CN111428386A - Elevator traction machine rotor fault diagnosis information fusion method based on complex network - Google Patents

Abstract

The invention relates to an elevator traction machine rotor fault diagnosis information fusion method based on a complex network. The correlation of the characteristic parameters is obtained by calculating the Pearson coefficients among the characteristic parameters, and the characteristic parameters are grouped according to the strength of the correlation; respectively calculating the similarity between sample data vectors according to the fault sample data of each group, and constructing a fault data complex network; acquiring a likelihood belief table and a reference center vector of each group by using a complex network community division method, and further acquiring corresponding reference evidence; and respectively activating the diagnostic evidences of each group aiming at the fault characteristic vector acquired on line, carrying out evidence fusion by using a Dempster combination rule, and making a fault decision by using the fused evidences to obtain a fault type corresponding to the on-line fault characteristic data. The invention carries out fusion reasoning of fault diagnosis evidence on the basis of a complex network and effectively improves the fault diagnosis precision of the elevator traction machine rotor by utilizing multi-source diagnosis information.

Description

Elevator traction machine rotor fault diagnosis information fusion method based on complex network

Technical Field

The invention relates to a complex network-based elevator traction machine rotor fault diagnosis information fusion method, and belongs to the technical field of elevator equipment state monitoring and fault diagnosis.

Background

The tractor is the power equipment of the elevator, is composed of a motor, a brake, a coupling, a reduction gearbox, a traction sheave, a frame and the like, and has the function of conveying and transmitting power to enable the elevator to normally run. Because the running working conditions of the elevator are complex and changeable, and uncertain factors in the installation, operation and maintenance processes are added, the traction machine serving as an elevator heart has frequent faults, the safe running of the elevator is influenced, even safety accidents are caused, and casualties and economic losses are caused. Therefore, it is imperative to monitor the state of the hoisting machine and accurately and efficiently diagnose the root cause of the failure.

The existing elevator traction machine mostly adopts a permanent magnet synchronous gearless traction machine, which comprises a machine base, a stator, a rotor body, a brake and the like, wherein the permanent magnet is fixed on the inner wall of the rotor body, the rotor body is arranged on a shaft through a key, and the shaft is arranged on a bilateral sealing deep groove ball bearing on a rear machine base and a self-aligning roller bearing on a front machine base. Identifying faults by vibration signals is a common mechanical equipment diagnosis strategy. During the operation of the traction machine, a plurality of components vibrate, and how to effectively utilize the vibration signals is the key for realizing accurate fault diagnosis. Considering the relevance and diversity among the collected vibration signal sample data, if such characteristics of the data can be deeply mined and utilized, the effectiveness of the diagnosis result must be improved.

As an emerging theoretical tool, the complex network can effectively model and analyze the incidence relation between entities (data), and provides support for solving the problem of data incidence analysis. Meanwhile, the information fusion method can properly dispose diversified data through integration of multi-source diagnosis data. Therefore, the invention provides an elevator traction machine rotor fault diagnosis information fusion method based on a complex network.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides an elevator traction machine rotor fault diagnosis information fusion method based on a complex network.

The correlation of the characteristic parameters is obtained by calculating the Pearson coefficients among the characteristic parameters, and the characteristic parameters are grouped according to the strength of the correlation; respectively calculating the similarity between sample data vectors according to the fault sample data of each group, and constructing a fault data complex network; acquiring a likelihood belief table and a reference center vector of each group by using a complex network community division method, and further acquiring corresponding reference evidence; the method comprises the steps of respectively activating diagnosis evidences of all groups aiming at fault feature vectors acquired on line, carrying out evidence fusion by using Dempster combination rules, making a fault decision by using the fused evidences, and obtaining a fault type corresponding to the on-line fault feature data.

The invention provides an elevator traction machine rotor fault diagnosis information fusion method based on a complex network, which comprises the following steps:

(1) setting a fault set theta of an elevator traction machine rotor as F₁,…,F_i,…,F_N|i＝1,2,…,N}，F_iAnd (3) representing the ith fault in the fault set theta, wherein N is the number of the fault modes contained in the elevator traction machine rotor.

(2) Let f_1,i，f_2,iAnd f_3,iTo be able to reflect each fault F in the set of faults Θ_iThe characteristic parameters are acceleration signals and are respectively provided by acceleration sensors at the fan end, the motor base and the motor shell driving end;

will f is_1,i，f_2,i，f_3,iAnd F_iExpressed as a set of samples M_i＝{[f_1,i,f_2,i,f_3,iF_i]|t＝1,2,3,…,S_iIn which [ f)_1,i(t),f_2,i(t),f_3,i(t),F_i]Is a sample vector, S_iIndicates that the fault is F_iSample data in the State, and denoted F_iNumber of samples in state, take S_iNot less than 100; respectively sampling sample data under each fault state and expressing the sample data in a set form

In total, a number of samples can be obtained,

| M | represents the number in the set M.

(3) Will fail F_iSample data f acquired in state_1,i(t)、f_2,i(t) and f_3,i(t) is expressed as a sample set M_i′＝{[f_1,i(t),f_2,i(t),f_3,i(t)]|t＝1,2,…,S_iRespectively sampling sample data in each fault state, and expressing the sample data in a set form

Satisfy | M | ═ M |)_i′|，|M_i' | denotes the set M_i' number of elements in.

(4) Carrying out correlation analysis on the sample fault characteristic parameters, and specifically calculating as follows:

two characteristic parameters with larger Pearson coefficients are grouped into the same group and marked as G_AThe characteristic parameters with less correlation are classified into another group, which is marked as G_B。

(5) For two groups of characteristic data G obtained in the step (4)_AAnd G_BRespectively constructing data complex network Net^AAnd Net^BDescribed in similarity matrices A and B, a_i,j，b_i,jAre elements in the matrices A and B, respectively, representing the sample x in each set of feature data_iAnd x_jDistance d between_i,jIs defined as formula (2):

a_i,j＝exp(-15×d_i,j)，b_i,j＝exp(-15×d_i,j) (2)

in the formula (d)_i,jUsing the euclidean distance metric, the similarity of the samples themselves is defined as 0, i.e., a when i ═ j_i,j＝0，b_i,j＝0。

(6) The complex network Net of data is divided by using the complex network community division principle^AAnd Net^BIs divided into K₁And K₂Class, is marked as

Including M sample vectors

To obtain K₁Class-corresponding sample Q_k1＝{[f_1,k1(u_k1),f_2,k1(u_k1),f_3,k1(u_k1),F_k1(u_k1)]|u_k1＝1,2,…,U_k1}，U_k1Represents T_k1Number of sample vectors in class, with Q_k1∈M，

F_k1(u_k1) ∈ theta, and can be formed from K₁Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₁A reference center vector, which may be denoted as C_k1＝[c_k1,1,c_k1,2,c_k1,3]Wherein K1 is 1,2, …, K₁。

The same can put the sample vector in M into

In (b) to obtain K₂Class-corresponding sample Q_k2＝{[f_2,k2(u_k2),f_2,k2(u_k2),f_3,k2(u_k2),F_k2(u_k2)]|u_k2＝1,2,…,U_k2}，U_k2Represents T_k2Number of sample vectors in class, with Q_k2∈M，

F_k2(u_k2) ∈ theta, and can be formed from K₂Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₂A reference center vector, which may be denoted as C_k2＝[c_k2,1,c_k2,2,c_k2,3]Wherein K2 is 1,2, …, K₂。

(7) F obtained according to the step (2), the step (3), the step (5) and the step (6)_iAnd T_k1、T_k2The relational table shown in Table 1 and Table 2 is constructed to show F_iAnd T_k1、F_iAnd T_k2The corresponding relation between the two; n is a radical of_k1,i、N_k2,iRespectively represents T_k1、T_k2The sample set corresponding to the class has a fault F_iIn combination with

And

and

wherein N is more than or equal to 0_k1,i、N_k2,i≤S_i。

TABLE 1F_iAnd T_k1Table of corresponding relationship between

TABLE 2F_iAnd T_k2Table of corresponding relationship between

(8) According to the corresponding relation table obtained in the step (7), when the fault is F_iWhen the sample data is classified into the k1 and k2, the likelihood function is:

and is provided with

Then the reference evidence of the i-th type fault corresponding to the k1 and k2 types can be defined as:

obtaining two parts of fault diagnosis reference evidence as

e_p1＝[e_p1,1,e_p1,2,…,e_p1,N]，e_p2＝[e_p2,1,e_p2,2,…,e_p2,N](5)

Constructing likelihood confidence tables as shown in tables 3 and 4 to describe T_k1、T_k2And F_iThe relationship between them.

TABLE 3T_k1Similar confidence table

TABLE 4T_k2Similar confidence table

(9) When the online monitoring obtains the fault characteristic parameter vector X (t) ═ f at the moment t_1,i(t),f_2,i(t),f_3,i(t)]Thereafter, an importance weight w of the evidence is defined_kDescription of evidence e_kRespectively solving the fault characteristic parameter vector and K according to the relative importance of the fault characteristic parameter vector and other evidences₁、K₂A reference center vector

And normalizing the Euclidean distance to obtain the Dis_k1And Dis_k2The calculation is as follows:

definition of w_k1＝Dis_k1，w_k2＝Dis_k2。

(10) Using importance weights w of evidence_kTo activate the reference evidence e_pThe calculation is as follows:

e_{p_1}＝w_k1*e_p1＝[e_p1,1′,e_p1,2′,…,e_p1,N′]，

e_{p_2}＝w_k2*e_p2＝[e_p2,1′,e_p2,2′,…,e_p2,N′](7)

e is to be_{p_1}、e_{p_2}Respectively carrying out normalization to obtain diagnosis evidence e_p1′、e_p2′。

(11) Combining rule pairs e by Dempster_p1′、e_p2' fusion is performed, and the fused diagnosis results are:

in the formula m₁、m₂Let reference evidence m for two quality functions defined on Θ₁＝e_p1′，m₂＝e_p2' definition

As a function of the combined mass,

indicating that Dempster combination rules may be applied to two or more quality functions, A, B,C is a fault type mode;

the diagnostic evidence after fusion was:

e_p＝[m(1),m(2),…,m(N)](9)

wherein m (1), m (2), m (N) respectively represent the failure type F after fusion and normalization₁、F₂、F_NAnd (7) reliability.

(12) Using the diagnostic evidence e obtained in step (11)_pAnd diagnosing the fault of the elevator traction machine rotor: e.g. of the type_pF corresponding to the maximum confidence value_iNamely the fault mode of the real fault characteristic parameter vector X (t).

In summary, the method comprises the steps of firstly determining a fault set and fault characteristic parameters of an elevator traction machine rotor, and respectively sampling sample data in each fault state to obtain a fault characteristic data sample set; the correlation of the characteristic parameters is obtained by calculating Pearson coefficients among the characteristic parameters, and the characteristic parameters are grouped according to the strength of the correlation; respectively calculating the similarity between sample data vectors according to the fault sample data of each group, and constructing a fault data complex network; acquiring a likelihood belief table and a reference center vector of each group by using a complex network community division method, and further acquiring corresponding reference evidence; and respectively activating the diagnostic evidences of each group aiming at the fault characteristic vector acquired on line, carrying out evidence fusion by using a Dempster combination rule, and making a fault decision by using the fused evidences to obtain a fault type corresponding to the on-line fault characteristic data. The program (compiling environment Matlab) compiled by the method can run on a monitoring computer, and is combined with hardware such as a sensor, a data collector and the like to form an online monitoring system for carrying out real-time monitoring and fault diagnosis on the rotor state of the elevator traction machine.

The invention has the beneficial effects that: 1. grouping the characteristic data by calculating a Pearson coefficient, so that the constructed complex network community is more accurate; 2. the acquired reference central point is more accurate by using the complex network community division, so that the acquisition of the reference evidence is more convenient and faster; 3. and defining the Euclidean distance from the sample vector to the community center as the evidence weight, thereby solving the defect of manually setting the evidence weight in the prior art.

Drawings

FIG. 1 is a block flow diagram of the method of the present invention.

Fig. 2 is a diagram of an elevator machine rotor fault diagnostic system.

Fig. 3 is a block diagram of an elevator traction machine rotor fault diagnosis system in an example of the method of the present invention.

Detailed Description

The invention relates to an elevator traction machine rotor fault diagnosis information fusion method based on a complex network, the flow chart of which is shown in figure 1, and the method comprises the following steps:

(1) setting a fault set theta of an elevator traction machine rotor as F₁,…,F_i,…,F_N|i＝1,2,…,N}，F_iAnd (3) representing the ith fault in the fault set theta, wherein N is the number of the fault modes contained in the elevator traction machine rotor.

(2) Let f_1,i，f_2,iAnd f_3,iTo be able to reflect each fault F in the set of faults Θ_iThe characteristic parameter is an acceleration signal, and the acceleration signals are respectively provided by acceleration sensors at the fan end, the motor base and the motor shell driving end, and f is measured by a sensor_1,i，f_2,i，f_3,iAnd F_iExpressed as a set of samples M_i＝{[f_1,i,f_2,i,f_3,iF_i]|t＝1,2,3,…,S_iIn which [ f)_1,i(t),f_2,i(t),f_3,i(t),F_i]Is a sample vector, S_iIndicates that the fault is F_iSample data in the State, and denoted F_iNumber of samples in state, take S_iNot less than 100; respectively sampling sample data under each fault state and expressing the sample data in a set form

In total, a number of samples can be obtained,

| M | represents the number in the set M.

(3) Will fail F_iIn the state of obtainingSample data f taken_1,i(t)、f_2,i(t) and f_3,i(t) is expressed as a sample set M_i′＝{[f_1,i(t),f_2,i(t),f_3,i(t)]|t＝1,2,…,S_iRespectively sampling sample data in each fault state, and expressing the sample data in a set form

Satisfy | M | ═ M |)_i′|，|M_i' | denotes the set M_i' number of elements in.

(4) Carrying out correlation analysis on the sample fault characteristic parameters, and specifically calculating as follows:

two characteristic parameters with larger Pearson coefficients are grouped into the same group and marked as G_AThe characteristic parameters with less correlation are classified into another group, which is marked as G_B。

(5) For two groups of characteristic data G obtained in the step (4)_AAnd G_BRespectively constructing data complex network Net^AAnd Net^BDescribed in similarity matrices A and B, a_i,j，b_i,jAre elements in the matrices A and B, respectively, representing the sample x in each set of feature data_iAnd x_jDistance d between_i,jIs defined as formula (2):

a_i,j＝exp(-15×d_i,j)，b_i,j＝exp(-15×d_i,j) (2)

in the formula (d)_i,jUsing the euclidean distance metric, the similarity of the samples themselves is defined as 0, i.e., a when i ═ j_i,j＝0，b_i,j＝0。

(6) The complex network Net of data is divided by using the complex network community division principle^AAnd Net^BIs divided into K₁And K₂Class, is marked as

Including M sample vectors

To obtain K₁Class-corresponding sample Q_k1＝{[f_1,k1(u_k1),f_2,k1(u_k1),f_3,k1(u_k1),F_k1(u_k1)]|u_k1＝1,2,…,U_k1}，U_k1Represents T_k1Number of sample vectors in class, with Q_k1∈M，

F_k1(u_k1) ∈ theta, and can be formed from K₁Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₁A reference center vector, which may be denoted as C_k1＝[c_k1,1,c_k1,2,c_k1,3]Wherein K1 is 1,2, …, K₁。

The same can put the sample vector in M into

In (b) to obtain K₂Class-corresponding sample Q_k2＝{[f_2,k2(u_k2),f_2,k2(u_k2),f_3,k2(u_k2),F_k2(u_k2)]|u_k2＝1,2,…,U_k2}，U_k2Represents T_k2Number of sample vectors in class, with Q_k2∈M，

F_k2(u_k2) ∈ theta, and can be formed from K₂Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₂A reference center vector, which may be denoted as C_k2＝[c_k2,1,c_k2,2,c_k2,3]Wherein K2 is 1,2, …, K₂。

(7) F obtained according to the step (2), the step (3), the step (5) and the step (6)_iAnd T_k1、T_k2The relational table shown in Table 1 and Table 2 is constructed to show F_iAnd T_k1、F_iAnd T_k2The corresponding relation between the two; n is a radical of_k1,i、N_k2,iRespectively represents T_k1、T_k2The sample set corresponding to the class has a fault F_iIn combination with

And

and

wherein N is more than or equal to 0_k1,i、N_k2,i≤S_i。

TABLE 1F_iAnd T_k1Table of corresponding relationship between

TABLE 2F_iAnd T_k2Table of corresponding relationship between

For the convenience of understanding the correspondence table shown in table 1 and table 2, this is exemplified here. The elevator traction machine rotor shown in fig. 2 has 3 failure modes F (N)_i: eccentric F of tractor rotor₁Imbalance of the rotor of the traction machine F₂Loosening of parts of tractor rotor F₃Then the fault set Θ is { F ═ F₁,F₂,F₃Their common characteristic parameter f_1,i，f_2,iAnd f_3,iAnd vibration signals are provided for acceleration sensors mounted at the fan end, the motor base and the motor housing drive end. Get S₁＝600,S₂＝300,S₃Obtaining sample data in each fault state through the step (2), wherein the total sample data is 1020 sample data, and the step (3) and the step (4) can be implementedCalculate f₁、f₂、f₃Has a Pearson coefficient of r (f)₁，f₂)＝0.0202，r(f₁，f₃)＝-0.0907，r(f₂，f₃) When the expression is-0.1092, it is known that₂And f₃Has strong correlation, can convert f₂、f₃Composition feature data G_A，f₁Is another set of characteristic data G_BObtaining Net through the calculation of the step (5)^AAnd Net^BNet is performed by the step (6)^AIs divided into 4 communities with the community center of C₁₁＝[0.0162，-0.115,0.0023],C₁₂＝[-0.0256，0.0196,0.064],C₁₃＝[0.0129，0.1764，-0.0044],C₁₄＝[-0.0191，0.0358，-0.0324]The samples in M are classified as T in Table 1 in step (6)₁、T₂、T₃、T₄Will Net^BIs divided into 6 communities with the community center of C₂₁＝[-0.0272,-0.0407,0.002],C₂₂＝[-0.0695,0.0028,0.006],C₂₃＝[0.0542,0.0286,0.008],C₂₄＝[1.6053,0.0178,0.031],C₂₅＝[-0.2823,0.0438,0.0023],C₂₆＝[-0.6934,0.0225,0.0119]The samples in M are classified as T in Table 2 in step (6)₁、T₂、T₃、T₄、T₅、T₆As follows:

TABLE 3F_iAnd T_k1Table of corresponding relationship between

TABLE 4F_iAnd T_k2Table of corresponding relationship between

(8) According to the corresponding relation table obtained in the step (7), when the fault is F_iWhen the sample data is classified into the k1 and k2, the likelihood function is:

and is provided with

Then the reference evidence of the i-th type fault corresponding to the k1 and k2 types can be defined as:

obtaining two parts of fault diagnosis reference evidence as

e_p1＝[e_p1,1,e_p1,2,…,e_p1,N]，e_p2＝[e_p2,1,e_p2,2,…,e_p2,N](5)

Constructing likelihood confidence tables as shown in tables 3 and 4 to describe T_k1、T_k2And F_iThe relationship between them.

TABLE 5T_k1Similar confidence table

TABLE 6T_k2Similar confidence table

For the convenience of understanding the correspondence table shown in table 5 and table 6, the description is given here by way of example. According to the corresponding relation table obtained in the step (7), the current fault state is obtained as F according to the formula (4) in the step (8)_iTime is put in Table 5₁、T₂、T₃、T₄Likelihood function value of class

T in Table 6₁、T₂、T₃、T₄、T₅、T₆The likelihood function value of a class is

Obtaining the fault F by the formula (4) in the step (8)₁Evidence e_p1,1＝[0.331,0.167,0.361,0.141]^T，e_p2,1＝[0.223,0.31,0.257,0,0.2083,0.0017]^TSimilarly, F can be obtained₂、F₃Corresponding reference evidence of failure e_p1,2＝[0.397,0.113,0.383,0.107]^T,e_p1,3＝[0.283,0.192,0.442,0.083]^T,e_p2,2＝[0.17,0.283,0.153,0.007,0.32,0.067]^T,e_p2,3＝[0.092,0.333,0.117,0.0167,0.283,0.158]^TMeanwhile, likelihood confidence tables such as table 7 and table 8 can be constructed to describe class T_k1、T_k2And F_iThe relationship between:

TABLE 7T_k1Similar confidence table

TABLE 8T_k2Similar confidence table

(9) When the online monitoring obtains the fault characteristic parameter vector X (t) ═ f at the moment t_1,i(t),f_2,i(t),f_3,i(t)]Thereafter, an importance weight w of the evidence is defined_kDescription of evidence e_kRespectively solving the fault characteristic parameter vector and K according to the relative importance of the fault characteristic parameter vector and other evidences₁、K₂A reference center vector

And normalizing the Euclidean distance to obtain the Dis_k1And Dis_k2The calculation is as follows:

definition of w_k1＝Dis_k1，w_k2＝Dis_k2。

(10) Using importance weights w of evidence_kTo activate the reference evidence e_pThe calculation is as follows:

e is to be_{p_1}、e_{p_2}Respectively carrying out normalization to obtain diagnosis evidence e_p1′、e_p2′。

(11) Combining rule pairs e by Dempster_p1′、e_p2' fusion is performed, and the fused diagnosis results are:

in the formula m₁、m₂Let reference evidence m for two quality functions defined on Θ₁＝e_p1′，m₂＝e_p2' definition

As a function of the combined mass,

indicating that the Dempster combination rule may act on two or more quality functions, A, B, C being fault type patterns;

the diagnostic evidence after fusion was:

e_p＝[m(1),m(2),…,m(N)](9)

wherein m (1), m (2), m (N) respectively represent the failure type F after fusion and normalization₁、F₂、F_NAnd (7) reliability.

(12) Using the diagnostic evidence e obtained in step (11)_pAnd diagnosing the fault of the elevator traction machine rotor: e.g. of the type_pF corresponding to the maximum confidence value_iNamely the fault mode of the real fault characteristic parameter vector X (t).

In order to deepen the importance weight w corresponding to the sample vector X (t)_kIt is understood that the fault characteristic parameter vector x (t) ═ 1 at the time of online acquisition is [0.0421,0.1695, -0.0864]Substituting the above into formula (6) in step (9) to obtain the fault characteristic vectors x (t) and K at the time when t is equal to 1₁、K₂The Euclidean distance between the central vectors is normalized to obtain the importance weight w_k1＝[0.2045,0.1139,0.2911,0.3905]，w_k2＝[0.2345,0.0884,0.4876,0.0527,0.0835,0.053]Obtaining the reference evidence e after activation through formula (7) in the step (10)_{p_1}′＝[0.2471,0.2473,0.2409]，e_{p_2}′＝[0.2225,0.1704,0.1409]. Normalized to obtain e_p1′＝[0.3362,0.3369,0.3269]，e_p2′＝[0.4168,0,3192,0.264]. E obtained in step (10)_p1' and e_p2' substitution of equation (8) into step (11) yields fused diagnostic evidence e_p＝[0.4155,0.3230,0.2615]And diagnosing the fault of the elevator traction machine rotor: e.g. of the type_pF corresponding to m (1) with maximum confidence coefficient value₁Namely the fault mode of the real fault characteristic parameter vector X (t).

The embodiments of the present invention will be described in detail below with reference to the accompanying drawings:

the flow chart of the method of the invention is shown in figure 1, and the core part is as follows: calculating Pearson coefficients of the characteristic parameters to obtain the correlation of the characteristic parameters, and grouping the characteristic parameters according to the strength of the correlation; respectively calculating the similarity between sample data vectors according to the fault sample data of each group, and constructing a fault data complex network; acquiring a likelihood belief table and a reference center vector of each group by using a complex network community division method, and further acquiring corresponding reference evidence; and respectively activating the diagnostic evidences of each group aiming at the fault characteristic vector acquired on line, carrying out evidence fusion by using a Dempster combination rule, and making a fault decision by using the fused evidences to obtain a fault type corresponding to the on-line fault characteristic data.

The steps of the method of the present invention will be described in detail below with reference to the preferred embodiment of the elevator machine rotor fault diagnosis system of fig. 2.

1. Example of setting up simulation experiment table of elevator traction device

As shown in the structure diagram of the elevator traction machine rotor fault diagnosis system in fig. 3, the vibration acceleration sensors are respectively installed at the drive end of the motor housing, the fan end and the motor base to collect vibration signals, the vibration signals collected by the three sensors are transmitted to the HG-8902 data collection box, processed by the signal conditioning circuit, and finally output to the monitoring computer through the a/D converter, and then the time domain vibration acceleration signals of the elevator traction machine rotor are obtained as fault characteristic signals by using HG-8902 data analysis software in L beyond view environment.

2. Elevator traction machine rotor fault setting and fault characteristic parameter selection

According to the specific characteristics of the experiment table, 3 typical failure modes are respectively set on the experiment table: the rotor of the traction machine is eccentric, the rotor of the traction machine is unbalanced, and parts of the rotor of the traction machine are loosened. And collecting vibration signals as fault characteristic parameters by mounting the vibration signals at a motor shell driving end, a fan end and a motor base.

3. Correlation analysis, complex network community division and corresponding relation table construction

Get S₁＝600,S₂＝300,S₃Obtaining sample data in each fault state through the step (2), wherein the total sample data is 1020 sample data, and f can be calculated through the steps (3) and (4)₁、f₂、f₃Has a Pearson coefficient of r (f)₁，f₂)＝0.0202，r(f₁，f₃)＝-0.0907，r(f₂，f₃) When the expression is-0.1092, it is known that₂And f₃Has strong correlation, can convert f₂、f₃Composition feature data G_A，f₁Is the characteristic data G_BCalculating to obtain Net through the step (5)^AAnd Net^BNet is performed by the step (6)^ADivided into 4 communities, community center C₁₁＝[0.0162-0.115,0.0023],C₁₂＝[-0.0256,0.0196,0.064],C₁₃＝[0.0129,0.1764,-0.0044],C₁₄＝[-0.0191,0.0358,-0.0324]. The samples in M are classified into T in Table 1 in step (6) in the same way₁、T₂、T₃、T₄Will Net^BIs divided into 6 communities with the community center of C₂₁＝[-0.0272,-0.0407,0.002]，C₂₂＝[-0.0695,0.0028,0.006]，C₂₃＝[0.0542，0.0286,0.008]，C₂₄＝[1.6053，0.0178,0.031],C₂₅＝[-0.2823，0.0438,0.0023],C₂₆＝[-0.6934,0.0225，0.0119]The samples in M are classified as T in Table 2 in step (6)₁、T₂、T₃、T₄、T₅、T₆。

As follows:

TABLE 9F_iAnd T_k1Table of corresponding relationship between

TABLE 10F_iAnd T_k2Table of corresponding relationship between

5. Constructing a likelihood belief table and obtaining diagnostic evidence

According to the corresponding relation table obtained in the step (7), when the fault state is F, the formula (4) in the step (8) can obtain_iTime is put in Table 5₁、T₂、T₃、T₄The likelihood function value of a class is

T in Table 6₁、T₂、T₃、T₄、T₅、T₆The likelihood function value of a class is

Obtaining corresponding F from the formula (4) in the step (8)₁Reference evidence of failure e_p1,1＝[0.331,0.167,0.361,0.141]^T，e_p2,1＝[0.223,0.31,0.257,0,0.2083,0.0017]^TSimilarly, F can be obtained₂、F₃Corresponding reference evidence of failure e_p1,2＝[0.397,0.113,0.383,0.107]^T，e_p1,3＝[0.0.283,0.192,0.442,0.083]^T，e_p2,2＝[0.17,0.283,0.153,0.007,0.32,0.067]^T，e_p2,3＝[0.092,0.333,0.117,0.0167,0.283,0.158]^TMeanwhile, likelihood confidence tables such as table 7 and table 8 can be constructed to describe class T_k1、T_k2And F_iThe relationship between:

TABLE 11T_k1Similar confidence table

TABLE 12T_k2Similar confidence table

Acquiring fault characteristic parameter vector X (t) ([ 0.0421,0.1695, -0.0864) at time t ═ 1 online]Substituting the above into formula (6) in step (8) to obtain the fault characteristic vectors x (t) and K at the time when t is equal to 1₁、K₂The Euclidean distance between the central vectors is normalized to obtain the importance weight

w_k1＝[0.2045,0.1139,0.2911,0.3905]，

w_k2＝[0.2345,0.0884,0.4876,0.0527,0.0838,0.0530]，

Obtaining the activated reference evidence through the formula (7) in the step (9)

Normalized to obtain e_p1′＝[0.3362,0.3369,0.3269]，e_p2′＝[0.4168,0,3192,0.264]。

6. Fault diagnosis

E obtained in step (10)_p1' and e_p2' substitution of equation (8) into step (11) yields fused diagnostic evidence e_p＝[0.4155,0.3230,0.2615]And diagnosing the fault of the elevator traction machine rotor: e.g. of the type_pF corresponding to m (1) with maximum confidence coefficient value₁The fault mode which is the real occurrence of the fault characteristic parameter vector X (t) is consistent with the real fault mode which is set by collecting the group of fault characteristic parameter vectors, and the decision result is correct.

Claims (3)

1. The elevator traction machine rotor fault diagnosis information fusion method based on the complex network is characterized by comprising the following steps:

(1) setting a fault set theta of an elevator traction machine rotor as F₁,…,F_i,…,F_N|i＝1,2,…,N}，F_iRepresenting the ith fault in the fault set theta, wherein N is the number of fault modes contained in the elevator traction machine rotor;

(2) let f_1,i，f_2,iAnd f_3,iTo be able to reflect each in the failure set ΘA fault F_iThe characteristic parameters are acceleration signals and are respectively provided by acceleration sensors at the fan end, the motor base and the motor shell driving end;

will f is_1,i，f_2,i，f_3,iAnd F_iExpressed as a set of samples M_i＝{[f_1,i,f_2,i,f_3,iF_i]|t＝1,2,3,…,S_iIn which [ f)_1,i(t),f_2,i(t),f_3,i(t),F_i]Is a sample vector, S_iIndicates that the fault is F_iSample data in the State, and denoted F_iNumber of samples in state, take S_iNot less than 100; respectively sampling sample data under each fault state and expressing the sample data in a set form

In total, a number of samples can be obtained,

| M | represents the number in the set M;

(3) will fail F_iF obtained under the state_1,i(t)、f_2,i(t) and f_3,i(t) is expressed as a sample set M_i′＝{[f_1,i(t),f_2,i(t),f_3,i(t)]|t＝1,2,…,S_iRespectively sampling sample data in each fault state, and expressing the sample data in a set form

Satisfy | M | ═ M |)_i′|，|M_i' | denotes the set M_i' number of elements in;

(4) carrying out correlation analysis on the sample fault characteristic parameters to obtain a Pearson coefficient; two characteristic parameters with larger Pearson coefficients are grouped into the same group and marked as G_AThe characteristic parameters with less correlation are classified into another group, which is marked as G_B；

(5) For two groups of characteristic data G obtained in the step (4)_AAnd G_BRespectively constructing data complex network Net^AAnd Net^BDescribed in similarity matrices A and B, a_i,j，b_i,jAre elements in the matrices A and B, respectively, representing the sample x in each set of feature data_iAnd x_jDistance d between_i,jA function of (a);

(6) the complex network Net of data is divided by using the complex network community division principle^AAnd Net^BIs divided into K₁And K₂Class, is marked as

Including M sample vectors

To obtain K₁Class-corresponding sample Q_k1＝{[f_1,k1(u_k1),f_2,k1(u_k1),f_3,k1(u_k1),F_k1(u_k1)]|u_k1＝1,2,…,U_k1}，U_k1Represents T_k1Number of sample vectors in class, with Q_k1∈M，

F_k1(u_k1) ∈ theta, and can be formed from K₁Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₁A reference center vector, which may be denoted as C_k1＝[c_k1,1,c_k1,2,c_k1,3]Wherein K1 is 1,2, …, K₁；

The same can put the sample vector in M into

In (b) to obtain K₂Class-corresponding sample Q_k2＝{[f_2,k2(u_k2),f_2,k2(u_k2),f_3,k2(u_k2),F_k2(u_k2)]|u_k2＝1,2,…,U_k2}，U_k2Represents T_k2Number of sample vectors in class, with Q_k2∈M，

F_k2(u_k2) ∈ theta, and can be formed from K₂Obtaining K corresponding to the sample set M by taking the mean value of each sample-like vector₂A reference center vector, which may be denoted as C_k2＝[c_k2,1,c_k2,2,c_k2,3]Wherein K2 is 1,2, …, K₂；

(7) F obtained according to the step (2), the step (3), the step (5) and the step (6)_iAnd T_k1、T_k2The relational table shown in Table 1 and Table 2 is constructed to show F_iAnd T_k1、F_iAnd T_k2The corresponding relation between the two; n is a radical of_k1,i、N_k2,iRespectively represents T_k1、T_k2The sample set corresponding to the class has a fault F_iIn combination with

And

and

wherein N is more than or equal to 0_k1,i、N_k2,i≤S_i；

TABLE 1F_iAnd T_k1Table of corresponding relationship between

TABLE 2F_iAnd T_k2Table of corresponding relationship between

(8) According to the corresponding relation table obtained in the step (7), when the fault is F_iWhen the sample data is classified into the k1 and k2, the likelihood function is:

and is provided with

Then the reference evidence defining the i-th type fault corresponding to the k1 and k2 types is:

obtaining two parts of fault diagnosis reference evidence as

e_p1＝[e_p1,1,e_p1,2,…,e_p1,N]，e_p2＝[e_p2,1,e_p2,2,…,e_p2,N]

Constructing likelihood confidence tables as shown in tables 3 and 4 to describe T_k1、T_k2And F_iThe relationship between;

TABLE 3T_k1Similar confidence table

TABLE 4T_k2Similar confidence table

(9) When the online monitoring obtains the fault characteristic parameter vector X (t) ═ f at the moment t_1,i(t),f_2,i(t),f_3,i(t)]Thereafter, an importance weight w of the evidence is defined_kDescription of evidence e_kRespectively solving the fault characteristic parameter vector and K according to the relative importance of the fault characteristic parameter vector and other evidences₁、K₂A reference center vector

And normalizing the Euclidean distance to obtain the Dis_k1And Dis_k2Definition of w_k1＝Dis_k1，w_k2＝Dis_k2；

(10) Using importance weights w of evidence_kTo activate the reference evidence e_pTo obtain e_{p_1}、e_{p_2}E is to be_{p_1}、e_{p_2}Respectively carrying out normalization to obtain diagnosis evidence e_p1′、e_p2′；

(11) Combining rule pairs e by Dempster_p1′、e_p2' fusion is performed, and the fused diagnosis results are:

in the formula m₁、m₂Let reference evidence m for two quality functions defined on Θ₁＝e_p1′，m₂＝e_p2' Definitions m ═ m₁⊕m₂For the combined quality function, ⊕ indicates that Dempster combination rules can act on two or more quality functions, both A, B, C are failure type patterns;

the diagnostic evidence after fusion was:

e_p＝[m(1),m(2),…,m(N)](9)

wherein m (1), m (2), m (N) respectively represent the failure type F after fusion and normalization₁、F₂、F_NReliability;

(12) using the diagnostic evidence e obtained in step (11)_pAnd diagnosing the fault of the elevator traction machine rotor: e.g. of the type_pF corresponding to the maximum confidence value_iNamely the fault mode of the real fault characteristic parameter vector X (t).

2. The complex network-based elevator traction machine rotor fault diagnosis information fusion method according to claim 1, whichIs characterized in that: in step (5), a_i,j，b_i,jThe definition is as follows:

a_i,j＝exp(-15×d_i,j)，b_i,j＝exp(-15×d_i,j)

in the formula (d)_i,jUsing the euclidean distance metric, the similarity of the samples themselves is defined as 0, i.e., a when i ═ j_i,j＝0，b_i,j＝0。

3. The complex network-based elevator traction machine rotor fault diagnosis information fusion method according to claim 1, characterized in that: step (10) e_{p_1}、e_{p_2}Is calculated as follows:

e_{p_1}＝w_k1*e_p1＝[e_p1,1′,e_p1,2′,…,e_p1,N′]，

e_{p_2}＝w_k2*e_p2＝[e_p2,1′,e_p2,2′,…,e_p2,N′]。

Images