CN103716012B - A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological - Google Patents

A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological Download PDF

Info

Publication number
CN103716012B
CN103716012B CN201310672036.XA CN201310672036A CN103716012B CN 103716012 B CN103716012 B CN 103716012B CN 201310672036 A CN201310672036 A CN 201310672036A CN 103716012 B CN103716012 B CN 103716012B
Authority
CN
China
Prior art keywords
filtering
structural
signal
result
adapting
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.)
Expired - Fee Related
Application number
CN201310672036.XA
Other languages
Chinese (zh)
Other versions
CN103716012A (en
Inventor
姜万录
李扬
董彩云
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.)
Yanshan University
Original Assignee
Yanshan University
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 Yanshan University filed Critical Yanshan University
Priority to CN201310672036.XA priority Critical patent/CN103716012B/en
Publication of CN103716012A publication Critical patent/CN103716012A/en
Application granted granted Critical
Publication of CN103716012B publication Critical patent/CN103716012B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention discloses a kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological, its content includes: after carrying out self-adapting multi-dimension morphological analysis filtering, adaptive approach is used to determine new structural element range scale last filter result, and with newly determined range scale, primary signal is carried out Multi-Scale Morphological Filtering, iteration said process, until determined by adaptive approach, structural element range scale is stable, finally to each yardstick results averaged as final result.The filtering method that the present invention analyzes based on iteration self-adapting Multiscale Morphological, self-adapting multi-dimension morphological analysis method has been carried out iteration use, can be when strong noise background and sample frequency be higher, remain to effectively suppress the interference of noise, harmonic wave etc., clearly extract fault signature, it is thus achieved that ideal filtering result.It compensate for the deficiency of self-adapting multi-dimension morphological analysis, enriches the theoretical method of shape filtering.

Description

A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological
Technical field
The present invention relates to the filtering method in a kind of rotating machinery fault signal processing field, specifically, be a kind of The filtering method analyzed based on iteration self-adapting Multiscale Morphological.
Background technology
Fault detection and diagnosis, Fault prediction techniques are the important classes of modern machinery and equipment operation maintenance and management Topic.In numerous plant equipment, rotating machinery has obtained substantial amounts of application, and its fault-signal is the most multiple Miscellaneous, often show the most non-linear, non-stationary.The most effectively suppress the interference of noise, harmonic wave etc., The fault signature extracting signal just becomes a key issue.
Traditional signal processing method is premised on the stationarity of signal, therefore, it is impossible to extract reflection event The signal of barrier characteristic information;In modern signal processing technology, wavelet transformation, Hilbert-Huang conversion, Although rarefaction representation, independent component analysis, cyclo-stationary signal analysis, manifold learning etc. all have non-linear Non-stationary Signal Analysis ability, but they all have respective limitation.
Mathematical morphology, as a kind of Nonlinear harmonic oscillator method, occurs the earliest and is developed in image procossing neck Territory, recently has scholar to be introduced into one-dimensional signal process field, and starts to carry out in mechanical fault diagnosis Application.In 441-445 page of " University of Science & Technology, Beijing's journal " the 4th phase of volume 30 in 2008, Zhang Lijun etc. " self-adapting multi-dimension morphological analysis and the application in bearing failure diagnosis thereof " literary composition proposes: from Adapt to Multiscale Morphological and analyze the method form spy compared with single scale morphological analysis method more conducively extraction signal Levy, it is to avoid the single scale morphological analysis blindness when structural element selects and to relevant priori Dependency, but when strong noise background and sample frequency are higher, noise and harmonic wave can not be effectively suppressed, Fault characteristic information can not be effectively extracted.
Summary of the invention
For when strong noise background and sample frequency are higher, through tradition self-adapting multi-dimension morphological analysis filtering After, extraction this problem inconspicuous of fault characteristic information.It is an object of the invention to provide a kind of use condition Broader, that filter effect is more stable signal filtering method.
To achieve these goals, the present invention is achieved by the following technical solutions:
A kind of based on iteration self-adapting Multiscale Morphological analysis (Iterative adaptive multiscale Morphology analysis, IAMMA) filtering method, its content comprises the steps:
Step one choice structure element shape
According to the impact on signal filtering result of the difformity structural element and the purpose of signal filtering, select Suitably structural element shape;Generally structural element length value is taken as odd number, in order to its midpoint is set to initial point, Thus ensure the symmetry of structural element;
Conventional structural element shape has triangular form, semi-circular, platypelloid type, and wherein triangular form structural element is fitted Closing and filter the interference of pulse signal, semi-circular structural element is suitable for filtering the interference of random noise, and platypelloid type It is zero that structural element generally takes its height value, only need to select length value, the easiest, be easily achieved, Thus be widely used in one-dimensional signal processes;
Step 2 self-adapting multi-dimension morphological analysis (adaptive multiscale morphology analysis, AMMA)
Primary signal is carried out self-adapting multi-dimension morphological analysis filtering, first to the discrete state signal gathered Carry out zero-mean process, look for after zero-meanization processes all positive peak points, then basis in discrete signal Each trunnion axis position, positive peak point place (abscissa), can obtain the interval of each adjacent positive peak point, root The minima being spaced according to adjacent positive peak and maximum determine structural element length dimension range lambdal;Basis afterwards Each vertical axes position, positive peak point place (vertical coordinate), according to the length little yardstick little yardstick of respective heights and length The computing of the policy setting of large scale respective heights large scale, can be calculated each length ruler of structural element The elevational dimension that degree is corresponding, so that it is determined that each mesostructure element shape;Calculate each mesostructure the most respectively Filter result under element, and using defeated as AMMA method of the meansigma methods of each scale topographical operation result Go out result;
Step 3 iteration uses self-adapting multi-dimension morphological analysis
Last filter result use adaptive approach determine new structural element range scale, and with new true Fixed range scale carries out Multi-Scale Morphological Filtering, iteration said process to primary signal;By adaptive approach Selected structural element range scale carries out multiple dimensioned differential filtering to signal, and each filtering all can be in signal Noise contribution filters, and extracts impact composition simultaneously;In so each iterative process, noise peak is gradually Reducing, the structural element yardstick chosen is gradually increased, and poor morphology value filtering is removed the ability of noise and gradually strengthened, Noise major part in signal is effectively filtered out by final poor morphology value filtering;The structural element now chosen is maximum Yardstick tends towards stability, and adaptive process terminates;Interval between the adjacent positive peak that thus self adaptation is chosen is more Be the interval between the adjacent positive peak of failure impact signal, be the most rational gap length;It is spaced according to this The structural element range scale that length obtains is more suitable, such that it is able to obtain more preferably signal filtering knot Really;
Step 4 terminates judging
Judge that the condition that iterative process terminates is: structural element yardstick maximum determined by adaptive approach is even Continuous three times identical;Judge that the new structural element yardstick every time using adaptive approach to determine in step 3 is maximum Whether value changes, if the new structural element yardstick maximum that determines of adaptive approach continuous three times identical, repeatedly Terminate for process;
Step 5 calculates iteration self-adapting Multiscale Morphological analysis result
Calculate the meansigma methods of differential filtering result under each yardstick, using this value as multiple dimensioned based on iteration self-adapting Morphological analysis filtering method processes the final result of signal;Under each yardstick, differential filtering result reflects letter respectively Characteristic information number under different scale, and the impact composition carrying machinery device fault information is present in each In individual yardstick, therefore the result sequence under each yardstick of gained is averaged as IAMMA method Eventually as a result, it is possible to more fully reflect the characteristic information of signal, and its impact composition prominent, effectively filter out Influence of noise.
IAMMA, AMMA method that is respectively adopted is to CWRU of the U.S. (Case Western Reserve University) the bearing vibration data of bearing test-bed collection are analyzed.It is respectively adopted two kinds of methods to obtain The frequency spectrum of the bearing inner race fault arrived is as shown in Figure 1.As seen from the figure, Fig. 1 (a) is original signal spectrum figure, IAMMA method filter result Fig. 1 (c) clearly presents rolling bearing inner ring fault characteristic frequency 162.2Hz And frequency multiplication, and the fault characteristic frequency in AMMA method filter result Fig. 1 (b) is submerged in noise.
Owing to using technique scheme, the one that the present invention provides is divided based on iteration self-adapting Multiscale Morphological The filtering method of analysis, compared with prior art has such beneficial effect:
The filtering method that the present invention analyzes based on iteration self-adapting Multiscale Morphological, to self-adapting multi-dimension form Analysis method has carried out iteration use, can remain to effectively suppress when strong noise background and sample frequency are higher The interference of noise, harmonic wave etc., clearly extracts fault signature, it is thus achieved that ideal filtering result.It compensate for adaptive Answer the deficiency that Multiscale Morphological is analyzed, enrich the theoretical method of shape filtering.
Accompanying drawing explanation
Fig. 1 is the frequency being respectively adopted AMMA and IAMMA method to rolling bearing inner ring failure recovery result Spectrum comparison diagram, wherein (a) is original signal spectrum figure, and (b) is AMMA method spectrogram, and (c) is IAMMA Method spectrogram;
Fig. 2 is the based on iteration self-adapting Multiscale Morphological analysis (IAMMA) filtering method concrete of the present invention Realize flow chart of steps.
Detailed description of the invention
Below in conjunction with the accompanying drawings the filtering method based on iteration self-adapting Multiscale Morphological analysis to the present invention make into One step ground explanation.
A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological, as in figure 2 it is shown, the method bag Containing specifically comprising the following steps that
Step one choice structure element shape
According to the impact on signal filtering result of the difformity structural element, and the purpose of signal filtering, choosing Select suitable structural element shape.
Step 2 self-adapting multi-dimension morphological analysis
First to gather discrete state signal f (n) carry out zero-mean process, obtain on this basis its from Dissipate sequence f1(n)={xn| n=1,2 ..., N}, looks for f1All positive peak point P={p in (n)m| m=1,2 ..., M}, According to each positive peak point pmTrunnion axis position, place (abscissa) pm,x, each adjacent positive peak point can be obtained Between be spaced apart: Ip={ip|ip=pm+1,x-pm,x,m=1,2,…,M-1}。
Interval I according to adjacent peak pointpTwo formulas below and:
Multiscale Morphological schoolmate degree range scale λ can be obtainedl: λl={λlminlmin+1,…,λlmax(j)-1,λlmax(j) }, Wherein j=1.In formula,For the operator that rounds up,Accord with for downward rounding operation.
By each positive peak point pmVertical axes position, place (vertical coordinate) pm,ySubstitution following formula:
λh=β·{min(pm,y)+α·[max(pm,y)-min(pm,y)]/(λlmaxlmin)} (3)
Structural element elevational dimension range lambda can be calculatedh, thus finally determine each mesostructure element shape Shape.Wherein, α=1,2 ..., λlmaxlmin+ 1, β are height ratio coefficient (0 < β < 1).
Under selected structural element yardstick, discrete signal is carried out poor morphology value filtering computing respectively, obtains each Filter result result_data (k, 1:1:N) under yardstick, wherein k=λlmin(1):1:λlmax(j), j=1.Calculate each Under yardstick, meansigma methods multi_result_data of filter result result_data (k, 1:1:N) filters as this Result.
Step 3 iteration uses self-adapting multi-dimension morphological analysis
Find out each positive peak point in the time domain form of consequential signal multi_result_data P′={p′m| m=1,2 ..., M ' }, according to each positive peak point p 'mTrunnion axis position, place p 'm,x, can obtain It is spaced apart between each adjacent positive peak point: I 'p={ip|ip=p′m+1,x-p′m,x,m=1,2,…,M′-1}.According to each peak value Between point, gap size selectes the maximum λ of structural element length dimensionlmax(j+1).Judge λlmax(j+1) with λlmaxThe size of (j), if λlmax(j+1) more than λlmaxJ (), then continue executing with, otherwise leap to step 4.
The structural element yardstick maximum λ chosen with the last timelmaxJ minima that () filters as this, this The structural element yardstick maximum λ chosenlmax(j+1) still as maximum, determine that this filters the knot used Constitutive element length dimension range lambdal' it is: λl′={λlmax(j),λlmax(j)+1,…,λlmax(j+1)-1,λlmax(j+1)}。 Structural element elevational dimension range lambda can be calculated according to formula (3)h', thus finally determine each mesostructure unit Element shape.
Under selected structural element yardstick, original discrete signal is carried out poor morphology value filtering computing respectively, Filter result result_data (k, 1:1:N) under each yardstick, wherein k=λlmax(j):1:λlmax(j+1).Meter Calculate under each yardstick meansigma methods multi_result_data of differential filtering result result_data (k, 1:1:N) as this The result of secondary filtering.
J, after adding 1, again initiates from step 3 using multi_result_data as object and processes.
Step 4 terminates judging
Judge λlmax(j+1) whether with λlmaxJ () is equal, if unequal, j jumps to step 3 after adding 1, no Then judge j whether equal to 1, as j=1, j after adding 1 using multi_result_data as object again from Step 3 is initial to be processed, and otherwise judges λlmax(j+1) whether with λlmax(j-1) equal, if equal, jump To step 5, otherwise j again initiates from step 3 using multi_result_data as object after adding 1 and carries out Process.
Step 5 calculates iteration self-adapting Multiscale Morphological analysis result
Calculate meansigma methods multi_result_data of differential filtering result result_data (k, 1:1:N) under each yardstick The final result of signal, wherein k=λ is processed as IAMMA methodlmin(1):1:λlmax(j+1).Each yardstick Lower differential filtering result result_data (k, 1:1:N) reflects signal characteristic information under different scale respectively, And the impact composition carrying machinery device fault information is present in each yardstick, therefore by each chi of gained Result sequence under Du is averaged the final result as IAMMA method, can more fully reflect letter Number characteristic information, and its impact composition prominent, effectively filter out influence of noise.

Claims (1)

1. the filtering method analyzed based on iteration self-adapting Multiscale Morphological, it is characterised in that: the method Content comprises the steps:
Step one choice structure element shape
According to the impact on signal filtering result of the difformity structural element and the purpose of signal filtering, select Suitably structural element shape;Structural element length value is taken as odd number, in order to its midpoint is set to initial point, from And ensure the symmetry of structural element;
Structural element shape has triangular form, semi-circular, platypelloid type, and wherein triangular form structural element is suitable for filtering The interference of pulse signal, semi-circular structural element is suitable for filtering the interference of random noise, platypelloid type structural element Taking its height value is zero, only need to select length value, the easiest, be easily achieved;
Step 2 self-adapting multi-dimension morphological analysis
Primary signal is carried out self-adapting multi-dimension morphological analysis filtering, first to the discrete state signal gathered Carry out zero-mean process, find out all positive peak points in the discrete signal after zero-meanization processes, then root According to each trunnion axis position, positive peak point place, the interval of each adjacent positive peak point can be obtained, according to phase Minima and the maximum at adjacent positive peak interval determine structural element length dimension scope;Afterwards according to each just Vertical axes position, peak point place, corresponding with length large scale high according to the length little yardstick little yardstick of respective heights The computing of the policy setting of degree large scale, can be calculated the height that each length dimension of structural element is corresponding Yardstick, so that it is determined that each mesostructure element shape;Calculate the morphology of each mesostructure element the most respectively Filter result, and using the meansigma methods of each scale topographical filter result as the output result of AMMA method;
Step 3 iteration uses self-adapting multi-dimension morphological analysis
Last filter result use adaptive approach determine new structural element range scale, and with new true Fixed range scale carries out Multi-Scale Morphological Filtering to primary signal, and is iterated this process;By certainly Adaptive method is selected structural element range scale and signal is carried out multiple dimensioned differential filtering, and each filtering all can be right Noise contribution in signal filters, and extracts impact composition simultaneously;In so each iterative process, noise Peak value gradually decreases, and the structural element yardstick chosen is gradually increased, and poor morphology value filtering removes the ability of noise Gradually strengthening, the noise major part in signal is effectively filtered out by final poor morphology value filtering;The knot now chosen Constitutive element out to out tends towards stability, and adaptive process terminates;Between the adjacent positive peak that thus self adaptation is chosen Interval be more the interval between the adjacent positive peak of failure impact signal, be the most rational gap length; The structural element range scale obtained according to this gap length is more suitable, such that it is able to obtain more preferably Signal filtering result;
Step 4 terminates judging
Judge that the condition that iterative process terminates is: structural element yardstick maximum determined by adaptive approach is even Continuous three times identical;Judge that the new structural element yardstick every time using adaptive approach to determine in step 3 is maximum Whether value changes, if the new structural element yardstick maximum that determines of adaptive approach continuous three times identical, repeatedly Terminate for process;
Step 5 calculates iteration self-adapting Multiscale Morphological analysis result
Calculate the meansigma methods of differential filtering result under each yardstick, using this value as multiple dimensioned based on iteration self-adapting Morphological analysis filtering method processes the final result of signal;Under each yardstick, differential filtering result reflects letter respectively Characteristic information number under different scale, and the impact composition carrying machinery device fault information is present in each In individual yardstick, therefore the result sequence under each yardstick of gained is averaged as IAMMA method Eventually as a result, it is possible to more fully reflect the characteristic information of signal, and its impact composition prominent, effectively filter out Influence of noise.
CN201310672036.XA 2013-12-11 2013-12-11 A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological Expired - Fee Related CN103716012B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310672036.XA CN103716012B (en) 2013-12-11 2013-12-11 A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310672036.XA CN103716012B (en) 2013-12-11 2013-12-11 A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological

Publications (2)

Publication Number Publication Date
CN103716012A CN103716012A (en) 2014-04-09
CN103716012B true CN103716012B (en) 2016-08-17

Family

ID=50408673

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310672036.XA Expired - Fee Related CN103716012B (en) 2013-12-11 2013-12-11 A kind of filtering method analyzed based on iteration self-adapting Multiscale Morphological

Country Status (1)

Country Link
CN (1) CN103716012B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106500991B (en) * 2016-10-14 2019-03-26 石家庄铁道大学 Bearing fault signal characteristic extracting methods based on self-adapting multi-dimension AVG-Hat transformation

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1604140A (en) * 2004-11-11 2005-04-06 上海交通大学 Morphologic filter automatic destination detecting method
CN101320996A (en) * 2008-05-27 2008-12-10 中山大学 Self-adapting noise elimination apparatus and method
CN102998118A (en) * 2012-11-29 2013-03-27 西安交通大学 Bearing quantitative diagnosis method based on morphological filtering and complexity measure

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060281068A1 (en) * 2005-06-09 2006-12-14 Chemimage Corp. Cytological methods for detecting a disease condition such as malignancy by Raman spectroscopic imaging
US7804986B2 (en) * 2005-09-13 2010-09-28 Siemens Medical Solutions Usa, Inc. System and method for detecting intervertebral disc alignment using vertebrae segmentation

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1604140A (en) * 2004-11-11 2005-04-06 上海交通大学 Morphologic filter automatic destination detecting method
CN101320996A (en) * 2008-05-27 2008-12-10 中山大学 Self-adapting noise elimination apparatus and method
CN102998118A (en) * 2012-11-29 2013-03-27 西安交通大学 Bearing quantitative diagnosis method based on morphological filtering and complexity measure

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
自适应形态滤波与局域波分解理论及滚动轴承故障诊断;崔宝珍;《中国博士学位论文全文数据库 工程科技Ⅱ辑》;20130815;全文 *

Also Published As

Publication number Publication date
CN103716012A (en) 2014-04-09

Similar Documents

Publication Publication Date Title
Ray et al. Support vector machine based fault classification and location of a long transmission line
Wang Some further thoughts about spectral kurtosis, spectral L2/L1 norm, spectral smoothness index and spectral Gini index for characterizing repetitive transients
Jiang et al. Feature extraction method of wind turbine based on adaptive Morlet wavelet and SVD
Urbanek et al. Application of averaged instantaneous power spectrum for diagnostics of machinery operating under non-stationary operational conditions
Lei et al. Fault diagnosis of rotating machinery using an improved HHT based on EEMD and sensitive IMFs
Jiang et al. An improved EEMD with multiwavelet packet for rotating machinery multi-fault diagnosis
CN103439110B (en) Rolling bearing early-stage weak fault diagnostic method
Miao et al. Improvement of kurtosis-guided-grams via Gini index for bearing fault feature identification
CN103675444B (en) A kind of high-precision Time-Frequency Analysis Method
Gautam et al. Overview of mathematical morphology in power systems—A tutorial approach
Bai et al. Fault diagnosis of rotating machinery using an intelligent order tracking system
CN103678299B (en) A kind of method and device of monitor video summary
CN103559888B (en) Based on non-negative low-rank and the sound enhancement method of sparse matrix decomposition principle
CN102323518B (en) Method for identifying local discharge signal based on spectral kurtosis
Li et al. Diagonal slice spectrum assisted optimal scale morphological filter for rolling element bearing fault diagnosis
Braun The synchronous (time domain) average revisited
CN103698687B (en) In integrated circuit hardware Trojan horse detection signal processing method and system
CN103645249B (en) Online fault detection method for reduced set-based downsampling unbalance SVM (Support Vector Machine) transformer
Powell et al. Classification methods for noise transients in advanced gravitational-wave detectors II: performance tests on Advanced LIGO data
CN103116875B (en) Self-adaptation bilateral filtering image de-noising method
CN106796716B (en) For providing the device and method of super-resolution for low-resolution image
CN104299229B (en) Infrared weak and small target detection method based on time-space domain background suppression
Rui et al. Online wavelet denoising via a moving window
CN106443335A (en) Lightning stroke fault identification method and system
CN104777442A (en) MRS (magnetic resonance sounding) FID (frequency identity) signal noise inhibition method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20160817

Termination date: 20181211