CN113049249A - Motor bearing fault diagnosis method and system - Google Patents
Motor bearing fault diagnosis method and system Download PDFInfo
- Publication number
- CN113049249A CN113049249A CN202011640103.6A CN202011640103A CN113049249A CN 113049249 A CN113049249 A CN 113049249A CN 202011640103 A CN202011640103 A CN 202011640103A CN 113049249 A CN113049249 A CN 113049249A
- Authority
- CN
- China
- Prior art keywords
- fault
- fault diagnosis
- particle
- library
- model
- 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
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 127
- 238000000034 method Methods 0.000 title claims abstract description 58
- 239000013598 vector Substances 0.000 claims abstract description 76
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000002372 labelling Methods 0.000 claims abstract description 37
- 238000013145 classification model Methods 0.000 claims abstract description 18
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000003213 activating effect Effects 0.000 claims abstract description 8
- 239000002245 particle Substances 0.000 claims description 173
- 230000007704 transition Effects 0.000 claims description 76
- 238000005457 optimization Methods 0.000 claims description 45
- 239000011159 matrix material Substances 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 25
- 238000012706 support-vector machine Methods 0.000 claims description 17
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 238000005096 rolling process Methods 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000002405 diagnostic procedure Methods 0.000 claims 1
- 238000004880 explosion Methods 0.000 abstract description 5
- 230000002401 inhibitory effect Effects 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 5
- 238000004445 quantitative analysis Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000004451 qualitative analysis Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012843 least square support vector machine Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Abstract
The invention provides a motor bearing fault diagnosis method and system, which can be used for: preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors; training the fault characteristic information vector to obtain a fault classification model; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model, and carrying out digital labeling on each output classification according to a preset labeling rule; analyzing the label obtained by digital labeling, and outputting a label representing the fault; building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network; training the built fault diagnosis model; and activating a corresponding library in the trained fault diagnosis model according to the output label, and then performing fault diagnosis reasoning according to the trained motor fault diagnosis model. The method is used for inhibiting the problem of space explosion of the fault diagnosis model of the complex system and improving the adaptability and accuracy of fault diagnosis of the fault diagnosis model of the bearing.
Description
Technical Field
The invention relates to the field of motor bearing fault diagnosis, in particular to a motor bearing fault diagnosis method and system.
Background
The development of scientific technology promotes the gradual miniaturization and integration of the motor, and the working state of a motor bearing directly influences the working efficiency and the anti-interference capability of the motor. Once the bearing fails, each functional module generates a nonlinear and fuzzy fault relation, which is not beneficial to equipment fault diagnosis and elimination. Therefore, motor bearing fault diagnosis is crucial to the normal operation of the motor.
The traditional motor bearing fault diagnosis mode mainly comprises a quantitative mode and a qualitative mode, wherein data driving based on quantitative analysis is a common fault diagnosis mode. For example, a method based on wavelet transformation, fuzzy entropy and a support vector machine is provided in the prior art, so that vibration interference signals of a bearing can be effectively removed, the fault diagnosis capability under strong noise is improved, but the frequency spectrum cannot be adaptively segmented, and the fault diagnosis lacks adaptability. The fault diagnosis method based on empirical mode decomposition and principal component analysis is provided in the prior art, a main bearing fault vibration signal is extracted through the empirical mode decomposition, the dimension of a feature vector is reduced by combining the principal component analysis, effective reduction of fault feature information is realized, and bearing fault classification is realized by combining a support vector machine. The prior art provides a fault diagnosis mode of a least square support vector machine, and solves the problem of low bearing fault classification accuracy by combining a nonlinear characteristic and time domain characteristic information fusion mode. In the prior art, a bearing fault diagnosis mode based on a deep belief network is provided, the network is formed by stacking a plurality of limited Boltzmann machines, characteristic information of data is found by combining low-level characteristics, real-time fault classification of continuous fault data is realized, and the accuracy and efficiency of fault diagnosis are improved. The above quantitative analysis-based method can perform signal analysis on the collected bearing fault data, and realize fault location by extracting the characteristic information, but the fault relation among the functional modules in the complex system has nonlinear and fuzzy characteristics, and the fault reasons of part of the functional modules cannot be diagnosed in a data-driven manner, so that the integrity is lacked.
Therefore, the invention provides a motor bearing fault diagnosis method and system, which are used for solving the problems.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a motor bearing fault diagnosis method and system, which are used for solving the problems that the traditional quantitative analysis mode fault diagnosis is lack of integrity and the qualitative analysis mode fault diagnosis is lack of timeliness, are used for inhibiting the problem of space explosion of a complex system fault diagnosis model, and are used for improving the adaptability of a bearing fault diagnosis model and the accuracy of fault diagnosis.
In a first aspect, the present invention provides a method for diagnosing a fault of a motor bearing, including:
step S1, collecting continuous vibration fault signals of a motor bearing;
step S2, preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors;
step S3, training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of a motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states;
step S4, analyzing the label obtained by digital labeling, and outputting a label representing the fault;
step S5, building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network;
step S6, training the set fault diagnosis model to obtain a trained fault diagnosis model;
and step S7, activating a corresponding library in the trained fault diagnosis model according to the output label, and then performing fault diagnosis reasoning according to the trained motor fault diagnosis model.
Further, the implementation method of step S2 includes:
step S21: EMD is adopted to carry out empirical mode decomposition on the continuous vibration fault signal;
step S22: taking the first 4 intrinsic mode components from the empirical mode decomposition result;
step S23: respectively calculating energy characteristics corresponding to the first 4 eigenmode components;
step S24: constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation;
step S25: and normalizing the energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
Further, the network structure of the CLPSO-FPN network is:
SCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in a bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix, which represents the mapping of P to T;
o is an output matrix which represents the mapping from T to P;
M=(m1,m2,...,mn) Representing the distribution vector identified by the library;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm) Representing a transition threshold distribution vector;
α=(α1,α,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a transition influence factor and represents the influence capacity of the transition on an output library of the transition, wherein r represents the number of all directional arcs of the transition to the output library of the transition;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
Further, training the parameters of the built motor fault diagnosis model based on a formula II-V in the step S6 until the training is finished when the average error F is minimum, and obtaining the trained motor fault diagnosis model;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error for the four operating conditions is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]Real numbers in the range, c being a learning factor of [0,2]Real numbers in the range, rand1i d、rand2i dRespectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing the speed and position of the d-dimensional parameter of the ith particle, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
Further, step S6 trains the parameters of the motor fault diagnosis model based on the formula (ii) - (v), the training process is:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjParticles L being globally optimal positionsg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
In a second aspect, the present invention provides a motor bearing fault diagnosis system, comprising:
the signal acquisition unit is used for acquiring continuous vibration fault signals of the motor bearing;
the signal preprocessing unit is used for preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors;
the continuous signal discretization unit is used for training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of the motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states;
the digital labeling unit is used for analyzing the label obtained by digital labeling and outputting a label representing the fault;
the model building module is used for building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network;
the model training module is used for training the built fault diagnosis model to obtain a trained fault diagnosis model;
and the fault diagnosis module is used for activating a corresponding library in the trained fault diagnosis model according to the output label and then carrying out fault diagnosis reasoning according to the trained motor fault diagnosis model.
Further, the signal pre-processing unit is configured to perform the following method steps:
step S21: EMD is adopted to carry out empirical mode decomposition on the continuous vibration fault signal;
step S22: taking the first 4 intrinsic mode components from the empirical mode decomposition result;
step S23: respectively calculating energy characteristics corresponding to the first 4 eigenmode components;
step S24: constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation;
step S25: and normalizing the energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
Further, the network structure of the CLPSO-FPN network is:
SCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in a bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix, which represents the mapping of P to T;
o is an output matrix which represents the mapping from T to P;
M=(m1,m2,...,mn) Representing the distribution vector identified by the library;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm) Representing a transition threshold distribution vector;
α=(α1,α,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a transition influence factor and represents the influence capacity of the transition on an output library of the transition, wherein r represents the number of all directional arcs of the transition to the output library of the transition;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
Further, the model training module trains the parameters of the built motor fault diagnosis model based on a formula II-V until the training is finished when the average error F is minimum, and a trained motor fault diagnosis model is obtained;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error for the four operating conditions is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]Real numbers in the range, c being a learning factor of [0,2]Real numbers in the range, rand1i d、rand2i dRespectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing the speed and position of the d-dimensional parameter of the ith particle, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
Further, the process of the model training module for carrying out the parameter of the motor fault diagnosis model based on the formula II-V comprises the following steps:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjFor global optimum bitParticles L placedg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
The beneficial effect of the invention is that,
the invention provides a fault diagnosis method and a system for a motor bearing, which aims at the problem that the traditional fault diagnosis method for the motor bearing lacks integrity and timeliness, provides a fault diagnosis method based on CLPSO-FPN, and is beneficial to improving the timeliness, the integrity and the accuracy of fault diagnosis of the motor bearing:
1) the method can effectively process the acquired continuous fault signals to obtain discrete fault classification signals to activate the fault library, and is favorable for solving the problems that the traditional quantitative analysis mode is lack of integrity in fault diagnosis and the traditional qualitative analysis mode is lack of timeliness in fault diagnosis;
2) the structure of the fault diagnosis model is optimized by adopting the transition influence factor, the problem of space explosion of the fault diagnosis model of the complex system is suppressed, the comprehensive particle swarm optimization is provided to optimize the parameters of the fault diagnosis model, and the adaptability of the bearing fault diagnosis model and the accuracy of fault diagnosis are improved;
3) the trained motor fault diagnosis model disclosed by the invention integrates CLPSO and FPN, and can be suitable for fault diagnosis of complex motor bearings to a certain extent.
In addition, the invention has reliable design principle, simple structure and very wide application prospect.
Drawings
In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present invention, the drawings used in the description of the embodiments or prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a schematic flow diagram of a method of one embodiment of the invention.
Fig. 2 is a signal diagram of a continuous vibration fault signal collected in the present invention.
FIG. 3 is a schematic diagram of a fault logic analysis process based on fault tree analysis according to the present invention.
FIG. 4 is an embodiment of the CLPSO-FPN-based bearing fault diagnosis model according to the present invention.
Fig. 5 shows a fault diagnosis model based on the FPN principle.
Fig. 6 is a schematic diagram of the four fault mean error curves of the present invention.
FIG. 7 is a schematic block diagram of a system of one embodiment of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiment of the present invention will be clearly and completely described below with reference to the drawings in the embodiment of the present invention, and it is obvious that the described embodiment is only a part of the embodiment of the present invention, and not all 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.
FIG. 1 is a schematic flow diagram of a method of one embodiment of the invention.
As shown in fig. 1, the method 100 includes:
and step S1, acquiring a continuous vibration fault signal of the motor bearing.
Specifically, the continuous vibration fault signal (raw data) of the motor bearing in the embodiment is derived from a continuous vibration fault signal of the motor bearing collected by an acceleration sensor at the driving end of a bearing of the university of kessensch laboratory SKF6205, which is shown in fig. 2. In FIG. 2, the horizontal axis represents the number of sampling points, and the vertical axis represents the amplitude (m/g) of the sampling points2)。
And step S2, preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors.
Step S2 is used to preprocess the continuous vibration fault signal of the motor bearing collected in step S1. Specifically, the pretreatment is as follows:
step S21: and performing empirical mode decomposition on the continuous vibration fault signal by adopting EMD.
In this embodiment, the continuous vibration fault signal is subjected to EMD decomposition to obtain n IMF components with different time characteristic scales, which are denoted as xi,i=1,2,…,n。
Step S22: decomposition of the result x from empirical modeiThe first 4 eigenmode components are taken (i ═ 1,2, …, n).
Wherein, the first 4 eigenmode components taken in this embodiment are x1、x2、x3And x4。
x1、x2、x3And x4IMF component 1, IMF component 2, IMF component 3, and IMF component 4 in that order.
Step S23: and respectively calculating energy characteristics corresponding to the first 4 eigenmode components.
Noting the eigenmode component xiThe energy characteristic corresponding to (i ═ 1,2,3,4) is Ei (i ═ 1,2,3,4), where:
step S24: and constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation.
In this embodiment, the constructed energy feature vector is [ E ]1,E2,E3,E4]。
Step S25: and normalizing the constructed energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
Specifically, if the fault characteristic information vector is represented by V, the respective calculation formulas of the fault characteristic information vector V are as follows:
V=[E1/E,E2/E,E3/E,E4/E]。
step S3, training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of a motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states.
And the actually measured bearing vibration signal is the collected bearing vibration signal of the motor bearing to be measured.
In the present embodiment, all the operating states of the motor bearing are classified into four types in advance, i.e., normal operation, rolling element failure, inner ring failure, and outer ring failure.
In the present embodiment, normal operation, rolling element failure, inner ring failure, and outer ring failure in four operation states are digitally labeled with arabic numerals 1,2,3, and 4 in sequence. After the digital labeling is completed, the arabic numerals 1,2,3, and 4 all become labels, which represent four different operating states of the motor bearing, that is, represent four different faults of the motor bearing, specifically, the label "1" represents that the operating state of the motor bearing is normal operation, the label "2" represents that the operating state of the motor bearing is a rolling element fault, the label "3" represents that the operating state of the motor bearing is an inner ring fault, and the label "4" represents that the operating state of the motor bearing is an outer ring fault, which are abbreviated as: 1-normal operation, 2-rolling element failure, 3-inner ring failure and 4-outer ring failure.
The fault classification model may output a corresponding operation state of the motor bearing according to a bearing vibration signal of the motor bearing input in the model, where each classification output by the fault classification model represents an operation state of the motor bearing, for example, a current bearing vibration signal of the motor bearing is a bearing vibration signal a, and after the bearing vibration signal a is input in the fault classification model, the fault classification model may output an operation state of the motor bearing corresponding to the bearing vibration signal a (the operation state is one of the four operation states).
The use of digital labeling realizes the discretization processing of continuous signals.
After the above digital tagging, an ordered tag sequence can be obtained.
In step S4, the digitally labeled tag is analyzed, and a tag indicating the occurrence of a failure is output.
Specifically, in the tags obtained by digital tagging, when the number of times of m tags continuously appearing is not less than 3 times, the occurrence of the fault represented by the m tags is confirmed, and the m tags are output and used for activating the corresponding fault library in the subsequent steps. Wherein, m label is any one of labels 1,2,3 and 4.
And step S5, building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network.
Before a fault diagnosis model is built, fault logic analysis and fault data statistics of a motor bearing are required to be carried out.
Fault logic analysis
The fault logic analysis is used for analyzing the motor bearing fault position location, fault type classification and fault reason, completing the establishment of a fault event table and providing a basis for the establishment of a fault diagnosis model.
The bearing structure and fault logic of the motor are analyzed in a fault tree analysis mode. The fault logic analysis flow based on the fault tree analysis method is shown in fig. 3, and specifically includes:
(1) analyzing a bearing structure, wherein the structure of the bearing structure is divided into an inner ring, an outer ring and a rolling body;
(2) analyzing the fault position, type and fault reason of the bearing;
(3) and (3) establishing a fault reason by means of downward step-by-step analysis in a mode of establishing a top-level event on the basis of the step (2) until the fault is ended when the fault cannot be analyzed or no further analysis is needed, wherein the obtained fault is called a bottom-level event.
(4) And establishing a bearing fault event table (namely a motor bearing fault event table).
For example, a bearing fault event table may be established as shown in table 1.
TABLE 1 bearing failure event table
(II) failure data statistics
The effect of the fault data statistics is to determine the confidence alpha (alpha) of the library in the bearing fault diagnosis model1,α2,...,αn),αi∈[0,1]Represents a depot PiConfidence of the represented fault event.
Based on the obtained classifications (each classification corresponds to a label) in the step S3, counting the fault occurrence probability of each fault event in the motor bearing fault event table; and then carrying out fuzzy processing on the obtained fault occurrence probability by adopting a fuzzy theory mode, and reasoning the fuzzy processing result by adopting a Bayesian mode to obtain the correlation between the confidence coefficient of each library in the motor bearing fault event table and the fault occurrence probability. The fuzzy definition of confidence of the library can be shown in table 2, and the statistical table of confidence of the fault event can be shown in table 3.
TABLE 2 confidence fuzzy definitions
Table 3 fault event confidence statistics table
Wherein the confidence vectors of the library are: α ═ α (0.658,0.750,0.785,0.810,0.875,0.756,0.850,0.775,0.880,0.710, 0.650).
And (4) combining the fault logic analysis and the fault data statistics, and building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network. Wherein the network structure of the CLPSO-FPN network is SCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
Wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in the established bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix which represents the mapping from P to T, namely an incidence matrix representing the library P and the transition T;
o is an output matrix which represents the mapping from T to P, namely represents the correlation matrix of the transition T and the library P;
M=(m1,m2,...,mn) The distribution vector identified by the library is represented, the identification of the activated library is represented by 1, and the identification of the inactivated library is represented by 0;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm) Representing a transition threshold distribution vector;
α=(α1,α2,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a transition influence factor and represents the influence capacity of the transition on an output library of the transition, wherein r represents the number of all directional arcs of the transition to the output library of the transition;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
Fig. 4 is a schematic structural diagram of an embodiment of the fault diagnosis model constructed by the present invention. Fig. 5 is a fault diagnosis model obtained based on the FPN principle. Comparing fig. 4 and 5, it can be seen that the white rectangles in fig. 5 are optimized transitions, and the transition influence factor is beneficial to suppressing the space explosion problem. According to the method, a Gaussian function is used for replacing the traditional transition reliability, a concept of a transition influence factor is provided, the transfer capability of the transition to an output library of the transition is converted into the fuzzy transfer capability reflected by the Gaussian function and the transition influence factor through a probability value obtained by traditional expert experience, the conversion from experience to fuzzification is realized, the expression capability of complex relations among modules is improved, unnecessary structural routes in a traditional Petri network model are reduced, and the modeling of a complex system is facilitated. The model in fig. 4 effectively reduces the propagation path of the transition represented by the white rectangle in fig. 5, and plays a role of optimizing the model, thereby being beneficial to inhibiting the problem of space explosion.
And step S6, training the built fault diagnosis model to obtain a trained fault diagnosis model.
Specifically, in the training process, training the parameters of the motor fault diagnosis model based on a formula II-V until the training is finished when the average error F is minimum, and obtaining the trained motor fault diagnosis model;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error under four operating states (normal operation, rolling element failure, inner ring failure, outer ring failure) is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]Real numbers in the range, c (including c)1And c2) Is a learning factor of [0,2]The real number within the range is,respectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing the speed and position of the d-dimensional parameter of the ith particle, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
The comprehensive particle swarm optimization is a new intelligent optimization algorithm obtained by combining a comprehensive learning strategy on the basis of a particle swarm optimization, is different from the particle swarm optimization which takes particles as optimization units, and the optimization units of the comprehensive particle swarm optimization are different dimensions in the particles, so that the problem of large difference of parameters of all dimensions in the particles can be solved. Meanwhile, the local optimal value of the local optimization target is changed into a local optimal value obtained by randomly selecting particles through comparison, and the local optimization capability of the algorithm is improved. Compared with a particle swarm algorithm, the method is more suitable for parameter optimization of a fault diagnosis model of a complex system.
CLPSO-FPN is a 12-tuple, and the step S6 trains (parameter optimization) the parameters of the motor fault diagnosis model based on the above formula (ii) -v, where the training process is as follows:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjParticles L being globally optimal positionsg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
In the present embodiment, the parameters of the bearing fault diagnosis model are trained according to four fault states of the motor bearing (i.e. the operating states of the four motor bearings, i.e. normal operation, rolling element fault, inner ring fault and outer ring fault, as described above), and the particle number S is defined as 200, the dimension D is defined as 43, and the parameters are overlappedThe generation number K is 2000, and the error formula used in the optimization process isIn the formula, alpha (p)i) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
The average error for the four operating conditions is: f ═ F1+F2+F3+F4。
To ensure the optimality of the training results, the average error curves of the four failure states shown in fig. 6 are obtained through 400 independent training (the horizontal axis represents the training times, and the vertical axis represents the error values). The average errors are all less than 4 multiplied by 10 according to the error curve-3The average error E when training to 11 th is the minimum: f2.695 × 10-8At this time, the training errors of the four fault states are respectively: f1=4.074×10-9,F2=3.6414×10-10,F3=5.034×10-8,F4=5.301×10-8。
The weight, the transition threshold and the transition influence factor of the fault diagnosis model obtained at this time are as follows:
ω1,1=0.5237、ω2,1=0.4763、ω3,2=0.2457、ω4,2=0.7543、ω5,3=0.2146、ω6,3=0.2577、ω7,3=0.5277、ω7,4=0.6497、ω8,4=0.3503、ω9,5=0.6175、ω10,5=0.3825、ω12,7=0.3397、ω13,7=0.6603、ω18,11=0.2083、ω14,110.7917, setting the other weights as 1 according to the fuzzy Petri net theory;
H=(0.2689,0.3174,0.5972,0.1839,0.8000,0.1619,0.6541,0.4163,0.2243,0.5121,0.0821,0.1101);
B=(0.0254,0.2060,0,0.1358,0.4891,0.1715,0.0883,0.0585,0,0.1016,0,0,0)。
and step S7, activating a corresponding library in the trained fault diagnosis model according to the output label, and then performing fault diagnosis reasoning according to the trained motor fault diagnosis model.
In the trained motor fault diagnosis model, parameters in the model are determined.
After the parameters in the model are determined, fault diagnosis deductive reasoning is carried out on the obtained fault information based on the CLPSO-FPN principle, the fault probability when a fault event occurs is obtained, and the fault positioning and predicting functions are achieved. The method specifically comprises a forward reasoning part and a reverse reasoning part.
The forward reasoning is that on the premise of positioning the bottom layer fault, the transmission of fault information is realized according to the running direction of the directed arcs in the Petri network, and the probability of the fault of the top layer fault module is predicted through confidence reasoning. And the backward reasoning is to carry out backward reasoning according to the direction of the directional arc on the premise of positioning the top layer fault, and predict the fault occurrence probability of the bottom layer module by combining the mode of the minimum cut set occurrence rate, thereby realizing the functions of positioning and predicting the fault cause.
Forward reasoning
The specific forward reasoning is as follows:
1) first, three operators are defined:
(1) competition operator A is an m × n matrix, C is an n-dimensional vector, then Cj=max(aij) Where i is 1,2, …, m, j is 1,2, …, n.
(2) Operator for taking large value A, B and C are all m × n matrices, then Cij=max(aij,bij) Where i is 1,2, …, m, j is 1,2, …, n.
(3) Direct multiplier A and C are m × n matrices, b is an m-dimensional vector, then Cij=aij×biWhere i is 1,2, …, m, j is 1,2, …, n.
2) Determining an initial identification vector M of the label activated library according to the label activated library0Initial transition trigger vector S0And initial library confidence vector alpha0。
3) Transition trigger discrimination
Defining: x ═ X1,x2,…xm) X is an m-dimensional vector of the equivalent sum of the confidence coefficient of the library and the corresponding weight value product, XkThe value produced for the k-th iteration.
Xk=(αk*Mk)·W (3.1)
In order to improve the reasoning efficiency of the algorithm, a Sigmoid function is introduced, and the transition triggering condition is judged in a matrix calculation mode.
Wherein d is infinity, and S is (S)1,s2,…,sm) Is a transition trigger vector such that when x>Lambda, satisfies the trigger condition si1 is ═ 1; when x is less than or equal to lambda, si=0,SkThe transition trigger vector generated for the kth iteration.
4) Fault propagation reasoning
The CLPSO-FPN forward reasoning process reflects the dynamic propagation path of the fault, wherein the Token reflects the fault occurrence condition of each module of the system, and the Token is transmitted from the input library to the output library along with the triggering of transition. The distribution of the tokens in the Petri network is called a library identifier and is represented by a vector M. The fault propagation reasoning equation is as follows.
Wherein M iskFor the bin-identified vectors generated for the kth iteration, the change in the vector reflects the change in the tobken in the bin. A. thenRepresenting an n-dimensional row vector with elements all 1.
5) Confidence inference
To determine confidence in the library in the CLPSO-FPN model, the Gaussian function 1/exp (10b (x-1)) is used in the method based on the traditional fuzzy Petri net2) And instead of the reliability of the transition, the influence of the transition on the output library is reflected by a transition influence factor b. Meanwhile, aiming at the competition relation encountered in the Petri network fault diagnosis process, a competition operator is introduced, a matrix reasoning mode is optimized, and an optimized confidence reasoning formula is shown as a formula (3.4).
Wherein alpha iskGenerating a library confidence coefficient vector for the kth iteration, and performing fault diagnosis reasoning according to the CLPSO-FPN algorithm when the alpha is higher than the alphak+1=αkWhen so, the inference ends.
(II) reverse reasoning
The specific reverse reasoning is as follows:
1) transition trigger discrimination
The transition triggering discrimination mode of the reverse reasoning is different from the forward reasoning, the transition triggering does not need to meet the condition of a transition threshold value, and the information transmission from a transition output library to an input library is inevitable. Therefore, the reverse transition trigger formula is defined as shown in the following formula (3.5).
Wherein S isk -Reverse transition trigger vector, I, generated for the kth iteration-Input matrices with backward reasoning, i.e. with forward reasoningAnd outputting the matrix.
2) Fault propagation reasoning
The fault propagation reasoning of the reverse reasoning is different from the forward reasoning, the forward fault propagation reasoning reflects the dynamic propagation path of the system fault, and the reverse reasoning reflects the bottom layer fault related to the top layer fault, namely various reasons for the fault generation. In order to improve the matrix reasoning capability of the algorithm, a reverse fault propagation reasoning formula is defined as shown in the following formula (3.6).
Wherein M isk -Vector identified by the inverse library, O, generated for the kth iteration-I is the output matrix of the backward inference, i.e. the input matrix of the forward inference. When in useWhen so, the inference ends.
3) Confidence inference
Because transition triggering and fault propagation in the reverse reasoning process are inevitable, the probability of the fault occurrence of each module cannot be obtained by using a forward confidence reasoning mode, and a minimum cut set ordering mode is provided for solving the problem that the confidence value cannot be obtained through reverse reasoning. When the top module breaks down, the bottom fault causing the top fault is obtained through reverse transition triggering and fault propagation reasoning, and finally the probability of each bottom fault module is obtained through a minimum cut set sequencing mode, so that the overhauling blindness is avoided, the efficiency of equipment fault removal and maintenance is improved, and the operation stability is improved. If the minimal cut set is G ═ p1,p2,…,pnAnd the probability formula of the occurrence of the minimal cut set is as follows:
FIG. 7 illustrates an embodiment of a motor bearing fault diagnostic system according to the present invention.
As shown in fig. 7, the system 200 includes:
the signal acquisition unit 201 is used for acquiring a continuous vibration fault signal of a motor bearing;
the signal preprocessing unit 202 is used for preprocessing the acquired continuous vibration fault signals to obtain corresponding fault characteristic information vectors;
the continuous signal discretization unit 203 is used for training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of the motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states;
a digital labeling unit 204, configured to analyze the label obtained by digital labeling and output a label representing a fault;
the model building module 205 is used for building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network;
the model training module 206 is configured to train the established fault diagnosis model to obtain a trained fault diagnosis model;
and the fault diagnosis module 207 is used for activating a corresponding library in the trained fault diagnosis model according to the output label, and then performing fault diagnosis reasoning according to the trained motor fault diagnosis model.
Optionally, as an embodiment of the present invention, the signal preprocessing unit 202 is configured to perform the following method steps:
step S21: EMD is adopted to carry out empirical mode decomposition on the continuous vibration fault signal;
step S22: taking the first 4 intrinsic mode components from the empirical mode decomposition result;
step S23: respectively calculating energy characteristics corresponding to the first 4 eigenmode components;
step S24: constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation;
step S25: and normalizing the energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
Optionally, as an embodiment of the present invention, a network structure of the CLPSO-FPN network in the model building module 205 is: sCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
Wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in a bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix, which represents the mapping of P to T;
o is an output matrix which represents the mapping from T to P;
M=(m1,m2,...,mn) Representing the distribution vector identified by the library;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm) Representing a transition threshold distribution vector;
α=(α1,α,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a transition influence factor and represents the influence capacity of the transition on an output library of the transition, wherein r represents the number of all directional arcs of the transition to the output library of the transition;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
Optionally, as an embodiment of the present invention, the model training module 206 trains parameters of the built motor fault diagnosis model based on a formula (ii) - (v), and the training is completed until the average error F is minimum, so as to obtain a trained motor fault diagnosis model;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error for the four operating conditions is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]Real numbers in the range, c being a learning factor of [0,2]Real numbers in the range, rand1i d、rand2i dRespectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing the speed and position of the d-dimensional parameter of the ith particle, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
Optionally, as an embodiment of the present invention, the process performed by the model training module 206 on the parameters of the motor fault diagnosis model based on the formula (ii) -v) includes the steps of:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjParticles L being globally optimal positionsg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
The same and similar parts in the various embodiments in this specification may be referred to each other. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to the description in the method embodiment.
It should be noted that, in this embodiment, the SVM is a Support Vector Machine (Support Vector Machine), the FPN is based on a Fuzzy Petri Net theory (Fuzzy Petri Net), and the CLPSO is a Comprehensive Particle Swarm Optimization (Comprehensive Learning Particle Swarm Optimization, which is provided by the invention for the CLPSO-FPN network).
Although the present invention has been described in detail by referring to the drawings in connection with the preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made on the embodiments of the present invention by those skilled in the art without departing from the spirit and scope of the present invention, and these modifications or substitutions are within the scope of the present invention/any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A motor bearing fault diagnosis method is characterized by comprising the following steps:
step S1, collecting continuous vibration fault signals of a motor bearing;
step S2, preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors;
step S3, training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of a motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states;
step S4, analyzing the label obtained by digital labeling, and outputting a label representing the fault;
step S5, building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network;
step S6, training the set fault diagnosis model to obtain a trained fault diagnosis model;
and step S7, activating a corresponding library in the trained fault diagnosis model according to the output label, and then performing fault diagnosis reasoning according to the trained motor fault diagnosis model.
2. The motor bearing fault diagnosis method according to claim 1, characterized in that the implementation method of step S2 comprises:
step S21: EMD is adopted to carry out empirical mode decomposition on the continuous vibration fault signal;
step S22: taking the first 4 intrinsic mode components from the empirical mode decomposition result;
step S23: respectively calculating energy characteristics corresponding to the first 4 eigenmode components;
step S24: constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation;
step S25: and normalizing the energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
3. The motor bearing fault diagnosis method of claim 1, characterized in that the network structure of the CLPSO-FPN network is: sCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
Wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in a bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix, which represents the mapping of P to T;
o is an output matrix which represents the mapping from T to P;
M=(m1,m2,...,mn) Representing the distribution vector identified by the library;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm),representing a transition threshold distribution vector;
α=(α1,α,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a transition influence factor and represents the influence capacity of the transition on an output library of the transition, wherein r represents the number of all directional arcs of the transition to the output library of the transition;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
4. The motor bearing fault diagnostic method of claim 3,
training the parameters of the built motor fault diagnosis model based on a formula II-V in the step S6 until the training is finished when the average error F is minimum, and obtaining the trained motor fault diagnosis model;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error for the four operating conditions is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]RangeInner real number, c is a learning factor of [0,2]Real numbers in the range, rand1i d、rand2i dRespectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing the speed and position of the d-dimensional parameter of the ith particle, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
5. The motor bearing fault diagnosis method of claim 4, wherein step S6 trains the parameters of the motor fault diagnosis model based on the formula (II) -V), the training process is:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjParticles L being globally optimal positionsg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
6. A motor bearing fault diagnostic system, comprising:
the signal acquisition unit is used for acquiring continuous vibration fault signals of the motor bearing;
the signal preprocessing unit is used for preprocessing the collected continuous vibration fault signals to obtain corresponding fault characteristic information vectors;
the continuous signal discretization unit is used for training a fault feature information vector by using a Support Vector Machine (SVM) to obtain a fault classification model for classifying the running state of the motor bearing; carrying out online classification on the actually measured bearing vibration signals by using a fault classification model obtained by training, and carrying out digital labeling on each output classification according to a preset labeling rule; the running states of the motor bearing are normal running, rolling body fault, inner ring fault and outer ring fault; the preset labeling rule is that Arabic numerals 1,2,3 and 4 are adopted to carry out digital labeling on the four running states;
the digital labeling unit is used for analyzing the label obtained by digital labeling and outputting a label representing the fault;
the model building module is used for building a fault diagnosis model of the motor bearing based on the CLPSO-FPN network;
the model training module is used for training the built fault diagnosis model to obtain a trained fault diagnosis model;
and the fault diagnosis module is used for activating a corresponding library in the trained fault diagnosis model according to the output label and then carrying out fault diagnosis reasoning according to the trained motor fault diagnosis model.
7. The motor bearing fault diagnostic system of claim 6, wherein the signal pre-processing unit is configured to perform the method steps of:
step S21: EMD is adopted to carry out empirical mode decomposition on the continuous vibration fault signal;
step S22: taking the first 4 intrinsic mode components from the empirical mode decomposition result;
step S23: respectively calculating energy characteristics corresponding to the first 4 eigenmode components;
step S24: constructing a one-dimensional energy feature vector by using all the energy features obtained by the calculation;
step S25: and normalizing the energy characteristic vector to obtain a normalized energy characteristic vector, wherein the normalized energy characteristic vector is the fault characteristic information vector.
8. The motor bearing fault diagnostic system of claim 6, wherein the network structure of the CLPSO-FPN network is: sCLPSO-FPN=(P,T,I,O,M,W,H,α,B,S,D,K);
Wherein, P ═ { P ═ P1,p2,...,pnP represents a set of libraries in a bearing fault event table;
T={t1,t2,...,tmt represents a set of transitions;
i is an input matrix, which represents the mapping of P to T;
o is an output matrix which represents the mapping from T to P;
M=(m1,m2,...,mn) Representing the distribution vector identified by the library;
W=(ωij) An n x m dimensional matrix formed by the weight values of the library represents the influence degree of the input library on the transition;
H=(λ1,λ2,...,λm) Representing a transition threshold distribution vector;
α=(α1,α,...,αn),αi∈[0,1]represents a depot PiConfidence representing a fault event, i ═ 1,2, …, n;
B=(b1,b2,…,br) B represents a cause of influence of transitionA son representing the influence ability of the transition on its output library, wherein r represents the number of all directional arcs of the transition to its output library;
s represents the number of particles in the comprehensive particle swarm algorithm;
d represents the dimension of the particles in the comprehensive particle swarm optimization;
and K represents the iteration number of the comprehensive particle swarm optimization.
9. The motor bearing fault diagnostic system of claim 7,
the model training module trains the parameters of the built motor fault diagnosis model based on a formula II-V until the training is finished when the average error F is minimum, and a trained motor fault diagnosis model is obtained;
wherein, the speed updating formula of the particle swarm optimization is synthesized:
the average error for the four operating conditions is: f ═ F1+F2+F3+F4 ⑤
In formulas (II) and (III), k is an inertia constant of 0,1]Real numbers in the range, c being a learning factor of [0,2]Real numbers in the range, rand1i d、rand2i dRespectively represent two independent random numbers at d-dimension of i-th particle, and the values are [0,1 ]]Random number in the range, Vi d、Li dRespectively representing d-dimensional parameters of i-th particleSpeed and position, Lp d、Lg dParticles representing a local optimum position and a global optimum position, respectively;
in the formula IV, alpha (p) in the formulai) And α' (p)i) Respectively represent a library office piLibrary confidence through confidence inference and data statistics.
10. The motor bearing fault diagnosis system of claim 9, wherein the process of the model training module on the parameters of the motor fault diagnosis model based on the formula (ii) -v comprises the steps of:
1) randomly generating S particles by a comprehensive particle swarm algorithm, wherein the particles comprise the three parameters of W, B and H;
2) inputting parameters of S particles into a fault diagnosis model, and obtaining a library p in a supervised learning modenThe error F of (2) is compared to obtain the minimum error F, at this time, the corresponding particle LjParticles L being globally optimal positionsg dThe local optimal position of the particle is its initial position;
3) determining dimension W to be updated, and randomly selecting local optimal positions L of particles at other positions2Is Lp dAccording to the update formula of CLPSO-FPN, the W parameter in the particle is directed to Lp d、Lg dUpdating the direction of the corresponding parameter;
4) repeating the above process until the parameter update of each dimension of the particle is completed;
5) comparing the iterative particle error value with the original error value, determining the position of the particle with the minimum error after the particle iteration as the local optimal position of the particle, repeating the steps until all the particles are updated, and updating the global optimal position of the particle swarm;
6) repeating the steps 3) to 5) until the iteration is finished to obtain the globally optimal particle L at the momentg dIs the set of parameters of the final model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011640103.6A CN113049249A (en) | 2020-12-31 | 2020-12-31 | Motor bearing fault diagnosis method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011640103.6A CN113049249A (en) | 2020-12-31 | 2020-12-31 | Motor bearing fault diagnosis method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113049249A true CN113049249A (en) | 2021-06-29 |
Family
ID=76508261
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011640103.6A Pending CN113049249A (en) | 2020-12-31 | 2020-12-31 | Motor bearing fault diagnosis method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113049249A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114943259A (en) * | 2022-07-22 | 2022-08-26 | 深圳市世坤科技实业有限公司 | Equipment fault diagnosis system and method for processing electric element by laser |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010019952A2 (en) * | 2008-08-15 | 2010-02-18 | Merrimack Pharmaceuticals, Inc. | Methods, systems and products for predicting response of tumor cells to a therapeutic agent and treating a patient according to the predicted response |
CN102721545A (en) * | 2012-05-25 | 2012-10-10 | 北京交通大学 | Rolling bearing failure diagnostic method based on multi-characteristic parameter |
CN105005644A (en) * | 2015-06-29 | 2015-10-28 | 山东科技大学 | Method for detecting fault of three-phase asynchronous motor |
CN106441888A (en) * | 2016-09-07 | 2017-02-22 | 广西大学 | High-speed train rolling bearing fault diagnosis method |
CN106596116A (en) * | 2016-11-29 | 2017-04-26 | 西安理工大学 | Vibration fault diagnosis method of wind generating set |
CN106908132A (en) * | 2017-01-20 | 2017-06-30 | 山东科技大学 | A kind of method that strain gauge load cell failure is detected based on improved Petri net |
CN108562853A (en) * | 2018-03-29 | 2018-09-21 | 上海交通大学 | Method of Motor Fault Diagnosis based on error correcting output codes support vector machines and system |
-
2020
- 2020-12-31 CN CN202011640103.6A patent/CN113049249A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010019952A2 (en) * | 2008-08-15 | 2010-02-18 | Merrimack Pharmaceuticals, Inc. | Methods, systems and products for predicting response of tumor cells to a therapeutic agent and treating a patient according to the predicted response |
CN102721545A (en) * | 2012-05-25 | 2012-10-10 | 北京交通大学 | Rolling bearing failure diagnostic method based on multi-characteristic parameter |
CN105005644A (en) * | 2015-06-29 | 2015-10-28 | 山东科技大学 | Method for detecting fault of three-phase asynchronous motor |
CN106441888A (en) * | 2016-09-07 | 2017-02-22 | 广西大学 | High-speed train rolling bearing fault diagnosis method |
CN106596116A (en) * | 2016-11-29 | 2017-04-26 | 西安理工大学 | Vibration fault diagnosis method of wind generating set |
CN106908132A (en) * | 2017-01-20 | 2017-06-30 | 山东科技大学 | A kind of method that strain gauge load cell failure is detected based on improved Petri net |
CN108562853A (en) * | 2018-03-29 | 2018-09-21 | 上海交通大学 | Method of Motor Fault Diagnosis based on error correcting output codes support vector machines and system |
Non-Patent Citations (4)
Title |
---|
XUEZHEN CHENG: "Adaptive Fault Diagnosis of Motors Using Comprehensive Learning Particle Swarm Optimizer with Fuzzy Petri Net", 《COMPUTING AND INFORMATICS》 * |
徐可: "基于经验模态分解和支持向量机的滚动轴承故障诊断", 《控制理论与应用》 * |
王英伟: "一种多种群综合学习粒子群优化算法", 《数学的实践与认识》 * |
闫玉涛: "《机械可靠性工程》", 华中科技大学出版社 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114943259A (en) * | 2022-07-22 | 2022-08-26 | 深圳市世坤科技实业有限公司 | Equipment fault diagnosis system and method for processing electric element by laser |
CN114943259B (en) * | 2022-07-22 | 2022-10-21 | 深圳市世坤科技实业有限公司 | Equipment fault diagnosis system and method for processing electric element by laser |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Feature-level attention-guided multitask CNN for fault diagnosis and working conditions identification of rolling bearing | |
CN110703057B (en) | Power equipment partial discharge diagnosis method based on data enhancement and neural network | |
CN109800875A (en) | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine | |
CN111967486A (en) | Complex equipment fault diagnosis method based on multi-sensor fusion | |
CN112633317A (en) | CNN-LSTM fan fault prediction method and system based on attention mechanism | |
WO2021257128A2 (en) | Quantum computing based deep learning for detection, diagnosis and other applications | |
CN112000084B (en) | Intelligent BIT design method of controller module based on 1D-CNN and GRU-SVM | |
CN114676742A (en) | Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network | |
CN114856941A (en) | Offshore wind power plant and unit fault diagnosis operation and maintenance system and diagnosis operation and maintenance method thereof | |
Agahi et al. | Decision fusion scheme for bearing defects diagnosis in induction motors | |
CN113049249A (en) | Motor bearing fault diagnosis method and system | |
Chen et al. | An enhanced artificial bee colony-based support vector machine for image-based fault detection | |
CN116501444B (en) | Abnormal cloud edge collaborative monitoring and recovering system and method for virtual machine of intelligent network-connected automobile domain controller | |
JP2019139651A (en) | Program, device and method for classifying unknown multi-dimensional vector data groups into classes | |
CN111881159A (en) | Fault detection method and device based on cost-sensitive extreme random forest | |
Zhu et al. | Fault diagnosis of wheelset bearings using deep bidirectional long short-term memory network | |
CN116503379A (en) | Lightweight improved YOLOv 5-based part identification method | |
CN108898157B (en) | Classification method for radar chart representation of numerical data based on convolutional neural network | |
CN115713162A (en) | Distributed low-carbon energy station health state early warning prediction method | |
CN112598186B (en) | Improved LSTM-MLP-based small generator fault prediction method | |
CN114062812A (en) | Fault diagnosis method and system for metering cabinet | |
Parri et al. | A hybrid GAN based autoencoder approach with attention mechanism for wind speed prediction | |
Xu et al. | Fast and robust neural network based wheel bearing fault detection with optimal wavelet features | |
Vaidya et al. | HyperEstimator: Evolving Computationally Efficient CNN Models with Grammatical Evolution. | |
Jin et al. | Research on multi-sensor information fusion algorithm of fan detection robot based on improved BP neural network |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210629 |
|
RJ01 | Rejection of invention patent application after publication |