CN112085062A - Wavelet neural network-based abnormal energy consumption positioning method - Google Patents
Wavelet neural network-based abnormal energy consumption positioning method Download PDFInfo
- Publication number
- CN112085062A CN112085062A CN202010796860.6A CN202010796860A CN112085062A CN 112085062 A CN112085062 A CN 112085062A CN 202010796860 A CN202010796860 A CN 202010796860A CN 112085062 A CN112085062 A CN 112085062A
- Authority
- CN
- China
- Prior art keywords
- neural network
- energy consumption
- wavelet neural
- value
- fitness
- 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
- 238000005265 energy consumption Methods 0.000 title claims abstract description 70
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 33
- 238000003062 neural network model Methods 0.000 claims abstract description 43
- 238000003745 diagnosis Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 16
- 230000002068 genetic effect Effects 0.000 claims abstract description 12
- 238000005457 optimization Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 27
- 238000004364 calculation method Methods 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 13
- 108090000623 proteins and genes Proteins 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 10
- 238000012937 correction Methods 0.000 claims description 6
- 230000005284 excitation Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/021—Gearings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Abstract
The invention provides an abnormal energy consumption positioning method based on a wavelet neural network, which comprises the following steps: s1: acquiring abnormal energy consumption data; s2: performing wavelet packet decomposition and reconstruction on the abnormal energy consumption data, and extracting an energy characteristic value of the abnormal energy consumption data to obtain a training sample; s3: constructing a wavelet neural network model; s4: inputting the training sample into a wavelet neural network model, and performing optimization training on the wavelet neural network model by combining a genetic algorithm to obtain an optimal wavelet neural network model; and performing abnormal energy consumption positioning diagnosis through the optimal wavelet neural network model to obtain a diagnosis result, thereby realizing accurate positioning of machine parts causing abnormal energy consumption. The invention provides an abnormal energy consumption positioning method based on a wavelet neural network, which improves the accuracy and stability of abnormal energy consumption positioning and solves the problem that machine parts with abnormal energy consumption are difficult to accurately position due to the fact that the accuracy of an abnormal energy consumption positioning technology is not high enough at present.
Description
Technical Field
The invention relates to the technical field of equipment detection, in particular to an abnormal energy consumption positioning method based on a wavelet neural network.
Background
The hydraulic machine is a general manufacturing device and is widely applied to various molding processes in the field of mechanical manufacturing. It has the advantages of high precision, high rigidity, large load capacity and the like. However, it also has the disadvantages of high energy consumption and low energy conversion efficiency. The number of chinese metal forming hydraulic machines is about 200 million, and they consume more than 2800 hundred million kWh of electrical energy per year, which corresponds to 3.3 million tons of carbon emissions. The production conditions of hydraulic machines are complex and they run at full load for long periods of time. Therefore, the probability of abnormal energy consumption is very high, when the energy consumption of the machine is abnormal, a large amount of energy loss can be caused, the mechanical energy efficiency is reduced, and even a machine halt and an inestimable safety accident are caused, so that the normal production process of the whole production line is influenced. However, the accuracy of the abnormal energy consumption positioning technology is not high enough at present, so that the machine part causing the abnormal energy consumption is difficult to accurately position.
In the prior art, as a chinese patent disclosed in 2019, 5, 3, a mechanical fault analysis method based on wavelet fuzzy recognition and an image analysis theory, which is disclosed as CN109708877A, adopts wavelet fuzzy recognition to quickly diagnose a rotary mechanical fault, and utilizes the image analysis theory to judge a specific position of the mechanical fault, but does not combine with a neural network for judgment, and the detection accuracy is not high enough.
Disclosure of Invention
The invention provides an abnormal energy consumption positioning method based on a wavelet neural network, aiming at overcoming the technical defect that machine parts with abnormal energy consumption are difficult to accurately position due to the fact that the accuracy of the conventional abnormal energy consumption positioning technology is not high enough.
In order to solve the technical problems, the technical scheme of the invention is as follows:
an abnormal energy consumption positioning method based on a wavelet neural network comprises the following steps:
s1: acquiring abnormal energy consumption data;
s2: performing wavelet packet decomposition and reconstruction on the abnormal energy consumption data, and extracting an energy characteristic value of the abnormal energy consumption data to obtain a training sample;
s3: constructing a wavelet neural network model;
s4: inputting the training sample into a wavelet neural network model, and performing optimization training on the wavelet neural network model by combining a genetic algorithm to obtain an optimal wavelet neural network model; and performing abnormal energy consumption positioning diagnosis through the optimal wavelet neural network model to obtain a diagnosis result, thereby realizing accurate positioning of machine parts causing abnormal energy consumption.
Preferably, step S2 specifically includes the following steps:
s2.1: carrying out wavelet packet multi-scale decomposition processing on the abnormal energy consumption data to obtain decomposed data;
s2.2: reconstructing the decomposed data by adopting a wavelet packet coefficient reconstruction algorithm to obtain reconstructed data;
s2.3: calculating the energy value of the reconstruction data so as to extract and obtain an energy characteristic value;
s2.4: performing dimension reduction processing on the energy characteristic value to obtain a dimension reduced energy characteristic value;
s2.5: and carrying out normalization processing on the dimension-reduced energy characteristic value to obtain a training sample.
Preferably, in step S2.2, the data length of the reconstructed data is consistent with the data length before the abnormal energy consumption data is decomposed.
Preferably, in step S2.4, a principal component analysis dimension reduction technique is used to perform dimension reduction processing on the energy characteristic value; the method specifically comprises the following steps:
assume that the initial vector X of energy eigenvalues contains n samples and m features, i.e.
X＝(x1,x2,...,xn)T；
The covariance matrix of the initial vector X is:
based on the K-L transform, the cumulative contribution rate c (q) of the principal component is:
using the eigenvector matrix U ═ U (U)1,U2,...,Um)TMapping the initial vector X to a new feature subspace to obtain a feature vector matrix Y, wherein the mapping formula is as follows:
Y＝UTX；
wherein the covariance matrix S is an n × n dimensional matrix and its eigenvalues satisfy λ1>λ2…>λm；x1,x2,...,xnAre all samples in an initial vector X, XiIs the ith component of the initial vector X; lambda [ alpha ]iIs the ith eigenvalue of the matrix S, q is the principal component of the matrix S, and S is the contribution ratio standard value.
Preferably, in step S2.5, the normalization formula is:
wherein y is a normalized interval, ymax、yminRespectively the maximum and minimum of the normalized interval, x is the data to be normalized in the eigenvector matrix Y, xmax、xminRespectively the maximum value and the minimum value in each row of data in the eigenvector matrix Y.
Preferably, in step S3, the number m of input layer nodes of the wavelet neural network model is determined according to the energy characteristic value, the number n of output layer nodes is determined according to the output result, and the number h of hidden layer nodes is determined according to the following formula:
h<m-1
h＝log2m
wherein alpha is a fitness coefficient.
Preferably, step S4 specifically includes the following steps:
step S4.1: inputting the training sample into a wavelet neural network model;
step S4.2: randomly generating a weight value and a threshold value of the wavelet neural network model, and setting a learning rate eta, an excitation function, an expected output value and an error preset value; wherein, the weight comprises a connection weight omega between each input layer and each hidden layerijConnecting each hidden layer with each output layer to obtain a weight value omegajkThe threshold value comprises a threshold value a of each hidden layerjThreshold b for each output layerk；
Step S4.3: optimizing weight and threshold by adopting a genetic algorithm;
step S4.4: calculating a network diagnosis error according to the weight and the threshold;
step S4.5: updating the weight and the threshold according to the network diagnosis error;
step S4.6: comparing the network diagnosis error with an error preset value; if the network diagnosis error is larger than the error preset value, returning to execute the step S4.4; and if the network diagnosis error is less than or equal to the error preset value, obtaining the optimal wavelet neural network model.
Preferably, in step S4.3, optimizing the weight and the threshold value by using a genetic algorithm specifically includes the following steps:
s4.3.1: population initialization: real number coding is carried out on individuals in the population, and a fitness standard value is set; wherein the individual consists of omegaij、ωjk、ajAnd bkComposition is carried out;
s4.3.2: obtaining the output value of the wavelet neural network according to the individual, and calculating the fitness F, wherein the calculation formula is as follows:
wherein, yiIs the expected output value, o, of the ith node of the wavelet neural networkiIs a small waveThe output value of the ith node of the network, alpha, is a fitness coefficient, and the smaller the fitness F, the better;
s4.3.3: selecting operation: selecting individuals with good fitness from the population by adopting a roulette method to form a new population A, wherein the selection probability p of each individual iiIs composed of
Wherein, FiFitness of individual i, FjThe fitness of the individual j is shown, and N is the number of individuals in the population;
s4.3.4: and (3) cross operation: crossing every two individuals in the new population by adopting a real number crossing method to obtain new individuals, so as to update the population and obtain a new population B;
s4.3.5: mutation operation: setting variation probability, randomly selecting an individual from the new population B to perform variation according to the variation probability to obtain a more excellent individual, wherein the calculation formula is as follows:
wherein, a'ijIs the j gene of the superior i individual, aijIs the jth gene of the ith individual, amaxIs gene aijUpper bound of aminIs gene aijLower boundary of r2Is a random number, f (g) is the probability of variation, r is the interval [0,1]Random number in G is the current iteration number, GmaxIs the maximum evolution algebra;
s4.3.6: obtaining a new output value of the wavelet neural network according to the more excellent individuals, and calculating to obtain new fitness;
s4.3.7: comparing the new fitness with a fitness standard value;
if the new fitness is greater than the fitness standard value, returning to step S4.3.3:
and if the new fitness is less than or equal to the fitness standard value, optimizing the weight and the threshold to obtain the optimal weight and threshold, and executing the step S4.4.
Preferably, in step S4.4, calculating the network diagnosis error according to the weight and the threshold specifically includes the following steps:
s4.4.1: calculating hidden layer output, wherein the calculation formula is as follows:
where φ (x) is the excitation function, x is the input vector, xiIs the input vector of the ith node, m is the number of nodes in the input layer, HjThe hidden layer output of the jth node;
s4.4.2: calculating output of an output layer, wherein the calculation formula is as follows:
wherein, OkIs the output layer output of the kth node;
s4.4.3: and calculating the network diagnosis error by the following calculation formula:
ek＝Yk-Ok；
wherein e iskDiagnosing the error for the network of the kth node, YkIs the desired output of the kth node.
Preferably, in step S4.5, optimizing the weight and the threshold according to the network diagnosis error specifically includes the following steps:
s4.5.1: and (3) adopting a gradient correction method as a learning algorithm of the weight, and optimizing the weight from the negative gradient direction of the network prediction error:
ωjk＝ωjk+ηHjek；
s4.5.2: and (3) optimizing the threshold from the negative gradient direction of the network prediction error by adopting a gradient correction method as a learning algorithm of the threshold:
b′k＝bk+ek。
compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides an abnormal energy consumption positioning method based on a wavelet neural network, which is characterized in that abnormal energy consumption data subjected to wavelet packet decomposition and reconstruction are input into the wavelet neural network optimized and improved by combining a genetic algorithm to obtain a diagnosis result, so that the accurate positioning of machine parts with abnormal energy consumption is realized, and the accuracy and the stability of the abnormal energy consumption positioning are improved.
Drawings
FIG. 1 is a flow chart of the steps for implementing the technical solution of the present invention;
fig. 2 is a flowchart illustrating an implementation step of step S4 according to the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, an abnormal energy consumption positioning method based on a wavelet neural network includes the following steps:
s1: acquiring abnormal energy consumption data;
s2: performing wavelet packet decomposition and reconstruction on the abnormal energy consumption data, and extracting an energy characteristic value of the abnormal energy consumption data to obtain a training sample;
s3: constructing a wavelet neural network model;
s4: inputting the training sample into a wavelet neural network model, and performing optimization training on the wavelet neural network model by combining a genetic algorithm to obtain an optimal wavelet neural network model; and performing abnormal energy consumption positioning diagnosis through the optimal wavelet neural network model to obtain a diagnosis result, thereby realizing accurate positioning of machine parts causing abnormal energy consumption.
In the specific implementation process, 2000 groups of energy consumption data under different conditions are collected, wherein the energy consumption data under each condition comprises four conditions of a normal state, an abnormal state of an electrical system, an abnormal state of a hydraulic system and an abnormal state of a mechanical system, and the energy consumption data under each condition comprises 500 groups, wherein 400 groups under each condition are used as training samples and input into a wavelet neural network model and are optimized by combining a genetic algorithm to obtain an optimal wavelet neural network model, then the remaining 100 groups under each condition are input into the optimal wavelet neural network model for testing, and the test result shows that the actual output of the optimal wavelet neural network model is basically consistent with the expected output, so that the effectiveness of the optimal wavelet neural network model is verified, and the accurate diagnosis and positioning of abnormal energy consumption can be realized.
More specifically, step S2 specifically includes the following steps:
s2.1: carrying out wavelet packet multi-scale decomposition processing on the abnormal energy consumption data to obtain decomposed data;
s2.2: reconstructing the decomposed data by adopting a wavelet packet coefficient reconstruction algorithm to obtain reconstructed data;
s2.3: calculating the energy value of the reconstruction data so as to extract and obtain an energy characteristic value;
s2.4: performing dimension reduction processing on the energy characteristic value to obtain a dimension reduced energy characteristic value;
s2.5: and carrying out normalization processing on the dimension-reduced energy characteristic value to obtain a training sample.
More specifically, in step S2.2, the data length of the reconstructed data is consistent with the data length before the abnormal energy consumption data is decomposed.
In the specific implementation process, the standard orthogonalized scale function
The set of functions w is generated by a two-scale difference equation as shown belown,j,k(t):＝2-j/2wn(2-jt-k)},n∈Z/Z-J is equal to Z, k is equal to Z and is called
The orthogonal wavelet packet of (a) is,
wherein the content of the first and second substances,
{hk}k∈Zand { gk}k∈ZIs formed by
A pair of derived conjugate quadrature filter coefficients. Wavelet packet multiscale analysis decomposes the signal according to frequency bands and divides each layer into 2j (j ═ 1,2, 3..) frequency bands. The method has good resolution in both high frequency band and low frequency band, and can provide more fine analysis for energy consumption data. The square integrable function L in Hilbert space is calculated according to different scale factors j (1,2, 3.. times.)2(R) decomposition into all wavelet subspaces Wj(j ∈ Z) (i.e., the closure of the wavelet function) is calculated.
Dimension subspace VjSum wavelet subspace WjUnified orthogonal decomposition into Vj+1
Spacing orthogonal wavelet packets
Is defined as a function w2nA subspace of (t). To convert W according to binary formjIs further subdivided, as used herein
Unified characterization VjAnd Wj。
Extend it to n ∈ Z+We can get
Wherein the content of the first and second substances,
and
is that
A subspace, we adopt
To represent
Then obtaining the decomposition formula of the subspace of the wavelet packet
That is to say
Energy consumption data obtained by multi-scale decomposition of wavelet packets is c0＝cJ+d1+d2+…+dJ(ii) a The wavelet packet coefficient reconstruction algorithm is as follows, and the length of the reconstructed energy consumption data is consistent with that of the original energy consumption data.
C′j＝H*C′j+1+G*D′j+1,j＝J-1,J-2...,0
Wherein H*And G*Are the dual operators of H and G, C'JAnd D'JLow frequency band and high frequency band, respectively, of the reconstructed time series of energy consumptions, pair cJAnd d1,d2,d3,…,dJRespectively reconstructing to obtain
X＝CJ+D1+D2+D3+...+DJ
Wherein, CJ＝{cJ,1,cJ,2,. is the approximate component after reconstruction of the J-th layer, D1＝{d1,1,d1,2,...},…,DJ＝{dJ,1,dJ,2,. is the reconstructed detail component from layer 1 to layer J.
Time series of energy consumptions s (t) epsilon L2(R) in the orthogonal wavelet packet space
Wavelet packet transformation coefficient p ofsThe formula for calculating (n, j, k) is as follows
Wherein, { h }k}k∈ZAnd { gk}k∈ZThe coefficients of the low-pass conjugate quadrature filter and the high-pass conjugate quadrature filter, respectively. For a given
The formula for calculating the energy value of s (t) in the time-frequency localization space is as follows:
at decomposition level j, E (j, n) represents the energy value of the nth wavelet node. The wavelet packet energy feature vector constructed by E (j, n) is
E(J,s)＝[E(J,0),E(J,1),...,E(J,2J-1)]
The calculation formula of the energy ratio of each node is:
where E (j, total) is the total energy at decomposition level j, and P (j, i) is the energy ratio of inode at decomposition level j.
More specifically, in step S2.4, a principal component analysis dimension reduction technique is used to perform dimension reduction processing on the energy characteristic value; the method specifically comprises the following steps:
assume that the initial vector X of energy eigenvalues contains n samples and m features, i.e.
X＝(x1,x2,...,xn)T；
The covariance matrix of the initial vector X is:
based on the K-L transform, the cumulative contribution rate c (q) of the principal component is:
and mapping the initial vector X to a new feature subspace by adopting a feature vector matrix to obtain a feature vector matrix Y, wherein the mapping formula is as follows:
Y＝UTX；
wherein the covariance matrix S is an n × n dimensional matrix and its eigenvalues satisfy λ1>λ2…>λm；x1,x2,...,xnAre all samples in an initial vector X, XiIs the ith component of the initial vector X; lambda [ alpha ]iIs the ith eigenvalue of the matrix S, q is the principal component of the matrix S, and S is the contribution ratio standard value.
In the specific implementation process, the dimension reduction processing is carried out on the energy characteristic value, so that the training time of the wavelet neural network model is greatly shortened, and the performance of the wavelet neural network model is optimized.
More specifically, in step S2.5, the normalization formula is:
wherein y is a normalized interval, ymax、yminRespectively the maximum and minimum of the normalized interval, x is the data to be normalized in the eigenvector matrix Y, xmax、xminRespectively the maximum value and the minimum value in each row of data in the eigenvector matrix Y.
More specifically, in step S3, the number m of input layer nodes of the wavelet neural network model is determined according to the energy characteristic value, the number n of output layer nodes is determined according to the output result, and the number h of hidden layer nodes is determined according to the following formula:
h<m-1
h＝log2m
wherein alpha is a fitness coefficient.
More specifically, as shown in fig. 2, step S4 specifically includes the following steps:
step S4.1: inputting the training sample into a wavelet neural network model;
step S4.2: randomly generating a weight value and a threshold value of the wavelet neural network model, and setting a learning rate eta, an excitation function, an expected output value and an error preset value; wherein, the weight comprises a connection weight omega between each input layer and each hidden layerijConnecting each hidden layer with each output layer to obtain a weight value omegajkThe threshold value comprises a threshold value a of each hidden layerjThreshold b for each output layerk；
Step S4.3: optimizing weight and threshold by adopting a genetic algorithm; the method specifically comprises the following steps:
s4.3.1: population initialization: real number coding is carried out on individuals in the population, and a fitness standard value is set; wherein the individual consists of omegaij、ωjk、ajAnd bkComposition is carried out;
s4.3.2: obtaining the output value of the wavelet neural network according to the individual, and calculating the fitness F, wherein the calculation formula is as follows:
wherein, yiIs the expected output value, o, of the ith node of the wavelet neural networkiThe output value of the ith node of the wavelet neural network is shown, alpha is a fitness coefficient, and the smaller the fitness F is, the better the fitness F is;
s4.3.3: selecting operation: selecting individuals with good fitness from the population by adopting a roulette method to form a new population A, wherein the selection probability p of each individual iiIs composed of
Wherein, FiFitness of individual i, FjThe fitness of the individual j is shown, and N is the number of individuals in the population;
s4.3.4: and (3) cross operation: crossing every two individuals in the new population by adopting a real number crossing method to obtain new individuals, so as to update the population and obtain a new population B;
s4.3.5: mutation operation: setting variation probability, randomly selecting an individual from the new population B to perform variation according to the variation probability to obtain a more excellent individual, wherein the calculation formula is as follows:
wherein, a'ijIs the j gene of the superior i individual, aijIs the jth gene of the ith individual, amaxIs gene aijUpper bound of aminIs gene aijLower boundary of r2Is a random number, f (g) is the probability of variation, r is the interval [0,1]Random number in G is the current iteration number, GmaxIs the maximum evolution algebra;
s4.3.6: obtaining a new output value of the wavelet neural network according to the more excellent individuals, and calculating to obtain new fitness;
s4.3.7: comparing the new fitness with a fitness standard value;
if the new fitness is greater than the fitness standard value, returning to step S4.3.3:
if the new fitness is less than or equal to the fitness standard value, optimizing the weight and the threshold to obtain the optimal weight and threshold, and executing the step S4.4;
step S4.4: calculating a network diagnosis error according to the weight and the threshold; the method specifically comprises the following steps:
s4.4.1: calculating hidden layer output, wherein the calculation formula is as follows:
where φ (x) is the excitation function, x is the input vector, xiIs the input vector of the ith node, m is the number of nodes in the input layer, HjThe hidden layer output of the jth node;
s4.4.2: calculating output of an output layer, wherein the calculation formula is as follows:
wherein, OkIs the output layer output of the kth node;
s4.4.3: and calculating the network diagnosis error by the following calculation formula:
ek＝Yk-Ok；
wherein e iskDiagnosing the error for the network of the kth node, YkIs the expected output of the kth node;
step S4.5: updating the weight and the threshold according to the network diagnosis error; the method specifically comprises the following steps:
s4.5.1: and (3) adopting a gradient correction method as a learning algorithm of the weight, and optimizing the weight from the negative gradient direction of the network prediction error:
ωjk＝ωjk+ηHjek；
s4.5.2: and (3) optimizing the threshold from the negative gradient direction of the network prediction error by adopting a gradient correction method as a learning algorithm of the threshold:
b′k＝bk+ek；
step S4.6: comparing the network diagnosis error with an error preset value; if the network diagnosis error is larger than the error preset value, returning to execute the step S4.4; and if the network diagnosis error is less than or equal to the error preset value, obtaining the optimal wavelet neural network model.
In the specific implementation process, the weight and the threshold of the wavelet neural network model are optimized by adopting a genetic algorithm, and the specific steps are as follows: and calculating the fitness of the individuals containing the weight and the threshold, and finding the individuals corresponding to the optimal fitness through selection, intersection and variation operations, thereby obtaining the optimal initial weight and the threshold of the wavelet neural network. And then training the wavelet neural network model to finally obtain an optimal wavelet neural network model, and performing abnormal energy consumption positioning diagnosis through the optimal wavelet neural network model to obtain a diagnosis result, thereby realizing accurate positioning of machine parts causing abnormal energy consumption.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. An abnormal energy consumption positioning method based on a wavelet neural network is characterized by comprising the following steps:
s1: acquiring abnormal energy consumption data;
s2: performing wavelet packet decomposition and reconstruction on the abnormal energy consumption data, and extracting an energy characteristic value of the abnormal energy consumption data to obtain a training sample;
s3: constructing a wavelet neural network model;
s4: inputting the training sample into a wavelet neural network model, and performing optimization training on the wavelet neural network model by combining a genetic algorithm to obtain an optimal wavelet neural network model; and performing abnormal energy consumption positioning diagnosis through the optimal wavelet neural network model to obtain a diagnosis result, thereby realizing accurate positioning of machine parts causing abnormal energy consumption.
2. The wavelet neural network-based abnormal energy consumption positioning method according to claim 1, wherein the step S2 specifically comprises the following steps:
s2.1: carrying out wavelet packet multi-scale decomposition processing on the abnormal energy consumption data to obtain decomposed data;
s2.2: reconstructing the decomposed data by adopting a wavelet packet coefficient reconstruction algorithm to obtain reconstructed data;
s2.3: calculating the energy value of the reconstruction data so as to extract and obtain an energy characteristic value;
s2.4: performing dimension reduction processing on the energy characteristic value to obtain a dimension reduced energy characteristic value;
s2.5: and carrying out normalization processing on the dimension-reduced energy characteristic value to obtain a training sample.
3. The wavelet neural network-based abnormal energy consumption positioning method according to claim 2, wherein in step S2.2, the data length of the reconstructed data is consistent with the data length before the decomposition of the abnormal energy consumption data.
4. The wavelet neural network-based abnormal energy consumption positioning method according to claim 2, characterized in that in step S2.4, a principal component analysis dimension reduction technique is adopted to perform dimension reduction processing on the energy characteristic value; the method specifically comprises the following steps:
assume that the initial vector X of energy eigenvalues contains n samples and m features, i.e.
X＝(x1,x2,...,xn)T；
The covariance matrix of the initial vector X is:
based on the K-L transform, the cumulative contribution rate c (q) of the principal component is:
using the eigenvector matrix U ═ U (U)1,U2,...,Um)TMapping the initial vector X to a new feature subspace to obtain a feature vector matrix Y, wherein the mapping formula is as follows:
Y＝UTX；
wherein the covariance matrix S is an n × n dimensional matrix and its eigenvalues satisfy λ1>λ2…>λm；x1,x2,...,xnAre all samples in an initial vector X, XiIs the ith component of the initial vector X; lambda [ alpha ]iIs the ith eigenvalue of the matrix S, q is the principal component of the matrix S, and S is the contribution ratio standard value.
5. The wavelet neural network-based abnormal energy consumption positioning method according to claim 4, wherein in step S2.5, the normalization formula is:
wherein y is a normalized interval, ymax、yminRespectively the maximum and minimum of the normalized interval, x is the data to be normalized in the eigenvector matrix Y, xmax、xminRespectively the maximum value and the minimum value in each row of data in the eigenvector matrix Y.
6. The wavelet neural network-based abnormal energy consumption location method according to claim 1, wherein in step S3, the number m of input layer nodes of the wavelet neural network model is determined according to the energy characteristic value, the number n of output layer nodes is determined according to the output result, and the number h of hidden layer nodes is determined according to the following formula:
wherein alpha is a fitness coefficient.
7. The wavelet neural network-based abnormal energy consumption positioning method according to claim 6, wherein the step S4 specifically comprises the following steps:
step S4.1: inputting the training sample into a wavelet neural network model;
step S4.2: randomly generating a weight value and a threshold value of the wavelet neural network model, and setting a learning rate eta, an excitation function, an expected output value and an error preset value; wherein, the weight comprises a connection weight omega between each input layer and each hidden layerijConnecting each hidden layer with each output layer to obtain a weight value omegajkThe threshold value comprises a threshold value a of each hidden layerjThreshold b for each output layerk；
Step S4.3: optimizing weight and threshold by adopting a genetic algorithm;
step S4.4: calculating a network diagnosis error according to the weight and the threshold;
step S4.5: updating the weight and the threshold according to the network diagnosis error;
step S4.6: comparing the network diagnosis error with an error preset value; if the network diagnosis error is larger than the error preset value, returning to execute the step S4.4; and if the network diagnosis error is less than or equal to the error preset value, obtaining the optimal wavelet neural network model.
8. The wavelet neural network-based abnormal energy consumption positioning method according to claim 7, wherein in step S4.3, optimizing the weight and the threshold value by using a genetic algorithm specifically comprises the following steps:
s4.3.1: population initialization: real number coding is carried out on individuals in the population, and a fitness standard value is set; wherein the individual consists of omegaij、ωjk、ajAnd bkComposition is carried out;
s4.3.2: obtaining the output value of the wavelet neural network according to the individual, and calculating the fitness F, wherein the calculation formula is as follows:
wherein, yiIs the expected output value, o, of the ith node of the wavelet neural networkiThe output value of the ith node of the wavelet neural network is shown, alpha is a fitness coefficient, and the smaller the fitness F is, the better the fitness F is;
s4.3.3: selecting operation: selecting individuals with good fitness from the population by adopting a roulette method to form a new population A, wherein the selection probability p of each individual iiIs composed of
Wherein, FiFitness of individual i, FjThe fitness of the individual j is shown, and N is the number of individuals in the population;
s4.3.4: and (3) cross operation: crossing every two individuals in the new population by adopting a real number crossing method to obtain new individuals, so as to update the population and obtain a new population B;
s4.3.5: mutation operation: setting variation probability, randomly selecting an individual from the new population B to perform variation according to the variation probability to obtain a more excellent individual, wherein the calculation formula is as follows:
wherein, a'ijIs the j gene of the superior i individual, aijIs the jth gene of the ith individual, amaxIs gene aijUpper bound of aminIs gene aijLower boundary of r2Is a random number, f (g) is the probability of variation, r is the interval [0,1]Random number in G is the current iteration number, GmaxIs the maximum evolution algebra;
s4.3.6: obtaining a new output value of the wavelet neural network according to the more excellent individuals, and calculating to obtain new fitness;
s4.3.7: comparing the new fitness with a fitness standard value;
if the new fitness is greater than the fitness standard value, returning to step S4.3.3:
and if the new fitness is less than or equal to the fitness standard value, optimizing the weight and the threshold to obtain the optimal weight and threshold, and executing the step S4.4.
9. The wavelet neural network-based abnormal energy consumption positioning method according to claim 7, wherein in step S4.4, calculating a network diagnosis error according to the weight and the threshold specifically comprises the following steps:
s4.4.1: calculating hidden layer output, wherein the calculation formula is as follows:
where φ (x) is the excitation function, x is the input vector, xiIs the input vector of the ith node, m is the number of nodes in the input layer, HjThe hidden layer output of the jth node;
s4.4.2: calculating output of an output layer, wherein the calculation formula is as follows:
wherein, OkIs the output layer output of the kth node;
s4.4.3: and calculating the network diagnosis error by the following calculation formula:
ek＝Yk-Ok；
wherein e iskDiagnosing the error for the network of the kth node, YkIs the desired output of the kth node.
10. The wavelet neural network-based abnormal energy consumption positioning method according to claim 7, wherein in step S4.5, optimizing the weight and the threshold according to the network diagnosis error specifically comprises the following steps:
s4.5.1: and (3) adopting a gradient correction method as a learning algorithm of the weight, and optimizing the weight from the negative gradient direction of the network prediction error:
ωjk＝ωjk+ηHjek；
s4.5.2: and (3) optimizing the threshold from the negative gradient direction of the network prediction error by adopting a gradient correction method as a learning algorithm of the threshold:
b'k＝bk+ek。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010796860.6A CN112085062A (en) | 2020-08-10 | 2020-08-10 | Wavelet neural network-based abnormal energy consumption positioning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010796860.6A CN112085062A (en) | 2020-08-10 | 2020-08-10 | Wavelet neural network-based abnormal energy consumption positioning method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112085062A true CN112085062A (en) | 2020-12-15 |
Family
ID=73735983
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010796860.6A Pending CN112085062A (en) | 2020-08-10 | 2020-08-10 | Wavelet neural network-based abnormal energy consumption positioning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112085062A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648343A (en) * | 2019-09-05 | 2020-01-03 | 电子科技大学 | Image edge detection method based on six-order spline scale function |
| CN114765574A (en) * | 2020-12-30 | 2022-07-19 | 中盈优创资讯科技有限公司 | Network anomaly delimitation positioning method and device |
| CN115375502A (en) * | 2022-08-16 | 2022-11-22 | 中国人民解放军海军指挥学院 | Intelligent overlapped community mining method and system based on dual-scale graph wavelet neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160299938A1 (en) * | 2015-04-10 | 2016-10-13 | Tata Consultancy Services Limited | Anomaly detection system and method |
| CN108052975A (en) * | 2017-12-12 | 2018-05-18 | 浙江大学宁波理工学院 | It is a kind of that real-time working condition Forecasting Methodology is run based on the vehicle of core pivot and neutral net |
| CN108435819A (en) * | 2018-05-29 | 2018-08-24 | 广东工业大学 | A kind of aluminum section extruder energy consumption method for detecting abnormality |
| CN111047732A (en) * | 2019-12-16 | 2020-04-21 | 青岛海信网络科技股份有限公司 | Equipment abnormity diagnosis method and device based on energy consumption model and data interaction |
-
2020
- 2020-08-10 CN CN202010796860.6A patent/CN112085062A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160299938A1 (en) * | 2015-04-10 | 2016-10-13 | Tata Consultancy Services Limited | Anomaly detection system and method |
| CN108052975A (en) * | 2017-12-12 | 2018-05-18 | 浙江大学宁波理工学院 | It is a kind of that real-time working condition Forecasting Methodology is run based on the vehicle of core pivot and neutral net |
| CN108435819A (en) * | 2018-05-29 | 2018-08-24 | 广东工业大学 | A kind of aluminum section extruder energy consumption method for detecting abnormality |
| CN111047732A (en) * | 2019-12-16 | 2020-04-21 | 青岛海信网络科技股份有限公司 | Equipment abnormity diagnosis method and device based on energy consumption model and data interaction |
Non-Patent Citations (2)
| Title |
|---|
| 代德宇: "基于GA-BP网络的铜管生产过程能耗预测模型研究及应用", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
| 吴正苗: "基于信息融合和极限学习机的模拟电路故障诊断", 《中国优秀硕士学位论文全文数据库 信息科技》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648343A (en) * | 2019-09-05 | 2020-01-03 | 电子科技大学 | Image edge detection method based on six-order spline scale function |
| CN110648343B (en) * | 2019-09-05 | 2022-09-23 | 电子科技大学 | Image edge detection method based on six-order spline scale function |
| CN114765574A (en) * | 2020-12-30 | 2022-07-19 | 中盈优创资讯科技有限公司 | Network anomaly delimitation positioning method and device |
| CN114765574B (en) * | 2020-12-30 | 2023-12-05 | 中盈优创资讯科技有限公司 | Network anomaly delimitation positioning method and device |
| CN115375502A (en) * | 2022-08-16 | 2022-11-22 | 中国人民解放军海军指挥学院 | Intelligent overlapped community mining method and system based on dual-scale graph wavelet neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110849627B (en) | Width migration learning network and rolling bearing fault diagnosis method based on same | |
| CN107066759B (en) | Steam turbine rotor vibration fault diagnosis method and device | |
| CN112085062A (en) | Wavelet neural network-based abnormal energy consumption positioning method | |
| CN110543860B (en) | Mechanical fault diagnosis method and system based on TJM (machine learning model) transfer learning | |
| CN110361778B (en) | Seismic data reconstruction method based on generation countermeasure network | |
| CN110334580A (en) | The equipment fault classification method of changeable weight combination based on integrated increment | |
| CN105678343A (en) | Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set | |
| CN110040594B (en) | Convolutional neural network-based elevator operation detection system and method | |
| CN111914705A (en) | Signal generation method and device for improving health state evaluation accuracy of reactor | |
| CN108491925A (en) | The extensive method of deep learning feature based on latent variable model | |
| CN112289391B (en) | Anode aluminum foil performance prediction system based on machine learning | |
| CN110363230A (en) | Stacking integrated sewage handling failure diagnostic method based on weighting base classifier | |
| CN110458189A (en) | Compressed sensing and depth convolutional neural networks Power Quality Disturbance Classification Method | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| CN112596016A (en) | Transformer fault diagnosis method based on integration of multiple one-dimensional convolutional neural networks | |
| CN115600088A (en) | Distribution transformer fault diagnosis method based on vibration signals | |
| CN114386452B (en) | Nuclear power circulating water pump sun gear fault detection method | |
| CN108805206A (en) | A kind of modified LSSVM method for building up for analog circuit fault classification | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| CN113807299B (en) | Sleep stage staging method and system based on parallel frequency domain electroencephalogram signals | |
| CN112380932B (en) | Vibration signal characteristic value selection method and elevator health state evaluation or fault diagnosis method | |
| CN114037014A (en) | Reference network clustering method based on graph self-encoder | |
| CN113869451A (en) | Rolling bearing fault diagnosis method under variable working conditions based on improved JGSA algorithm | |
| CN113780160A (en) | Electric energy quality disturbance signal classification method and system | |
| CN111275109A (en) | Power equipment state data characteristic optimization method and system based on self-encoder |
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: 20201215 |
|
| RJ01 | Rejection of invention patent application after publication |