CN115060497B - Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN - Google Patents
Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN Download PDFInfo
- Publication number
- CN115060497B CN115060497B CN202210655047.6A CN202210655047A CN115060497B CN 115060497 B CN115060497 B CN 115060497B CN 202210655047 A CN202210655047 A CN 202210655047A CN 115060497 B CN115060497 B CN 115060497B
- Authority
- CN
- China
- Prior art keywords
- neural network
- bearing
- fault diagnosis
- probabilistic neural
- pnn
- 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.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 49
- 241000287127 Passeridae Species 0.000 claims abstract description 67
- 238000013528 artificial neural network Methods 0.000 claims abstract description 39
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 28
- 238000010845 search algorithm Methods 0.000 claims abstract description 25
- 239000011159 matrix material Substances 0.000 claims abstract description 19
- 238000005457 optimization Methods 0.000 claims abstract description 18
- 238000003062 neural network model Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000000295 complement effect Effects 0.000 claims abstract description 7
- 238000012216 screening Methods 0.000 claims abstract description 6
- 238000009499 grossing Methods 0.000 claims description 18
- 210000002569 neuron Anatomy 0.000 claims description 18
- 238000009826 distribution Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000000739 chaotic effect Effects 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000012360 testing method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 244000062645 predators Species 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004904 shortening Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization, which comprises the following steps: s1, collecting bearing vibration signals under different fault states; s2, decomposing the bearing vibration signal by adopting a complementary set empirical mode decomposition algorithm to obtain an intrinsic mode component; s3, calculating correlation coefficients of the eigenmode components and bearing vibration signal data, and screening to obtain effective eigenmode components; s4, extracting energy entropy of effective eigenvector components to form a feature vector matrix; s5, constructing a probabilistic neural network model, inputting the feature vector matrix into the probabilistic neural network model, training and optimizing the probabilistic neural network model, and obtaining a probabilistic neural network bearing fault diagnosis model; and S6, inputting the feature vector matrix into the probabilistic neural network bearing fault diagnosis model, constructing the probabilistic neural network fault diagnosis model optimized by the improved sparrow search algorithm, and completing fault identification and classification.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis, and particularly relates to a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization.
Background
The bearing is used as an important basic component of mechanical equipment and often operates under the complex working conditions of heavy load, impact, speed change and the like, and faults in the forms of rolling body deformation, abrasion, corrosion and the like are easy to occur. The bearing failure will seriously affect the normal operation of the mechanical equipment and even cause safety accidents. Therefore, the timely, accurate and reliable detection and fault diagnosis of the bearing state are realized, and the method has important significance for guaranteeing the performance efficiency, the service life and the safe operation of mechanical equipment.
Due to vibration coupling of a mechanical system and complex environmental influence, vibration signals have the characteristics of nonlinearity and non-stability, and conventional signal processing methods based on time domain, frequency domain or time-frequency domain principles are difficult to process. Empirical Mode Decomposition (EMD) is an important method for analyzing nonlinear and unsteady signals, and can adaptively decompose the signals into a series of intrinsic mode components (IMFs) under different time scales, but has the problem of mode aliasing, and reduces the signal decomposition accuracy. In order to improve the defect of EMD, the method of integrated empirical mode decomposition (EEMD) changes the distribution state of extreme points in the signal by adding Gaussian white noise with different amplitudes into the original signal, so that the problem of modal aliasing can be effectively solved, but EEMD needs to rely on increasing operation times to eliminate the influence of the added Gaussian white noise on the decomposition result, and obviously has the defects of low calculation efficiency and larger reconstruction error. The Variational Modal Decomposition (VMD) method has a complete mathematical basis as a non-recursive signal decomposition method, can realize a better signal decomposition effect on the basis of improving modal aliasing, but has the problem that parameters such as decomposition number, penalty factor and the like are difficult to select. Currently, the information entropy is used as a measure of uncertainty of a signal or a system, can reflect vibration characteristics of the bearing in different fault states, is used as a characteristic extraction method of the information entropy, and can effectively represent the change of signal characteristics from the angle of energy change. In the field of fault diagnosis, a Probabilistic Neural Network (PNN) has the advantages of simple calculation process and high convergence speed compared with other artificial intelligence methods, and has extremely high stability and fault tolerance on individual abnormal data. The smoothing factor is used as the unique input parameter in the PNN and directly influences the identification and diagnosis performance of the PNN. However, the PNN smoothing factor is often selected by means of manual experience, and lacks of adaptivity, and conventional optimization methods such as Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) have the disadvantages of being prone to being trapped in local optimum, being too slow in convergence speed and the like. Compared with the traditional intelligent optimization algorithm, the algorithm has advantages in searching precision, convergence speed, stability and the like, but due to the problems that the SSA is difficult to ensure enough randomness during population initialization, an edge searching mechanism is not stable enough and the like, the situation that local optimization cannot be jumped out and global optimization cannot be realized still occurs.
Disclosure of Invention
The invention aims to provide a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization, which improves the feature extraction precision of bearing vibration signals and the fault diagnosis accuracy.
In order to achieve the above object, the present invention provides a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization, comprising the steps of:
s1, collecting bearing vibration signals under different fault states;
s2, decomposing the bearing vibration signal by adopting a complementary set empirical mode decomposition algorithm to obtain an intrinsic mode component;
s3, calculating correlation coefficients of the eigenmode components and the bearing vibration signal data, and screening to obtain effective eigenmode components;
s4, extracting energy entropy of the effective eigenmode components to form a feature vector matrix;
s5, constructing a probabilistic neural network model, inputting the feature vector matrix into the probabilistic neural network model, training and optimizing the probabilistic neural network model, and obtaining a probabilistic neural network bearing fault diagnosis model;
and S6, inputting the feature vector matrix into the probabilistic neural network bearing fault diagnosis model, constructing a probabilistic neural network fault diagnosis model optimized by the improved sparrow search algorithm, and completing the identification and classification of the improved sparrow search algorithm optimized probabilistic neural network fault.
Optionally, in the step S2, the method for decomposing the bearing vibration signal by using a complementary set empirical mode decomposition algorithm to obtain an intrinsic mode component includes:
s2.1, setting Gaussian white noise group numbers, and obtaining Gaussian white noise by using a random function;
s2.2, adding a pair of Gaussian white noise with opposite signs and equal amplitude values into the bearing vibration signals to obtain two new bearing vibration signals;
s2.3, respectively carrying out empirical mode decomposition on the two new bearing vibration signals to obtain two groups of intrinsic mode components;
s2.4, iteratively repeating the S2.2 and the S2.3 by taking the calculation times as 1 unit until the calculation times are the Gaussian white noise group number;
s2.5, decomposing the Gaussian white noise group for several times, and then obtaining the intrinsic mode component by integrating the average value of all the intrinsic mode components.
Optionally, the method for obtaining the effective eigenmode component in S3 includes:
and arranging the eigenmode components, and screening out the effective eigenmode components by calculating the correlation coefficient of the eigenmode components and the bearing vibration signal data.
Optionally, the method for extracting the energy entropy of the effective eigenmode component in S4 and forming the feature vector matrix includes:
based on the screened effective eigenmode components, the total energy of the effective eigenmode components is obtained, the proportion of the energy of each effective eigenmode component to the total energy is calculated, the energy entropy of each effective eigenmode component is obtained, the energy entropy of the effective eigenmode components of vibration signals of different fault states of the bearing is extracted, and a bearing fault feature vector matrix is formed.
Optionally, constructing the probabilistic neural network model in S5 includes:
the probabilistic neural network model comprises an input layer, a mode layer, a summation layer and an output layer;
the input layer receives the bearing vibration signal, and the eigenvectors in the eigenvector matrix correspond to the energy entropy of the effective eigenvector component;
the mode layer calculates and inputs the characteristic vector to be matched with the bearing vibration signal to obtain the quantity of neurons, and the mode layer is connected with the input layer through connection weights;
the number of the neurons in the summation layer is the same as the total number of the mode layers, and the neurons in the summation layer are connected with the corresponding neurons in the mode layers;
the output layer is composed of the same number of competing neurons as the summing layer.
Optionally, the method for obtaining the improved sparrow search algorithm optimization probabilistic neural network bearing fault diagnosis model in S5 includes:
optimizing initialization conditions of sparrow populations by utilizing the characteristic of randomness and ergodic property of the improved Logistic chaotic map;
calculating and sequencing the fitness value of each sparrow, and selecting the current optimal fitness value and the corresponding position thereof, and the worst fitness value and the corresponding position thereof;
after one iteration is completed, the fitness value of each sparrow is recalculated, and the optimal position and the fitness of the whole sparrow population, and the fitness of the worst position and the worst position are updated according to the current state of the sparrow population.
Optionally, introducing a t distribution strategy, obtaining a new solution at the optimal position of the solution by using a disturbance operator, calculating individual fitness values of the original population and the new variant population, respectively determining the optimal solutions of the two populations before and after the variant, and recording the corresponding positions.
Optionally, constructing the probabilistic neural network fault diagnosis model optimized by the improved sparrow search algorithm includes: and optimizing the smoothing factor of the probabilistic neural network by adopting an improved sparrow search algorithm, obtaining the optimized smoothing factor, inputting the optimized smoothing factor into the probabilistic neural network model, and constructing a probabilistic neural network fault diagnosis model optimized by the improved sparrow search algorithm.
The invention has the technical effects that: the invention discloses a bearing fault diagnosis method based on CEEMD energy entropy and optimizing PNN, which adopts the CEEMD method to adaptively decompose a bearing vibration signal into a series of intrinsic mode components, effectively reduces reconstruction signal errors caused by adding white noise, better inhibits mode aliasing phenomenon, and improves the decomposition precision of nonlinear and nonstationary signals; the correlation coefficient criterion is utilized to effectively remove some pseudo-modal components caused by noise, and effective intrinsic modal components representing the characteristics of the vibration signal are reserved to the maximum extent; the energy entropy of the effective eigenmode components is used as a feature vector, fault state information is effectively reflected, the difference between different states is more obvious, and the data dimension is reduced on the basis of improving the fault recognition precision; by improving the traditional sparrow searching algorithm, the problems of uneven initial population distribution and local optimum sinking are effectively solved, the global searching capability is enhanced, and the convergence rate is improved; the accuracy of bearing fault diagnosis is further improved by constructing an ISSA-PNN diagnosis model of the probability neural network optimized by the improved sparrow search algorithm.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:
FIG. 1 is a schematic flow chart of a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization in an embodiment of the invention;
fig. 2 is a schematic structural diagram of a probabilistic neural network PNN in a bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to an embodiment of the present invention;
FIG. 3 is a flowchart of ISSA optimized PNN in a bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to an embodiment of the present invention;
FIG. 4 is a time domain waveform diagram of vibration fault signals of 6 different bearings in a bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to an embodiment of the present invention;
FIG. 5 is a graph showing the correlation coefficient distribution of each eigen mode component and the original signal after CEEMD decomposition in the bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to the embodiment of the present invention;
FIG. 6 is an energy entropy distribution diagram of an intrinsic mode component after CEEMD decomposition of various fault signals in a bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to an embodiment of the present invention;
fig. 7 is a diagram showing the effect of fault classification of a test set in a bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN according to an embodiment of the present invention.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
As shown in fig. 1-7, the present embodiment provides a bearing fault diagnosis method based on CEEMD energy entropy and PNN optimization, which includes the following steps:
step 1: collecting vibration signal data a (t) of the bearing in different fault states to form a bearing vibration signal data set;
step 2: the bearing vibration signal is adaptively decomposed into a series of eigenmode components by using a complementary set empirical mode decomposition (CEEMD) algorithm, specifically according to the following method:
step 2.1: setting a Gaussian white noise group number M, obtaining Gaussian white noise by using a random function, setting the Gaussian white noise amplitude to be 0.1-0.3 times of the standard deviation of an original vibration signal a (t), wherein the selection range is derived from related literature and experience, and the operation frequency i=1;
step 2.2: adding a pair of Gaussian white noise with opposite signs and equal amplitude to the original vibration signal to obtain two new strings of signals:
wherein a (t) is the original vibration signal, n i (t) is the i-th added Gaussian white noise,
and->
The i-th signal to which white gaussian noise is added is each signal.
Step 2.3: for a pair of
And->
Empirical Mode Decomposition (EMD) was performed separately to obtain two sets of eigenmode components:
wherein,,
and->
The j th intrinsic mode components after the i th EMD decomposition are respectively, K is the number of the intrinsic mode components after the decomposition, and r i + (t) and r i - (t) the residual amounts obtained by the ith EMD decomposition.
Step 2.4: let i=i+1 and repeat steps 2.2 and 2.3 until i=m.
Step 2.5: the integrated average value of all the eigenmode components after M times of decomposition is calculated as each eigenmode component decomposed by CEEMD:
wherein c j (t) is the j-th eigenmode component after CEEMD decomposition, j=1, 2, …, K.
Step 3: calculating the correlation coefficient between each intrinsic mode component and an original vibration signal, selecting the intrinsic mode component with a larger correlation coefficient as an effective mode component to represent the vibration characteristic of the original signal, and removing the pseudo mode component with a smaller correlation coefficient, wherein the method specifically comprises the following steps of:
the eigenmode components obtained after CEEMD decomposition are arranged from high frequency to low frequency, where the high frequency components may contain random noise, the low frequency components may contain trend terms and residual components due to interpolation errors and boundary effects, only a portion of the eigenmode components can characterize the nature of the original signal, and the remaining eigenmode components are some spurious mode components caused by noise. Therefore, the invalid eigenmode components need to be removed to preserve the original signal features to the maximum. By calculating the respective eigenmode component c j (t) correlation coefficient with the original vibration signal a (t), selecting an eigenmode component with relatively large correlation coefficient to represent effective information in the original signal, and selecting a correlation coefficient C r (j) Obtained as follows:
wherein c j,l Is the j-th eigenmode component c j The first data point in (t), a l For the first data point, N, in the original vibration signal a (t) l The total number of data points is the number of data points,
and->
C respectively j Average of the data in (t) and a (t).
If the correlation coefficient is larger, the correlation between the intrinsic mode component and the original signal is stronger, and the characteristics of the working state of the bearing contained in the intrinsic mode component are rich. Instead, this indicates that the bearing operating state characteristics contained in the eigenmode component are less, and possibly even spurious components should be removed.
Step 4: extracting the energy entropy of each effective eigenmode component to form a feature vector matrix capable of quantitatively reflecting the running state of the bearing, wherein the method specifically comprises the following steps of:
the h effective eigenmode components are obtained after the correlation coefficient screening, and the energy E carried by each eigenmode component j Can be obtained as follows:
the total energy E of the h effective eigenmode components is then:
calculating the proportion P of the energy of each effective eigenmode component to the total energy j The method comprises the following steps:
obtaining the energy entropy H of each effective eigenmode component j The method comprises the following steps:
extracting effective eigenvalues of vibration signals of different fault states of bearingThe energy entropy value of the modal component forms a bearing fault characteristic vector matrix x= [ x ] 1 ,x 2 ,…,x h ] T A part of the data set is randomly selected as a training set, and the rest of the data set is used as a test set.
Step 5: a Probabilistic Neural Network (PNN) model neural network is constructed, a Probabilistic Neural Network (PNN) model and Sparrow Search Algorithm (SSA) parameters are initialized. Selecting the positions of sparrow individuals as smoothing factors of a probabilistic neural network, selecting the sum of the bearing fault diagnosis error rate of a training set and the bearing fault diagnosis error rate of a test set as an fitness function of a sparrow search algorithm to calculate the fitness value of each sparrow individual position, and optimizing the probabilistic neural network PNN, wherein the method specifically comprises the following steps:
step 5.1: a Probabilistic Neural Network (PNN) model is constructed.
The probabilistic neural network is a feedforward neural network developed based on a radial basis network, and as shown in fig. 2, the network includes four layers, an input layer, a pattern layer, a summation layer and an output layer.
The input layer receives the value from the data set, transmits the feature vector to the network, and the number of neurons and the feature vector dimension are equal, namely the number h of the effective eigenvector components obtained in the step 4 are equal, and correspond to the energy entropy of the h effective eigenvector components in the bearing fault feature vector;
the mode layer calculates the matching relation between the input characteristic vector and each mode in the data set, the mode layer is connected to the input layer through the connection weight, the number of neurons in the mode layer is equal to the product of the fault type and the number of samples, and the Gaussian function of each sample and the output function of the mode layer are expressed as:
wherein i=1, 2, …, b, k=1, 2, …, m i B is the total number of fault types in the data set, m i For the number of samples of the ith fault type in the dataset, h is the dimension of the input dataset, and sigma is the probability godSmoothing factor, x, of PNN model over network ik Is the center value of the kth sample in the class i fault type in the dataset.
The summation layer sums up probabilities belonging to a certain fault type to obtain an estimated probability density function of the classification mode. Each type has only one summing layer element which is connected to the mode layer elements belonging to the own type only and not to other elements in the mode layer, so that the summing layer elements simply add the outputs of the mode layer elements belonging to the own type independently of the outputs of the mode layer elements belonging to the other types. Output weights for class i fault types of the summation layer:
wherein g i (x) Is the output of the summing layer for the type i fault type.
The number of neurons in the summation layer is the same as the total number of pattern layers, and the neurons in this layer are connected only to the corresponding neurons in the pattern layers and not to other neurons.
The output layer is composed of competing neurons of the same number as the summation layer, where each neuron corresponds to a type of failure. The function is to receive the output generated by the summation layer and set the type corresponding to the highest probability in the network summation layer as the output result:
step 5.2: sparrow Search Algorithm (SSA) parameters are initialized.
Setting the sparrow population scale as R, wherein the proportion of discoverers and early-warning persons in the whole population is P respectively D And P S The dimension of the objective function is D, and the upper limit and the lower limit of the initial value are U respectively b And L b Maximum number of iterations t max 。
Selecting the position of the sparrow individual as a PNN smoothing factor of the Probabilistic Neural Network (PNN)The positions of r sparrows are X r ＝[X r,1 ,…,X r,d ,…,X r,D ]，r＝1,2,…,R，X r,d Indicates the position of the r-th sparrow in the D-th dimension, d=1, 2, …, D.
Selecting the sum of the bearing fault diagnosis error rate of the training set and the bearing fault diagnosis error rate of the test set as the fitness function of the sparrow search algorithm to calculate the fitness value of the r-th sparrow of the fitness value of each sparrow individual position as
Fitness is a Fitness function defined as follows:
Fitness＝argmin(Train_ErrorRate+Test_ErrorRate) (12)
wherein train_error is the bearing fault diagnosis error rate of the training set, and test_error is the bearing fault diagnosis error rate of the Test set.
The sparrow search algorithm optimizes the smoothing factor of the probabilistic neural network, namely, optimizes the position of the sparrow population with the scale of R in the D-dimensional space.
Step 6: according to the problems that the sparrow search algorithm has insufficient randomness of an initialization position and an edge search mechanism is not stable enough and falls into local optimum in actual operation, the sparrow search algorithm is improved, as shown in fig. 3, and the method specifically comprises the following steps:
step 6.1: in order to effectively maintain diversity of the population, improve global searching capability of a sparrow searching algorithm, optimize initializing conditions of the sparrow population by utilizing characteristics of randomness and ergodicity of improved Logistic chaotic mapping, and define the improved Logistic chaotic mapping by the following formula:
wherein, the initial value X 0 ∈[-1,1]，m∈[0,2]。
Step 6.2: calculating fitness value of each sparrow
And sorting, and selecting the current optimal fitness and the corresponding position thereof, and the worst fitness value and the corresponding position thereof. Selecting R.P with the best adaptability in each generation of population D Sparrows were found, the remainder of R- (1-P D ) The sparrow is used as follower with the quantity of R.P S Sparrow individuals at the edges of the population can rapidly move to the safe area when early warning is performed as early warning persons, so that a better position is obtained, and position updating is performed on discoverers, joiners and early warning persons.
The finder location is updated by:
wherein t represents the current iteration number, t max The number of iterations of the maximum is indicated,
represents the position information of the t iteration of the r-th sparrow in the d-th dimension, λε (0, 1)]Is a random number, V W ∈[0,1]For the early warning value, V ST ∈[0.5,1]For the security value, Q is a random number subject to normal distribution, and L represents a 1×d-order matrix in which each element is 1. When V is W ＜V ST When it is indicated that no predators are found to be present, the discoverer can continue with a broader search operation, but when V W ≥V ST When the predators are found by sparrows, the sparrows send out alarms which are larger than the safety value to other sparrows in the population, and the found sparrows can be led to other safe areas to quickly fly to find food.
The location update of the enrollee follows the following rules:
wherein,,
represents the optimal position, X, of the current finder at the t+1st iteration W Representing the global worst position at this time, A is a 1 XD dimensional matrix and each element is randomly assigned a value of-1 or 1, A + ＝A T (AA T ) -1 . When->
When the r-th participant is shown not to acquire food, the participant is required to fly to other places to find food in order to meet the energy demand of the participant. When->
Meaning that the nth participant will randomly seek a location to feed around the current optimal location.
The position of the precaution person is updated according to the following rule:
wherein X is B Representing the global current optimal position, beta is a step control parameter, is a random number obeying a normal distribution with a mean value of 0 and a variance of 1, and u E < -1,1]To represent a random number of sparrow movement direction, f r Is the fitness value of the current sparrow, f B And f W Epsilon is a small constant that is not 0 in order to meet denominators, respectively the current global optimum and worst fitness values. When f r ＝f B Individuals, who are present in the middle of the sparrow population at this time, are aware of the risk of being predated, and should immediately approach other sparrows to reduce their risk. When f r ＞f B And the current sparrow is positioned at the edge of the population, so that the sparrow is very easy to prey by predators.
The existence of the early warning person can enhance the global searching capability of the algorithm, but the quantity of the early warning person is kept unchanged in the iterative process, so that the later convergence speed of the algorithm is reduced. Therefore, the number of alertors is adaptively changed in the algorithm operation process, so that the method is beneficial to global search in the early stage of the algorithm and numerical convergence in the later stage of the algorithm, and an early warning releasing mechanism is set according to the following rules:
wherein S is D The proportion of the early warning persons needed when the iteration times are t.
The random sparrow individuals can cause the sparrow individuals to fly beyond the edges, which can have the consequence of being trapped in a local optimum. The edge exploration mechanism is thus set up according to the following rules:
where q.epsilon.0, 1 is a random number and w is a multiple factor set by the boundary size.
Step 6.3: after one iteration is completed, the fitness value of each sparrow is recalculated
And updating the optimal position and the fitness thereof of the whole population, and the worst position and the fitness thereof according to the current state of the sparrow population.
Step 6.4: in order to further improve the optimizing performance of the sparrow searching algorithm and improve the defect that the algorithm falls into local optimum, a t distribution strategy is introduced, and a disturbance operator is used for obtaining a new solution at the optimum position of the solution. the t distribution can fully utilize population information in the current iteration, has stronger global searching capability similar to Cauchy variation when the iteration times t are smaller, has stronger local searching capability similar to Gaussian variation when the iteration times t are larger, thereby effectively enhancing the iterative optimizing performance of an SSA algorithm, enabling the whole population individuals to quickly search the global optimal value, shortening the convergence time and updating the target position of the solution according to the following rules:
if rand (0, 1) < p s Then
Wherein,,
the variation probability of t distribution, t d And (t) represents t distribution subject to the current iteration number t as a free parameter.
And calculating individual fitness values of the original population and the new variant population, respectively determining optimal solutions of the two populations before and after the variant, if the mutated optimal solution is more optimal, updating the global optimal solution, and recording the corresponding positions.
If rand (0, 1) is greater than or equal to p s And directly executing the step 6.5 according to the t distribution strategy not executed.
Step 6.5: judging whether the maximum iteration times are reached, if not, continuing to iterate and update, otherwise, outputting the optimal smoothing factor to the probabilistic neural network PNN model, and constructing an ISSA-PNN bearing fault diagnosis model.
Step 7: optimizing a smoothing factor of the probabilistic neural network by using an improved sparrow search algorithm, obtaining an optimized smoothing factor, inputting the optimized smoothing factor into a Probabilistic Neural Network (PNN) model, and constructing a probabilistic neural network (ISSA-PNN) fault diagnosis model optimized by the improved sparrow search algorithm;
step 8: taking training set data as input and bearing fault type as output, and completing training of an ISSA-PNN bearing fault diagnosis model;
step 9: and inputting the test set data into a trained bearing fault diagnosis model of the ISSA-PNN (inverse neural network) optimized by an improved sparrow search algorithm, finishing the mapping from fault data to fault states, outputting a bearing fault type result, and realizing the fault diagnosis of the bearing.
Inspection of the method
The invention selects bearing vibration acceleration signal data provided by bearing data center of electric engineering laboratory of Kassi university of America. The vibration acceleration signal is from the driving end of the rolling bearing, the type of the bearing is SKF6025, the sampling frequency of the signal is 12kHz, the rotating speed of the bearing is 1797r/min, the load of the motor is 0HP, and the total 6 types of bearing fault types are extracted, and are shown in the table 1. Each type of fault data sample intercepts signals with the duration of 0.2s, each sample intercepts 2400 data points, 50 groups of samples are taken, and fig. 4 shows time domain signal waveforms of 6 types of fault types in the first group of samples.
The number of Gaussian white noise groups M=200 is set, the amplitude of Gaussian white noise is set to be 0.2 times of the standard deviation of an original vibration signal, and a complementary set empirical mode decomposition (CEEMD) algorithm is adopted to adaptively decompose the bearing vibration signal into a series of intrinsic mode components. The effective eigenmode components are screened by using a correlation coefficient criterion, as shown in fig. 5, the correlation coefficients between almost all the 6 th eigenmode components and the original signal are obviously reduced, which means that the correlation between the first 5 eigenmode components and the original signal is strong, and the main fault information in the signal is covered, so that the first 5 IMF components after CEEMD decomposition are selected as analysis objects to extract the energy entropy characteristics of the signal, and the fault state of the bearing is shown in table 1.
TABLE 1
The energy entropy of each effective eigenmode component is extracted to form a feature vector data set, and the distribution of the energy entropy in the 6 different bearing states, taking the first set of sample signals as an example, is shown in fig. 6. It can be seen that the energy entropy values of each eigenmode component differ significantly when the bearing is operated in different states. The value of each intrinsic mode component of the bearing in a normal state is far larger than that of other fault states, the energy entropy distribution of each intrinsic mode component in different fault states is different, and the condition that the capacity entropy value of one intrinsic mode component is larger than that of other intrinsic mode components always exists. The difference of the energy entropy of different eigenvalues can reflect the characteristic of the bearing in different working states, and can be used as the characteristic vector of the bearing state.
30 groups of samples are randomly selected and divided into training sets, and the rest 20 groups of samples are divided into test sets. Constructing a probabilistic neural network PNN model and a sparrow search SSA algorithm,
setting the scale of the sparrow population as R=20, randomly generating initial sparrow positions, wherein the proportion of discoverers and early-warning persons in the whole population is P respectively D =70% and P S =20%, the objective function dimension is d=30, and the upper and lower limits of the initial value are U respectively b =5 and L b =0.0001, maximum number of iterations t max Security value V =30 ST =0.6. And optimizing the smoothing factor by using an improved sparrow search algorithm ISSA, and obtaining the optimal smoothing factor of 0.0067. And assigning the optimal smoothing factor to the probabilistic neural network PNN model to complete training, and obtaining the ISSA-PNN fault diagnosis model. The test set data is input into an ISSA-PNN diagnosis model for fault diagnosis, and the accuracy reaches 99.17 percent, as shown in figure 7.
The foregoing is merely a preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (6)
1. A bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN, characterized by comprising the steps of:
s1, collecting bearing vibration signals under different fault states;
s2, decomposing the bearing vibration signal by adopting a complementary set empirical mode decomposition algorithm to obtain an intrinsic mode component;
s3, calculating correlation coefficients of the eigenmode components and the bearing vibration signal data, and screening to obtain effective eigenmode components;
s4, extracting energy entropy of the effective eigenmode components to form a feature vector matrix;
s5, constructing a probabilistic neural network model, inputting the feature vector matrix into the probabilistic neural network model, training and optimizing the probabilistic neural network model, and obtaining a probabilistic neural network bearing fault diagnosis model;
s6, inputting the feature vector matrix into the probabilistic neural network bearing fault diagnosis model, constructing a probabilistic neural network fault diagnosis model optimized by an improved sparrow search algorithm, and completing the identification and classification of the improved sparrow search algorithm optimized probabilistic neural network fault; the method for obtaining the improved sparrow search algorithm optimization probability neural network bearing fault diagnosis model comprises the following steps: optimizing initialization conditions of sparrow populations by utilizing the characteristic of randomness and ergodic property of the improved Logistic chaotic map; calculating and sequencing the fitness value of each sparrow, and selecting the current optimal fitness value and the corresponding position thereof, and the worst fitness value and the corresponding position thereof; after one iteration is completed, calculating the fitness value of each sparrow again, and updating the optimal position and the fitness of the whole population and the fitness of the worst position and the worst position according to the current state of the sparrow population; and introducing a t distribution strategy, obtaining a new solution at the optimal position of the solution by using a disturbance operator, calculating individual fitness values of the original population and the new variant population, respectively determining the optimal solutions of the two populations before and after the variant, and recording the corresponding positions.
2. The CEEMD energy entropy and PNN optimization based bearing fault diagnosis method according to claim 1, wherein the method for decomposing the bearing vibration signal by adopting a complementary set empirical mode decomposition algorithm in S2 to obtain an intrinsic mode component comprises the following steps:
s2.1, setting Gaussian white noise group numbers, and obtaining Gaussian white noise by using a random function;
s2.2, adding a pair of Gaussian white noise with opposite signs and equal amplitude values into the bearing vibration signals to obtain two new bearing vibration signals;
s2.3, respectively carrying out empirical mode decomposition on the two new bearing vibration signals to obtain two groups of intrinsic mode components;
s2.4, iteratively repeating the S2.2 and the S2.3 by taking the calculation times as 1 unit until the calculation times are the Gaussian white noise group number;
s2.5, decomposing the Gaussian white noise group for several times, and then obtaining the intrinsic mode component by integrating the average value of all the intrinsic mode components.
3. The CEEMD energy entropy and PNN optimization based bearing fault diagnosis method according to claim 2, wherein the method of obtaining effective eigenmode components in S3 comprises:
and arranging the eigenmode components, and screening out the effective eigenmode components by calculating the correlation coefficient of the eigenmode components and the bearing vibration signal data.
4. The CEEMD energy entropy and PNN optimization-based bearing fault diagnosis method according to claim 3, wherein the method for extracting the energy entropy of the effective eigenmode component in S4 and forming a eigenvector matrix includes:
based on the screened effective eigenmode components, the total energy of the effective eigenmode components is obtained, the proportion of the energy of each effective eigenmode component to the total energy is calculated, the energy entropy of each effective eigenmode component is obtained, the energy entropy of the effective eigenmode components of vibration signals of different fault states of the bearing is extracted, and a bearing fault feature vector matrix is formed.
5. The CEEMD energy entropy and PNN optimization-based bearing fault diagnosis method according to claim 1, wherein constructing a probabilistic neural network model in S5 comprises:
the probabilistic neural network model comprises an input layer, a mode layer, a summation layer and an output layer;
the input layer receives the bearing vibration signal, and the eigenvectors in the eigenvector matrix correspond to the energy entropy of the effective eigenvector component;
the mode layer calculates and inputs the characteristic vector to be matched with the bearing vibration signal to obtain the quantity of neurons, and the mode layer is connected with the input layer through connection weights;
the number of the neurons in the summation layer is the same as the total number of the mode layers, and the neurons in the summation layer are connected with the corresponding neurons in the mode layers;
the output layer is composed of the same number of competing neurons as the summing layer.
6. The CEEMD energy entropy and PNN optimization based bearing fault diagnosis method of claim 1, wherein constructing a probabilistic neural network fault diagnosis model optimized by a modified sparrow search algorithm comprises: and optimizing the smoothing factor of the probabilistic neural network by adopting an improved sparrow search algorithm, obtaining the optimized smoothing factor, inputting the optimized smoothing factor into the probabilistic neural network model, and constructing a probabilistic neural network fault diagnosis model optimized by the improved sparrow search algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210655047.6A CN115060497B (en) | 2022-06-10 | 2022-06-10 | Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210655047.6A CN115060497B (en) | 2022-06-10 | 2022-06-10 | Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115060497A CN115060497A (en) | 2022-09-16 |
| CN115060497B true CN115060497B (en) | 2023-07-11 |
Family
ID=83200720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210655047.6A Active CN115060497B (en) | 2022-06-10 | 2022-06-10 | Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115060497B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115688017A (en) * | 2022-12-29 | 2023-02-03 | 南昌工程学院 | FRCMDE-based transformer core fault voiceprint diagnosis method and device |
| CN116660639B (en) * | 2023-03-10 | 2024-05-03 | 国网宁夏电力有限公司电力科学研究院 | Transformer fault diagnosis method and system based on CCGOA-SVDD |
| CN116127300B (en) * | 2023-04-01 | 2023-06-27 | 常熟天地煤机装备有限公司 | Coal cutter load identification method and device and computer readable storage medium |
| CN116520990B (en) * | 2023-04-28 | 2023-11-24 | 暨南大学 | Sign language identification method and system based on lightweight neural network and glove |
| CN116626548B (en) * | 2023-05-31 | 2024-02-23 | 国网甘肃省电力公司经济技术研究院 | Transformer winding deformation diagnosis method based on ISSA-PNN |
| CN117150379B (en) * | 2023-08-02 | 2024-08-06 | 中国船舶集团有限公司第七一九研究所 | Printed circuit board fault diagnosis method |
| CN117574057B (en) * | 2023-11-24 | 2024-05-28 | 昆明理工大学 | Intelligent fault diagnosis method for vertical water pump unit |
| CN118277874A (en) * | 2024-06-04 | 2024-07-02 | 沈阳顺义科技股份有限公司 | Diesel engine crankshaft bearing fault diagnosis method based on FEEMD-PE-PNN |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108168890A (en) * | 2018-02-10 | 2018-06-15 | 北京工业大学 | A kind of method of rolling bearing fault diagnosis |
| CN110307981B (en) * | 2019-06-17 | 2021-01-26 | 西安工程大学 | Bearing fault diagnosis method based on PNN-IFA |
| CN110595765A (en) * | 2019-08-26 | 2019-12-20 | 西安理工大学 | Wind turbine generator gearbox fault diagnosis method based on VMD and FA _ PNN |
| CN111796342A (en) * | 2020-06-05 | 2020-10-20 | 山东科技大学 | Tiny fault diagnosis method and device for cam type absolute gravimeter |
| CN112345252A (en) * | 2020-11-22 | 2021-02-09 | 国家电网有限公司 | Rolling bearing fault diagnosis method based on EEMD and improved GSA-SOM neural network |
| CN112733603B (en) * | 2020-12-11 | 2024-06-11 | 江苏大学 | Variable frequency scroll compressor fault diagnosis method based on improved VMD and SVM |
| CN112989734B (en) * | 2021-02-25 | 2022-05-03 | 中国人民解放军海军航空大学 | Equipment analog circuit fault diagnosis method based on probabilistic neural network |
| CN113869145B (en) * | 2021-09-10 | 2024-07-26 | 武汉大学 | Circuit fault diagnosis method and system for lightweight gradient elevator and sparrow search |
| CN113850181A (en) * | 2021-09-23 | 2021-12-28 | 华北电力大学 | Gas turbine inlet guide vane system fault diagnosis method based on feature information fusion |
| CN114166509A (en) * | 2022-02-10 | 2022-03-11 | 中国人民解放军海军工程大学 | Motor bearing fault prediction method |
-
2022
- 2022-06-10 CN CN202210655047.6A patent/CN115060497B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115060497A (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115060497B (en) | Bearing fault diagnosis method based on CEEMD energy entropy and optimized PNN | |
| Long et al. | Evolving deep echo state networks for intelligent fault diagnosis | |
| CN110006649B (en) | Bearing fault diagnosis method based on improved ant lion algorithm and support vector machine | |
| An et al. | Options for Prognostics Methods: A review of data-driven and physics-based prognostics | |
| CN110009030B (en) | Sewage treatment fault diagnosis method based on stacking meta-learning strategy | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN112001270A (en) | Ground radar automatic target classification and identification method based on one-dimensional convolutional neural network | |
| CN111785329A (en) | Single-cell RNA sequencing clustering method based on confrontation automatic encoder | |
| CN114399032B (en) | Method and system for predicting metering error of electric energy meter | |
| CN111915246B (en) | Parallel detection method for grain storage quantity of granary | |
| CN113705396B (en) | Motor fault diagnosis method, system and equipment | |
| CN115841278B (en) | Method, system, equipment and medium for evaluating running error state of electric energy metering device | |
| CN116542382A (en) | Sewage treatment dissolved oxygen concentration prediction method based on mixed optimization algorithm | |
| CN115597869A (en) | Bearing fault diagnosis method based on MTF-SDAE-LightGBM | |
| CN114220164B (en) | Gesture recognition method based on variational modal decomposition and support vector machine | |
| CN116881841A (en) | Hybrid model fault diagnosis method based on F1-score multistage decision analysis | |
| CN114118138A (en) | Bearing composite fault diagnosis method based on multi-label field self-adaptive model | |
| Badriyah et al. | Improving stroke diagnosis accuracy using hyperparameter optimized deep learning | |
| CN114905335A (en) | Cutter wear prediction method combining domain confrontation and convolution neural network | |
| Saufi et al. | Machinery fault diagnosis based on a modified hybrid deep sparse autoencoder using a raw vibration time-series signal | |
| CN112966770B (en) | Fault prediction method and device based on integrated hybrid model and related equipment | |
| Akinwale Adio et al. | Translated Nigeria stock market price using artificial neural network for effective prediction | |
| CN117371321A (en) | Internal plasticity depth echo state network soft measurement modeling method based on Bayesian optimization | |
| CN116186586A (en) | Rolling bearing fault diagnosis method based on improved empirical mode decomposition algorithm and optimized deep confidence network | |
| CN116226612A (en) | Bearing failure mode identification method based on AVMD and SPNN models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220909 Address after: 289 No. 330099 Jiangxi city of Nanchang Province, high tech Development Zone Avenue Applicant after: NANCHANG INSTITUTE OF TECHNOLOGY Address before: 289 No. 330099 Jiangxi city of Nanchang Province, high tech Development Zone Avenue Applicant before: NANCHANG INSTITUTE OF TECHNOLOGY Applicant before: NANJING BOILER & PRESSURE VESSEL SUPERVISION AND INSPECTION INSTITUTE |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |