensional NumPy array; processing the array, dividing the values in the first column by the values in the second column respectively, keeping two decimal places to calculate the mode; and drawing the clustering results.

[0014] In a preferred embodiment, the core point data includes: ensuring that the array has at least two columns; extracting the first column and the second column data; calculating the result of the first column divided by the second column; sorting the result; creating an empty list to store the division result; traversing the sorted array, starting from the second element, dividing the current element by the previous element, and adding the result to the list; and filtering out the results in the list whose values are between 1.1 and 1.6.

[0015] In a preferred embodiment, calculating the possible transmission ratio ratio value includes: keeping two decimal places for each result; and finding the mode using the mode function in the scipy library.

[0016] In a preferred embodiment, the condensed hierarchical clustering includes: loading data; extracting data; only keeping rows whose Epm_nEng(rpm) column values are between 1200 and 1500; converting the nested list to a two-dimensional NumPy array; generating simulation data; calculating the silhouette coefficient for different cluster numbers; selecting the cluster number with the highest silhouette coefficient; clustering using the optimal cluster number; calculating the centroid of each cluster; plotting the clustering results; and plotting the cluster centers.

[0017] Compared with the prior art, the commercial vehicle gear recognition method based on operating condition driving data has the following beneficial effects: the present scheme uses multiple modal machine learning clustering algorithms to respectively infer the gear speed / vehicle speed ratio values in the operating condition data points, and multiple clustering algorithms jointly cluster the operating condition points, avoiding the problem of one-sidedness of single algorithm calculation results, and improving the gear recognition accuracy; the commercial vehicle gear recognition method introduces the user input total gear number of the vehicle combined with the design rules of the transmission to infer and reconstruct the calculation results, and only one round of clustering calculation is needed to derive the speed / vehicle speed ratio values of all gears, without the need for multiple rounds of calculation, greatly saving the consumption of calculation resources, improving the recognition accuracy of the scheme, and without the need for additional external equipment, without additional user use cost. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flowchart of a commercial vehicle gear recognition method according to an embodiment of the present application;

[0019] Figure 2 is an example data screenshot of a csv file of a dbscan clustering algorithm according to an embodiment of the present application;

[0020] Figure 3 is a comparison chart of calculation data and official parameters of a commercial vehicle gear recognition method according to an embodiment of the present application;

[0021] Figure 4This is a schematic diagram comparing the results of the dbscan clustering algorithm and the agglomerative hierarchical clustering algorithm according to an embodiment of the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0023] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0024] like Figure 1 As shown, a preferred embodiment of the present invention provides a method for commercial vehicle gear identification based on driving condition data, comprising: an onboard intelligent terminal collecting driving condition data of the vehicle and sending the driving condition data to a big data management platform; the big data management platform reading the historical driving data file of the vehicle and comparing it with the driving condition data; using a multi-modal clustering algorithm to perform point clustering on the speed / vehicle speed data of the driving condition data, and finding the cluster centers calculated by all clustering algorithms; considering the ratio of the coordinates of the center points of multiple cluster centers as a typical value that can reflect the transmission ratio level, multiple typical values ​​will generate multiple Y / X coordinate values, arranging the multiple Y / X coordinate values ​​from small to large, and then successively comparing them. The calculation yields n-1 comparison values, where engine speed is the Y-coordinate, vehicle speed is the X-coordinate, and n represents the coordinates of multiple center points. Based on the principle that the transmission ratios are designed as a geometric series, the mode of the n-1 comparison values ​​for Y / X at each cluster center is found, and the mode is inferred to be the transmission ratio. The engine speed / vehicle speed ratio at high vehicle speeds is read, and it is inferred that this ratio represents the highest gear. Based on the total number of gears and the transmission ratio, the engine speed / vehicle speed ratios for the remaining gears are recursively derived from the highest gear. Finally, by combining the recursively derived engine speed / vehicle speed ratios with nearest neighbor matching of the operating data at any point, the gear value at that operating point can be obtained.

[0025] In some implementations, the total number of gears is the total number of gears for the current vehicle entered by the user.

[0026] In some implementations, the vehicle speed during high-speed driving is between 110 and 120 km / h.

[0027] In some implementations, multimodal clustering algorithms mainly include K-Means, mean-shift clustering, DBSCAN clustering, GMM EM clustering, agglomerative hierarchical clustering, or graph community detection clustering, etc.

[0028] In some implementations, DBSCAN includes: loading data; user input of the known total number of gears; retaining only rows where the Epm_nEng (rpm) column value is between 1200 and 1500; converting the nested list into a two-dimensional NumPy array; calculating the mean of each column; calculating the standard deviation of each column; standardizing the data in each column; performing clustering using DBSCAN; obtaining cluster labels and core sample indexes; outputting core point data; calculating possible gear ratios; outputting the mode as the inferred gear ratio; extracting rows where the vehicle speed is between a certain value and calculating the minimum gear ratio; converting the nested list into a two-dimensional NumPy array; processing the array, dividing the value of the first column by the value of the second column, retaining two decimal places to calculate the mode; and plotting the clustering results.

[0029] In some implementations, outputting core data also includes calculation steps such as ensuring the array has at least two columns, extracting data from the first and second columns, checking if the second column has zeros to avoid division by zero errors, calculating the result of dividing the first column by the second column, sorting the result, creating an empty list to store the division result, traversing the sorted array, starting from the second element, dividing the current element by the previous element and adding the result to the list, and filtering out results in the list whose values ​​are between 1.1 and 1.6.

[0030] In some implementations, calculating possible gear ratios also includes steps such as rounding each result to two decimal places and finding the mode using the mode function in the scipy library.

[0031] like Figure 2 The image shown is a screenshot of sample data from a script input in a CSV file generated by the dbscan clustering algorithm. In the above example, the total number of gears input is 12, and the inferred gear ratio is 1.29. The highest gear has a speed / vehicle speed ratio of 24.67, and the subsequent gears are as follows: Figure 2 The string of numbers below the screenshot is the recursive transmission ratio value for each gear.

[0032] In some implementations, agglomerative hierarchical clustering includes: loading data; extracting data; retaining only rows where the Epm_nEng(rpm) column value is between 1200 and 1500; converting the nested list into a two-dimensional NumPy array; generating simulated data; calculating the silhouette coefficients for different numbers of clusters; selecting the number of clusters with the highest silhouette coefficient; clustering using the optimal number of clusters; calculating the centroid of each cluster; plotting the clustering results; and plotting the cluster centers.

[0033] like Figure 3The script shown is based on a hierarchical clustering algorithm, the principle of recursive gear is similar, but the clustering algorithm is different, according to the example data of the csv file running out of the algorithm, using the actual inspection results, and according to the parameters provided by the official, the ratio of the gear ratio is close to the actual inferred result of 1.29.

[0034] As Figure 4 shown, and referring to Figure 2 , in some embodiments, the above two clustering algorithms are different in the mechanism of searching for clusters, and the difference in effect is that different shape distributions of scattered points have different clustering effects. The above two clusters are relatively suitable for bar-shaped scattered point distribution. If the working condition data is drawn, the distribution of gear working condition points is like Figure 2 , Figure 4 The two contrastive figures can illustrate the different applicability of different clustering algorithms when facing different scattered point distributions.

[0035] In some embodiments, the commercial vehicle gear recognition method based on working condition driving data of the application cannot be completely realized by relying on a certain clustering algorithm. More accurately, it should be a "design rule experience + clustering algorithm" combined gear calculation mode. The clustering algorithm includes but is not limited to the foregoing clustering algorithms.

[0036] In summary, the commercial vehicle gear recognition method based on working condition driving data of the application has the following advantages: the scheme adopts multiple modal machine learning clustering algorithms to respectively infer the gear speed / vehicle speed ratio in the working condition data points, and multiple clustering algorithms cluster the working condition points, avoiding the problem of one-sidedness of single algorithm calculation results, and improving the gear recognition accuracy; the commercial vehicle gear recognition method introduces the user input total gear number of the vehicle combined with the design rule of the gearbox to infer and reconstruct the calculation results, which only needs one round of clustering calculation to derive the speed / vehicle speed ratio of all gears, without multiple rounds of calculation, greatly saving the consumption of calculation resources, improving the recognition accuracy of the scheme, and without additional external equipment, without additional user cost.

[0037] The foregoing description of specific exemplary embodiments of the application is for purposes of illustration and example. These descriptions are not intended to limit the application to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teachings. The specific exemplary embodiments were chosen and described in order to explain the principles of the application and its practical application to thereby enable one of ordinary skill in the art to best utilize the application and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the application be defined by the claims and their equivalents.

Claims

1. A method for identifying a gear of a commercial vehicle based on operating condition driving data, characterized in that, The application relates to a method for inferring the gear ratio of a vehicle. The method comprises: a vehicle-mounted intelligent terminal collecting working condition data of vehicle driving and sending the working condition data to a big data management platform, the big data management platform reading historical driving data files of the vehicle driving and comparing the working condition data; using a plurality of modal clustering algorithms to perform point cluster clustering on the speed / vehicle speed ratio data of the working condition data, and finding out the cluster centers calculated by all the clustering algorithms; taking the ratio of the center point coordinates of the plurality of cluster centers as a typical value capable of reflecting the gear ratio level, the plurality of typical values generating a plurality of Y / X ratio coordinate values, arranging the plurality of Y / X ratio coordinate values from small to large, and then sequentially calculating the ratios to obtain n-1 comparison values, wherein the speed is the Y coordinate, the vehicle speed is the X coordinate, and n is the number of center point coordinates; according to the design of the gear ratio of the gearbox as an equal ratio series, finding out the mode of the n-1 comparison values of the Y / X ratio of the cluster centers, and inferring that the mode is the gear ratio ratio; reading the speed / vehicle speed ratio in high-speed driving, and inferring that the speed / vehicle speed ratio in this state is the speed / vehicle speed ratio of the highest gear; according to the total number of gears and the gear ratio ratio, recursively deriving the speed / vehicle speed ratio of the remaining gears from the speed / vehicle speed ratio of the highest gear; and 2. The method of claim 1, wherein, combining the gear ratio of each gear derived by recursion with the working condition data of any point to perform nearest neighbor matching, so as to obtain the gear value under the working condition point.

3. The method of claim 1, wherein, The total number of gears is the total number of gears of the current vehicle input by the user.

4. The operating condition based drive data method of claim 1, wherein, The vehicle speed during high-speed driving is between 110 and 120 km / h.

5. The operating mode based drive data driven commercial vehicle gear identification method of claim 4, wherein, The plurality of modal clustering algorithms include K-Means, mean shift clustering, DBSCAN, GMM EM clustering, agglomerative hierarchical clustering or graph community detection. The DBSCAN comprises: loading data; the user inputting the known total number of gears; only keeping the rows with the Epm_nEng column values between 1200 and 1500; converting the nested list into a two-dimensional NumPy array; calculating the mean value of each column; calculating the standard deviation of each column; standardizing each column of data; using DBSCAN to perform clustering; obtaining the clustering label and core sample index; outputting the core point data; calculating the possible gear ratio ratio; outputting the mode as the inferred gear ratio ratio; extracting the rows with the vehicle speed between a certain value, and calculating the minimum gear ratio; converting the nested list into a two-dimensional NumPy array; processing the array, dividing the values in the first column by the values in the second column respectively, and keeping two decimal places calculating the mode; and 6. The operating mode based drive data method of claim 5, wherein, drawing the clustering results. The core point data output comprises: ensuring that the array has at least two columns; extracting the first column and the second column data; calculating the result of the first column divided by the second column; sorting the result; creating an empty list to store the division results; traversing the sorted array, starting from the second element, dividing the current element by the previous element, and adding the result to the list; and 7. The operating mode based drive data driven commercial vehicle gear identification method of claim 5, wherein, filtering out the results with the values between 1.1 and 1.6 in the list. The calculation of the possible gear ratio ratio comprises: keeping two decimal places for each result; and using the mode function in the scipy library to find the mode.

8. The operating mode based drive data driven commercial vehicle gear identification method of claim 4, wherein, The agglomerative hierarchical clustering comprises: Loading data; Extracting data; Keeping only rows with Epm_nEng column values between 1200-1500; Converting nested lists to a two-dimensional NumPy array; Generating simulated data; Computing silhouette coefficients for different numbers of clusters; Selecting the number of clusters with the highest silhouette coefficient; Clustering using the optimal number of clusters; Computing the centroid of each cluster; Plotting the clustering results; and Plotting the cluster centers.

Images