CN116124460A - Bearing life prediction method and system based on health index construction - Google Patents
Bearing life prediction method and system based on health index construction Download PDFInfo
- Publication number
- CN116124460A CN116124460A CN202211675864.4A CN202211675864A CN116124460A CN 116124460 A CN116124460 A CN 116124460A CN 202211675864 A CN202211675864 A CN 202211675864A CN 116124460 A CN116124460 A CN 116124460A
- Authority
- CN
- China
- Prior art keywords
- index
- health
- degradation
- bearing
- feature
- 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.)
- Granted
Links
- 230000036541 health Effects 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 72
- 238000010276 construction Methods 0.000 title claims abstract description 9
- 238000006731 degradation reaction Methods 0.000 claims abstract description 63
- 230000015556 catabolic process Effects 0.000 claims abstract description 62
- 230000008569 process Effects 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 6
- 238000012216 screening Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 27
- 238000005259 measurement Methods 0.000 claims description 26
- 238000005096 rolling process Methods 0.000 claims description 16
- 230000009467 reduction Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 239000013598 vector Substances 0.000 claims description 10
- 238000007405 data analysis Methods 0.000 claims description 7
- 238000012544 monitoring process Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 238000007476 Maximum Likelihood Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 16
- 239000011159 matrix material Substances 0.000 description 9
- 238000001228 spectrum Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 230000014509 gene expression Effects 0.000 description 4
- 230000035772 mutation Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036449 good health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000006467 substitution reaction Methods 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
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a bearing life prediction method constructed based on health indexes, which comprises the following steps: establishing a health index based on an analytic hierarchy process and a differential algorithm, and establishing a residual service life prediction model based on a Weibull proportional risk (WPHM) model; the analytic hierarchy process is used for screening sensitive degradation characteristics, the differential algorithm is used for carrying out weight optimization on the sensitive degradation characteristic set, and a linear weighting method is adopted for constructing one-dimensional health indexes of residual service life prediction; the WPHM model is used for parameter optimization and constructing a prediction function of the residual service life, and the constructed one-dimensional health index is substituted into the WPHM prediction function to obtain the residual service life of the bearing. The invention also discloses a bearing life prediction system constructed based on the health index, which is developed by using a LabVIEW platform, and is used for predicting the residual service life of the bearing based on the construction of the health index and the establishment of a prediction model.
Description
Technical Field
The invention relates to the technical field of mechanical life prediction, in particular to a bearing life prediction method and system constructed based on health indexes.
Background
Rolling bearings are one of the most critical parts in rotary machines, and are extremely prone to failure due to complex working conditions, thereby causing equipment downtime, economic loss and even casualties. In view of this, carry out accurate prediction to antifriction bearing's remaining life and arrange the maintenance plan rationally according to the prediction result, it is all significant to improve equipment reliability, reduce equipment shut down and maintenance cost, avoid the emergence of major incident.
In general, the prediction of the remaining service life of a rolling bearing is briefly described as three steps: data acquisition, health index construction and residual service life prediction model establishment. Health indexes play an important role in the prediction of the residual service life, and a good health index can simplify the prediction model and obtain an accurate prediction result. Generally, when constructing a bearing health index, sensitive degradation features are first screened by constructing a mixed metric index, and then a plurality of screened features are fused to obtain the health index. However, when constructing the mixed measurement index, each measurement index is usually set with a weight directly and manually, and has strong subjectivity. Meanwhile, when the health index is built, most models use a linear method to perform feature fusion, but the mechanical degradation process is generally nonlinear, so that the method has certain limitation.
Disclosure of Invention
The invention aims to improve the accuracy of predicting the residual service life of a rolling bearing and provides a method and a system for predicting the service life of the bearing constructed based on health indexes.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a bearing life prediction method constructed based on health indexes comprises the following steps:
s101, constructing health indexes based on an analytic hierarchy process and a differential algorithm;
s102, constructing a WPHM life prediction function based on the Weibull proportion risk model, and substituting the processed health index into the WPHM life prediction function to obtain the residual service life of the bearing.
Preferably, the specific process for constructing the health index based on the analytic hierarchy process and the differential algorithm comprises the following steps:
s201, collecting vibration signals of the rolling bearing, extracting degradation characteristics of three aspects of time domain, frequency domain and time-frequency domain according to the vibration signals, and carrying out noise reduction and normalization treatment on each degradation characteristic;
s202, calculating weights of four measurement indexes of monotonicity, robustness, trend and consistency of each degradation characteristic by using an analytic hierarchy process, and constructing a mixed measurement index based on the weights of the four measurement indexes;
s203, constructing the sensitive degradation characteristic set based on the mixed measurement index;
s204, reducing the dimension of the sensitive degradation feature set to the intrinsic dimension by using an equidistant feature mapping algorithm through an estimation method based on manifold hypothesis;
and S205, carrying out weight optimization on the reduced sensitive degradation feature set by using a differential algorithm, and constructing a one-dimensional health index of residual service life prediction by using a linear weighting method.
Preferably, the noise reduction and normalization process specifically includes: carrying out noise reduction treatment on the original degradation characteristics by using a 7-point moving average method so as to enable the characteristic curve to be relatively smooth; normalizing all the features to ensure that the range is between 0 and 1, wherein the 7-point sliding average method formula is as follows:
wherein x is n Is the original characteristic signal; n is the data length of the vibration signal; x is x n MA Is a new characteristic signal after noise reduction;
the normalization formula is:
wherein X is norm Is a normalization result; x is a characteristic sequence; x is X min ，X max Is the minimum and maximum in the feature sequence.
Preferably, the weights of the four metrics of monotonicity, robustness, trend and consistency are calculated by a pairwise comparison method, in particular,
the monotonicity index formula is:
wherein X= { X k } k＝1:K To degenerate the characteristic sequence, x k Indicated at t k The feature value of the moment, K is the total number of feature values in the degraded feature sequence;
representing differences in the feature sequence;
And->
The number of the positive and negative derivatives is respectively represented;
the robustness index formula is:
wherein x is k Indicated at t k A characteristic value of the moment;
is characterized by t k Average trend value of time;
the trend index formula is:
in the method, in the process of the invention,
and->
Respectively the degenerate feature sequence { x } k } k＝1:K Sum time { t k } k＝1:K Is a sequence of sequences ordered by (a);
the consistency index formula is:
wherein P is EOL Is a vector formed by degradation characteristic values of the bearing at failure time, P O Is a vector composed of degradation eigenvalues at the initial time.
Preferably, the mixed measurement index is obtained by linearly weighting four evaluation indexes of monotonicity, robustness, trend and consistency of degradation characteristics, and the calculation formula is as follows:
wherein alpha is i Is the weight of each index.
Preferably, the method for constructing the sensitive degradation characteristic set comprises the following steps: after the mixed measurement index of each feature is calculated, a mixed measurement index threshold is set, and the degradation feature of the feature mixed measurement index exceeding the threshold is selected as the sensitive degradation feature to construct a sensitive degradation feature set.
Preferably, the method for constructing the health index comprises the following steps: and optimizing the weight of the dimensionality reduced sensitive degradation characteristic set by using a differential algorithm, and multiplying each optimized weight coefficient by the corresponding sensitive degradation characteristic by using a linear weighting method so as to construct a one-dimensional health index of residual service life prediction.
Preferably, the step S102 specifically includes:
s701, calculating a parameter estimation value of the WPHM model by using a maximum likelihood estimation method, and substituting the parameter estimation value into the WPHM model to obtain a WPHM life prediction function;
s702, reconstructing the phase space of the health index, substituting the health index into an extreme learning machine for training to obtain a pseudo health index at a predicted time point, and substituting the pseudo health index into a WPHM life prediction function to obtain the residual service life of the rolling bearing.
The invention also provides a bearing life prediction system constructed based on health indexes, which is developed based on LabVIEW, and comprises a state monitoring system and a life prediction system, wherein the state monitoring system comprises a data acquisition function, a data analysis and processing function and a state alarm function; the life prediction system comprises screening of sensitive degradation characteristics, construction of health indexes, and prediction results and prediction errors of residual service life.
The invention has the following advantages:
1. the health index constructed based on the analytic hierarchy process and the differential algorithm can discover early faults of the bearing earlier, and the excellent monotonicity can objectively and effectively reflect the degradation trend of the bearing;
2. the prediction accuracy is improved by the residual service life prediction model established based on the Weibull proportion risk model, and the obtained residual service life prediction result of the bearing is basically consistent with the actual service life;
3. in the residual service life prediction system based on LabVIEW development, functions of each module are complete, the operation is simple, the interface is neat and visual, the visual operation of the system enables the system to be more complete and convenient in interaction with a user, and each operation result can be reflected on a program panel, so that each step is visual and easy to understand.
Drawings
Fig. 1 is a schematic diagram of a process for constructing a health index according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of an analytic hierarchy process according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of a geodesic distance and a euclidean distance in a three-dimensional space according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a differential evolution algorithm flow provided in an embodiment of the present invention.
Fig. 5 is a schematic diagram of a flow of establishing a residual life prediction model according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of an extreme learning machine network structure according to an embodiment of the present invention.
Fig. 7 is a schematic structural diagram of a rolling bearing state monitoring and life predicting system according to an embodiment of the present invention.
Fig. 8 is a schematic diagram of a data acquisition module of the system according to an embodiment of the present invention.
Fig. 9 is a schematic diagram of a time domain analysis module for data analysis and processing of a system according to an embodiment of the present invention.
Fig. 10 is a schematic diagram of a frequency domain analysis module for data analysis and processing of a system according to an embodiment of the present invention.
Fig. 11 is a schematic diagram of a health indicator construction module of the system according to an embodiment of the present invention.
Fig. 12 is a schematic diagram of a life prediction module of the system according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the following description refers to a specific embodiment, and further details of a method and a system for predicting bearing life constructed based on health indicators according to the present invention are described in the following, wherein the described embodiment is only a part of the embodiments of the present invention, but not all the embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention is further described below with reference to the accompanying drawings.
A bearing life prediction method constructed based on health indexes comprises the following steps:
s101, constructing health indexes based on an analytic hierarchy process and a differential algorithm;
s102, constructing a WPHM life prediction function based on the Weibull proportion risk model, and substituting the processed health index into the WPHM life prediction function to obtain the residual service life of the bearing.
As shown in fig. 1, the method for constructing health indexes based on analytic hierarchy process and differential algorithm comprises the following steps:
(1) A set of sensitive degradation features is established. And acquiring vibration signals of the rolling bearing, extracting original degradation characteristics from the time domain, the frequency domain and the time domain, and carrying out noise reduction and normalization processing on the original degradation characteristics to obtain candidate degradation characteristics. Calculating weights of four measurement indexes such as monotonicity, robustness, trend and consistency by an analytic hierarchy process, obtaining a mixed measurement index by linear weighting of four evaluation indexes, designing a mixed measurement index threshold based on the mixed measurement index of each characteristic, screening sensitive degradation characteristics, and constructing a sensitive degradation characteristic set;
(2) And constructing a health index. Carrying out intrinsic dimension estimation on the sensitive degradation feature set based on manifold hypothesis, reducing the dimension of the sensitive degradation feature set to the intrinsic dimension by using an equidistant feature mapping algorithm, carrying out weight optimization by using a differential evolution algorithm, and constructing a health index of the rolling bearing residual service life prediction by using a linear weighting method;
referring to fig. 2, which is a schematic diagram of a flow chart of an analytic hierarchy process, the specific steps for calculating weights of four metrics are as follows:
(1) Firstly, comparing and scoring importance degrees between any two indexes of four measurement indexes such as monotonicity, robustness, trend and consistency, wherein the importance levels are shown in table 1, and meanwhile, a judgment matrix A of each index is constructed as shown in formula (15):
wherein A is ij Representing the i element as compared to the j elementImportance of (3).
Table 1 importance level
(2) And (3) carrying out consistency test on the judgment matrix A, defining a consistency index CI as shown in a formula (16), and a consistency ratio CR as shown in a formula (17), wherein RI is a random consistency index, and can be obtained through table lookup 2.
When the consistency ratio CR <0.1, the check is passed, otherwise the judgment matrix needs to be reconstructed.
Wherein lambda is max For judging the maximum value in the matrix A; m is the number of elements, where m=4.
TABLE 2 random consistency index
(3) The judgment matrix A is subjected to consistency test, a feature vector corresponding to the maximum feature value of the judgment matrix is calculated, and normalization processing is carried out on the feature vector by using the formula (2) to obtain a weight vector [ alpha ] of four measurement indexes 1 ，α 2 ，α 3 ，α 4 ]。
The specific steps of dimension reduction of the sensitive degradation feature set by using the equidistant feature mapping algorithm are described in conjunction with the schematic diagram of geodesic distance and euclidean distance in three-dimensional space as shown in fig. 3. As shown in fig. 3, there are two expressions of the distance between two points A, B in three-dimensional space, and the first expression is represented by a conventional euclidean distance, namely, a dashed line distance in the figure; the second is represented by the sum of the finite Euclidean distances between the point and the nearest neighbor, i.e., each data point is connected to its nearest neighbor, on which basis the shortest path of all data pairs in the neighborhood graph is calculated to approximate the actual distance between them, i.e., the solid line distance in the graph, which is also known as geodesic distance. The specific steps of dimension reduction by using the equidistant feature mapping algorithm are as follows:
first, assume that the sensitive degradation feature set is a D-dimensional high-dimensional space in which there are n sample points x= { X 1 ，x 2 ，…，x n }，x i ∈R D ，i＝1，2，…，n。
(1) And constructing an undirected graph G (V, E) of the original high-dimensional data, wherein V is a set of points in the undirected graph, and E is a set of edges in the undirected graph. Calculating sample points x in high-dimensional space i Euclidean distance between K neighboring points, K value can be determined or calculated according to K-neighboring principle to obtain sample point x i And epsilon is the number of points in the radius circle. Wherein the sample point x i Corresponding node v i E V, if x j Is x i Is connected with x by undirected edge i And x j Is denoted as edge (x) i ，x j )。
(2) Calculate the shortest path distance, and edge (x i ，x j ) The weight of E is set to d (x i ，x j ) Wherein d (x i ，x j ) Is x i And x j If x is the Euclidean distance of i And x j D (x) is a non-adjacent point i ，x j ) = infinity. Calculating shortest path distance between any two points in undirected graph G (V, E) using Floyd algorithm instead of their geodesic distance d G (x i ，x j ) The method comprises the following steps:
(3) Calculating low-dimensional embedded coordinates assuming all data points x i The data point obtained when the D-dimensional high-dimensional original space is mapped into the D-dimensional low-dimensional space is y j ∈R d ，j＝1，2，…，n，d<D. The loss function is defined as follows:
solving (19) by using classical MDS algorithm to obtain data Y= { Y after dimensionality reduction 1 ，y 2 ，…，y n The dimension of the sensitive degradation feature set is reduced from D to D, which is the intrinsic dimension, and an estimated value is obtained based on manifold assumptions.
The steps of constructing a health indicator are described in conjunction with the differential evolution algorithm flow diagram as shown in fig. 4.
(1) After the sensitive degradation feature set is reduced in dimension to the intrinsic dimension, in order to obtain a health index better reflecting the degradation state of the bearing, an objective function mixing measurement index HM and a variable omega are firstly constructed i The constraints of (2) are as follows:
in the method, in the process of the invention,
for the j-th feature, ω, of the d-dimensional low-dimensional intrinsic features j The weight of the j element in the sensitive degradation feature set F.
Health index F p Can be expressed as:
for the optimization problem, a differential evolution algorithm is adopted to solve, and the flow is shown in fig. 4.
(2) Initializing a population to produce an initial population
NP indicates populationSize, n is the dimension of the solution space. Constraints exist for the present method
The initial solution is no longer entirely random but is given by equation (22): />
Where rand (0, 1) represents a random number taken in the interval (0, 1).
(3) The initial solution generated by the method can meet the constraint condition, the population meeting the constraint condition is called a standard population (normal omega, n omega), in order to enrich the diversity of the population and improve the convergence rate, a standard reverse population (normal negative omega, no omega) is given on the basis of generating the initial population, and the standard reverse population no omega is generated in the following way:
(1) first, the standard population is inverted by equation (23) to obtain an inverted population (opsite ω, oω):
(2) obviously, the reverse population oω does not necessarily all satisfy the constraint condition, so the oω needs to be normalized to satisfy the constraint condition, and the normalization step is shown in the formula (24):
(3) after the standard reverse population no omega is obtained, combining the standard population n omega and the standard reverse population no omega to be used as initial populations, calculating the fitness value of the combined populations, and selecting the first NP populations with large fitness values to continue subsequent mutation, intersection and selection operations. In order to ensure that the population generated during the mutation, crossover process is also a standard population, the scaling factor F is set to 1 during the mutation process and the generated population is normalized after crossover operation.
The flow of predicting the remaining life of the rolling bearing will be described with reference to fig. 5 and 6. FIG. 5 is a flowchart of a method for establishing a residual service life prediction model, wherein a maximum likelihood estimation method is utilized to perform parameter estimation on a Weibull proportion risk model, and the obtained parameter estimation value is substituted into a WPHM model to obtain a WPHM service life prediction function; in order to improve prediction accuracy, phase space reconstruction is carried out on the health index, the reconstructed health index is input into an extreme learning machine for training, a pseudo health index at a prediction time point is obtained, and finally the pseudo health index is substituted into a WPHM life prediction function to obtain the residual service life of the rolling bearing. The method comprises the following specific steps:
(1) First, a WPHM lifetime prediction function is constructed. The weibull ratio failure rate model expression is:
wherein h (t, z (t)) is a fault rate function; beta is a shape parameter, and theta is a scale parameter; z (t) is the covariate, herein the value of the health indicator at time t, and γ is the regression parameter of z (t). Parameters to be estimated are beta, theta and gamma, and the maximum likelihood estimation method is used for obtaining the estimated values of the parameters
Obtaining T 0 The RUL prediction function for time is as follows: />
Wherein T' =t+t 0 ；R(T 0 ,z(T 0 ) As a reliability function, the expression is:
(2) And phase space reconstruction is carried out on the health index, so that the prediction accuracy is improved. Before this, two key parameters have to be determined: delay time τ and embedding dimension m.
The mutual information method is adopted to determine the delay time tau: let { X, Y } = { X (i), X (i+τ): i e [1, n- τ ] }, i.e., X is a time series, Y is a delay time series, assuming that a certain phase space vector after reconstruction, Y (i) = { X (i), X (i+τ), …, X (i+ (m-1) τ) }. Knowledge of the information theory shows that the delay time corresponding to the first minimum value point of the mutual information entropy of the X and Y systems is the optimal delay time tau.
Determining an embedding dimension m by adopting a Cao method: assume that a certain phase space vector Y after reconstruction m (i)＝{x(i)，x(i+τ)，…，x(i+(m-1)τ)}，Y n ， m (i) Is Y m (i) The Cao method may be represented by the following formula:
E 1 (m)＝E(m+1)/E(m) (30)
E 2 (m)＝E * (m+1)/E * (m) (32)
in ║. ║ ∞ Is the maximum modulus norm; a (i, m) is a parameter for determining nearest neighbor; n is the number of data in the time series. E (E) 1 And E is 2 Is an index for selecting an appropriate embedding dimension, when E 1 Stable and E 2 The corresponding embedding dimension m is the minimum embedding dimension when approaching 1.
(3) The extreme learning machine is a single hidden layer fuzzy neural network, and the network structure diagram is shown in fig. 6, wherein n, l and m are the number of nodes of an input layer, a hidden layer and an output layer respectively. The extreme learning machine algorithm steps are as follows:
(1) given N pieces of prediction data [ X ] 1 ，X 2 ，…，X N ] T ∈R Nxn And corresponding data tag [ Y ] 1 ，Y 2 ，…，Y N ] T ∈R Nx1 Will [ X ] 1 ，X 2 ，…，X N ] T As input to the model [ Y ] 1 ，Y 2 ，…，Y N ] T As an output of the model;
(2) determining the number of hidden layer nodes l and an activation function g (x), and randomly determining an input weight W and a bias b by a network model;
(3) solving an output matrix H;
(4) the output weight β is solved by equation (33).
Wherein T is a time matrix,
assume health index { F p (i) I=1, 2, …, n } is reconstructed by phase space reconstruction into n- (m-1) τ m-dimensional vectors { a i :i＝1，…，n-(m-1)τ}：
Defining a matrix { B } i ＝F p (i+1+(d-1)τ):i＝1，…，n-1-(d-1)τ}:
(4) Assuming that the health index after the time t is predicted, the remaining service life prediction steps are as follows:
step one: health index { A at time t-1 before i ，B i :i＝1，…，t-1-(m-1)τ，A i For input B i Output is used as training data to be input into a prediction model of the extreme learning machine for training;
step two: after the data training is completed, A is t-(m-1)τ ＝[F p (t-(m-1)τ)，F p (t-mτ)，…，F p (t)]As input to the predictive model, output B t-(m-1)τ From B i The definition of (1) indicates that the health index F at time t+1 p (t+1) can be represented by B t-(m-1)τ To express: f (F) p (t+1)＝B t-(m-1)τ ；
Step three: output result F of step 2 p (t+1) brought to A t-(m-1)τ+1 In (A) t+1-(m-1)τ ＝[F p (t+1-(m-1)τ)，F p (t+1-mτ)，…，F p (t+1)]Will A t+1-(m-1)τ As input to the predictive model, output B t+1-(m-1)τ Also F p (t+2)＝B t+1-(m-1)τ ；
Step four: repeating the second and third steps until the prediction is finished;
step five: after the health index after the time t is obtained, the health index is input into the system (26) to obtain the residual service life.
Fig. 7 and fig. 8 show a bearing life prediction system constructed based on health indexes, which is divided into a rolling bearing state monitoring and life prediction system, and specifically includes: and the four modules comprise data acquisition, data analysis and processing, health index construction, life prediction and the like.
The rolling bearing state monitoring system is described by combining with figures 8, 9 and 10, and the temperature and vibration signals can be acquired by the data acquisition module shown in figure 8, and as the temperature change is less obvious and the change speed is slower, a large number of temperature samples are not required to be acquired every moment, only one temperature sample is required to be acquired every time, and a user only needs to set a sampling interval. The recording mode is set in the 'save setting', only the collected data is displayed in the 'off' state without saving, and the collected data is saved in the TDMS file in the 'record and read' state. In order to judge whether a fault or failure occurs, an alarm system is further arranged, when the signal value exceeds a threshold value, the alarm indicator lamp is lighted to be red, and meanwhile, the alarm times are set in the alarm system at the temperature.
As shown in fig. 9, the time domain analysis module for data analysis and processing mainly performs filtering processing on the introduced vibration signal, and obtains the autocorrelation after filtering and each time domain characteristic parameter.
As shown in fig. 10, the frequency domain analysis module for data analysis and processing mainly obtains the amplitude spectrum and the power spectrum of the signal, analyzes the frequency bands of the signal mainly distributed by the amplitude spectrum, and analyzes the energy distribution on each frequency band by the power spectrum.
The rolling bearing life prediction system is described with reference to fig. 11 and 12, and the rolling bearing life prediction system is shown in fig. 11 as a health index construction module, introduces vibration signals, clicks a 'calculate mixed metric index' button, calculates 41 degradation characteristics of three aspects of time domain, frequency domain, time domain and frequency domain, and calculates and displays mixed metric index values of the degradation characteristics. The 41 degradation features include:
(1) 14 time domain degradation characteristics, as shown in Table 3, assume that the collected vibration signal is x i (i=1, 2, …, N), N representing the number of sampling points.
TABLE 3 time domain degradation characterization
(2) 13 frequency domain degradation features: for time domain signal x i (i=1, 2, …, N) frequency domain amplitude spectra can be obtained using FFT:
the frequency domain degradation characteristics are shown in table 4, where f k For the frequency at time kA value of the frequency.
TABLE 4 frequency domain degradation characterization
(3) And 8 wavelet packet energy characteristics obtained by decomposing the three-layer wavelet packet and 6 energy characteristics obtained by decomposing the empirical mode.
The mixed measurement index threshold is designed according to the degree of distinction among the mixed measurement index values of the degradation features, the 'build health index' button is clicked, and the system screens out the sensitive degradation feature set and fuses the sensitive degradation feature set to display the health index.
As shown in fig. 12, the life prediction module sets the extreme learning machine parameters, selects the activation function, and sets the number of hidden neurons. And clicking a 'start prediction' button, and inputting the health index into the built WPHM prediction model by the system to obtain a residual service life prediction result and a prediction error of the bearing.
The foregoing examples have shown only the preferred embodiments of the invention, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that modifications, improvements and substitutions can be made by those skilled in the art without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (8)
1. A bearing life prediction method constructed based on health indexes is characterized by comprising the following steps:
s101, constructing health indexes based on an analytic hierarchy process and a differential algorithm;
s102, constructing a WPHM life prediction function based on the Weibull proportion risk model, and substituting the constructed one-dimensional health index into the WPHM life prediction function to obtain the residual service life of the bearing.
2. The method for predicting the life of the bearing constructed based on the health index according to claim 1, wherein the specific process for constructing the health index based on the analytic hierarchy process and the differential algorithm comprises the following steps:
s201, collecting vibration signals of the rolling bearing, extracting degradation characteristics of three aspects of time domain, frequency domain and time-frequency domain according to the vibration signals, and carrying out noise reduction and normalization treatment on each degradation characteristic;
s202, calculating weights of four measurement indexes of monotonicity, robustness, trend and consistency of each degradation characteristic by using an analytic hierarchy process, and constructing a mixed measurement index based on the weights of the four measurement indexes;
s203, constructing the sensitive degradation characteristic set based on the mixed measurement index;
s204, carrying out eigenvector estimation on the sensitive degradation feature set based on manifold assumptions, and reducing the dimension of the sensitive degradation feature set to the eigenvector by using an equidistant feature mapping algorithm;
and S205, carrying out weight optimization on the reduced sensitive degradation feature set by using a differential algorithm, and constructing a one-dimensional health index of residual service life prediction by using a linear weighting method.
3. The method for predicting the life of a bearing constructed based on health indexes according to claim 2, wherein the noise reduction and normalization process is specifically: carrying out noise reduction treatment on the original degradation characteristics by using a 7-point moving average method so as to enable the characteristic curve to be relatively smooth; normalizing all the features to ensure that the range is between 0 and 1, wherein the 7-point sliding average method formula is as follows:
wherein x is n Is the original characteristic signal; n is the data length of the vibration signal;
is a new characteristic signal after noise reduction;
the normalization formula is:
wherein X is norm Is a normalization result; x is a characteristic sequence; x is X min ，X max Is the minimum and maximum in the feature sequence.
4. The method for predicting the life of a bearing constructed based on health indicators according to claim 2, wherein weights of the four metrics of monotonicity, robustness, trend and consistency are calculated by a pairwise comparison method, in particular,
the monotonicity index formula is:
wherein X= { X k } k＝1:K To degenerate the characteristic sequence, x k Indicated at t k The feature value of the moment, K is the total number of feature values in the degraded feature sequence;
representing differences in the feature sequence;
And->
The number of the positive and negative derivatives is respectively represented;
the robustness index formula is:
wherein x is k Indicated at t k A characteristic value of the moment;
is characterized by t k Average trend value of time;
the trend index formula is:
in the method, in the process of the invention,
and->
Respectively the degenerate feature sequence { x } k } k＝1:K Sum time { t k } k＝1:K Is a sequence of sequences ordered by (a); the consistency index formula is:
wherein P is EOL Is a vector formed by degradation characteristic values of the bearing at failure time, P O Is a vector composed of degradation eigenvalues at the initial time.
5. The method for predicting the life of a bearing constructed based on health indexes according to claim 2, wherein the mixed metric index is obtained by linearly weighting four evaluation indexes of monotonicity, robustness, trend and consistency of degradation characteristics, and the calculation formula is as follows:
wherein alpha is i Is the weight of each index.
6. The method for predicting the life of a bearing constructed based on health indexes according to claim 2, wherein the method for constructing the sensitive degradation characteristic set is as follows: after the mixed measurement index of each feature is calculated, a mixed measurement index threshold is set, and the degradation feature of the feature mixed measurement index exceeding the threshold is selected as the sensitive degradation feature to construct a sensitive degradation feature set.
7. The method for predicting the life of a bearing constructed based on health indicators according to claim 1, wherein the step S102 specifically comprises:
s701, calculating a parameter estimation value of the WPHM model by using a maximum likelihood estimation method, and substituting the parameter estimation value into the WPHM model to obtain a WPHM life prediction function;
s702, reconstructing the phase space of the health index, substituting the health index into an extreme learning machine for training to obtain a pseudo health index at a predicted time point, and substituting the pseudo health index into a WPHM life prediction function to obtain the residual service life of the rolling bearing.
8. The bearing life prediction system constructed based on the health index is characterized by being developed based on LabVIEW, and comprises a state monitoring system and a life prediction system, wherein the state monitoring system comprises a data acquisition function, a data analysis and processing function and a state alarm function; the life prediction system comprises screening of sensitive degradation characteristics, construction of health indexes, and prediction results and prediction errors of residual service life.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211675864.4A CN116124460B (en) | 2022-12-26 | 2022-12-26 | Bearing life prediction method and system based on health index construction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211675864.4A CN116124460B (en) | 2022-12-26 | 2022-12-26 | Bearing life prediction method and system based on health index construction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116124460A true CN116124460A (en) | 2023-05-16 |
| CN116124460B CN116124460B (en) | 2024-01-30 |
Family
ID=86300152
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211675864.4A Active CN116124460B (en) | 2022-12-26 | 2022-12-26 | Bearing life prediction method and system based on health index construction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116124460B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592383A (en) * | 2024-01-19 | 2024-02-23 | 四川晟蔚智能科技有限公司 | Method, system, equipment and medium for predicting equipment health life |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102313577A (en) * | 2011-06-24 | 2012-01-11 | 上海大学 | Equipment health state evaluation and recession prediction method based on multi-channel sensing signals |
| US20160169769A1 (en) * | 2014-12-10 | 2016-06-16 | Eric Robert Bechhoefer | Rotating body monitoring and alert system and method |
| WO2018041046A1 (en) * | 2016-08-27 | 2018-03-08 | 国网冀北电力有限公司电力科学研究院 | Output characteristic calculation method and tolerance allocation method for electromagnetic mechanism of load switch |
| CN109187025A (en) * | 2018-09-19 | 2019-01-11 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique of integrated KELM |
| CN110222437A (en) * | 2019-06-12 | 2019-09-10 | 中国神华能源股份有限公司 | Appraisal procedure, device and the storage medium of train car team health status |
| CN110930052A (en) * | 2019-12-02 | 2020-03-27 | 国网山东省电力公司高密市供电公司 | Method, system, equipment and readable storage medium for predicting failure rate of power transformation equipment |
| CN111399474A (en) * | 2020-02-29 | 2020-07-10 | 中南大学 | Health index-based life prediction method and device for balance control module |
| WO2021088346A1 (en) * | 2019-11-09 | 2021-05-14 | 北京工业大学 | Performance degradation dictionary construction and similarity-based bearing lifetime prediction method |
| CN114841078A (en) * | 2022-05-26 | 2022-08-02 | 重庆大学 | Method and system for predicting remaining service life of bearing |
| CN114925476A (en) * | 2022-06-08 | 2022-08-19 | 安徽农业大学 | Method for predicting quaternion of rolling bearing degradation track augmentation and storage medium |
| CN115046764A (en) * | 2022-05-05 | 2022-09-13 | 燕山大学 | Early fault diagnosis method for rolling bearing |
| CN115271238A (en) * | 2022-08-10 | 2022-11-01 | 广东电网有限责任公司 | Method and device for predicting service life of transformer |
-
2022
- 2022-12-26 CN CN202211675864.4A patent/CN116124460B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102313577A (en) * | 2011-06-24 | 2012-01-11 | 上海大学 | Equipment health state evaluation and recession prediction method based on multi-channel sensing signals |
| US20160169769A1 (en) * | 2014-12-10 | 2016-06-16 | Eric Robert Bechhoefer | Rotating body monitoring and alert system and method |
| WO2018041046A1 (en) * | 2016-08-27 | 2018-03-08 | 国网冀北电力有限公司电力科学研究院 | Output characteristic calculation method and tolerance allocation method for electromagnetic mechanism of load switch |
| CN109187025A (en) * | 2018-09-19 | 2019-01-11 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique of integrated KELM |
| CN110222437A (en) * | 2019-06-12 | 2019-09-10 | 中国神华能源股份有限公司 | Appraisal procedure, device and the storage medium of train car team health status |
| WO2021088346A1 (en) * | 2019-11-09 | 2021-05-14 | 北京工业大学 | Performance degradation dictionary construction and similarity-based bearing lifetime prediction method |
| CN110930052A (en) * | 2019-12-02 | 2020-03-27 | 国网山东省电力公司高密市供电公司 | Method, system, equipment and readable storage medium for predicting failure rate of power transformation equipment |
| CN111399474A (en) * | 2020-02-29 | 2020-07-10 | 中南大学 | Health index-based life prediction method and device for balance control module |
| CN115046764A (en) * | 2022-05-05 | 2022-09-13 | 燕山大学 | Early fault diagnosis method for rolling bearing |
| CN114841078A (en) * | 2022-05-26 | 2022-08-02 | 重庆大学 | Method and system for predicting remaining service life of bearing |
| CN114925476A (en) * | 2022-06-08 | 2022-08-19 | 安徽农业大学 | Method for predicting quaternion of rolling bearing degradation track augmentation and storage medium |
| CN115271238A (en) * | 2022-08-10 | 2022-11-01 | 广东电网有限责任公司 | Method and device for predicting service life of transformer |
Non-Patent Citations (1)
| Title |
|---|
| 谷梦瑶;陈友玲;王新龙;: "多退化变量下基于实时健康度的相似性寿命预测方法", 计算机集成制造系统, no. 02 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592383A (en) * | 2024-01-19 | 2024-02-23 | 四川晟蔚智能科技有限公司 | Method, system, equipment and medium for predicting equipment health life |
| CN117592383B (en) * | 2024-01-19 | 2024-03-26 | 四川晟蔚智能科技有限公司 | Method, system, equipment and medium for predicting equipment health life |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116124460B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107480341B (en) | A kind of dam safety comprehensive method based on deep learning | |
| CN111967486A (en) | Complex equipment fault diagnosis method based on multi-sensor fusion | |
| CN108874959A (en) | A kind of user's dynamic interest model method for building up based on big data technology | |
| CN115278741B (en) | Fault diagnosis method and device based on multi-mode data dependency relationship | |
| CN109858140B (en) | Fault diagnosis method for water chilling unit based on information entropy discrete Bayesian network | |
| CN110674858B (en) | Traffic public opinion detection method based on space-time correlation and big data mining | |
| CN116124460B (en) | Bearing life prediction method and system based on health index construction | |
| CN116977708B (en) | Bearing intelligent diagnosis method and system based on self-adaptive aggregation visual view | |
| CN114118789A (en) | Radar transmitter state evaluation method based on fuzzy comprehensive evaluation and comprehensive weighting | |
| CN117034143B (en) | Distributed system fault diagnosis method and device based on machine learning | |
| CN113269384B (en) | Method for early warning health state of river system | |
| CN115185937A (en) | SA-GAN architecture-based time sequence anomaly detection method | |
| CN114444910A (en) | Electric power Internet of things-oriented edge network system health degree evaluation method | |
| CN115130499A (en) | Vibration signal-based online measurement and prediction method for slip ratio of bearing retainer under variable working conditions | |
| Wang et al. | Construction of fuzzy control charts based on weighted possibilistic mean | |
| CN117493930A (en) | Multi-element time sequence anomaly detection method based on contrast generation countermeasure network | |
| CN110728377B (en) | Intelligent fault diagnosis method and system for electromechanical equipment | |
| Fogliatto et al. | A hierarchical method for evaluating products with quantitative and sensory characteristics | |
| CN115876467A (en) | Pseudo label transfer type two-stage field self-adaptive rolling bearing fault diagnosis method | |
| CN110084423A (en) | A kind of link prediction method based on local similarity | |
| CN117473430A (en) | Non-invasive load classification method and device | |
| CN107437112A (en) | A kind of mixing RVM model prediction methods based on the multiple dimensioned kernel function of improvement | |
| CN116596836A (en) | Pneumonia CT image attribute reduction method based on multi-view neighborhood evidence entropy | |
| US11481267B2 (en) | Reinforcement learning approach to root cause analysis | |
| CN118378178B (en) | Transformer fault identification method and system based on residual map convolution 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |