CN115374811A - Novel fault state diagnosis method for rolling bearing - Google Patents
Novel fault state diagnosis method for rolling bearing Download PDFInfo
- Publication number
- CN115374811A CN115374811A CN202210882129.4A CN202210882129A CN115374811A CN 115374811 A CN115374811 A CN 115374811A CN 202210882129 A CN202210882129 A CN 202210882129A CN 115374811 A CN115374811 A CN 115374811A
- Authority
- CN
- China
- Prior art keywords
- rolling bearing
- fault
- adopting
- signal
- bearing
- 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
- 238000005096 rolling process Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000003745 diagnosis Methods 0.000 title claims abstract description 22
- 241000544061 Cuculus canorus Species 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 15
- 230000009467 reduction Effects 0.000 claims abstract description 13
- 238000012706 support-vector machine Methods 0.000 claims abstract description 13
- 230000036541 health Effects 0.000 claims abstract description 9
- 230000001133 acceleration Effects 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000006731 degradation reaction Methods 0.000 abstract description 5
- 230000015556 catabolic process Effects 0.000 abstract description 3
- 238000005070 sampling Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 238000005457 optimization Methods 0.000 description 9
- 238000001228 spectrum Methods 0.000 description 8
- 235000005770 birds nest Nutrition 0.000 description 7
- 230000008859 change Effects 0.000 description 7
- 230000000630 rising effect Effects 0.000 description 7
- 235000005765 wild carrot Nutrition 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 238000013016 damping Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 241000272177 Cuculiformes Species 0.000 description 1
- 241001327403 Kernia Species 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 240000000111 Saccharum officinarum Species 0.000 description 1
- 235000007201 Saccharum officinarum Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012631 diagnostic technique Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000011068 loading method Methods 0.000 description 1
- 230000001050 lubricating effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 239000010453 quartz Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 1
- 238000003756 stirring Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 238000003466 welding 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
Abstract
The invention provides a novel method for diagnosing the fault state of a rolling bearing, which comprises the following steps: arranging a vibration signal sensor on the surface of the bearing to be tested, and connecting the vibration signal sensor with a testing device to obtain a vibration signal of the rolling bearing as an original diagnostic signal; evaluating the health state of the acquired vibration signals of the rolling bearing by adopting a Kernel index, and finishing the definition of a fault stage; deconvoluting the signal of the fault stage by adopting the maximum correlation kurtosis of the optimal parameter to carry out noise reduction pretreatment and highlight the impact component; and extracting the characteristics of the vibration signals of all stages by adopting a hierarchical entropy, and finishing the classification of the fault states of the rolling bearings by adopting a support vector machine optimized by a cuckoo algorithm. The invention accurately describes the bearing degradation by using the Gini index, provides a method for selecting the length and the kurtosis of the permutation entropy filter, has strong self-adaptability, overcomes the limitation of manual experience selection, improves the diagnosis accuracy by using the cuckoo algorithm, and has the accuracy reaching 100 percent compared with the prior art.
Description
Technical Field
The invention relates to a fault identification technology in a mechanical fault operation state, in particular to a novel fault state diagnosis method for a rolling bearing.
Background
The rolling bearing is an important part in rotary machinery, can reduce the friction loss between a shaft and a shaft in the operation process, and is also one of parts which are easy to fail. The running condition of the rolling bearing directly influences the normal operation of mechanical equipment, so that the condition fault diagnosis of the rolling bearing has important engineering significance. The Gini index is an index used for measuring the income distribution uniformity degree in economics, and the index is used for processing bearing degradation data and judging the early fault range of the bearing.
Because the production environment is complex, the transient impact generated by the rolling bearing fault is easily submerged by other signal components, and therefore, the noise reduction pretreatment of the acquired original signal is necessary. The deconvolution of the maximum correlation kurtosis can eliminate interference components in signals, enhance pulse impact in the signals and improve the signal-to-noise ratio, and the deconvolution method has been widely focused by researchers in the field of rolling bearings in recent years. The extraction of early fault characteristics of the rolling bearing is realized by people based on MCKD and 1.5 dimensional spectrum; and deconvolution is carried out by using the maximum correlation kurtosis to be applied to the gear of the cutting part of the coal mining machine, so as to extract weak fault characteristics. The quality of the maximum correlation kurtosis deconvolution noise reduction effect depends on the length of the filterLAnd period of impactTBut, however, doIt is from the lack of adaptability of personal experience that both parameters were chosen in the above study. The permutation entropy is commonly used for measuring the complexity of the time sequence and is highly sensitive to the mutation signal, and the invention selects the length of the filter by taking the permutation entropy as a standardLSelecting the impact period based on the kurtosisTTo achieve parameter optimization of the MCKD. The rolling bearing signal is nonlinear and non-stable, and whether accurate fault characteristic information is extracted from the rolling bearing signal or not influences subsequent intelligent diagnosis effects. The hierarchical entropy is an entropy proposed in 2011 by predecessors, and can simultaneously analyze low-frequency components and high-frequency components in a vibration signal through a construction operator, so that hidden bearing fault characteristic signals can be more comprehensively extracted. Researchers also combine wavelet packet energy and hierarchical entropy for fault diagnosis of the rolling bearing, and feature extraction of the vibration signal of the electric spindle is achieved by applying the wavelet packet hierarchical entropy. The support vector machine is widely applied to pattern recognition, and is also applied to the detection and recognition of the sugarcane seed caned bud. And identifying the surface defects of the friction stir welding based on the SVM. But its accuracy is affected by the kernel function g and the penalty parameter c.
It can be seen that although there are many existing diagnostic methods for rolling bearing failure, there are some limitations and they are not satisfactory. There is a need to explore new approaches to the development of better diagnostic techniques.
Disclosure of Invention
Therefore, the invention provides a new method for diagnosing the fault state of the rolling bearing, which can more conveniently and more accurately realize the non-operation and non-stop control diagnosis on whether the rolling bearing in the running state has the fault hidden danger or not, acquire the fault information in the rolling bearing and visually display the hidden fault information through technical means, thereby providing a new technical choice for the prior art of fault diagnosis of the rolling bearing.
The new method for diagnosing the fault state of the rolling bearing is characterized by comprising the following steps:
arranging a vibration signal sensor on the surface of a bearing to be tested, and connecting the vibration signal sensor with a testing device to obtain a vibration signal of a rolling bearing as an original diagnostic signal;
step (II): and evaluating the health state of the acquired vibration signals of the rolling bearing by adopting a Kernig index, and finishing the definition of a fault stage.
Step (three): and deconvoluting the signal in the fault stage by adopting the maximum correlation kurtosis of the optimal parameter to carry out noise reduction preprocessing and highlight the impact component of the signal.
Step (IV): and extracting the characteristics of the vibration signals at each stage by adopting the hierarchical entropy, and finishing the classification of the fault state of the rolling bearing by adopting a support vector machine optimized by a cuckoo algorithm.
The specific acquisition mode of the vibration signals of the rolling bearing in the step (I) is to arrange acceleration sensors in the horizontal and vertical directions on the bearing to be detected and connect the acceleration sensors with a detection device.
Dividing the Keyny index corresponding to the fault state in the step (I) into:
in the normal operation stage, the curve of the Gini index fluctuates slightly between 0.415 and 0.428;
in the early failure stage, the Gini index curve is gradually increased between 0.42 and 0.51;
in the gradual failure stage, the curve of the Gini index fluctuates up and down greatly between 0.2 and 0.515.
The data operation of each step is carried out in the computer.
The technical effects of the invention are embodied in the following aspects:
(1) Aiming at the health detection problem of the rolling bearing, the invention provides a new method for constructing a health index by adopting a Kernia index. Compared with the root mean square value of the common index, the Gini index can well describe the degradation process of the bearing and is more sensitive to early failure.
(2) The goodness of the noise reduction effect for MCKD depends mainly on the filter lengthLAnd period of failureTThe optimal parameter selection method for selecting the length of the filter and selecting the fault period by using the permutation entropy is provided for solving the problem of two parameters, so that the MCKD is more adaptive, and the limitation of parameters selected by manual experience is overcome.
(3) Parameter optimization is carried out on the fault diagnosis classification model through the cuckoo algorithm, and the diagnosis accuracy of the fault diagnosis model can be effectively improved.
(4) The method has the advantage that the diagnosis accuracy rate of the method for the rolling bearing correlation state reaches 100%, and compared with the existing diagnosis method, the method has obvious superiority.
Drawings
FIG. 1 is a flow chart of the method steps of the present invention.
Fig. 2 is a schematic diagram of the signal acquisition of the vibration original signal of the rolling bearing adopted by the invention.
FIG. 3 is a root mean square trend graph.
FIG. 4 is a plot of the trend of the Kini index for the examples.
Fig. 5 is a time domain waveform diagram at time 5280 min.
FIG. 6 is an envelope spectrum at time 5280 min.
FIG. 7 is a time domain waveform diagram at time 5330 min.
FIG. 8 is the envelope spectrum at time 5330 min.
Fig. 9 is a diagram of the MCKD filter length L iteration.
Fig. 10 is a graph of MCKD fault cycle iterations.
FIG. 11 is an envelope spectrum of the denoised signal at time 5280 min.
Fig. 12 is a state recognition result diagram.
In fig. 2, the components are numbered as follows:
1-a motor; 2-a transmission pair; 3-a rotating shaft; 4-bearing number one; 5-bearing II; 6-bearing number three; 7-bearing number four; 8-a thermocouple; 9-acceleration sensor.
FIG. 3 illustrates: the abscissa:tx 10min (unit: min), ordinate: root mean square value/(m.s) -2 ) (value of root mean square, unit: m.s -2 ). The duration time of the second group of tests is 164h, 984 data files are collected totally, the sampling interval is 10min, the root mean square value of each data file is calculated (each data file corresponds to one root mean square value), and a root mean square value curve is obtained by forming lines by connecting points. K is the change of the bearing state by the least square method (from 5330 min)With a rising trend) of the data are fitted to a straight line.
FIG. 4 illustrates: the abscissa:t×10min (unit: min), ordinate: coefficient of kini/(m.s) -2 ) (value of the kini coefficient, unit: m.s -2 ). The second group of test duration is 164h, 984 data files are collected in total, the sampling interval is 10min, the damping coefficient of each data file is calculated (each data file corresponds to one damping coefficient value), and a damping coefficient value curve is obtained by connecting points in a line. K is a slope obtained by fitting a straight line to the rising edge data of the beginning of the change in the bearing state (rising from 5280 min) by the least square method.
FIG. 5 illustrates: the abscissa:t/s(sampling time, unit: second), ordinate:A/mV (amplitude, unit: millivolts).
FIG. 6 illustrates: the abscissa:fin/Hz (sampling frequency, unit: hertz), ordinate:A/mV (amplitude, unit: millivolts.
FIG. 7 illustrates: the abscissa:t/s(sampling time, unit: second), ordinate:Aand/mV (amplitude, unit: millivolt).
FIG. 8 illustrates: the abscissa:fin/Hz (sampling frequency, unit: hertz), ordinate:Athe frequency of the outer ring fault characteristic frequency is 1 multiplied by 230.7Hz, and the frequency of the outer ring fault characteristic frequency is 2 multiplied by 460.8 Hz.
FIG. 9 illustrates: the abscissa: the length L of the MCKD filter, the number of iterations (unit: times), ordinate: the value of the permutation entropy PE (span [ 01 ]). Parameter optimization is carried out on the parameters (filter length L) of the MCKD by taking the permutation entropy as a standard. The value of L is selected over an interval of [2,300]. L =11 identified in the figure is the maximum value of the permutation entropy found in this interval, i.e. the value of the optimum parameter for the filter length L.
FIG. 10 illustrates: the abscissa: MCKD fault cycleTNumber of iterations (unit: times), ordinate: the value of kurtosis. Using kurtosis as standard, for MCKD parameters (fault period)T) And optimizing the parameters. T has a value selection interval of [1,300 ]]. T =5 identified in the figure is the maximum value of the kurtosis found in this interval, i.e. the fault cycleTThe optimization parameter value of (2).
FIG. 11 illustrates: the abscissa:fin/Hz (sampling frequency, unit: hertz), ordinate:Athe frequency of 230.7Hz marked in the figure is multiplied by 1 frequency of the bearing outer ring fault characteristic frequency.
FIG. 12 illustrates: the abscissa: test set sample (unit: one), ordinate: class tags (tag 1 is a normal phase signal, tag 2 is an early failure phase signal, and tag 3 is a gradual failure phase signal).
Detailed Description
The positive technical effects of the invention are further explained in the following by combining the drawings and examples.
Gini's index (Gini), maximum Correlation Kurtosis Deconvolution (MCKD), and permutation entropy (g: (G:)H p ) Hierarchical entropy of) And cuckoo algorithm and its optimization support vector are existing mathematical concepts, and their algorithms are also prior art. For convenience of illustrating the application process of the present invention, their relevant definitions and operation processes are briefly described as follows:
index of kini
The kini index is an index in economics and is commonly used for reflecting the fairness degree of social wealth distribution and monitoring the stability degree of society. When a rolling bearing fails, the vibration signal energy is usually concentrated at certain positions of the signal sequence, and the failure signal presents sparsity similar to the imbalance of the income distribution of residents described in the economics. The invention provides a method for processing bearing degradation data by using a Gini index, and representing the development trend of rolling bearing faults by using the Gini index as a standard to evaluate the health state. The kini index is defined as:
whereinBIs a positive sequence arrangement of the time series,Fis the sum of the time series.
The value range of the Gini index is [0,1].
Maximum correlation kurtosis deconvolution
The related kurtosis is a concept developed by taking kurtosis as a base stone, and can enhance periodic impact components submerged by noise. MCKD is to detect the periodicity of the impact in the acquired original signal by taking the maximum correlation kurtosis as an objective function. The correlation kurtosis is defined by formula (1):
in the formula:y n is an input signal;Tis the period of the impact signal;Mis the displacement number;Lis the filter length; filter coefficient
The following can be obtained:
wherein:
the flow of the MCKD algorithm implemented by iteration is as follows:
step 1: determining parameters of MCKDL,M,T;
and 3, step 3: obtaining an output signal;
Satisfy the signal before and after filteringThen the iteration is ended and step 3 is executed.εIs the iteration termination threshold, which is a small positive number.
Permutation entropy
The permutation entropy has higher sensitivity to the change of the signal and is an index for measuring the complexity of the signal time series. The more disordered the time series, the smaller the permutation entropy. The more random the time series, the greater the permutation entropy. The entropy of the permutation is defined as follows:
(1) The phase space reconstruction is carried out on the sampling sequence to obtain:
x(i)=[x(i),x(i+τ),…,x(i+(m-1)τ)] (10)
in the formula:mandτrespectively embedding dimension and delay time. (2) To pairx(i) Is/are as followsmThe data are sorted in ascending order to obtain:x(i+(j 1 -1)τ)≤x(i+(j 2 -1)τ)≤…≤x(i+(j m -1)τ) (11) in the formula:j 1 ,j 2 ,…,j m is composed ofx (i) And re-ordering the element position indexes. (3) If it isx (i) There are two elements of equal size, namely:x(i+(j 1 -1)τ)=x(i+(j 2 -1)τ) (12) sorting by the size of the position index value ifj 1 <j 2 Then, thenx(i+(j 1 -1)τ)<x(i+(j 2 -1)τ) (13) (4) for any of the reconstructed signalsx (i) A set of sequences in ascending order is available:s(g)=(j 1 ,j 2 ,…,j m ) (14) in the formula:g=1,2,…,kand is andk≤m| A And therefore, the first and second electrodes are,mthe different sequences reconstructed by the dimensional phase space havem| A Arrangement, sequences(g) Is one of the arrangements.
(5) The probability of occurrence of each sequence is calculated, i.e.:
P i =I/K (15)
in the formula: i iss(g) The number of occurrences.
(7) To pairH p (m,τ) Normalization processing can obtain:
as can be seen from the definition of the permutation entropy,H p has a value range of [0,1]]。
The invention measures the lengths of different filters of the MCKD according to the permutation entropyLThe resulting noise reduction effect is determinedLSelecting the value of (1). Considering the factors of the loss of the original signal of the signal after noise reduction caused by the overlarge calculated amount and the filter length, etc., the value of the filter length is limited to [2,300]]. Setting initial filter lengthL=100。
Hierarchical entropy
The operation process of the hierarchical entropy is as follows:
When the value of j is taken as 0,the low-frequency component is obtained after one-layer decomposition. When j is taken to be 1, the value of j,corresponding to the obtained high frequency component.
(2) Define onenDimension vectorWhereinIs 0 or 1, and there is a unique integereCorresponding to it.eThe expression of (a) is:
thus, the hierarchical components resulting from the time series decomposition can be expressed as:
cuckoo algorithm
The cuckoo algorithm is a heuristic algorithm which is used for enhancing the global searching capability by simulating the brooding behavior of cuckoos in nature and combining with the Layvin flight. The cuckoo algorithm flow is as follows:
(1) Setting distributionNInitial position of individual bird nest and cuckooY i ,iThe value of (a) is [1,N]. And calculating the fitness value of each nest, and taking the nest with the maximum fitness as the optimal nest and reserving the optimal nest.
(2) And updating the positions of other bird nests and calculating a fitness value. The location status update formula is shown in equation (22):
is a firstiA bird nest ist+1 of the new location after the update,in order to be the step-size factor,the search path for the flight of the Levis meets the principles of short-distance walking with small step length and long-distance walking with large step length。uIs a step size factor for the levitational flight,vis the Laevir flight index.u,vAre all variables that satisfy a gaussian distribution.
(3) And comparing the current maximum fitness value with the reserved fitness value, and if the current maximum fitness value is larger than the reserved fitness value, updating the optimal bird nest and reserving the optimal bird nest.
(4) After the position is updated, random number is adoptedr∈[0,1]And discovery probabilityPA comparison is made.r<PAnd then the position of the bird nest is updated,r≥pthe position of the bird nest is kept unchanged.
(5) And if the set iteration number is not reached, returning to the step (2), and if not, outputting the optimal nest position and fitness value.
Cuckoo algorithm optimization support vector machine
The support vector machine is a machine learning algorithm based on VC dimension theory and satisfying the principle of the minimum state of structural risk, has the advantages of being suitable for small samples, strong in generalization capability and the like, and mainly solves the problem that the prior art can not meet the requirement of a formulaThe optimal hyperplane is used for distinguishing the fault states of the rolling bearing, and the quality of the classification effect is mainly influenced by parametersc,gThe influence of (c). Therefore, in order to ensure that the support vector machine has higher classification accuracy, the method adopts the cuckoo optimization algorithm to carry out parameter optimization on the support vector machine.
The fault diagnosis steps for the rolling bearing are as follows:
arranging a vibration signal sensor on the surface of a bearing to be tested, and connecting the vibration signal sensor with a testing device to obtain a vibration signal of a rolling bearing as an original diagnostic signal;
step (II): and evaluating the health state of the acquired vibration signals of the rolling bearing by adopting a Gini index, and finishing the definition of the fault stage.
Step (III): and deconvoluting the signal in the fault stage by adopting the maximum correlation kurtosis of the optimal parameter to carry out noise reduction preprocessing and highlight the impact component of the signal.
Step (IV): and extracting the characteristics of the vibration signals of all stages by adopting a hierarchical entropy, and finishing the classification of the fault states of the rolling bearings by adopting a support vector machine optimized by a cuckoo algorithm.
The above steps are shown in fig. 1.
Fault detection
In order to prove the effectiveness of the invention in diagnosing the fault state of the rolling bearing, the following processing procedure of the vibration signal of the known fault is used for illustration:
experimental signals come from a full-life cycle acceleration test of a rolling bearing in an NSFI/UCR intelligent maintenance system center (see early fault diagnosis [ J ]. Vibration of the rolling bearing with Tang Guiji, wang Xiaolong and IVMD fusion singular value difference spectrums, and test and diagnosis 2016.36 (04): 700-707+ 810). The test stand configuration is shown in fig. 2. The test bed comprises an alternating current motor 1, a main shaft 3, a transmission pair 2, a first bearing 4, a second bearing 5, a third bearing 6, a fourth bearing 7, a vibration sensor 9, a thermocouple, a loading device, a lubricating system, a testing device and the like. The rotating speed of the rotating shaft is 2000r/min, acceleration sensors are respectively arranged in the horizontal direction and the vertical direction of the rolling bearing to acquire signals, and the parameters of the rolling bearing are shown in table 1. In the experiment, 4 sets of data were collected, and the second set of data was selected for analysis. When the second group of data is collected, the bearing No. 1 has an outer ring fault. The sampling frequency is 20kHz, and the outer ring fault frequency is 230Hz. The duration of the second group of tests is 164h, 984 data files are collected in total, the sampling interval is 10min, and the number of sampling points is 20480 points.
The equipment and the using process for the experiment are as follows:
an AC motor is rotated at a constant rotation speed of 2000 rpm, power is transmitted to a rotating shaft through a friction belt, four double row roller bearings of type ZA-115 are mounted on the rotating shaft, and a radial load of 6000 lbs is applied to a rolling bearing through a spring mechanism so as to accelerate life loss of the bearing. Vibration signals of the bearing in the horizontal direction (X direction) and the vertical direction (Y direction) are measured through a high-precision quartz acceleration sensor, and data are acquired by adopting a DAQ 6062E data acquisition card. The experiment totally collects 4 groups of data, and each group of data records the whole process of the bearing state from normal to failure. The failure of the second set of data records is the outer ring failure of bearing number 1.
When the data of the whole service life of the rolling bearing is analyzed, the state is defined by adopting a root mean square value. Fig. 3 is a trend graph of the operating state of the experimental signal analyzed by root mean square, and it can be seen from the graph that the operating state of the rolling bearing starts to change at the time of 5330min (from 5330min, the root mean square value curve has a rising trend).
Fig. 4 is a trend graph of the operating state of the experimental signal analyzed by using the economic kini index, and it can be seen from the trend graph that the operating state of the rolling bearing starts to change at 5280min (from 5280min, the kini index curve has a rising trend).
The signals at times 5280min and 5330 are extracted and subjected to envelope processing, the results of which are shown in fig. 5-8. It can be known from the figure that the envelope spectrum at the time of 5280min is seriously interfered by noise, the characteristic frequency is difficult to extract, and the characteristic frequency and the frequency multiplication at the time of 5330min are obvious.
And performing noise reduction processing on the signal at the time of 5280min by adopting the optimal parameter MCKD provided by the invention. Filter length selected using permutation entropyL=11, failure cycle screening by kurtosisT=5, as shown in fig. 9-10.
Fig. 11 is an envelope spectrum of a signal at the time of 5280min after noise reduction by using the optimal parameter MCKD, and the fault characteristic frequency of the rolling bearing can be clearly observed from the envelope spectrum. Therefore, the validity of the Kernel index in judging the state change of the rolling bearing and the noise reduction effect of the optimal parameter MCKD can be proved, and the rising edge data of the bearing beginning to change is subjected to linear fitting by adopting the least square method to carry out linear fitting on the rising edge data so as to obtain the linear slopeKThe comparison of the values shows that the kini index is more sensitive to changes in the condition of the rolling bearing than the root mean square value.
As shown in fig. 4, the operating state of the full-life rolling bearing can be divided into 3 stages by the kini index: the running time is 0min-5280min, and the Gini index fluctuates up and down slightly by taking 0.42 as a mean value; 5280min-7020min is the early failure stage, and the Gini index is increased between 04.2-0.51; 7020min-9840min is the gradual failure stage, the Gini index fluctuates greatly, and the lower value is reduced to 0.2.
At present, 100 groups of data of 3 stages are respectively extracted, 300 groups of data are totally used for fault state diagnosis, 30 groups of data are selected to form a training set in each stage, and the rest 70 groups of data form a testing set. Feature extraction is carried out on data of a training set and a test set by adopting the hierarchical entropy (the number of decomposition layers n =1 of the hierarchical entropy is adopted) to form a feature matrix, the feature matrix of the training set is input into a support vector machine optimized by a cuckoo algorithm for training, a trained model is used for prediction, a recognition result is shown in figure 12, and the accuracy reaches 100%.
The results of comparing the method proposed by the present invention with other conventional methods are shown in table 1, and the superiority of the diagnostic method of the present invention can be seen.
TABLE 1 Experimental comparison results table
Algorithm | Percent classification accuracy% |
SVM | 88.09% |
ELM | 86.19% |
PSO-SVM | 98.57% |
CS- |
100% |
Conclusion of this example:
(1) Aiming at the health detection problem of the rolling bearing, the invention provides a novel method for constructing a health index by adopting a Gini index. Compared with the root mean square value of the common index, the Gini index can well describe the degradation process of the bearing and is more sensitive to early failure.
(2) The goodness of the noise reduction effect for MCKD depends mainly on the filter lengthLAnd period of failureTThe optimal parameter selection method for selecting the length of the filter and selecting the fault period by using the permutation entropy is provided for solving the problem of two parameters, so that the MCKD is more adaptive, and the limitation of parameters selected by manual experience is overcome.
(3) Parameter optimization is carried out on the fault diagnosis classification model through the cuckoo algorithm, so that the diagnosis accuracy of the fault diagnosis model can be effectively improved.
The English abbreviation term noun explanation in the text:
MCKD: deconvoluting the maximum correlation kurtosis; gini: a kini index;H p : arranging entropy;Q j : a hierarchical entropy; SVM: a support vector machine; CS-SVM: a support vector machine optimized by a cuckoo algorithm; PSO-SVM is a support vector machine optimized by a particle swarm algorithm; ELM: an extreme learning machine.
Claims (4)
1. A new method for diagnosing the fault state of a rolling bearing is characterized by comprising the following steps:
arranging a vibration signal sensor on the surface of a bearing to be tested, and connecting the vibration signal sensor with a testing device to obtain a vibration signal of a rolling bearing as an original diagnosis signal;
step (II): evaluating the health state of the acquired vibration signals of the rolling bearing by adopting a Gini index, and finishing the definition of a fault stage;
step (three): performing noise reduction preprocessing on the signal in the fault stage by adopting maximum correlation kurtosis deconvolution of an optimal parameter and highlighting an impact component of the signal;
step (IV): and extracting the characteristics of the vibration signals of all stages by adopting a hierarchical entropy, and finishing the classification of the fault states of the rolling bearings by adopting a support vector machine optimized by a cuckoo algorithm.
2. The method for diagnosing a failure state of a rolling bearing according to claim 1, wherein the vibration signals of the rolling bearing in the step (a) are collected by setting acceleration sensors in horizontal and vertical directions for the bearing to be tested, and connecting the acceleration sensors to the detection device.
3. The new method for diagnosing a failure state of a rolling bearing according to claim 1, wherein the kuni index corresponding to the failure state in the step (one) is divided into:
the variation of the Gini index between 0.415 and 0.428 is normal;
fluctuations in the Gini index between 0.42 and 0.51 are early failures;
the variation of the Gini index between 0.2 and 0.515 was considered to be a failure.
4. The new method for diagnosing the failure state of the rolling bearing according to claim 1, wherein the data operation of each step is performed in a computer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210882129.4A CN115374811A (en) | 2022-07-26 | 2022-07-26 | Novel fault state diagnosis method for rolling bearing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210882129.4A CN115374811A (en) | 2022-07-26 | 2022-07-26 | Novel fault state diagnosis method for rolling bearing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115374811A true CN115374811A (en) | 2022-11-22 |
Family
ID=84063996
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210882129.4A Pending CN115374811A (en) | 2022-07-26 | 2022-07-26 | Novel fault state diagnosis method for rolling bearing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115374811A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106844826A (en) * | 2016-12-02 | 2017-06-13 | 上海电机学院 | A kind of method for the diagnosis of gearbox of wind turbine failure predication |
CN107657088A (en) * | 2017-09-07 | 2018-02-02 | 南京工业大学 | Fault Diagnosis of Roller Bearings based on MCKD algorithms and SVMs |
CN108364021A (en) * | 2018-02-08 | 2018-08-03 | 西北工业大学 | A kind of bearing fault characteristics extracting method arranging entropy based on level |
CN110595751A (en) * | 2019-09-19 | 2019-12-20 | 华东理工大学 | Early fault characteristic wavelet reconstruction method guided by Gini index and application thereof |
CN113468688A (en) * | 2021-07-05 | 2021-10-01 | 西安交通大学 | Bearing fault diagnosis method based on parameter optimization VMD and weighted Gini index |
CN113702044A (en) * | 2021-08-13 | 2021-11-26 | 华中科技大学 | Bearing fault detection method and system |
-
2022
- 2022-07-26 CN CN202210882129.4A patent/CN115374811A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106844826A (en) * | 2016-12-02 | 2017-06-13 | 上海电机学院 | A kind of method for the diagnosis of gearbox of wind turbine failure predication |
CN107657088A (en) * | 2017-09-07 | 2018-02-02 | 南京工业大学 | Fault Diagnosis of Roller Bearings based on MCKD algorithms and SVMs |
CN108364021A (en) * | 2018-02-08 | 2018-08-03 | 西北工业大学 | A kind of bearing fault characteristics extracting method arranging entropy based on level |
CN110595751A (en) * | 2019-09-19 | 2019-12-20 | 华东理工大学 | Early fault characteristic wavelet reconstruction method guided by Gini index and application thereof |
CN113468688A (en) * | 2021-07-05 | 2021-10-01 | 西安交通大学 | Bearing fault diagnosis method based on parameter optimization VMD and weighted Gini index |
CN113702044A (en) * | 2021-08-13 | 2021-11-26 | 华中科技大学 | Bearing fault detection method and system |
Non-Patent Citations (2)
Title |
---|
刘波 等: "MCKD与改进的LSSVM在滚动轴承故障诊断中的应用", 《测控技术与仪器仪表》, pages 81 - 85 * |
杨斌 等: "基于CEEMD和自适应MCKD诊断滚动轴承早期故障", 《北京工业大学学报》, 28 February 2019 (2019-02-28), pages 111 - 118 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Pan et al. | LiftingNet: A novel deep learning network with layerwise feature learning from noisy mechanical data for fault classification | |
CN109460618B (en) | Rolling bearing residual life online prediction method and system | |
CN110276416B (en) | Rolling bearing fault prediction method | |
Yang et al. | Bearing fault automatic classification based on deep learning | |
Lin et al. | Hyper-spherical distance discrimination: A novel data description method for aero-engine rolling bearing fault detection | |
CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
CN112906644A (en) | Mechanical fault intelligent diagnosis method based on deep migration learning | |
CN112729834B (en) | Bearing fault diagnosis method, device and system | |
CN115901249B (en) | Rolling bearing performance degradation evaluation method combining feature optimization and multi-strategy optimization SVDD | |
CN111076934A (en) | Method for diagnosing potential fault of bearing based on S transformation | |
CN114755017B (en) | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network | |
Shi et al. | Generalized variable-step multiscale Lempel-Ziv complexity: a feature extraction tool for bearing fault diagnosis | |
Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
CN114091525A (en) | Rolling bearing degradation trend prediction method | |
Li et al. | An orthogonal wavelet transform-based K-nearest neighbor algorithm to detect faults in bearings | |
Wei et al. | WSAFormer-DFFN: A model for rotating machinery fault diagnosis using 1D window-based multi-head self-attention and deep feature fusion network | |
Xu et al. | Fault diagnosis of subway traction motor bearing based on information fusion under variable working conditions | |
CN116401950B (en) | Rolling bearing performance degradation prediction method based on time chart convolution neural network | |
CN113283028A (en) | Fault diagnosis method for gear of gear box | |
Wan et al. | Adaptive asymmetric real Laplace wavelet filtering and its application on rolling bearing early fault diagnosis | |
Hu et al. | Incipient mechanical fault detection based on multifractal and MTS methods | |
Zhang et al. | Gearbox health condition identification by neuro-fuzzy ensemble | |
CN115374811A (en) | Novel fault state diagnosis method for rolling bearing | |
Qin et al. | Application of sensitive dimensionless parameters and PSO–SVM for fault classification in rotating machinery |
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 |