CN110763466A - Adaboost algorithm combined GABP rolling bearing diagnosis method - Google Patents
Adaboost algorithm combined GABP rolling bearing diagnosis method Download PDFInfo
- Publication number
- CN110763466A CN110763466A CN201911081380.5A CN201911081380A CN110763466A CN 110763466 A CN110763466 A CN 110763466A CN 201911081380 A CN201911081380 A CN 201911081380A CN 110763466 A CN110763466 A CN 110763466A
- Authority
- CN
- China
- Prior art keywords
- gabp
- formula
- classifier
- adaboost
- diagnosis
- 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
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 discloses a GABP rolling bearing diagnosis method combined by Adaboost algorithm, which processes bearing fault signals, selects a characteristic sample set by combining a factor analysis method, takes a BP neural network optimized by genetic algorithm as a basic classifier, and realizes the diagnosis of the bearing fault by a strong classifier by utilizing the enhancement capability of the Adaboost algorithm.
Description
Technical Field
The invention relates to a fault diagnosis technology, in particular to a GABP rolling bearing diagnosis method based on Adaboost algorithm combination.
Background
In a rotating machine system, a bearing plays an important role in bearing and transmission capacity, and the performance state of the bearing can determine whether the machine can operate well or not. The bearing has larger service life dispersion due to bearing materials, preparation technology, working environment and the like, and is very easy to break down in the long-term use process. In some rotating machines, it is difficult to diagnose the bearings in time because the bearings are mounted at special positions or are soaked in oil for a long time.
In the early fault diagnosis of the bearing, the fault diagnosis of the bearing is often carried out according to the size, materials and the like of oil abrasive particles, but the method based on the oil analysis is often greatly influenced by the skill level of a tester. With the intensive research on rotor dynamics, the diagnosis of faults can be realized according to vibration signals of an object.
In early fault diagnosis, partial signal statistical characteristic parameters are extracted, and bearing faults are diagnosed by combining a support vector machine, an extreme learning machine, a probabilistic neural network and the like, so that good results are obtained. However, in practice, the effect of simply diagnosing the network is poor, and on the other hand, optimizing the diagnostic network by using the optimization algorithm can result in the multiplied diagnostic time. In the real-time diagnosis process, the requirements on diagnosis time and accuracy are high. The Adaboost algorithm can improve the diagnosis effect of the weak classifier, is applied to the aspects of train key part fault identification and airborne fuel pump fault diagnosis, and has a less ideal effect.
Disclosure of Invention
The invention aims to solve the technical problem of providing a GABP rolling bearing diagnosis method combined with Adaboost algorithm, which processes bearing fault signals, selects a characteristic sample set by combining a factor analysis method, takes a BP neural network optimized by genetic algorithm as a basic classifier, and realizes the diagnosis of the strong classifier on the bearing fault by utilizing the enhancement capability of the Adaboost algorithm.
In order to solve the above problems, the present invention provides the following technical solutions: a GABP rolling bearing diagnosis method combined by Adaboost algorithm specifically comprises the following steps:
(1) acquiring rolling bearing signals by using a sensor:
and acquiring normal conditions of the bearing and outer ring scratch fault signals by adopting an acceleration sensor.
(2) And (3) normalizing the time domain characteristic parameters:
and (3) carrying out normalization processing on the time domain characteristic parameters by adopting a formula (1):
in the formula (1), xmax、xminRepresenting the maximum, minimum and mean values, x, of the characteristic parameteriThe ith value of a certain characteristic parameter is represented, and x represents the normalized value.
(3) Dimension reduction processing is carried out on time domain characteristic parameters
α factor analysis is adopted in the factor analysis to extract features, and the first 5 factors with the cumulative variance reaching 99.463% are selected as input parameters of a diagnostic model.
(4) Building an ELM-Adaboost-based fault diagnosis prediction model
A plurality of basic classifier GABP models are combined into a strong classifier through an Adaboost lifting algorithm, and the diagnosis result obtained by the strong classifier is expected to be the fault type label of the diagnosis target directly.
Further, the basic classifier in the step (4) is a genetic algorithm optimized BP neural network, which comprises a genetic algorithm optimizing part and a neural network part, and the specific process of the genetic algorithm optimized BP neural network is as follows:
step 1: according to sample data to be diagnosed, determining the structure of the BP network and parameters to be optimized, and dividing the data samples into the following parts in a non-repeated way: training set, checking set and testing set;
step 2: setting the population size and the iteration step number in the GA according to the parameters to be optimized in the BP network;
and step 3: selecting, crossing and mutating the population to generate new individuals, and calculating the fitness value of the population according to a formula (2):
n in the formula (2) is the number of check set samples, yiFor desired output of the network, OiIs a real output;
and 4, step 4: storing the optimal solution according to the fitness value, repeating the step 3, recording the optimal solution, and replacing the local optimal solution with the global optimal solution;
and 5: judging whether a termination condition is met, and if not, repeating the step 3 and the step 4; if yes, outputting the optimal weight and threshold of the BP network.
Further, in the step (4), a plurality of basic classifiers GABP models are combined into a strong classifier through an Adaboost algorithm, and in order to better allocate the weight proportion of each basic classifier, the Adaboost algorithm needs to be appropriately adjusted, and a multi-classified training data set T { (x)1,y1),(x2,y2),..,(xN,yN) In which xi∈Rn,y i1,2,3,4, the specific process is as follows:
step 1: initializing the weight distribution of training data;
wi=1/N,i=1,2,...,N
step 2: cycle M1: M (M represents the number of basic classifiers)
(1) Using basic classifiers BPm(x) Training data, and calculating a weight error through the following formula (3);
(2) calculating the weight of the mth basic classifier by adopting a formula (4) (k represents the classification number);
(3) updating the weights of all sample data by adopting a formula (5);
and step 3: after all classifiers are circulated, the number a of basic classifiers is calculatedmCarrying out normalization processing, and outputting a GABP-Adaboost classifier through a formula (6);
advantageous effects
The invention relates to a GABP rolling bearing diagnosis method combined with Adaboost algorithm, which comprises the steps of processing a bearing fault signal, selecting a characteristic sample set by combining a factor analysis method, taking a BP neural network optimized by genetic algorithm as a basic classifier, and realizing diagnosis of the bearing fault by a strong classifier by utilizing the enhancement capability of the Adaboost algorithm. The basic classifier GABP model mainly comprises a genetic algorithm optimizing part and a neural network part, firstly, an initial population is used as a weight and a threshold of a BP network, an error output by the BP network is used as a fitness function, and the weight and the threshold of the BP network are adjusted by carrying out crossing and variation operations on the initial population, so that the optimal parameters of the BP network are selected. In conclusion, after the basic classifier BP models form the strong classifier model, the fault recognition rate and stability of the classifier can be improved, and therefore the bearing fault can be effectively recognized.
Drawings
FIG. 1 is a flow chart of a strong classifier GABP-Adaboost model.
Fig. 2 is a flow chart of a basic classifier GABP model algorithm.
DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION
The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments.
A GABP rolling bearing diagnosis method combined by Adaboost algorithm specifically comprises the following steps:
(1) acquiring rolling bearing signals by using a sensor:
and acquiring normal conditions of the bearing and outer ring scratch fault signals by adopting an acceleration sensor.
(2) And (3) normalizing the time domain characteristic parameters:
and (3) carrying out normalization processing on the time domain characteristic parameters by adopting a formula (1):
in the formula (1), xmax、xminRepresenting the maximum, minimum and mean values, x, of the characteristic parameteriThe ith value of a certain characteristic parameter is represented, and x represents the normalized value.
(3) Dimension reduction processing is carried out on time domain characteristic parameters
α factor analysis is adopted in the factor analysis to extract features, and the first 5 factors with the cumulative variance reaching 99.463% are selected as input parameters of a diagnostic model.
(4) Building an ELM-Adaboost-based fault diagnosis prediction model
A plurality of basic classifier GABP models are combined into a strong classifier through an Adaboost lifting algorithm, and the diagnosis result obtained by the strong classifier is expected to be the fault type label of the diagnosis target directly.
According to the method, bearing fault signals are processed, a characteristic sample set is selected by combining a factor analysis method, a BP neural network optimized by a genetic algorithm is used as a basic classifier, and diagnosis of the bearing fault by a strong classifier is realized by utilizing the enhancement capability of an Adaboost algorithm.
The following experimental data are compiled from the bearing data center website of western university of storage, usa, and 6205-2RS-JEM-SKF type rolling bearings are adopted in the rolling bearing fault test, and the bearing size parameters are shown in table 1:
TABLE 1
In the test, the rotation frequency of the motor is 1730r/min, the frequency is 48kHz, and the whole period sampling is adopted to obtain 60 groups of data in total of vibration signals of inner ring faults, outer ring faults and normal states.
Because the acquired vibration signals have certain drift, the drift processing is carried out by adopting a method of fitting and eliminating a trend term by a least square method. Meanwhile, the aliasing phenomenon of weak noise is often generated during acquisition, and the vibration signal is subjected to smoothing processing.
Considering the error brought to the diagnosis result by the order of magnitude and dimension problems among the time domain characteristic parameters, the time domain characteristic parameters need to be normalized.
The factor analysis adopts α factor analysis to extract features, and the first 5 factors with accumulated variance reaching 99.463% are selected as input parameters of the diagnosis model, namely, the average value, the absolute value, the peak-to-peak value, the root mean square value and the skewness, and the time domain parameters after data normalization and factor dimension reduction are shown in table 2.
TABLE 2
Collecting 100 groups of sample data of each type of fault, totaling 300 groups of data, and randomly and repeatedly distributing the data samples as follows: 240 training set samples, 60 test set samples. In order to avoid irreproducibility brought to the diagnosis result by the genetic algorithm and the weight value and the threshold value of the randomly generated initial value of the BP network, the diagnosis result is an average value after 30 times.
The GABP, the BP network and the BP-Adaboost are used as comparison algorithms, the average diagnosis result is shown in table 3, the simple BP neural network has the worst diagnosis effect, the diagnosis accuracy of a training set and a test set is less than 80%, the diagnosis effect of the BP network improved by the Adaboost algorithm is superior to that of the GABP, the diagnosis time is shortened by nearly 70 times, the GABP diagnosis effect improved by the Adaboost algorithm is the best, the diagnosis rate reaches 90%, and the GA optimization BP part exists in the algorithm, so that the diagnosis time is the longest.
TABLE 3
By comprehensively comparing the data of 30 repeated diagnosis results, the GABP-Adaboost has better repeated diagnosis effect, weaker fluctuation and better error convergence effect.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are also included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (3)
1. A GABP rolling bearing diagnosis method combined by Adaboost algorithm is characterized by comprising the following steps:
(1) acquiring rolling bearing signals by using a sensor:
and acquiring normal conditions of the bearing and outer ring scratch fault signals by adopting an acceleration sensor.
(2) And (3) normalizing the time domain characteristic parameters:
and (3) carrying out normalization processing on the time domain characteristic parameters by adopting a formula (1):
in the formula (1), xmax、xminRepresenting the maximum, minimum and mean values, x, of the characteristic parameteriThe ith value of a certain characteristic parameter is represented, and x represents the normalized value.
(3) Dimension reduction processing is carried out on time domain characteristic parameters
α factor analysis is adopted in the factor analysis to extract features, and the first 5 factors with the cumulative variance reaching 99.463% are selected as input parameters of a diagnostic model.
(4) Building an ELM-Adaboost-based fault diagnosis prediction model
A plurality of basic classifier GABP models are combined into a strong classifier through an Adaboost lifting algorithm, and the diagnosis result obtained by the strong classifier is expected to be the fault type label of the diagnosis target directly.
2. The method for diagnosing the GABP rolling bearing combined by the Adaboost algorithm according to claim 1, wherein the basic classifier in the step (4) is a BP neural network optimized by the genetic algorithm, which comprises a genetic algorithm optimizing part and a neural network part, and the specific process of the BP neural network optimized by the genetic algorithm is as follows:
step 1: according to sample data to be diagnosed, determining the structure of the BP network and parameters to be optimized, and dividing the data samples into the following parts in a non-repeated way: training set, checking set and testing set;
step 2: setting the population size and the iteration step number in the GA according to the parameters to be optimized in the BP network;
and step 3: selecting, crossing and mutating the population to generate new individuals, and calculating the fitness value of the population according to a formula (2):
n in the formula (2) is the number of check set samples, yiFor desired output of the network, OiIs a real output;
and 4, step 4: storing the optimal solution according to the fitness value, repeating the step 3, recording the optimal solution, and replacing the local optimal solution with the global optimal solution;
and 5: judging whether a termination condition is met, and if not, repeating the step 3 and the step 4; if yes, outputting the optimal weight and threshold of the BP network.
3. The method for diagnosing the GABP rolling bearing combined by the Adaboost algorithm according to claim 1, wherein the step (4) combines a plurality of basic classifier GABP models into a strong classifier by the Adaboost algorithm, and the Adaboost algorithm needs to be adjusted appropriately to better distribute the weight ratio of each basic classifier, and a multi-classification training data set T { (x)1,y1),(x2,y2),..,(xN,yN) In which xi∈Rn,yi1,2,3,4, the specific process is as follows:
step 1: initializing the weight distribution of training data;
wi=1/N,i=1,2,...,N
step 2: cycle M1: M (M represents the number of basic classifiers)
(1) Using basic classifiers BPm(x) Training data, and calculating a weight error through the following formula (3);
(2) calculating the weight of the mth basic classifier by adopting a formula (4) (k represents the classification number);
(3) updating the weights of all sample data by adopting a formula (5);
and step 3: after all classifiers are circulated, the number a of basic classifiers is calculatedmCarrying out normalization processing, and outputting a GABP-Adaboost classifier through a formula (6);
round (x) in the said formula (6) represents a rounding function, and
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911081380.5A CN110763466A (en) | 2019-11-07 | 2019-11-07 | Adaboost algorithm combined GABP rolling bearing diagnosis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911081380.5A CN110763466A (en) | 2019-11-07 | 2019-11-07 | Adaboost algorithm combined GABP rolling bearing diagnosis method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110763466A true CN110763466A (en) | 2020-02-07 |
Family
ID=69336765
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911081380.5A Pending CN110763466A (en) | 2019-11-07 | 2019-11-07 | Adaboost algorithm combined GABP rolling bearing diagnosis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110763466A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111508517A (en) * | 2020-04-28 | 2020-08-07 | 电子科技大学中山学院 | Intelligent micro-motor product control method based on noise characteristics |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105677564A (en) * | 2016-01-04 | 2016-06-15 | 中国石油大学(华东) | Adaboost software defect unbalanced data classification method based on improvement |
CN106529577A (en) * | 2016-10-20 | 2017-03-22 | 北京工业大学 | Synchronization selection, weighting and sorting method for weak classifiers in AdaBoost algorithm |
CN110307982A (en) * | 2019-06-19 | 2019-10-08 | 电子科技大学 | Bearing fault classification method based on CNN and Adaboost |
-
2019
- 2019-11-07 CN CN201911081380.5A patent/CN110763466A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105677564A (en) * | 2016-01-04 | 2016-06-15 | 中国石油大学(华东) | Adaboost software defect unbalanced data classification method based on improvement |
CN106529577A (en) * | 2016-10-20 | 2017-03-22 | 北京工业大学 | Synchronization selection, weighting and sorting method for weak classifiers in AdaBoost algorithm |
CN110307982A (en) * | 2019-06-19 | 2019-10-08 | 电子科技大学 | Bearing fault classification method based on CNN and Adaboost |
Non-Patent Citations (3)
Title |
---|
张坤: "城轨列车走行部滚动轴承故障诊断算法研究", 《中国优秀硕士学位论文全文数据库工程科技II辑》 * |
徐腊梅: "《质子交换膜燃料电池模拟与优化》", 31 August 2012, 国防工业出版社 * |
阮敬: "《Python数据分析基础》", 30 September 2017, 北京:中国统计出版社 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111508517A (en) * | 2020-04-28 | 2020-08-07 | 电子科技大学中山学院 | Intelligent micro-motor product control method based on noise characteristics |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6947219B2 (en) | Abnormal sound detection device, abnormal model learning device, abnormal detection device, abnormal sound detection method, abnormal sound generation device, abnormal data generation device, abnormal sound generation method, and program | |
CN110598851A (en) | Time series data abnormity detection method fusing LSTM and GAN | |
CN110647867B (en) | Bearing fault diagnosis method and system based on self-adaptive anti-noise neural network | |
CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
RU2488815C2 (en) | Method and apparatus for classifying sound-generating processes | |
CN113780056A (en) | Rolling bearing fault diagnosis method based on vibration signal | |
CN110132554B (en) | Rotary machine fault diagnosis method based on deep Laplace self-coding | |
CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
CN110297479B (en) | Hydroelectric generating set fault diagnosis method based on convolutional neural network information fusion | |
CN111538311B (en) | Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining | |
CN113532829A (en) | Reciprocating compressor fault diagnosis method based on improved RCMDE | |
CN114266339A (en) | Rolling bearing fault diagnosis method based on small sample and GAF-DCGAN | |
CN110763466A (en) | Adaboost algorithm combined GABP rolling bearing diagnosis method | |
CN115034137A (en) | RVM and degradation model-based two-stage hybrid prediction method for residual life of bearing | |
CN113758709A (en) | Rolling bearing fault diagnosis method and system combining edge calculation and deep learning | |
CN114169460A (en) | Sample screening method, sample screening device, computer equipment and storage medium | |
CN113469252A (en) | Extra-high voltage converter valve operation state evaluation method considering unbalanced samples | |
CN109508350B (en) | Method and device for sampling data | |
CN116735198A (en) | Fan gear box fault diagnosis method and system based on speEMD and double index screening | |
CN115901255A (en) | Bearing fault diagnosis method based on parameter optimization VMD-SVM | |
CN113409213B (en) | Method and system for enhancing noise reduction of time-frequency diagram of fault signal of plunger pump | |
CN115876467A (en) | Pseudo label transfer type two-stage field self-adaptive rolling bearing fault diagnosis method | |
CN114861349A (en) | Rolling bearing RUL prediction method based on model migration and wiener process | |
CN114705431A (en) | Rolling bearing fault diagnosis method based on multi-parameter screening criterion and GWO-PNN | |
CN113051809A (en) | Virtual health factor construction method based on improved restricted Boltzmann machine |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200207 |
|
RJ01 | Rejection of invention patent application after publication |