CN114611604A - User screening method based on electric drive assembly load characteristic fusion and clustering - Google Patents

User screening method based on electric drive assembly load characteristic fusion and clustering Download PDF

Info

Publication number
CN114611604A
CN114611604A CN202210237252.0A CN202210237252A CN114611604A CN 114611604 A CN114611604 A CN 114611604A CN 202210237252 A CN202210237252 A CN 202210237252A CN 114611604 A CN114611604 A CN 114611604A
Authority
CN
China
Prior art keywords
drive assembly
electric drive
load
user
clustering
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210237252.0A
Other languages
Chinese (zh)
Inventor
王震
周驰
赵礼辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Linui Shanghai Intelligent Technology Co ltd
Original Assignee
Linui Shanghai Intelligent Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Linui Shanghai Intelligent Technology Co ltd filed Critical Linui Shanghai Intelligent Technology Co ltd
Priority to CN202210237252.0A priority Critical patent/CN114611604A/en
Publication of CN114611604A publication Critical patent/CN114611604A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Power Engineering (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a user screening method based on electric drive assembly load characteristic fusion and clustering, which comprises the following steps: acquiring user sample load data, and analyzing and obtaining failure leading load of the electric drive assembly based on the user sample load data; acquiring key load characteristic parameters of the electric drive assembly based on the failure dominant load of the electric drive assembly; and processing the key load characteristic parameters of the electric drive assembly, and realizing typical user screening of the reliability key load of the associated electric drive assembly based on the processed load characteristic parameters. The method realizes typical user screening by associating the key load of the reliability of the electric drive assembly, can effectively cover the regional distribution, the driving behavior and the driving working condition of actual overall users, can provide load input for accelerating the endurance specification compilation of the electric drive assembly in the stages of product development and test verification, accelerates the development process of products, and provides certain technical support for high-quality product development.

Description

User screening method based on electric drive assembly load characteristic fusion and clustering
Technical Field
The invention belongs to the field of load data analysis of electric drive systems, and particularly relates to a user screening method based on electric drive assembly load feature fusion and clustering.
Background
The electric drive system is used as a core component of a new energy automobile, and in the development and verification stage of an electric drive assembly product, the reliability design and accelerated durability check of the electric drive assembly under the service condition of an actual user are very important. The method is oriented to a large amount of user data and comprises multidimensional load characteristics, a typical user sample is rapidly screened to serve as the basis for design and verification of an electric drive assembly product, the product research and development period can be shortened, and the product reliability level can be improved. At present, a random sampling method can be used for fast sampling, but due to the randomness of sampling, the sampling result deviates from the overall characteristics, the multi-dimensional load information of an electric drive assembly is difficult to cover by infinite random sampling under the single characteristic, and a user label corresponding to sample data cannot be obtained by adopting a pseudo-random number sampling-based method.
Disclosure of Invention
The invention relates to electric drive assembly failure dominant load construction characteristic parameters, and provides a user screening method based on electric drive assembly load characteristic fusion and clustering, which is used for realizing rapid sampling of typical users and providing small samples and representative load input for accelerated endurance assessment of an electric drive assembly, thereby accelerating the product research and development process.
In order to achieve the purpose, the invention provides the following scheme: a user screening method based on electric drive assembly load characteristic fusion and clustering comprises the following steps:
acquiring user sample load data, and analyzing and obtaining failure dominant load of the electric drive assembly based on the user sample load data;
acquiring a key load characteristic parameter of the electric drive assembly based on the failure dominant load of the electric drive assembly;
and processing the key load characteristic parameters of the electric drive assembly, and realizing typical user screening of the reliability key load of the associated electric drive assembly based on the processed load characteristic parameters.
Preferably, the sample load data at least includes load data in different regions, different driving behaviors and different driving conditions;
the failure leading load of the electric drive assembly at least comprises motor rotating speed, motor torque, acceleration, thermal stress and electric stress.
Preferably, the process of obtaining the critical load characteristic parameters of the electric drive assembly comprises,
determining key load characteristics of the electric drive assembly through load amplitude distribution and combined distribution based on the failure dominant load of the electric drive assembly;
analyzing user characteristic differences based on the key load characteristics of the electric drive assembly to obtain typical characteristic parameters of the failure load of the associated electric drive assembly under user conditions;
the typical characteristic parameters include, but are not limited to, user run time, mileage, speed mean, speed variance, motor torque mean, motor torque variance, acceleration mean, acceleration variance.
Preferably, the processing of the critical load characteristic parameters of the electric drive assembly comprises the steps of performing multi-feature fusion analysis, user feature cluster analysis, minimum sample size determination, classified random sampling, hypothesis testing, establishment of objective functions and constraint conditions, and determination of a typical user optimal solution.
Preferably, the processing of the key load characteristic parameters of the electric drive assembly further comprises user classification by multi-feature fusion analysis and user clustering analysis; determining a sampling minimum sample size based on the total sample size and the expected estimation error; analyzing the difference between the extracted sample characteristics and the overall sample characteristics by adopting a classified random sampling method and a hypothesis testing method; and establishing an objective function and a constraint condition, and realizing typical user screening of the reliability key load of the associated electric drive assembly.
Preferably, the multi-feature fusion analysis is based on a principal component analysis method, the high-dimensional variable and correlated initial feature parameters are converted into a plurality of low-dimensional mutually-irrelevant principal component feature parameters, and the electrical drive assembly key load feature parameters are subjected to dimensionality reduction to obtain a dimensionality reduction feature matrix; the principal component feature parameter retains parameter information of the initial feature parameter.
Preferably, the user clustering analysis is based on a dimension reduction characteristic matrix, and a clustering algorithm is adopted to classify the users to obtain a user characteristic clustering classification result;
and the classified random sampling carries out random sampling on each type of users based on the user characteristic clustering classification result.
Preferably, the hypothesis test analyzes the extracted sample using a Z test;
the Z test statistic was:
Figure BDA0003542763920000031
wherein the content of the first and second substances,
Figure BDA0003542763920000032
is the sample mean, μ0Is the overall mean, σ is the overall standard deviation, and n is the number of samples.
The invention discloses the following technical effects:
the invention provides a user screening method based on electric drive assembly load feature fusion and clustering, which plays an important role in links such as product design, simulation, test verification and the like based on typical user load data in the electric drive assembly product research and development process. According to the invention, the key load of the reliability of the electric drive assembly is associated, typical user screening is realized by adopting technical methods such as multi-feature fusion, cluster analysis, minimum sample size determination, classified random sampling, hypothesis test and the like, the regional distribution, driving behaviors and driving conditions of actual overall users can be effectively covered, the key load information of the reliability of the electric drive assembly is considered in the screened user load characteristics, the load input can be provided for accelerating the endurance specification compilation of the electric drive assembly in the stages of product development and test verification, the product research and development process is accelerated, and a certain technical support is provided for the high-quality product development.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a method according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
As shown in FIG. 1, the invention provides a user screening method based on electric drive assembly load feature fusion and clustering, which comprises the following steps:
acquiring user sample load data, and analyzing and obtaining failure leading load of the electric drive assembly based on the user sample load data;
acquiring key load characteristic parameters of the electric drive assembly based on the failure dominant load of the electric drive assembly;
and processing the key load characteristic parameters of the electric drive assembly, and realizing typical user screening of the reliability key load of the associated electric drive assembly based on the processed load characteristic parameters.
The sample load data at least comprises load data in different regions, different driving behaviors and different driving working conditions;
the failure leading load of the electric drive assembly at least comprises motor rotating speed, motor torque, acceleration, thermal stress and electric stress.
The process of obtaining the critical load characteristic parameters of the electric drive assembly includes,
determining key load characteristics of the electric drive assembly through load amplitude distribution and combined distribution based on the failure dominant load of the electric drive assembly;
and analyzing the user characteristic difference based on the key load characteristics of the electric drive assembly to obtain the typical characteristic parameters of the failure load of the associated electric drive assembly under the user condition.
The typical characteristic parameters include, but are not limited to, user run time, mileage, speed mean, speed variance, motor torque mean, motor torque variance, acceleration mean, acceleration variance.
The method comprises the steps of processing the key load characteristic parameters of the electric drive assembly, including multi-feature fusion analysis, user feature cluster analysis, minimum sample size determination, classified random sampling, hypothesis testing, target function and constraint condition establishment, and typical user optimal solution determination.
Processing the key load characteristic parameters of the electric drive assembly further comprises the steps of carrying out user classification by adopting multi-characteristic fusion analysis and user clustering analysis; determining a sampling minimum sample size based on the total sample size and the expected estimation error; analyzing the difference between the extracted sample characteristics and the overall sample characteristics by adopting a classified random sampling method and a hypothesis testing method; and establishing an objective function and a constraint condition, and realizing typical user screening of the reliability key load of the associated electric drive assembly.
The multi-feature fusion analysis is based on a principal component analysis method, the high-dimensional variable and correlated initial feature parameters are converted into a plurality of low-dimensional mutually-irrelevant principal component feature parameters, and the electrical drive assembly key load feature parameters are subjected to dimension reduction to obtain a dimension reduction feature matrix; the principal component feature parameter retains parameter information of the initial feature parameter.
The user clustering analysis is based on a dimension reduction characteristic matrix, and a clustering algorithm is adopted to classify users to obtain a user characteristic clustering classification result; classified random sampling each class of users is randomly sampled based on the user feature clustering classification results. Hypothesis testing the samples taken were analyzed using the Z test.
Example one
Further, the invention provides a user screening method based on electric drive assembly load feature fusion and clustering, which comprises the following steps: the method comprises the steps of analyzing failure dominant loads of an electric drive assembly, determining key load characteristics of the electric drive assembly from the load amplitude distribution and combined distribution angles, and determining the optimal solution of a typical user by multi-characteristic fusion analysis, user characteristic cluster analysis, minimum sample size determination, classified random sampling, significance test, target function and constraint condition establishment. The specific implementation steps are as follows:
step 1, load data of 3 thousands of new energy vehicles in different regions, different driving behaviors and different driving conditions are contained on the basis of user overall sample data, failure leading loads of an electric drive assembly are analyzed, and load factors influencing reliability of the electric drive assembly in the user data are determined.
In this embodiment, step 1 is based on user overall sample data, and because a lot of load channels are included in the actual user load collection process, the critical load affecting the failure of the electric drive assembly component needs to be concerned for the reliability evaluation of the electric drive assembly, so that representative characteristic parameters are constructed. The failure leading load of the electric drive assembly component mainly comprises load characteristics such as motor rotating speed, motor torque, acceleration, thermal stress, electric stress and the like.
And 2, determining the key load characteristics of the electric drive assembly from the load amplitude distribution and combined distribution angles. The speed mean value, the speed variance, the torque mean value, the torque variance, the acceleration mean value and the acceleration variance of each user in a certain time can be counted for different users. In addition, in order to reflect the use conditions of different users, the invention uses the characteristics of running time, driving mileage and the like to distinguish the difference between different users. And finally, analyzing the characteristic difference among users based on the load characteristic amplitude distribution and the combined distribution so as to select the typical characteristic parameters of the related electric drive assembly failure load under the user condition.
In the embodiment, the load amplitude distribution and the combined distribution angle in the step 2 determine the key load characteristics of the electric drive assembly. By counting the load characteristic amplitude distribution and the combined distribution under different users, the behavior characteristics and the difference of the users can be reflected, and a basis is provided for load characteristic selection in the typical user screening process. The test statistics shows that the total driving mileage is 1 ten thousand kilometers in 8 months, and the total driving mileage of more than 90 percent of users is less than 28000 kilometers; the calculation shows that the maximum number of users can run for about 1.4h per day in 8 months on average; the average speed of the user is between 10km/h and 70 km/h. The mean distribution of the torques is mainly divided into positive torques and negative torques, the mean of the positive torques being substantially 10Nm to 25Nm on average, and the negative torques being substantially-40 Nm to 0Nm on average; the average vehicle speed is 10 km/h-70 km/h, and the torque average characteristic is divided into a positive torque and a negative torque. From the view of load fluctuation, the larger the vehicle speed variance is, the larger the rear motor torque variance is, the certain linear relationship between the vehicle speed variance and the rear motor torque variance is, and 16 characteristic parameters are selected to screen typical users, as shown in table 1.
TABLE 1
Serial number Characteristic parameter Means of
1 day Days of operation
2 mileage Total mileage
3 mile_per_day Average daily mileage
4 speed_cnt Total time of operation
5 speed_mean Mean value of velocity
6 speed_Variance Variance of velocity
7 Torque_F_cnt Front motor run time
8 Toque_F_mean Front motor torque mean
9 Torque_F_Variance Front motor torque variance
10 Torque_R_cnt Run time of rear motor
11 Torque_R_mean Mean value of torque of rear motor
12 Torque_R_Variance Rear motor torque variance
13 Acc_X_mean Mean value of acceleration in X direction
14 Acc_X_Variance Variance of acceleration in X direction
15 Acc_Y_mean Mean value of acceleration in Y direction
16 Acc_Y_Variance Variance of acceleration in Y direction
And 3, performing multi-feature fusion analysis. Due to the large total user sample amount and multiple parameter dimensions, the extraction of samples based on single feature distribution is difficult to cover the whole feature information. Extracting samples separately considering a plurality of feature distributions increases the amount of calculation. Therefore, the invention provides a dimension reduction method considering multi-feature fusion, which converts complex feature indexes with high dimensional variables and certain correlation into a plurality of low-dimensional independent principal components for replacement and reserves a large amount of information in original feature parameters. The invention adopts a principal component analysis method to realize the key load characteristic dimension reduction of the electric drive assembly, and carries out principal component extraction when the accumulated information contribution rate reaches more than 80 percent, and the extracted principal component can cover most of information content in the original characteristic.
In this embodiment, the multi-feature fusion analysis method in step 3 mainly adopts a principal component analysis method, and assumes that there are n users, each user has p feature parameters, and the calculated feature parameter matrix of all users is denoted as X. Setting a random variable XiHas a mean value of μiIf the covariance matrix is Σ i, then the p feature parameters are linearly transformed through principal component analysis, and the generated comprehensive index is the principal component and is marked as y1,y2,…,yp. Wherein the characteristic parameter matrix X can be expressed as:
Figure BDA0003542763920000091
the expression of the transformed principal component is shown as formula (2):
Figure BDA0003542763920000092
wherein, y1,y2,…,ypIs each main component, x1,x2,…,xpIs an original characteristic parameter l'1,l'2,…,l'pThe load factor is a characteristic parameter in each principal component.
Solving each principal component y from correlation coefficient matrix or covariance matrixpWherein the variance and covariance calculation formula is as follows:
var(yj)＝l'j∑lj j＝1,2,…,p (3)
cov(yj,yk)＝l'j∑lk j,k＝1,2,…,p (4)
before solving the eigenvalue and the corresponding eigenvector through the covariance matrix, the dispersion degree difference is large due to the values of each variable caused by different dimensions due to the influence of the result dimensions of the principal component analysis. In order to eliminate some unreasonable influence caused by different dimensions, each original variable is usually standardized. The normalized variables are then:
Figure BDA0003542763920000101
wherein, vector Z ═ (Z)1,z2,…,zp) The covariance matrix of' is the correlation coefficient matrix ρ ═ ρ (x)i,xj)]p×p. After principal component analysis, the covariance matrix of the principal components is a diagonal matrix Lambda, and the characteristic root of which the diagonal elements are correlation coefficients rho is Lambda1,λ2,…λp. Wherein the matrix of correlation coefficients
Figure BDA0003542763920000102
Ith principal component yiWith the original characteristic parameter variable xjThe correlation coefficient between them is called factor load quantity, it reflects the original characteristic parameter variable and mainThe closer the absolute value of the closeness between the components is to 1, the closer the relationship is.
The total cumulative contribution to each principal component can be represented by r:
Figure BDA0003542763920000103
in the formula (6), the contribution ratio of the jth principal component is λjR, the cumulative variance contribution rate of the first m principal components is
Figure BDA0003542763920000104
When the cumulative contribution rate reaches more than 80%, the first m principal components are extracted to replace most of the characteristic information quantity in the original variables.
And 4, clustering and analyzing the user characteristics. And (3) classifying the users by adopting a clustering algorithm based on the 5-dimensional feature matrix after the dimension reduction obtained in the step (3), and taking the distance from each user in each class to the clustering center after clustering as a basis for distinguishing the user difference, thereby realizing the independent sampling of each class.
In the embodiment, aiming at the user characteristic clustering analysis in the step 4, based on the principal component obtained after the dimensionality reduction in the step 3, the user categories are divided by adopting a K-Means clustering algorithm, and the distance from each user in each category to the clustering center after clustering is used as a basis for distinguishing the user difference, so that each category is independently sampled.
Assume the sample data is matrix Xn×mWherein n is the number of samples, and m is the corresponding characteristic dimension. Initializing k cluster centers { C1,C2,…,CkAnd calculating the Euclidean distance from each sample to the center of each cluster, wherein the formula is as follows:
Figure BDA0003542763920000111
in the formula (7), XiRepresenting the ith sample, i is more than or equal to 1 and less than or equal to n; cjJ is more than or equal to 1 and less than or equal to k, and represents the jth clustering center;Xitrepresents the t-th characteristic of the ith sample, wherein t is more than or equal to 1 and less than or equal to m; cjtRepresenting the t-th attribute of the j-th cluster center.
For the selection of the cluster number k, the elbow rule is that when the k value increases, the inflection point of the drop of the loss function is calculated, the damage function is the square sum of errors in the cluster (SSE), the distance from the sample point in the cluster to the cluster center reflects the degree of agglomeration in the cluster, and then the calculation formula for the square sum of errors in the cluster is as follows:
Figure BDA0003542763920000112
in the formula (8), CiIs a cluster of type i, p is CiSample point of (1), miIs CiCluster center point of (2).
And dividing the users into three categories through k-means clustering, wherein the first category only comprises two users, and the total driving mileage and the running time of the two users are minimum and deviate from the total characteristics of the users. Therefore, in this embodiment, only the second class and the third class are sampled separately.
And 5, determining the minimum sample size. The determination of the minimum sample size in a sample is related to the sampling error, the sample mean value fluctuates above and below the true mean value in the process of multiple sampling, and the fluctuation of the deviation relative to the overall mean value is the sampling error. The confidence interval for the overall mean is made up of the sample mean x and the estimation error. Under the conditions of repeated sampling and infinite overall sampling, the estimation error is
Figure BDA0003542763920000121
zα/2Together with the sample size n determines the magnitude of the estimation error. Once the confidence level 1-alpha, z is determinedα/2The value of (c) is determined. For a given zα/2The value of (c) and the total standard deviation sigma, the amount of samples needed for any desired estimation error can be determined. Assuming the expected estimation error is ε, then
Figure BDA0003542763920000122
From this, the formula for determining the sample size can be derived as follows:
Figure BDA0003542763920000123
where ε is the acceptable estimation error at a given confidence level, zα/2Is determined by the confidence level used in the interval estimation.
And 6, classifying and randomly sampling. Because the Euclidean distance from each user to the cluster center of the cluster is determined after the user characteristics are clustered, the invention respectively carries out random sampling on each class by taking the mean value of the Euclidean distance of each class of overall samples as a target, so that the distance mean value and the variance of the extracted samples are close to the mean value and the variance of the whole class.
Step 7, hypothesis testing. Based on the sample information, the difference between the overall index and the sample index is deduced according to a certain probability level. For known overall variance, the extracted samples can be analyzed using the Z-test, with Z-test statistics:
Figure BDA0003542763920000124
in the formula (11), the reaction mixture is,
Figure BDA0003542763920000131
is the sample mean, μ0Is the overall mean, σ is the overall standard deviation, and n is the number of samples.
When each type of user is randomly sampled, the characteristic mean value of the extracted sample is assumed to be equal to the overall mean value: h0; the alternative assumption is that: the feature mean of the extracted samples deviates from the overall mean: H1. assuming that the probability of the occurrence is P, the larger the value of P, the closer the feature mean of the extracted sample is to the overall feature mean.
In this embodiment, the overall variance described in step 7 is the variance of the distance from each user to the center of the cluster.
Step (ii) ofAnd 8, establishing an objective function and a constraint condition. In order to make the user sample mean extracted from each class of users approximately equal to the class of overall sample mean, and the sample variance is close to the overall variance. Taking the probability P value of the establishment of the hypothesis in the Z test as an optimization parameter, and assuming that the overall variance of each type of user characteristics is
Figure BDA0003542763920000132
The variance of the extracted samples is
Figure BDA0003542763920000133
Thus establishing the objective function and constraints in the random sampling process. An objective function: p>0.99 or max (P); constraint conditions are as follows:
Figure BDA0003542763920000134
and 9, determining the optimal solution of the typical user. And (4) respectively carrying out infinite sampling on each type of users based on the objective function and the constraint condition established in the step (8) until the condition is met. In addition, the screened typical users need to further consider the regional distribution of the rate, and because different regions and road conditions have certain differences in the damage to the electric drive assembly, the screened typical users need to be adjusted in combination with the regional distribution of the users so as to cover users in multiple regions.
In this embodiment, the minimum sample size is 200 users, and it is determined that 119 users are extracted from the second class and 81 users are extracted from the third class according to the ratio of the second class and the third class users after the cluster analysis. And obtaining various typical user samples through infinite classification sampling, wherein the mean value and the variance of the second type sample are shown in table 2, and the mean value and the variance of the third type sample are shown in table 3. And finally, determining 200 typical users including 18 provincial regions, 4 direct municipalities and 2 autonomous region users by combining the regional distribution conditions of the users. The method has certain representativeness in areas of northeast, northwest, China, east China, southwest and south China.
TABLE 2
Sample mean Global mean value Mean difference rate
2.5695 2.5701 -0.023％
Sample variance Total variance Variance ratio of difference
1.9665 1.9671 0.031％
TABLE 3
Sample mean Global mean value Mean difference rate
2.4492 2.4509 -0.069％
Sample variance Total variance Variance ratio of difference
1.9550 1.9575 0.128％
The method comprises the steps of analyzing failure dominant loads of the electric drive assembly based on user overall sample data, determining load factors influencing the reliability of the electric drive assembly, analyzing user characteristic differences from load amplitude distribution and combined distribution, determining key load characteristic parameters of the electric drive assembly, classifying users by adopting a multi-characteristic fusion analysis and cluster analysis method, determining minimum sample size of sampling based on overall sample size and expected estimation errors, and realizing typical user screening of the reliability critical loads of the associated electric drive assembly by adopting a classified random sampling method, a hypothesis testing method, an established objective function and constraint conditions. Typical users screened according to the technical method can effectively cover the actual overall user regional distribution, driving behaviors and driving conditions, wherein the screened user load characteristics take the reliability key load information of the electric drive assembly into consideration, so that load input can be provided for accelerating the endurance specification compilation of the electric drive assembly in the product development and test verification stages, and the product research and development process is accelerated.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.

Claims (8)

1. A user screening method based on electric drive assembly load feature fusion and clustering is characterized by comprising the following steps:
acquiring user sample load data, and analyzing and obtaining failure dominant load of the electric drive assembly based on the user sample load data;
acquiring a key load characteristic parameter of the electric drive assembly based on the failure dominant load of the electric drive assembly;
and processing the key load characteristic parameters of the electric drive assembly, and realizing typical user screening of the reliability key load of the associated electric drive assembly based on the processed load characteristic parameters.
2. The electric drive assembly load signature fusion and clustering-based user screening method of claim 1,
the sample load data at least comprises load data in different regions, different driving behaviors and different driving conditions;
the failure leading load of the electric drive assembly at least comprises motor rotating speed, motor torque, acceleration, thermal stress and electric stress.
3. The electric drive assembly load signature fusion and clustering-based user screening method of claim 1,
the process of obtaining the critical load characteristic parameters of the electric drive assembly includes,
determining key load characteristics of the electric drive assembly through load amplitude distribution and combined distribution based on the failure dominant load of the electric drive assembly;
analyzing user characteristic differences based on the key load characteristics of the electric drive assembly to obtain typical characteristic parameters of the failure load of the associated electric drive assembly under user conditions;
the typical characteristic parameters include, but are not limited to, user running time, mileage, speed mean, speed variance, motor torque mean, motor torque variance, acceleration mean, and acceleration variance.
4. The electric drive assembly load signature fusion and clustering-based user screening method of claim 1,
the processing of the key load characteristic parameters of the electric drive assembly comprises the steps of performing multi-feature fusion analysis, user feature cluster analysis, minimum sample size determination, classified random sampling, hypothesis testing, establishment of a target function and constraint conditions, and determination of a typical user optimal solution.
5. The electric drive assembly load signature fusion and clustering based user screening method of claim 4,
processing the key load characteristic parameters of the electric drive assembly further comprises the steps of carrying out user classification by adopting multi-characteristic fusion analysis and user clustering analysis; determining a sampling minimum sample size based on the total sample size and the expected estimation error; analyzing the difference between the extracted sample characteristics and the overall sample characteristics by adopting a classified random sampling method and a hypothesis testing method; and establishing an objective function and a constraint condition, and realizing typical user screening of the reliability key load of the associated electric drive assembly.
6. The electric drive assembly load signature fusion and clustering-based user screening method of claim 5,
the multi-feature fusion analysis is based on a principal component analysis method, the high-dimensional variable and correlated initial feature parameters are converted into a plurality of low-dimensional mutually-irrelevant principal component feature parameters, and the electrical drive assembly key load feature parameters are subjected to dimension reduction to obtain a dimension reduction feature matrix; the principal component feature parameter retains parameter information of the initial feature parameter.
7. The electric drive assembly load signature fusion and clustering-based user screening method of claim 5,
the user clustering analysis is based on a dimension reduction characteristic matrix, and a clustering algorithm is adopted to classify users to obtain a user characteristic clustering classification result;
and the classified random sampling carries out random sampling on each type of users based on the user characteristic clustering classification result.
8. The electric drive assembly load signature fusion and clustering-based user screening method of claim 5,
the hypothesis test adopts a Z test to analyze the extracted sample;
the Z test statistic was:
Figure FDA0003542763910000031
wherein the content of the first and second substances,
Figure FDA0003542763910000032
is the sample mean, μ0Is the overall mean, σ is the overall standard deviation, and n is the number of samples.
CN202210237252.0A 2022-03-11 2022-03-11 User screening method based on electric drive assembly load characteristic fusion and clustering Pending CN114611604A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210237252.0A CN114611604A (en) 2022-03-11 2022-03-11 User screening method based on electric drive assembly load characteristic fusion and clustering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210237252.0A CN114611604A (en) 2022-03-11 2022-03-11 User screening method based on electric drive assembly load characteristic fusion and clustering

Publications (1)

Publication Number Publication Date
CN114611604A true CN114611604A (en) 2022-06-10

Family

ID=81863340

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210237252.0A Pending CN114611604A (en) 2022-03-11 2022-03-11 User screening method based on electric drive assembly load characteristic fusion and clustering

Country Status (1)

Country Link
CN (1) CN114611604A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115062702A (en) * 2022-06-16 2022-09-16 四川大学 PCA-E based product perceptual semantic vocabulary extraction method
US11868432B1 (en) 2022-06-16 2024-01-09 Sichuan University Method for extracting kansei adjective of product based on principal component analysis and explanation (PCA-E)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115062702A (en) * 2022-06-16 2022-09-16 四川大学 PCA-E based product perceptual semantic vocabulary extraction method
CN115062702B (en) * 2022-06-16 2023-09-08 四川大学 Product perceptual semantic vocabulary extraction method based on PCA-E
US11868432B1 (en) 2022-06-16 2024-01-09 Sichuan University Method for extracting kansei adjective of product based on principal component analysis and explanation (PCA-E)

Similar Documents

Publication Publication Date Title
CN110213222B (en) Network intrusion detection method based on machine learning
CN114611604A (en) User screening method based on electric drive assembly load characteristic fusion and clustering
CN108681742B (en) Analysis method for analyzing sensitivity of driver driving behavior to vehicle energy consumption
CN111160401A (en) Abnormal electricity utilization judging method based on mean shift and XGboost
CN110633729A (en) Driving risk hierarchical clustering method for intelligent networking vehicle group test
CN111401785A (en) Power system equipment fault early warning method based on fuzzy association rule
CN112488226B (en) Terminal abnormal behavior identification method based on machine learning algorithm
CN108268886B (en) Method and system for identifying plug-in operation
CN113688558A (en) Automobile driving condition construction method and system based on large database samples
CN110562261B (en) Method for detecting risk level of driver based on Markov model
CN112633337A (en) Unbalanced data processing method based on clustering and boundary points
CN112365060B (en) Preprocessing method for network Internet of things sensing data
CN113489685A (en) Secondary feature extraction and malicious attack identification method based on kernel principal component analysis
CN113064976A (en) Accident vehicle judgment method based on deep learning algorithm
CN113569462A (en) Distribution network fault level prediction method and system considering weather factors
CN116796271A (en) Resident energy abnormality identification method
CN113569920A (en) Second neighbor anomaly detection method based on automatic coding
CN114756420A (en) Fault prediction method and related device
CN115797044B (en) Credit wind control early warning method and system based on cluster analysis
CN116244657A (en) Train axle temperature abnormality identification method based on generation of countermeasure network and ensemble learning
CN113705920B (en) Method for generating water data sample set for thermal power plant and terminal equipment
CN113792141A (en) Feature selection method based on covariance measurement factor
CN113591780A (en) Method and system for identifying driving risk of driver
CN113837865A (en) Method for extracting multi-dimensional risk feature strategy
CN111127184B (en) Distributed combined credit evaluation method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination