CN115438301A - Equipment operation trend prediction method based on ICEEMDAN quadratic decomposition coupling informar model - Google Patents

Equipment operation trend prediction method based on ICEEMDAN quadratic decomposition coupling informar model Download PDF

Info

Publication number
CN115438301A
CN115438301A CN202211315575.3A CN202211315575A CN115438301A CN 115438301 A CN115438301 A CN 115438301A CN 202211315575 A CN202211315575 A CN 202211315575A CN 115438301 A CN115438301 A CN 115438301A
Authority
CN
China
Prior art keywords
decomposition
component
wavelet
model
signal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202211315575.3A
Other languages
Chinese (zh)
Other versions
CN115438301B (en
Inventor
田乐
王荟芸
常明煜
郭茂祖
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Civil Engineering and Architecture
Original Assignee
Beijing University of Civil Engineering and Architecture
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing University of Civil Engineering and Architecture filed Critical Beijing University of Civil Engineering and Architecture
Priority to CN202211315575.3A priority Critical patent/CN115438301B/en
Publication of CN115438301A publication Critical patent/CN115438301A/en
Application granted granted Critical
Publication of CN115438301B publication Critical patent/CN115438301B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G06F17/148—Wavelet transforms
    • 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)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Algebra (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Complex Calculations (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides an equipment operation trend prediction method based on ICEEMDAN secondary decomposition coupling operator model, firstly VMD decomposition is carried out on an original signal; calculating fuzzy entropy of each component; judging the value of each component fuzzy entropy, if the value is larger than a set threshold, carrying out ICEEMDAN secondary decomposition, and if the value is smaller than the set threshold, carrying out wavelet soft threshold denoising; performing cross-correlation function analysis on the component subjected to ICEEMDAN secondary decomposition, eliminating the component with the correlation coefficient smaller than a specified threshold, calculating an autocorrelation function on the component meeting the condition, determining a high-frequency component, and performing wavelet soft-hard threshold combined denoising on the high-frequency component; reconstructing the effective signal; and inputting the reconstructed signal into an inner model to obtain a prediction result. The invention solves the problem that the prediction result has large error due to the fact that the equipment operation process signal recorded by the sensor contains a large amount of noise.

Description

Equipment operation trend prediction method based on ICEEMDAN quadratic decomposition coupling informar model
Technical Field
The invention relates to a device operation trend prediction problem, in particular to a device operation trend prediction method based on an ICEEMDAN secondary decomposition coupling operator model.
Background
The equipment operation trend prediction is based on the operation state prediction equipment operation condition trend such as equipment vibration, so that the equipment maintenance planning or fault risk early warning and the like are facilitated, and the equipment operation trend prediction method is widely used in the field temporary establishment of facility reliability prediction and equipment reliability and safety prediction. The existing device operation trend prediction technology generally adopts an Empirical Mode Decomposition (EMD) method, but has the problems of end point effect and modal component aliasing. In addition, a Transformer model is often adopted in the prior art, and although the Transformer model has obviously superior performance in capturing long-term dependence compared with RNN, the secondary calculation complexity of a self-attention mechanism is high; the problem of memory bottleneck of the lower stack layer of long sequence input and slow reasoning speed when predicting long output. Therefore, the problems of overlarge signal-to-noise ratio, poor denoising effect and overlarge prediction error of the traditional prediction model of the denoising method in the prior art exist.
Disclosure of Invention
In order to solve the problems of overlarge signal-to-noise ratio, poor denoising effect and overlarge prediction error of a traditional prediction model in the denoising method in the prior art, the invention provides an equipment operation trend prediction method based on an ICEEMDAN quadratic decomposition coupling informar model, which comprises the following steps:
s10: performing VMD decomposition on the original signal;
s20: calculating fuzzy entropy of the components of the primary decomposition;
s30: setting a threshold value of the fuzzy entropy;
s40: judging the fuzzy entropy value of each component obtained by the primary decomposition, wherein the component A is as follows: the fuzzy entropy value is less than a specified threshold, and the component B is: the fuzzy entropy value is larger than a specified threshold value;
s50: if the component of the primary decomposition meets the component A, carrying out wavelet soft threshold denoising on the component to obtain a denoised signal of the primary decomposition;
s60: if the component of the first decomposition meets the component B, performing ICEEMDAN secondary decomposition on the component to obtain a secondary decomposed component;
s70: calculating the correlation coefficient of the component of ICEEMEDAN quadratic decomposition, wherein the X component is as follows: the component with the correlation coefficient less than 0.1, and the Y component is the component with the correlation coefficient more than 0.1;
s80: if the component of the quadratic decomposition meets the X component, directly rejecting the component;
s90: if the component of the secondary decomposition meets the Y component, calculating an autocorrelation function, determining a high-frequency component, and carrying out wavelet soft-hard threshold combined denoising on the high-frequency component to obtain a denoised signal of the secondary decomposition;
s100: performing waveform reconstruction on the denoising signal subjected to the primary decomposition and the denoising signal subjected to the secondary decomposition to obtain a final denoising signal;
s110: and inputting the obtained denoising signal into an informar model to obtain prediction data.
Preferably, the VMD decomposition algorithm in S10 is:
wherein, S11: initialization
Figure 754393DEST_PATH_IMAGE001
，
Figure 916384DEST_PATH_IMAGE002
，
Figure 977881DEST_PATH_IMAGE003
And
Figure 262231DEST_PATH_IMAGE004
；
wherein, S12: and (3) execution period:
Figure 674758DEST_PATH_IMAGE005
;
wherein, S13: when the temperature is higher than the set temperature
Figure 702757DEST_PATH_IMAGE006
Then, the data is updated according to the following formula
Figure 884340DEST_PATH_IMAGE007
：
Figure 762428DEST_PATH_IMAGE008
Wherein, S14: updating
Figure 662251DEST_PATH_IMAGE009
：
Figure 228362DEST_PATH_IMAGE010
Wherein, S15: updating
Figure 530030DEST_PATH_IMAGE011
：
Figure 890604DEST_PATH_IMAGE012
Wherein, S16: steps S12 to S15 are repeated until the iteration stop condition is satisfied.
Figure 277723DEST_PATH_IMAGE013
in the formula ,
Figure 585208DEST_PATH_IMAGE014
representing the components of the IMF after decomposition,
Figure 475803DEST_PATH_IMAGE015
representing the center frequency of each component.
Figure 272858DEST_PATH_IMAGE016
Representing the lagrange multiplier, is used to represent,
Figure 881694DEST_PATH_IMAGE017
which represents a second-order penalty factor,
Figure 55186DEST_PATH_IMAGE018
which represents the frequency of the radio signal,
Figure 65868DEST_PATH_IMAGE019
，
Figure 220774DEST_PATH_IMAGE020
，
Figure 316906DEST_PATH_IMAGE021
are respectively corresponding to
Figure 28510DEST_PATH_IMAGE022
，
Figure 893698DEST_PATH_IMAGE023
，
Figure 32555DEST_PATH_IMAGE024
The fourier transform of (d).
Figure 615984DEST_PATH_IMAGE019
Is that
Figure 803382DEST_PATH_IMAGE025
The residual after wiener filtering.
Preferably, the fuzzy entropy calculation method in S20 is as follows:
wherein, S21: for one M-point sampling sequence definition:
Figure 788656DEST_PATH_IMAGE026
wherein, S22: reconstructing according to the continuous sequence of the sequence numbers to generate a group of n-dimensional vectors,
Figure 98415DEST_PATH_IMAGE027
wherein
Figure 169139DEST_PATH_IMAGE028
Represents the value of n consecutive u starting at the jth point,
Figure 222545DEST_PATH_IMAGE029
means of mean value, see formula
Figure 62325DEST_PATH_IMAGE030
Wherein, S23: defining two n-dimensional vectors
Figure 228471DEST_PATH_IMAGE031
And
Figure 520912DEST_PATH_IMAGE032
the distance between
Figure 378010DEST_PATH_IMAGE033
Is the one of the two corresponding elements with the largest difference, i.e.
Figure 72296DEST_PATH_IMAGE034
Figure 723858DEST_PATH_IMAGE035
Wherein, S24: by fuzzy functions
Figure 769174DEST_PATH_IMAGE036
Defining two vectors
Figure 164383DEST_PATH_IMAGE031
And
Figure 650859DEST_PATH_IMAGE032
similarity between them
Figure 738901DEST_PATH_IMAGE037
I.e. by
Figure 5934DEST_PATH_IMAGE038
In the above formula, function
Figure 204834DEST_PATH_IMAGE036
And m and r are respectively the gradient and the width of the boundary of the exponential function.
Wherein, S25: defining functions
Figure 608134DEST_PATH_IMAGE039
Wherein, S26: repeating the steps S22-S25, and reconstructing a group of n + 1-dimensional vectors according to the sequence number, wherein the function is defined as follows:
Figure 867077DEST_PATH_IMAGE040
wherein, S27: the fuzzy entropy is defined as:
Figure 808357DEST_PATH_IMAGE041
when the M value is a finite value, the fuzzy entropy estimation is carried out when the sequence number length is M according to the seven steps
Figure 545369DEST_PATH_IMAGE042
Preferably, the threshold formula of wavelet soft threshold denoising in S50 is:
Figure 68754DEST_PATH_IMAGE043
the wavelet base is cB10, and the number of wavelet layers is set to be 3;
Figure 498598DEST_PATH_IMAGE044
is the detail coefficient of the first layer decomposition, N is the data length, and j is the number of decomposition layers.
Preferably, the ICEEMDAN decomposition algorithm in S60 is:
wherein, S61: adding a set of white noises to an original sequence
Figure 474645DEST_PATH_IMAGE045
Structural sequence
Figure 15347DEST_PATH_IMAGE046
To obtain a first groupResidual error
Figure 658818DEST_PATH_IMAGE047
Wherein, S62: calculating a first modal component
Figure 931668DEST_PATH_IMAGE048
Wherein, S63: continuously adding white noise, and calculating a second group of residual errors by using local mean decomposition
Figure 660590DEST_PATH_IMAGE049
Defining a second modal component
Figure 739404DEST_PATH_IMAGE050
：
Figure 237381DEST_PATH_IMAGE051
Wherein, S64: computing the Kth residual
Figure 743449DEST_PATH_IMAGE052
And modal component
Figure 959667DEST_PATH_IMAGE053
；
Wherein, S65: and obtaining all modes and residual numbers until the calculation decomposition is finished.
x is the signal to be decomposed and x is,
Figure 530588DEST_PATH_IMAGE054
representing the k-order modal component resulting from EMD decomposition,
Figure 883072DEST_PATH_IMAGE055
which represents the generation of a local mean value of the signal,
Figure 825620DEST_PATH_IMAGE045
representing white gaussian noise.
Preferably, the relational number calculation formula in S70 is:
Figure 263554DEST_PATH_IMAGE056
wherein ,
Figure 684171DEST_PATH_IMAGE057
，
Figure 156741DEST_PATH_IMAGE058
the function of the mean function is to average the columns, a representing the original signal and B the decomposed components.
Preferably, the formula of the autocorrelation function calculated in S90 is as follows:
Figure 942294DEST_PATH_IMAGE059
wherein T is a signal
Figure 867525DEST_PATH_IMAGE060
The time of observation of (a) is,
Figure 91833DEST_PATH_IMAGE061
is that
Figure 418909DEST_PATH_IMAGE060
And
Figure 703260DEST_PATH_IMAGE062
the correlation between them.
Preferably, the wavelet soft-hard threshold denoising algorithm in S90 is:
wherein, S91: and carrying out wavelet decomposition on the noisy signals. And selecting a sym8 wavelet base, setting the number of wavelet layers to be 5, and performing wavelet decomposition to obtain a group of wavelet coefficients.
Wherein, S92: carrying out threshold quantization processing on each layer of high-frequency coefficients of wavelet decomposition to obtain an estimated value of the wavelet coefficients:
Figure 115787DEST_PATH_IMAGE063
wherein ,
Figure 65157DEST_PATH_IMAGE044
for the detail coefficients of the first layer decomposition,
Figure 246740DEST_PATH_IMAGE064
is the data length.
Wherein, S93: and performing inverse wavelet transform on the wavelet coefficients subjected to threshold quantization processing to reconstruct signals to obtain de-noised signals.
Preferably, the encoder of the inner model in S110 receives a long sequence input, and obtains the feature representation through a ProbSparse self-attention module and a self-attention distillation module. The ProbSparse Self-attention mechanism replaces the original attention matrix with a sparse matrix, greatly reduces the calculation power requirement while maintaining good performance, highlights the dominant factor in the Self-attention by halving the cascade layer input, and effectively processes overlong input sequences. The decoder receives the long sequence input, interacts with the coding features through multi-head attention, and finally directly predicts and outputs the target part.
The invention provides an optimized device operation trend prediction method based on an ICEEMDAN secondary decomposition coupling informar model, wherein a VMD and a fuzzy entropy are combined to process facility vibration signals collected by a sensor, a wavelet soft threshold method is used for denoising high-frequency noise, the effectiveness of decomposed components is ensured, the ICEEMDAN and an autocorrelation coefficient are combined to screen the high-frequency noise, and the wavelet soft and hard threshold method is used for denoising the high-frequency noise, so that the task operation integrity and the prediction accuracy are improved; and an informar model is adopted to predict the processed data, so that the prediction error is reduced, the model operation efficiency is improved, and the prediction precision is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary, and that other embodiments can be derived from the drawings provided by those of ordinary skill in the art without inventive effort.
FIG. 1 is a schematic view of a work flow provided by the present invention;
FIG. 2 shows the present invention an informer model structure diagram.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced otherwise than as specifically described herein, and thus the scope of the present invention is not limited by the specific embodiments disclosed below.
The embodiment provides a method for realizing device operation trend prediction based on an ICEEMDAN quadratic decomposition coupling informar model, as shown in fig. 1, the method specifically includes the following steps:
step S10: VMD decomposition is performed on the original signal.
The decomposition algorithm is as follows:
step S11: initialization
Figure 701992DEST_PATH_IMAGE001
，
Figure 601815DEST_PATH_IMAGE002
，
Figure 167925DEST_PATH_IMAGE003
And
Figure 204014DEST_PATH_IMAGE004
；
step S12: and (3) execution period:
Figure 830168DEST_PATH_IMAGE005
;
step S13: when the temperature is higher than the set temperature
Figure 154970DEST_PATH_IMAGE006
Then, update according to the following formula
Figure 259192DEST_PATH_IMAGE007
Figure 415367DEST_PATH_IMAGE008
Step S14: updating
Figure 212422DEST_PATH_IMAGE009
Figure 821258DEST_PATH_IMAGE010
Step S15: updating
Figure 994750DEST_PATH_IMAGE011
Figure 690917DEST_PATH_IMAGE012
Step S16: steps S12 to S15 are repeated until the iteration stop condition is satisfied.
Figure 658873DEST_PATH_IMAGE013
in the formula ,
Figure 755005DEST_PATH_IMAGE014
representing the components of the IMF after decomposition,
Figure 201030DEST_PATH_IMAGE015
representing the center frequency of each component.
Figure 331797DEST_PATH_IMAGE016
Representing the lagrange multiplier, is used to represent,
Figure 470654DEST_PATH_IMAGE017
representing a second order penalty factor
Figure 991766DEST_PATH_IMAGE065
Which represents the frequency of the radio signal,
Figure 241481DEST_PATH_IMAGE019
，
Figure 226755DEST_PATH_IMAGE020
，
Figure 536513DEST_PATH_IMAGE021
are respectively corresponding to
Figure 607238DEST_PATH_IMAGE022
，
Figure 660644DEST_PATH_IMAGE023
，
Figure 500424DEST_PATH_IMAGE024
The fourier transform of (d).
Figure 168035DEST_PATH_IMAGE019
Is that
Figure 460476DEST_PATH_IMAGE025
The residual after wiener filtering.
VMD (spatial mode decomposition) is an adaptive, completely non-recursive method of modal composition and signal processing. The method has the advantages that the number of modal decompositions can be determined, the self-adaptability is shown in that the number of modal decompositions of a given sequence is determined according to actual conditions, the optimal center frequency and the limited bandwidth of each mode can be matched in a self-adaptive mode in the subsequent searching and solving processes, effective separation of inherent modal components (IMF) and frequency domain division of signals can be achieved, effective decomposition components of given signals are obtained, and finally the optimal solution of the variation problem is obtained. Therefore, the problem that the EMD method has an end point effect and aliasing of modal components is solved, a firmer mathematical theory basis is provided, the time sequence non-stationarity with high complexity and strong nonlinearity can be reduced, and the relatively stable subsequence containing a plurality of different frequency scales is obtained through decomposition, so that the method is suitable for the non-stationarity sequence. The ICEEMDAN method is an improved algorithm of the CEEMDAN method, gaussian white noise corresponding to a modal order after decomposition by the EMD method is added in an original sequence, the modal aliasing problem existing in the EMD method can be effectively solved, and interference of other signals is prevented.
Step S20: the fuzzy entropy of the once decomposed components is calculated.
The fuzzy entropy calculation formula is as follows:
step S21: for one M-point sample sequence definition:
Figure 317574DEST_PATH_IMAGE026
step S22: a set of n-dimensional vectors is generated by successive sequential reconstruction of the sequence numbers,
Figure 11860DEST_PATH_IMAGE027
wherein
Figure 663421DEST_PATH_IMAGE028
Represents the value of n consecutive u starting at the jth point,
Figure 708738DEST_PATH_IMAGE029
means of mean value, see formula
Figure 41630DEST_PATH_IMAGE030
Step S23: defining two n-dimensional vectors
Figure 590423DEST_PATH_IMAGE031
And
Figure 678465DEST_PATH_IMAGE032
the distance between
Figure 945498DEST_PATH_IMAGE033
The one of the two corresponding elements having the largest difference, i.e.
Figure 878819DEST_PATH_IMAGE034
Figure 547698DEST_PATH_IMAGE035
Step S24: by fuzzy functions
Figure 806641DEST_PATH_IMAGE036
Defining two vectors
Figure 983806DEST_PATH_IMAGE031
And
Figure 986398DEST_PATH_IMAGE032
similarity between them
Figure 509783DEST_PATH_IMAGE037
I.e. by
Figure 674048DEST_PATH_IMAGE038
In the above formula, function
Figure 915673DEST_PATH_IMAGE036
For an exponential function, m, r are the gradient and width of the boundary of the exponential function, respectively.
Step S25: defining functions
Figure 456376DEST_PATH_IMAGE039
Step S26: repeating the steps S22-S25, and reconstructing a group of n + 1-dimensional vectors according to the sequence number, wherein the function is defined as follows:
Figure 771951DEST_PATH_IMAGE040
step S27: the fuzzy entropy is defined as:
Figure 372697DEST_PATH_IMAGE041
and when the value M is a finite value, estimating the fuzzy entropy when the sequence number length obtained according to the seven steps is M:
Figure 836039DEST_PATH_IMAGE042
step S30: a prescribed threshold value of blur entropy is set. The predetermined threshold value may be set to 0.05.
Step S40: and judging the fuzzy entropy value of each component obtained by the primary decomposition, wherein the component A is as follows: the fuzzy entropy value is less than a specified threshold, and the component B is: the fuzzy entropy value is greater than a prescribed threshold.
Step S50: and if the component of the primary decomposition meets the component A, performing wavelet soft threshold denoising on the component to obtain a denoising signal of the primary decomposition.
The wavelet soft threshold denoising algorithm comprises the following steps:
step S51: and carrying out wavelet decomposition on the noisy signals. Selecting cB10 wavelet base, setting the number of wavelet layers to be 3, and performing wavelet decomposition to obtain a group of wavelet coefficients.
Step S52: carrying out threshold quantization processing on each layer of high-frequency coefficients of wavelet decomposition to obtain an estimated value of the wavelet coefficients, wherein a threshold formula is as follows:
Figure 180433DEST_PATH_IMAGE043
wherein ,
Figure 678410DEST_PATH_IMAGE044
is the detail coefficient of the first layer decomposition, N is the data length, j is the scoreAnd (5) the number of solution layers.
Step S53: and performing inverse wavelet transform on the wavelet coefficients subjected to threshold quantization processing to reconstruct signals to obtain de-noised signals.
Step S60: if the component of the first decomposition meets the component B, performing ICEEMDAN second decomposition on the component to obtain a component of the second decomposition, wherein an ICEEMDAN decomposition algorithm is as follows:
step S61: adding a set of white noises to an original sequence
Figure 184478DEST_PATH_IMAGE045
Structural sequence
Figure 587646DEST_PATH_IMAGE046
To obtain a first set of residuals
Figure 470151DEST_PATH_IMAGE047
Step S62: calculating a first modal component
Figure 822635DEST_PATH_IMAGE048
Step S63: continuously adding white noise, and calculating a second group of residual errors by using local mean decomposition
Figure 499604DEST_PATH_IMAGE049
Defining a second modal component
Figure 937539DEST_PATH_IMAGE050
。
Figure 623735DEST_PATH_IMAGE051
Step S64: computing the Kth residual
Figure 768409DEST_PATH_IMAGE052
And modal component
Figure 881858DEST_PATH_IMAGE053
Step S65: and obtaining all modes and residual numbers until the calculation decomposition is finished.
x is the signal to be decomposed and x is,
Figure 807089DEST_PATH_IMAGE054
representing the k-order modal component resulting from EMD decomposition,
Figure 31397DEST_PATH_IMAGE055
which represents the generation of a local mean value of the signal,
Figure 358473DEST_PATH_IMAGE045
representing white gaussian noise.
Step S70: the correlation coefficient of the components of the icemedan quadratic decomposition is calculated. The X component is: the component with the correlation coefficient less than 0.1, and the Y component with the correlation coefficient greater than 0.1.
In step S70, the relational number calculation formula is:
Figure 642824DEST_PATH_IMAGE056
wherein ,
Figure 789771DEST_PATH_IMAGE066
the function of the mean function is to average the columns, a representing the original signal and B the decomposed components.
Step S80: and if the component of the second decomposition meets the X component, directly rejecting the component.
Step S90: and if the component of the secondary decomposition meets the Y component, calculating an autocorrelation function, determining a high-frequency component, and carrying out wavelet soft-hard threshold combined denoising on the high-frequency component to obtain a denoised signal of the secondary decomposition.
The formula of the autocorrelation function calculated in step S90 is:
Figure 503256DEST_PATH_IMAGE059
wherein T is a signal
Figure 684839DEST_PATH_IMAGE060
The time of observation of (a) is,
Figure 140091DEST_PATH_IMAGE061
is that
Figure 39914DEST_PATH_IMAGE060
And
Figure 606024DEST_PATH_IMAGE062
the correlation between them.
The wavelet soft-hard threshold denoising algorithm in the step S90 is as follows:
step S91: and carrying out wavelet decomposition on the noisy signals. And selecting a sym8 wavelet base, setting the number of wavelet layers to be 5, and performing wavelet decomposition to obtain a group of wavelet coefficients.
Step S92: carrying out threshold quantization processing on each layer of high-frequency coefficients of wavelet decomposition to obtain an estimated value of the wavelet coefficients:
Figure 579796DEST_PATH_IMAGE063
wherein ,
Figure 205950DEST_PATH_IMAGE044
for the detail coefficients of the first layer decomposition,
Figure 327490DEST_PATH_IMAGE064
is the data length.
Step S93: and performing inverse wavelet transform on the wavelet coefficients subjected to threshold quantization processing to reconstruct signals to obtain de-noised signals.
Step S100: and performing waveform reconstruction on the denoising signal subjected to the primary decomposition and the denoising signal subjected to the secondary decomposition to obtain a final denoising signal.
Step S110: and inputting the obtained denoising signal into an informar model to obtain prediction data.
The encoder of the inventive enhancer model is shown in FIG. 2, and receives a long sequence input, and features the input through a ProbSparse self-attention module and a self-attention distillation module. The ProbSparse Self-attention mechanism replaces the original attention matrix with a sparse matrix, greatly reduces the computational power requirements while maintaining good performance, and effectively processes overlong input sequences by halving the cascade layer input to highlight the leading factor in the Self-attention mechanism (Self-attention). The decoder receives the long sequence input, interacts with the coding features through multi-head attention, and finally directly predicts and outputs the target part. The rest unexplained parts are the conventional settings of the inner model, and are not described in detail herein.
The ProbSparse autocorrelation mechanism of the Informmer model enables the time complexity and the memory utilization rate to reach
Figure 697291DEST_PATH_IMAGE067
(ii) a An auto-correlation distillation operation, which highlights the feature of height separation on the J stacked layers and greatly reduces the spatial complexity, which helps the model to receive long sequence input; the generator decoder (decoder) directly performs multi-step prediction at one time, avoids error accumulation generated by single-step prediction, improves prediction precision and reduces prediction time.
The method combines the VMD and the fuzzy entropy to process the facility vibration signal collected by the sensor, utilizes the wavelet soft threshold method to denoise the high-frequency noise, ensures the effectiveness of the decomposed component, combines the ICEEMDAN and the autocorrelation coefficient to screen the high-frequency noise, and utilizes the wavelet soft threshold and the wavelet hard threshold method to denoise the high-frequency noise, thereby improving the task operation integrity and the prediction accuracy; and the processed data is creatively combined and predicted by adopting an informar model, so that the prediction error is reduced, the model operation efficiency is improved, and the prediction precision is improved. Therefore, the invention solves the problem that the prediction result has large error due to a large amount of noise contained in the equipment operation process signal recorded by the sensor. The invention improves the accuracy of the prediction task, reduces the error caused by noise and improves the prediction precision.
The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the embodiments. It will be apparent, however, to one skilled in the art that the embodiments may be practiced without the specific details. Thus, the foregoing descriptions of specific embodiments described herein are presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the embodiments to the precise forms disclosed. It will be apparent to those skilled in the art that many modifications and variations are possible in light of the above teaching. Further, as used herein to refer to the position of a component, the terms above and below, or their synonyms, do not necessarily refer to an absolute position relative to an external reference, but rather to a relative position of the component with reference to the drawings.
Moreover, the foregoing drawings and description include many concepts and features that may be combined in various ways to achieve various benefits and advantages. Thus, features, components, elements and/or concepts from various different figures may be combined to produce embodiments or implementations not necessarily shown or described in this specification. Furthermore, not all features, components, elements and/or concepts shown in a particular figure or description are necessarily required to be in any particular embodiment and/or implementation. It is to be understood that such embodiments and/or implementations fall within the scope of the present description.

Claims (12)

1. An equipment operation trend prediction method based on an ICEEMDAN quadratic decomposition coupling informar model is characterized by comprising the following steps:
step S10: performing VMD decomposition on the original signal;
step S20: calculating fuzzy entropy of the components of the primary decomposition;
step S30: setting a specified threshold value of the fuzzy entropy;
step S40: judging the fuzzy entropy value of each component obtained by the primary decomposition, wherein the component A is as follows: the fuzzy entropy value is less than a specified threshold, and the component B is: the fuzzy entropy value is larger than a specified threshold value;
step S50: if the component of the primary decomposition meets the component A, performing wavelet soft threshold denoising on the component to obtain a denoising signal of the primary decomposition;
step S60: if the component of the first decomposition meets the component B, performing ICEEMDAN secondary decomposition on the component to obtain a secondary decomposed component;
step S70: calculating the correlation coefficient of the component of ICEEMEDAN quadratic decomposition, wherein the X component is as follows: the component with the correlation coefficient less than 0.1, and the Y component is the component with the correlation coefficient more than 0.1;
step S80: if the component of the secondary decomposition meets the X component, directly rejecting the component;
step S90: if the component of the secondary decomposition meets the Y component, calculating an autocorrelation function, determining a high-frequency component, and carrying out wavelet soft-hard threshold combined denoising on the high-frequency component to obtain a denoising signal of the secondary decomposition;
step S100: performing waveform reconstruction on the denoising signal subjected to the primary decomposition and the denoising signal subjected to the secondary decomposition to obtain a final denoising signal;
step S110: and inputting the final de-noising signal into an interpolator model to obtain prediction data.
2. The ICEEMDAN quadratic decomposition coupling informar model-based device operation trend prediction method according to claim 1, wherein: the VMD decomposition algorithm in the step S10 is as follows:
step S11: initialization
Figure 550327DEST_PATH_IMAGE001
，
Figure 432832DEST_PATH_IMAGE002
，
Figure 722999DEST_PATH_IMAGE003
And
Figure 665547DEST_PATH_IMAGE004
；
step S12: and (3) execution period:
Figure 103482DEST_PATH_IMAGE005
；
step S13: when in use
Figure 789678DEST_PATH_IMAGE006
Then, the data is updated according to the following formula
Figure 996669DEST_PATH_IMAGE007
：
Figure 110118DEST_PATH_IMAGE008
Step S14: updating
Figure 723764DEST_PATH_IMAGE009
：
Figure 948072DEST_PATH_IMAGE010
Step S15: updating
Figure 275149DEST_PATH_IMAGE011
：
Figure 293920DEST_PATH_IMAGE012
Step S16: repeating steps S12 to S15 until the iteration stop condition is satisfied:
Figure 706447DEST_PATH_IMAGE013
in the formula ,
Figure 734446DEST_PATH_IMAGE014
representing the components of the IMF after decomposition,
Figure 853712DEST_PATH_IMAGE015
represent each groupA center frequency of the portion;
Figure 308964DEST_PATH_IMAGE016
representing the lagrange multiplier, is used to represent,
Figure 943207DEST_PATH_IMAGE017
which represents a second-order penalty factor,
Figure 509318DEST_PATH_IMAGE018
which represents the frequency of the radio signal,
Figure 810986DEST_PATH_IMAGE019
，
Figure 437140DEST_PATH_IMAGE020
，
Figure 558679DEST_PATH_IMAGE021
respectively correspond to
Figure 115432DEST_PATH_IMAGE022
，
Figure 271606DEST_PATH_IMAGE023
，
Figure 68661DEST_PATH_IMAGE024
Fourier transform of (3);
Figure 677497DEST_PATH_IMAGE019
is that
Figure 585410DEST_PATH_IMAGE025
The residual after wiener filtering.
3. The method of claim 1, wherein the method for predicting the device operation trend based on the icemdan quadratic decomposition coupling informar model is characterized in that: the fuzzy entropy calculation method in the step S20 comprises the following steps:
step S21: for one M-point sample sequence definition:
Figure 596091DEST_PATH_IMAGE026
，
step S22: reconstructing according to the continuous sequence of the sequence numbers to generate a group of n-dimensional vectors,
Figure 501731DEST_PATH_IMAGE027
wherein
Figure 597863DEST_PATH_IMAGE028
Represents the value of n consecutive u starting at the jth point,
Figure 309467DEST_PATH_IMAGE029
the mean value is represented by the average value,
Figure 174654DEST_PATH_IMAGE029
the calculation is as follows:
Figure 313512DEST_PATH_IMAGE030
step S23: defining two n-dimensional vectors
Figure 896940DEST_PATH_IMAGE031
And
Figure 121158DEST_PATH_IMAGE032
a distance therebetween
Figure 840852DEST_PATH_IMAGE033
Is the one of the two corresponding elements with the largest difference, i.e.
Figure 150611DEST_PATH_IMAGE034
Figure 221335DEST_PATH_IMAGE035
Step S24: by fuzzy functions
Figure 274742DEST_PATH_IMAGE036
Defining two vectors
Figure 114522DEST_PATH_IMAGE031
And
Figure 532865DEST_PATH_IMAGE032
similarity between them
Figure 90885DEST_PATH_IMAGE037
I.e. by
Figure 682404DEST_PATH_IMAGE038
In the above formula, function
Figure 376690DEST_PATH_IMAGE036
Is an exponential function, and m and r are respectively the gradient and the width of the boundary of the exponential function;
step S25: defining functions
Figure 293831DEST_PATH_IMAGE039
Step S26: repeating the steps S22-S25, and reconstructing a group of n + 1-dimensional vectors according to the sequence number, wherein the function is defined as follows:
Figure 260518DEST_PATH_IMAGE040
step S27: the fuzzy entropy is defined as:
Figure 655728DEST_PATH_IMAGE041
when the value of M is a finite value, the fuzzy entropy estimation is carried out according to the sequence number with the length of M obtained in the steps S21-S27:
Figure 470100DEST_PATH_IMAGE042
。
4. the method of claim 1, wherein the method for predicting the device operation trend based on the icemdan quadratic decomposition coupling informar model is characterized in that: the prescribed threshold value is set to 0.05 in step S30.
5. The ICEEMDAN quadratic decomposition coupling informar model-based device operation trend prediction method according to claim 1, wherein: the wavelet soft threshold denoising algorithm in the step S50 is as follows:
step S51: performing wavelet decomposition on the noisy signals, selecting cB10 wavelet basis, setting the number of wavelet layers to be 3, and performing wavelet decomposition to obtain a group of wavelet coefficients;
step S52: carrying out wavelet soft threshold quantization processing on each layer of high-frequency coefficients of wavelet decomposition to obtain an estimated value of the wavelet coefficients, wherein a wavelet soft threshold formula is as follows:
Figure 292562DEST_PATH_IMAGE043
wherein ,
Figure 559596DEST_PATH_IMAGE044
is the detail coefficient of the first layer decomposition, N is the data length, j is the number of decomposition layers;
step S53: and performing inverse wavelet transform on the wavelet coefficient subjected to the wavelet soft threshold quantization processing to reconstruct a signal to obtain a de-noised signal.
6. The ICEEMDAN quadratic decomposition coupling informar model-based device operation trend prediction method according to claim 1, wherein: the icemdan decomposition algorithm in step S60 is:
step S61: adding a set of white noises to an original sequence
Figure 758496DEST_PATH_IMAGE045
Structural sequence
Figure 365058DEST_PATH_IMAGE046
To obtain a first set of residuals
Figure 358421DEST_PATH_IMAGE047
，
Step S62: calculating a first modal component
Figure 112751DEST_PATH_IMAGE048
，
Step S63: continuously adding white noise, and calculating a second group of residual errors by using local mean decomposition
Figure 849763DEST_PATH_IMAGE049
Defining a second modal component
Figure 373148DEST_PATH_IMAGE050
，
Figure 802992DEST_PATH_IMAGE051
Step S64: computing the Kth residual
Figure 733033DEST_PATH_IMAGE052
And modal component
Figure 8157DEST_PATH_IMAGE053
，
Step S65: obtaining all modes and residual numbers until the calculation decomposition is finished;
wherein, x is the signal to be decomposed,
Figure 651628DEST_PATH_IMAGE054
representing the k-order modal component resulting from the EMD decomposition,
Figure 986794DEST_PATH_IMAGE055
which represents the generation of a local mean value of the signal,
Figure 715716DEST_PATH_IMAGE045
representing white gaussian noise.
7. The method of claim 1, wherein the method for predicting the device operation trend based on the icemdan quadratic decomposition coupling informar model is characterized in that: the formula for calculating the relation number in step S70 is:
Figure 794530DEST_PATH_IMAGE056
wherein ,
Figure 292508DEST_PATH_IMAGE057
，
Figure 1838DEST_PATH_IMAGE058
the function of the mean function is to average the columns, a representing the original signal and B the decomposed components.
8. The ICEEMDAN quadratic decomposition coupling enhancer model-based device operation trend prediction method according to claim 7, wherein a correlation coefficient threshold is defined to be 0.1.
9. The ICEEMDAN quadratic decomposition coupling informar model-based device operation trend prediction method according to claim 1, wherein: the formula of the autocorrelation function calculated in step S90 is:
Figure 952476DEST_PATH_IMAGE059
wherein T is a signal
Figure 834981DEST_PATH_IMAGE060
The time of observation of (a) is,
Figure 187465DEST_PATH_IMAGE061
is that
Figure 130014DEST_PATH_IMAGE060
And
Figure 567948DEST_PATH_IMAGE062
the correlation between them.
10. The method of claim 9, wherein the method for predicting the device operation trend based on the icemdan quadratic decomposition coupling informar model is characterized in that: the autocorrelation function of the high-frequency noise component at the zero point reaches the maximum value, and the autocorrelation functions of the high-frequency noise component at other times approach 0; the autocorrelation function of the low-frequency noise component is maximum at the zero point, and the low-frequency noise component does not completely conform to the noise characteristics at the rest of the time.
11. The ICEEMDAN quadratic decomposition coupling informar model-based device operation trend prediction method according to claim 1, wherein: the wavelet soft-hard threshold denoising algorithm in the step S90 is as follows:
step S91: carrying out wavelet decomposition on the noisy signal, selecting sym8 wavelet basis, setting the number of wavelet layers to be 5, and carrying out wavelet decomposition to obtain a group of wavelet coefficients;
step S92: carrying out threshold quantization processing on each layer of high-frequency coefficients of wavelet decomposition to obtain an estimated value of the wavelet coefficients, wherein a threshold formula is as follows:
Figure 175516DEST_PATH_IMAGE063
wherein ,
Figure 648086DEST_PATH_IMAGE044
for the detail coefficients of the first layer decomposition,
Figure 761535DEST_PATH_IMAGE064
is the data length;
step S93: and performing inverse wavelet transform on the wavelet coefficients subjected to threshold quantization processing to reconstruct signals to obtain de-noised signals.
12. The method of claim 1, wherein the method for predicting the device operation trend based on the icemdan quadratic decomposition coupling informar model is characterized in that:
in step S110, an encoder of the interpolator model receives a long sequence input, and obtains a feature representation through a ProbSparse self-attention module and a self-attention distillation module;
the ProbSparse Self-attention mechanism replaces the attention matrix with a sparse matrix and highlights the dominant factor in the Self-attention by halving the cascade layer input;
the decoder receives the long sequence input, interacts with the coding features through multi-head attention, and finally directly predicts and outputs the target part.
CN202211315575.3A 2022-10-26 2022-10-26 Equipment operation trend prediction method based on ICEEMDAN (information and communication technology) secondary decoupling index model Active CN115438301B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211315575.3A CN115438301B (en) 2022-10-26 2022-10-26 Equipment operation trend prediction method based on ICEEMDAN (information and communication technology) secondary decoupling index model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211315575.3A CN115438301B (en) 2022-10-26 2022-10-26 Equipment operation trend prediction method based on ICEEMDAN (information and communication technology) secondary decoupling index model

Publications (2)

Publication Number Publication Date
CN115438301A true CN115438301A (en) 2022-12-06
CN115438301B CN115438301B (en) 2023-06-20

Family

ID=84252828

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211315575.3A Active CN115438301B (en) 2022-10-26 2022-10-26 Equipment operation trend prediction method based on ICEEMDAN (information and communication technology) secondary decoupling index model

Country Status (1)

Country Link
CN (1) CN115438301B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116578844A (en) * 2023-06-07 2023-08-11 广东工业大学 Cold quantity prediction method, system and related equipment for heating ventilation air conditioner

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107688120A (en) * 2016-11-17 2018-02-13 丹阳华神电器有限公司 Signals and associated noises processing method and iteration singular spectrum Soft-threshold Denoising Method based on fuzzy entropy
CN108491355A (en) * 2018-02-05 2018-09-04 南京邮电大学 A kind of ultrasonic signal noise-reduction method based on CEEMD and wavelet packet
CN109299430A (en) * 2018-09-30 2019-02-01 淮阴工学院 The short-term wind speed forecasting method with extreme learning machine is decomposed based on two stages
CN111553513A (en) * 2020-04-14 2020-08-18 太原理工大学 Medium-and-long-term runoff prediction method based on quadratic decomposition and echo state network
CN113011669A (en) * 2021-03-30 2021-06-22 北京科技大学 Method for predicting monthly stock quantity of live pigs
US20220045509A1 (en) * 2020-08-05 2022-02-10 Wuhan University Method and system of predicting electric system load based on wavelet noise reduction and emd-arima
CN114886378A (en) * 2022-04-11 2022-08-12 广州芯相随科技有限公司 Improved complementary set modal decomposition-based joint denoising method and system

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107688120A (en) * 2016-11-17 2018-02-13 丹阳华神电器有限公司 Signals and associated noises processing method and iteration singular spectrum Soft-threshold Denoising Method based on fuzzy entropy
US20180138896A1 (en) * 2016-11-17 2018-05-17 Hwashen Electronics Co., Ltd. Fuzzy entropy based noisy signal processing method and iterative singular spectrum analysis soft threshold de-noising method
CN108491355A (en) * 2018-02-05 2018-09-04 南京邮电大学 A kind of ultrasonic signal noise-reduction method based on CEEMD and wavelet packet
CN109299430A (en) * 2018-09-30 2019-02-01 淮阴工学院 The short-term wind speed forecasting method with extreme learning machine is decomposed based on two stages
CN111553513A (en) * 2020-04-14 2020-08-18 太原理工大学 Medium-and-long-term runoff prediction method based on quadratic decomposition and echo state network
US20220045509A1 (en) * 2020-08-05 2022-02-10 Wuhan University Method and system of predicting electric system load based on wavelet noise reduction and emd-arima
CN113011669A (en) * 2021-03-30 2021-06-22 北京科技大学 Method for predicting monthly stock quantity of live pigs
CN114886378A (en) * 2022-04-11 2022-08-12 广州芯相随科技有限公司 Improved complementary set modal decomposition-based joint denoising method and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
余帆，王磊，江巧永，闫群民，皇金锋: "基于特征筛选的VMD-MIC-SSA-Informer短期负荷预测" *
董永生: "小波变换与阈值函数在数字信号去噪中的应用" *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116578844A (en) * 2023-06-07 2023-08-11 广东工业大学 Cold quantity prediction method, system and related equipment for heating ventilation air conditioner
CN116578844B (en) * 2023-06-07 2024-01-26 广东工业大学 Cold quantity prediction method, system and related equipment for heating ventilation air conditioner

Also Published As

Publication number Publication date
CN115438301B (en) 2023-06-20

Similar Documents

Publication Publication Date Title
CN110598166B (en) Wavelet denoising method for adaptively determining wavelet layering progression
US7197453B2 (en) System and method for optimizing the operation of an oversampled discrete Fourier transform filter bank
CN110598615B (en) Data noise reduction method and system for monitoring bridge structure
CN115438301A (en) Equipment operation trend prediction method based on ICEEMDAN quadratic decomposition coupling informar model
CN111950759A (en) Short-term wind speed prediction method based on two-stage decomposition, LSTM and AT
EP1146649A1 (en) Tyem-series predicting method using wavelet number series and device therefor
CN114886378A (en) Improved complementary set modal decomposition-based joint denoising method and system
CN111275244A (en) Vehicle speed time series frequency division prediction method
CN109724693B (en) Fusion spectrum denoising method based on stationary wavelet
Aminghafari et al. Forecasting time series using wavelets
Weickert et al. Analytic wavelet packets—combining the dual-tree approach with wavelet packets for signal analysis and filtering
CN114330459A (en) Vibration source response signal-noise separation method and system based on multi-layer decomposition
JP2023545820A (en) Generative neural network model for processing audio samples in the filter bank domain
CN107133921B (en) Image super-resolution reconstruction method and system based on multilevel neighborhood embedding
CN111582205A (en) Noise reduction method based on multi-resolution singular value decomposition model
Yang et al. SAR image denoising algorithm based on Bayes wavelet shrinkage and fast guided filter
Kulkarni et al. Periodicity-aware signal denoising using Capon-optimized Ramanujan filter banks and pruned Ramanujan dictionaries
CN107315713B (en) One-dimensional signal denoising and enhancing method based on non-local similarity
Xu et al. An adaptive threshold method for image denoising based on wavelet domain
CN109614937A (en) A kind of wavelet field signal enhancing method based on group sparsity
CN110927678B (en) Adaptive sparse fractional order fuzzy function clutter suppression and moving target detection method
Cengiz Signal reconstruction with adaptive multi-rate signal processing algorithms
Richter et al. Adaptive Signal Processing
Zoican Speech de-noising system with non local means algorithm
Chang et al. Maneuvering Target Tracking with High-Order Correlated Noise–A Multirate Kalman Filtering Approach

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant