CN103528836A - Rotary machine fault diagnosis method based on symbolic dynamics disable word mode - Google Patents
Rotary machine fault diagnosis method based on symbolic dynamics disable word mode Download PDFInfo
- Publication number
- CN103528836A CN103528836A CN201310432071.4A CN201310432071A CN103528836A CN 103528836 A CN103528836 A CN 103528836A CN 201310432071 A CN201310432071 A CN 201310432071A CN 103528836 A CN103528836 A CN 103528836A
- Authority
- CN
- China
- Prior art keywords
- parameter
- class
- sequence
- sample
- disable
- 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
Links
Images
Abstract
The invention discloses a rotary machine fault diagnosis method based on a symbolic dynamics disable word mode. The method is characterized by comprising the following steps: I, building time domain signal samples of equipment needing to be diagnosed in a normal state and various fault states; II, symbolizing each sample, namely, changing the samples into value domain sequences; III, changing the value domain sequences into symbolic sequences; IV, generating feature vectors through the symbolic sequences, wherein the feature vectors specifically include four parameter groups; V, standardizing and normalizing the feature vectors, and performing PCA (Principal Component Analysis) dimension reduction process; VI, generating new feature vectors after dimension reduction, and standardizing and normalizing the new vectors; VII, diagnosing by adopting hierarchical recursive LibSVM, acquiring a parameter g and a penalty parameter c by using a grid method, and performing modelling training, VIII, testing a test sample by using a model. The method can be widely applied to fault diagnosis of various rotary machines.
Description
Technical field
The present invention relates to method for diagnosing faults, be specifically related to prohibit based on symbolic dynamics the rotary machinery fault diagnosis method of word pattern.
Background technology
Current, rotating machinery has been widely applied in the fields such as Aero-Space, automobile boats and ships, metallurgical petrochemical industry, weapons equipment and energy generating.Comparatively typical rotating machinery as gear, rotary main shaft, bearing etc., the fault diagnosis of rotating machinery is the hot issue of engineering circles application always.Rotary machinery fault diagnosis process is generally: signals collecting, signal are processed and fault diagnosis.Fault diagnosis is actually a kind of mode identification procedure.
The method for diagnosing faults of rotating machinery has a lot.Generally be divided into based on kinetic model, based on dynamic signal analysis and the mode based on data-driven.Diagnostic method based on kinetic model, is generally to set up complete kinetics equation, by math equation, describes the dynamic characteristic of rotating machinery.Generally, kinetic model is not easy to set up very much, and while breaking down, its model can change; Dynamic signal analysis method is that current application obtains a kind of method a lot, and its method is that time domain or frequency-region signal are carried out to signal processing, extracts and wherein can characterize the special composition of defect or fault, by mode identification method, has judged whether fault; Method based on data-driven, is the various sample of signal based on obtaining, and adopts supervision or non-supervisory method, by extracting the feature of data the inside, sets up the relation between failure modalities and feature, the method for implementation pattern identification fault diagnosis.
The difficulty that current method for diagnosing faults exists is: one. be difficult to set up the kinetic model of accurate rotating machinery.The dynamic characteristic of fault is the fundamental way of carrying out fault diagnosis.Because system is crossed binding feature and inner complicated driving source, making machinery is that vibration coupling is comparatively outstanding in work; Meanwhile, face the fluctuation variability operating modes such as shock load, intermittent load during work, this has directly caused the non-stationary of response signal.These two factor interactions, making fault is that response signal has obvious nonlinear characteristic.This has caused rotating machinery essence difficulty in diagnostic techniques.Its two: be difficult to effectively overcome in row actual condition response signal signal to noise ratio (S/N ratio) low.The noise of the rotating machinery of actual condition is all more serious, and its characteristic frequency of rotating machinery is distributed in the more of low frequency medium frequency more, and these characteristic frequencies are as easy as rolling off a log by pollution from environmental noise.Its three: characteristic parameter need to be optimized with diagnostic model robustness to be needed to improve.Although extract at fault signature, have a lot of meaningful achievements in research aspect Fault Pattern Recognition, in some rotating machinery, its characteristic frequency is very complicated and be difficult to determine.In actual diagnosis, the optimization of characteristic parameter and the robustness of diagnostic model are the important paths that improves diagnosis and prediction accuracy, reduces reject rate and misclassification rate.
Rotating machinery is actually a complicated dynamical system, and its concrete detailed dynamical structure is difficult to understand.In rotating machinery normal operation, its dynamical system is stable, namely its kinetic parameter is should be stable within a stable scope, when machinery breaks down, its dynamic characteristic will change, different change directions, and different change degree, corresponding certain fault.Particularly importantly, when system breaks down, the Nonlinear Dynamical Characteristics of system will there will be, different Nonlinear Dynamical Characteristics corresponding to fault, if can characterize to a certain extent these Nonlinear Dynamical Characteristics, method by data-driven is set up fault model, just can the fine fault diagnosis of carrying out.Accordingly, this method proposes a kind of diagnostic method based on time-domain signal symbolic dynamics disable word pattern.
Symbolic dynamics is the Yi Ge branch of system dynamics, and eighties of last century is since the seventies, from pure research, be applied to engineering reality gradually in the middle of.The thought of symbolic dynamics is carried out " coarse grain " to time-domain signal sequence exactly and is changed processing, forms symbol sebolic addressing.In the situation that considering to embed peacekeeping delay parameter, system is simplified with abstract.According to Bants& The symbolism method of Pompe, time-domain signal can be decomposed into ergodic word pattern sequence.Conventionally the method for symbolization dynamics research can be divided into word pattern statistical information, ordinary word pattern and disable word pattern.The dynamic characteristic of symbol sebolic addressing energy retention system, can not necessarily meet embedding theorems, be easy to carry out data modeling processing, so symbolic dynamics is at aspects such as fault diagnosis and predictions, has started to be subject to people's concern and has carried out correlative study simultaneously.The method of word pattern statistical information is by research symbol sebolic addressing, to express the parameter of uncertainty, as entropy, complexity etc., carrys out the dynamic characteristic of analytic system.Nineteen ninety-five, the Werner Ebeling symbolization entropy of Germany is studied the statistical distribution information in a large amount of texts; The Mihajlo Grbovic of U.S. Temple university etc. 2012 propose a kind of distribution sensor method for diagnosing faults based on PCA and maximum entropy fuzzy decision; Argentine L.Zunino is by the random and chaotic characteristic of multiple dimensioned complexity-entropy plane discrimination system; The Shuen-De Wu of the state-run normal university in Taiwan in 2012 builds the characteristic parameter of diagnosis by multiple dimensioned sequence entropy, then utilize SVM to carry out the fault diagnosis of automobile bearing; Domestic aspect: acting like a bully etc. of Institute of Automation, CAS in 2005 proposed the division methods in a kind of self-adaptive symbol interval, and the diagnosing malfunction to motor thus; Within 2007, Xi'an hands over large Li Chongsheng etc. to utilize the information entropy of chaos and sequence number sequence to carry out rolling bearing fault diagnosis; The Chen Xiao equality of Jiangsu University has proposed the Yongyong He of the Fault Diagnosis of Roller Bearings Tsing-Hua University based on symbol entropy and SVM, waits 2012 and proposes approximate entropy as fault diagnosis nonlinear characteristic parameters; The Bing Yu of Nanjing space flight and aviation university proposed Wavelet Entropy as the characteristic parameter of sensor fault diagnosis in 2011; The Zheng Jinde of Hunan University etc. propose a kind of Fault Diagnosis of Roller Bearings based on multi-scale entropy, and by obtaining the Sample Entropy of the sequence of the difference embedding peacekeeping tolerance limit after coarse, recycling SVM carries out fault diagnosis; The Tang Youfu of Shanghai Communications University in 2012 etc. has proposed the compressor Representative Faults Diagnosis method of non linear complexity, by signal is carried out after binary symbol sequence processing, the LZC of the sequence of calculation, the characteristic interval of the LZC by different faults state is divided and is diagnosed; The multi-scale information entropy of the EMD decomposed signal of the extraction such as the Xue Tinggang of Xi'an University of Technology hydraulic turbine tail pipe, as the characteristic parameter of Fault Pattern Recognition, has been analyzed complicated pulsating pressure signal, etc.
Disable word pattern, refers to the word pattern that does not occur in the ergodic process of symbol sebolic addressing or seldom occur.Research based on disable word pattern, is mainly used in system features identification at present, and research is not still deeply.2000, had literature research related notion and the expression method of disable word in symbolic dynamics; 2007, the distribution situation of true, pseudo-disable word pattern in definite stochastic system that had literature research; The application of disable word pattern in stock market that had literature research, has studied feature and the impact on stock behavior of system disable word pattern under different delay; 2008, the distribution situation of disable word pattern in a large amount of daylight saving time order sequenced data that had document analysis; 2012, the people such as Osvaldo A.Rosso of Brazil studied in entropy-complexity plane, have studied the sensitivity of disable word pattern to sequence length and noise.Up to the present, there is not yet and adopt the method for disable word pattern to carry out the document of fault diagnosis.
The characteristic present of the disable word pattern during symbol is ergodic the non-linear dynamic feature of system.Therefore,, while adopting disable word model to there is the fault of different Nonlinear Dynamical Characteristics as research, there is natural undeniable advantage.From in essence, when system breaks down, its dynamic behavior will change, and nonlinear characteristic will occur or generation state changes, and now the disable word model of sequence number sequence also changes thereupon; From the sensitivity to noise, when the symbol sebolic addressing length of analyzing meets certain condition, disable word pattern is to Gaussian noise, coloured noise and f
-λnoise pollution is immunity completely; Even length is inadequate, when noise exists certain level, noise also can be processed by statistical study the impact of disable word pattern, eliminates the impact of noise; From operability or algorithm complex, the quantity of disable word pattern is far smaller than ordinary pattern, therefore adopt the mode of data-driven to study the relation between disable word pattern and fault, carry out characteristic parameter optimization, it is practicable building sane diagnostic model.
Summary of the invention
Problem to be solved by this invention is to provide the rotary machinery fault diagnosis method of prohibiting word pattern based on symbolic dynamics.
In order to solve the problems of the technologies described above, first technical scheme of the present invention is that the rotary machinery fault diagnosis method based on symbolic dynamics taboo word pattern, is characterized in that: comprise the steps:
The first step: set up and need the time-domain signal sample of diagnostic device under normal condition and various malfunction to be respectively no less than 30, the sampling length of each sample is not less than 4096 points;
Second step: to each sample, carry out symbolism processing, be about to time domain sequences and become codomain sequence samples; Be about to each sample and be divided into equably M region by its peak-to-peak value scope, M is 6; Each point on each sample is carried out to region and choose, the region at the value place on sample is the value in codomain sequence;
Thus, time domain sequences becomes codomain sequence; M: signal coarse degree discrimination; Namely between signal codomain peak-to-peak value, the interval number of equidistant segmentation; Coarse degree is different, also different to the simulation degree of original signal; If too little, some signal detail will be submerged, if too large, coarse has little significance; According to the previous experiments research of related data and this method, think that the situation of M=6 is applicable to the signal coarse process in rotary machinery fault diagnosis;
The 3rd step, symbol sebolic addressing form:
The formation of symbol sebolic addressing is by delay parameter τ and go through length I and determine, point from sampling instant minimum in the codomain sequence obtaining in second step is started at, before the sampled point of sample is got according to sampling instant order from small to large, (L-I* τ+1) individual point is respectively as start point signal, start point signal is added to the starting point time delay τ with signal, 2 τ, 3 τ ... (I-1) point that τ is corresponding forms respectively one group of sequence, and distinguish time delay τ by start point signal in sequence and with start point signal, 2 τ, 3 τ ... (I-1) point that τ is corresponding carries out symbol definition successively, again according to time-domain signal amplitude putting in order from big to small, symbol sebolic addressing set or word set, wherein
L: the length of sample of signal;
τ: delay parameter, τ gets 1 to 6 natural number; τ was 1 one cycle of expression time delay, and last starting point and a rear starting point are adjacent sampled point, and τ is 2 expression time delay two cycles, the sampled point that has been last starting point and a rear starting point interval, the like, τ was 6 six cycles of expression time delay, last starting point and a rear starting point interval five sampled points;
Go through length I and get 6;
The 4th step: generating feature vector, comprising:
The 4th. a step: generate parameter group one, be specially and calculate total taboo word rate f
rinformation entropy S with symbol sebolic addressing
h; Wherein:
Wherein:
Total disable word rate:
N
ffor all disable word pattern-word quantity that occur;
Mt is word pattern sum; M
t=M unequal to 720;
The 4th. two steps: generate parameter group two, be specially:
For in whole word pattern sum, be divided in order 30 groups, each group has 24 patterns, to every class symbol sequence, adds up respectively in 24 patterns the number N that disable word pattern occurs
i; The N that each group the inside is occurred
idivided by total disable word quantity, obtain 30 disable word rate f
1, f
2..., f
30;
Parameter group two is actually and represents that disable word pattern is in a kind of distribution of certain scale;
The 4th. three steps: generate parameter group three, be specially:
In whole word pattern sum, be divided in order six groups, each group has 120 patterns, to every class symbol sequence, adds up respectively the disable word pattern count N occurring 120 pattern the insides
i', then divided by total number of words N that forbids
f, obtain six disable word rate f
b1, f
b2..., f
b6; Wherein:
In fact parameter group three represents a kind of distribution of disable word pattern on larger scale;
The 4th. four steps: generate parameter group four, be specially:
Every class symbol sequence is divided into two sections of front and back, and every section consists of three characters; According to the big or small position relationship of these three characters, namely every three characters have four kinds of position relationships, and front and back are two sections so, have 16 kinds of patterns, in whole word pattern sum the inside, to every class symbol sequence, add up respectively the disable word pattern quantity N occurring under 16 kinds of patterns
ci, then divided by total disable word pattern quantity, form disable word rate f
ci;
In word set, M
t, also referred to as dictionary table.Each sample of signal is generated to the word pattern set obtaining, in length L, statistics does not appear at the pattern of mod table the inside, or the pattern occurring with very low probability, is called disable word pattern.Disable word pattern is not only relevant with the dynamic characteristic of signal, also relevant with the noise signal being loaded on signal.But no matter be white noise, pink noise or f
-λnoise, is loaded on signal, at L, reaches after certain length, and to the quantity of the disable word pattern of signal, distribution influence, not large.Therefore, adopt after the method for disable word pattern, support antimierophonic very capable.
The 5th step, proper vector is carried out after standardization and normalized, and carry out PCA dimension-reduction treatment;
New proper vector after the 6th step, generation dimensionality reduction, and new proper vector is carried out to standardization and normalized;
The LibSVM of the 7th step, employing layering recurrence diagnoses, and kernel function adopts RBF, utilizes gridding method to obtain RBF nuclear parameter g parameter and the penalty parameter c parameter of LibSVM, and trains modeling;
The 8th step: utilize model to test test sample book, judge whether every class accuracy is greater than 90%; If every class accuracy is to be greater than 90%, judged whether merging class; If do not merge class, model construction is complete, can carry out diagnostic work; If any merging class, π=X is set, τ=1, M=6, returns to second step; If there is accuracy to be less than 90%, judgement τ < 6? as τ < 6, establish τ=τ+1, return to second step; As τ=6, then judgement has an accuracy > 90% classification, as do not have, this is diagnosed unsuccessfully, cannot modeling, can not diagnose; As there is accuracy τ=6 > 90% classification, in τ=1, o'clock observe, the class of mutually obscuring that does not reach accuracy 90% is merged into a class, and M=6 is set, π=π-X+1, τ=1, returns to second step; Wherein: π is identifying object, N is total fault category number to be sorted; X is the classification number of the class of merging.
The beneficial effect of the rotary machinery fault diagnosis method based on symbolic dynamics taboo word pattern of the present invention is: the disable word pattern of symbolization dynamics of the present invention the inside, adopting multiple disable word pattern statistical parameter is proper vector parameter, utilize layering recurrence LibSvm method, carry out multistage fault diagnosis, can be the in the situation that of strong noise pollution, mechanical fault is identified, and recognition correct rate is higher, can be widely used in rotary machinery fault diagnosis.
Accompanying drawing explanation
Fig. 1 is the rotary machinery fault diagnosis method process flow diagram of prohibiting word pattern based on symbolic dynamics of the present invention.
Fig. 2 is second and third step exemplary plot of rotary machinery fault diagnosis method of prohibiting word pattern based on symbolic dynamics of the present invention.
Fig. 3 (a) is the wherein a kind of position relationship that generates three characters in parameter group four.
Fig. 3 (b) is the wherein the second position relationship that generates three characters in parameter group four.
Fig. 3 (c) is wherein the third position relationship that generates three characters in parameter group four.
Fig. 3 (d) is wherein the 4th kind of position relationship that generates three characters in parameter group four.
Embodiment
Referring to Fig. 1, based on symbolic dynamics, prohibit the rotary machinery fault diagnosis method of word pattern, carry out as follows:
The first step: set up and need the time-domain signal sample of diagnostic device under normal condition and various malfunction to be respectively no less than 30, the sampling length L of each sample is not less than 4096 points;
Second step, to each sample, carry out symbolism processing, be about to sampled point and become codomain sequence; Be about to each sample and be divided into equably M region by its peak-to-peak value scope, M is 6; Each point on each sample is carried out to region and choose, the region at the value place on sample is the value in symbol sebolic addressing; Thus, time domain sequences becomes codomain sequence; M: signal coarse degree discrimination; Namely between signal codomain peak-to-peak value, the interval number of equidistant segmentation;
Specifically, referring to Fig. 2, get M=6, between peak-to-peak value, divided 6 equal altitudes intervals; The codomain sequence of 11 sampled points in time domain waveform foremost is 15462534661;
The 3rd step: codomain sequence is become to symbol sebolic addressing:
The formation of symbol sebolic addressing is by delay parameter τ and go through length I and determine; Delay parameter τ is not more than 6 natural number; Before the sampled point of sample is got according to sampling instant order from small to large, (L-I* τ+1) individual point is respectively as start point signal; Start point signal is added to starting point time delay τ, 2 τ, 3 τ with signal ... (I-1) point that τ is corresponding forms respectively one group of sequence, and distinguish time delay τ, 2 τ, 3 τ by start point signal in sequence and with start point signal ... (I-1) point that τ is corresponding carries out symbol definition successively, again according to time-domain signal amplitude putting in order from big to small, symbol sebolic addressing set or word set; Wherein, τ: delay parameter, τ gets 1 to 6 natural number; τ was 1 one cycle of expression time delay, and last starting point and a rear starting point are adjacent sampled point, and τ is 2 expression time delay two cycles, the sampled point that has been last starting point and a rear starting point interval, the like, τ was 6 six cycles of expression time delay, last starting point and a rear starting point interval five sampled points;
Go through length I and get 6 points;
In the specific implementation, definable is gone through length I=M, in the codomain sequence that is about to obtain in second step minimum sampling instant corresponding o'clock as the starting point of first group of sequence, the starting point of first group of sequence is added to this starting point time delay τ, 2 τ, 3 τ ... (I-1) point that τ is corresponding forms first group of sequence, again using the starting point time delay τ with first group of sequence corresponding o'clock as the starting point of second group of sequence, the starting point of the second class symbol sequence is added to starting point time delay τ, 2 τ, 3 τ with the second class symbol sequence ... (I-1) point that τ is corresponding forms second group of sequence, the like, finally using the starting point time delay with first group of sequence (L-I* τ+1) τ corresponding o'clock as (L-I* τ+1) group sequence starting point, the starting point of (L-I* τ+1) class symbol sequence is added to the starting point time delay τ of (L-I* τ+1) class symbol sequence, 2 τ, 3 τ ... (I-1) point that τ is corresponding forms (L-I* τ+1) group sequence, and distinguish time delay τ by start point signal in sequence and with start point signal, 2 τ, 3 τ ... (I-1) point that τ is corresponding carries out symbol definition successively, again according to time-domain signal amplitude putting in order from big to small, final (L-I* τ+1) the individual symbol sebolic addressing that forms, be that symbol sebolic addressing set or word set are { s
i,
Specifically referring to Fig. 2, totally 11 points, the delay parameter between 2 of front and back is τ; Length I=M=6 is gone through in definition; τ=1; The t1 of usining corresponding o'clock as the starting point of first group of sequence, six points that t1 and t2 to t6 are corresponding form first group of sequence, and the codomain of these six points is 154625; T1 to t6 is set to " 1 ", " 2 ", " 3 ", " 4 ", " 5 ", " 6 "; Also can be set to other symbols; And according to time-domain signal amplitude putting in order from big to small, obtain the first class symbol sequence s
1=426321; In like manner, by t
2the corresponding some starting point as second group of sequence, six points that t2 and t3 to t7 are corresponding form second group of sequence, and the codomain of these six points is 546253; T2 to t7 is set to " 1 ", " 2 ", " 3 ", " 4 ", " 5 ", " 6 "; And according to time-domain signal amplitude putting in order from big to small, obtain the second class symbol sequence S2=315264; The like, the codomain of the 3rd group of sequence is 462534, obtains the 3rd class symbol sequence S3=241653; The codomain of the 4th group of sequence is 625346, obtains the 4th class symbol sequence S4=163542; After be not repeated.In Fig. 2, as τ=2, should form first group of sequence with t1, t3, t5, t7, point that t9, t11 are corresponding; Second group of sequence should be take t3 as starting point is with t3, t5, t7, t9 ... Deng six corresponding points, form second group of sequence, remaining is tired stating not.The 4th step, generating feature vector, comprising:
The 4th. a step, generation parameter group one: parameter group one is considered the overall objective of symbol sebolic addressing, and one has two, is respectively total taboo word rate f
rinformation entropy S with sequence
h:
Total disable word rate:
N
ffor all disable word pattern-word quantity that occur;
Mt is word pattern sum; M
t=M unequal to 720;
S
hinformation entropy for symbol sebolic addressing:
The probability that wherein p (π) occurs for sampled point; π ∈ ∏ (L) is the arrangement of certain word arbitrarily;
S
halgorithm consistent with general Shannon entropy algorithm;
The 4th. two steps, generation parameter group two:
Parameter group two is considered, in whole word pattern sum, to be divided in order 30 groups, and each group has m
t1=M
t/ 30=240 pattern; To every class symbol sequence, add up respectively in 240 patterns the number N that disable word pattern occurs
i; The N that every group of the inside occurred
idivided by total disable word quantity, obtain 30 disable word rate f
1, f
2..., f
30;
Parameter group two is actually and represents that disable word pattern is in a kind of distribution of certain scale;
The 4th. three steps, generation parameter group three: parameter group three is considered, in whole word pattern sum, to be divided in order six groups, and each group has m
t2=M
t/ 6=120 pattern, to every class symbol sequence, adds up respectively the disable word pattern count N occurring in 120 patterns
i', then divided by total number of words N that forbids
f, obtain six disable word rate f
b1, f
b2..., f
b6;
In fact parameter group three represents a kind of distribution of disable word pattern on larger scale;
The 4th. four steps, generation parameter group four: parameter group four considers that, in whole dictionary table, the morphology of the disable word pattern of appearance distributes; Every class symbol sequence is divided into two sections of front and back, and every section consists of three characters; According to the big or small position relationship of these three characters, namely every three characters have four kinds of position relationships, and front and back are two sections so, have 16 kinds of patterns,
In whole word pattern sum the inside, to every class symbol sequence, add up respectively the disable word pattern quantity N occurring under 16 kinds of patterns
ci, then divided by total disable word pattern quantity, form disable word rate f
ci;
The morphology of disable word pattern characterizes i.e. four kinds of position relationships of three characters and sees that Fig. 3 (a) is to Fig. 3 (d);
The design of above-mentioned four groups of parameters is that the disable word mode profile, morphology of integral power characteristic from sequence, different resolution level, three aspects of disable word mode profile are investigated.But in fact between these characteristic parameters, have some coupling, therefore, need to carry out dimension-reduction treatment to characteristic parameter, the method for processing is PCA method.After dimensionality reduction, the mapping according to proper vector in new dimension space, obtains new proper vector, adopts SVM method to carry out model training, finally obtains training pattern.Utilize this model just can carry out the diagnosis of new samples.
The 5th step, proper vector is carried out after standardization and normalized, and carry out PCA dimension-reduction treatment; The vector parameter dimension of certain signal of sign that above-mentioned four parameter group obtain is 54.These parameters, from Global Information, segmentation disable word pattern information, 3 aspects of disable word morphology of signal, have been described the Nonlinear Dynamical Characteristics of system comprehensively.But between these features, more or less exist coupling.This method adopts the method for principal component analysis (PCA) PCA to carry out feature vectors dimensional down.PCA is the universal method of a standard, and in PCA method, total contribution rate of take is more than or equal to 85% as standard, generates the proper vector of new dimension space.Before carrying out PCA, because parameters dimension is different, this 54 dimension parameter is carried out to standardization, normalized, after processing, all parameters are all between-1 to 1.
New proper vector after the 6th step, generation dimensionality reduction, and new proper vector is carried out to standardization and normalized;
The 7th step, adopt multi-level SVM recognition methods, obtain the parameter of LibSVM, be mainly nearest RBF nuclear parameter g parameter and penalty parameter c parameter, and train modeling;
In training process, for given normal condition, fault 1, fault 2 ... fault N, has altogether N+1 pattern.Correspondence has m*(N+1) group training sample, have n*(N+1) group test sample book.This method requires number of training to be respectively no less than 30.
The major parameter that LibSVM carries out fault diagnosis and pattern-recognition is set to:
SVM type selecting: select C-SVM;
Kernel function type: select RBF kernel function;
Degree parameter in kernel function: the default parameters 3 that is set to SVM;
Penalty parameter c and recently RBF nuclear parameter g: the scope that arranges is for [2^ (10), 2^ (10)];
CV process: CV process is default value 3 foldings.The stepping amount of c parameter and g parameter is 0.5
Due to the fault diagnosis of the applicable small sample of this method, adopt the method for trellis traversal to obtain optimum parameter c and g;
Cross-verification model n-fold: be set to 3;
According to the penalty parameter c obtaining and nearest RBF nuclear parameter g, to m*(N+1) group training sample trains, and obtains model.
In rotary machinery fault diagnosis process, the phenomenon of a lot of faults is like comparing class, as rotor unbalance with misalign; Inner ring fault in rolling bearing and normal condition etc.Especially these faults are at the fault initial stage, and the Nonlinear Dynamical Characteristics showing is not clearly.Therefore, although according to this method, a plurality of parameters of disable word pattern have been extracted, and carried out PCA Dimensionality reduction, in our experiment, find, if directly adopt the multiclass method of many minutes to carry out SVM training and identification, when it is tested test sample book, usually be difficult to the recognition result that reaches satisfied, if discrimination is not higher than 90% etc.Therefore,, as a kind of general method, need to adopt layering recurrence SVM recognition methods.
The 8th step, utilize model to test test sample book, judge whether every class accuracy is greater than 90%; If every class accuracy is to be greater than 90%, judgement has the class of merging; If do not merge class, model construction is complete, can carry out diagnostic work; If any merging class, π=X is set, τ=1, M=6, returns to second step; If there is accuracy to be less than 90%, judgement τ < 6? as τ < 6, establish τ=τ+1, return to second step; As τ=6, then judgement has an accuracy > 90% classification, as do not have, this is diagnosed unsuccessfully, cannot modeling, can not diagnose; As there is accuracy τ=6 > 90% classification, in τ=1, o'clock observe, the class of mutually obscuring that does not reach accuracy 90% is merged into a class, and M=6 is set, τ=π-X+1, τ=1, returns to second step; Wherein: π is identifying object, N is total fault category number to be sorted; τ is that time delay is controlled parameter; X is the classification number of the class of merging.
This method is actually a kind of modeling pattern of recursion by different level, utilizes disable word pattern to extract proper vector, utilizes the LibSVM modeling method of layering recurrence, finally can realize modeling and diagnosis under complex object.
Below, utilize the rolling bearing fault data (http://csegroups.case.edu/bearingdatacenter/home) in U.S.'s Case Western Reserve University electrical engineering laboratory to carry out checking and the test of said new method.
In its Mishap Database the inside, measuring and calculation object: motor Hp1, FE end, rpm:1797; Data n ormal_0(normal condition), inner ring fault (IR007_1.mat), outer ring fault (B007_1.mat) and rolling body fault (OR007@3_1.mat).Totally four kinds of sample modes.Be that total identifying object is π=4, comprise three kinds of malfunctions and a kind of normal condition; Because these sample data amounts are very large, the sample of single type has 120,000 multiple spots.Adopting the first half data is training sample, and later half is as test sample book.
Concrete diagnostic procedure is as follows:
The first step: training sample and test sample book are respectively in 60,000 data points, and first three ten thousand as training sample region, rear 30,000 as volumes sample areas; In region separately, generate at random starting point, get continuous 4096 points as sample; Training and testing sample number is 40 of every kind of patterns;
Second step: adopt M=6, all test sample books and training sample are carried out to symbolism processing, become codomain sequence samples;
The 3rd step: adopt τ=1, go through length I=M; Codomain sequence samples is formed to symbol sebolic addressing set;
The 4th step: 320 sampling feature vectors that symbol sebolic addressing set generated to 54 dimensions; Comprise:
Generate parameter group one: parameter group one is respectively total taboo word rate fr and the information entropy Sh of sequence;
Wherein:
S
halgorithm consistent with general Shannon entropy algorithm;
Generate parameter group two: in whole word pattern sum, be divided in order 30 groups, each group has 24 patterns, to every class symbol sequence, add up respectively in 24 patterns, the number N that disable word pattern occurs
i; The N that every group of the inside occurred
idivided by total disable word quantity, obtain 30 disable word rate f
1, f
2..., f
30;
Generate parameter group three: in whole word pattern sum, be divided in order six groups, each group has 120 patterns, to every class symbol sequence, add up respectively the disable word pattern count N occurring 120 pattern the insides
i', then divided by total number of words N that forbids
f, obtain six disable word rate f
b1, f
b2..., f
b6; Wherein:
Generate parameter group four: every class symbol sequence is divided into two sections of front and back, and every section consists of three characters; According to the big or small position relationship of these three characters, namely every three characters have four kinds of position relationships, and front and back are two sections so, have 16 kinds of patterns, in whole word pattern sum the inside, to every class symbol sequence, add up respectively the disable word pattern quantity N occurring under 16 kinds of patterns
ci, then divided by total disable word pattern quantity, form disable word rate f
ci;
The 5th step: these 320 sampling feature vectors are carried out to standardization and normalized;
The 6th step: adopt PCA method to carry out dimensionality reduction to 320 samples, contribution rate is set to 85%.Now obtaining new dimension is 26; According to mapping relations, generate the sampling feature vectors of 320 26 dimensions;
The 7th step: according to the method to set up of LibSVM parameter, obtain RBF nuclear parameter g and penalty parameter c in the parameter of LibSVM, generated thus training pattern model1;
The 8th step: according to model1, test sample book is tested, test accuracy result is: normal 87%, inner ring 80%, outer ring 91%, rolling body 93%;
Adjust time control parameter τ=2, until 6, test accuracy table is table 1:
Table 1: the accuracy table of four classes under different time control parameter values
From table 1, find, cannot reach desirable diagnostic result.The analysis found that, normal and inner ring fault is can discrimination not high, namely the most mutually obscures.
The 9th step: normal and inner ring fault are merged into a class, separately the sample after 20 training samples of random choose and symbolism serializing of test sample book.Now, overall number of categories is 3.
Generate 240 54 dimensional feature vectors, standardization normalization, PCA dimensionality reduction.Now dimensionality reduction drops to 30 dimensions, regeneration 30 dimensional feature vectors, standardization normalization.Generate c and g parameter.Generated thus training pattern model2; 3 class objects are tested, and result is as table 2:
The accuracy table of table 2 three classes under different time control parameter values
| τ | Normally+inner ring fault (%) | Outer ring fault (%) | Rolling body fault (%) |
| τ＝1 | 91 | 95 | 94 |
From table 2, find, reached requirement.Success is three classes separately.Now merge class and be 2. and will distinguish normal and inner ring fault.
The 9th step: select respectively 30, sample after normal and inner ring fault symbolism serializing, generate the proper vector of 54 dimension parameters.Standardization, normalization.PCA dimensionality reduction, this stylish dimension is 33, generates the proper vector under new dimension; Training generates parameter c and the g that LibSVM needs, generation model model3.Test sample book is tested, and test result is in Table 3:
The accuracy table of table 3 two classes under different time control parameter values
Diagnosis is found, is got τ=2 and just can reach requirement.
Note: the recognition correct rate of table 3 be actually three classes normal+inner ring recognition correct rate a1, be multiplied by two classes identification LibSVM recognition correct rate a2.A1.a2. namely
Thus, whole diagnostic model can with have two, be respectively in training pattern model2 and model model3 in the model of τ=2 o'clock; To any one new test sample book, first τ=1 is set, according to model2, judgement is normal+inner ring fault, or outer ring fault or rolling body fault.If belong to outer ring fault and rolling body fault, directly can obtain diagnostic result; If normal+inner ring fault, arranges τ=2, again obtain proper vector, bring model3 into and differentiate, just can know normal or inner ring fault.According to this thinking, total diagnostic result accuracy is 95.2%.
In order to verify the noise resisting ability of this method, we respectively to every kind of test sample book add signal to noise ratio (S/N ratio) be 5,2,1,0.1,0.01 and 0.001 white Gaussian noise in Table 4, and add that pink noise is in Table 5.And training sample remains the sample that does not add any noise, the overall accuracy of identification following (doing the mean value after identification 10 times)
Table 4:
| Signal to noise ratio (S/N ratio) 5 | 2 | 1 | 0.1 | 0.01 | 0.001 |
| 100% | 100% | 100% | 95.2% | 93.4% | 86.1% |
Table 5:
| Signal to noise ratio (S/N ratio) 5 | 2 | 1 | 0.1 | 0.01 | 0.001 |
| 96.1% | 95.2% | 95.1% | 95.2% | 93.4% | 86.1% |
Between signal to noise ratio (S/N ratio) 5 to 0.1, little on identification impact.After signal to noise ratio (S/N ratio) 0.001, identification stability decreases.
Therefore, this method shows, the disable word pattern of symbolization dynamics the inside, and adopting multiple disable word pattern statistical parameter is proper vector parameter, utilizes layering recurrence LibSvm method, carries out multistage fault diagnosis.Can, the in the situation that of strong noise pollution, mechanical fault be identified.Recognition correct rate is higher.
Above the specific embodiment of the present invention is described, still, the scope that is not limited only to embodiment of the present invention's protection.
Claims (1)
1. based on symbolic dynamics, prohibit the rotary machinery fault diagnosis method of word pattern, it is characterized in that: comprise the steps:
The first step: set up and need the time-domain signal sample of diagnostic device under normal condition and various malfunction to be respectively no less than 30, the sampling length L of each sample is not less than 4096 points;
Second step: to each sample, carry out symbolism processing, be about to time domain sequences and become codomain sequence samples; Be about to each sample and be divided into equably M region by its peak-to-peak value scope, M is 6; Each point on each sample is carried out to region and choose, the region at the value place on sample is the value in codomain sequence;
The 3rd step: codomain sequence is become to symbol sebolic addressing:
The formation of symbol sebolic addressing is by delay parameter τ and go through length I and determine; Before the sampled point of sample is got according to sampling instant order from small to large, (L-I* τ+1) individual point is respectively as start point signal, start point signal is added to starting point time delay τ, 2 τ, 3 τ with signal ... (I-1) point that τ is corresponding forms respectively one group of sequence, and distinguish time delay τ, 2 τ, 3 τ by start point signal in sequence and with start point signal ... (I-1) point that τ is corresponding carries out symbol definition successively, according to time-domain signal amplitude putting in order from big to small, obtain symbol sebolic addressing set or word set again; Wherein:
L: the length of sample of signal;
τ: delay parameter, τ gets 1 to 6 natural number;
The 4th step: generating feature vector, comprising:
The 4th. a step: generate parameter group one, be specially and calculate total taboo word rate f
rinformation entropy S with symbol sebolic addressing
h; Wherein:
Wherein:
Total disable word rate:
N
ffor all disable word pattern-word quantity that occur;
Mt is word pattern sum; M
t=M unequal to 720;
The 4th. two steps: generate parameter group two, be specially:
In whole word pattern sum, be divided in order 30 groups, each group has 24 patterns, to every class symbol sequence, adds up respectively in 24 patterns the number N that disable word pattern occurs
i; The N that every group of the inside occurred
idivided by total disable word quantity, obtain 30 disable word rate f
1, f
2..., f
30;
The 4th. three steps: generate parameter group three, be specially:
In whole word pattern sum, be divided in order six groups, each group has 120 patterns, to every class symbol sequence, adds up respectively the disable word pattern count N occurring 120 pattern the insides
i', then divided by total number of words N that forbids
f, obtain six disable word rate f
b1, f
b2..., f
b6; Wherein:
The 4th. four steps: generate parameter group four, be specially:
Every class symbol sequence is divided into two sections of front and back, and every section consists of three characters; According to the big or small position relationship of these three characters, namely every three characters have four kinds of position relationships, and front and back are two sections so, have 16 kinds of patterns, in whole word pattern sum the inside, to every class symbol sequence, add up respectively the disable word pattern quantity N occurring under 16 kinds of patterns
ci, then divided by total disable word pattern quantity, form disable word rate f
ci;
The 5th step: proper vector is carried out after standardization and normalized, then carry out PCA dimension-reduction treatment;
The 6th step: generate the new proper vector after dimensionality reduction, and new proper vector is carried out to standardization and normalized;
The LibSVM of the 7th step, employing layering recurrence diagnoses, and kernel function adopts RBF, utilizes gridding method to obtain RBF nuclear parameter g parameter and the penalty parameter c parameter of LibSVM, and trains modeling;
The 8th step: utilize model to test test sample book, judge whether every class accuracy is greater than 90%; If every class accuracy is to be greater than 90%, judgement has the class of merging; Judge whether the merging class of setting; If do not merge class, model construction is complete, can carry out diagnostic work; If any merging class, τ=X is set, τ=1, M=6, returns to second step; If there is accuracy to be less than 90%, judgement τ < 6? as τ < 6, establish τ=τ+1, return to second step; As τ=6, then judge whether accuracy > 90% classification, as do not have, this is diagnosed unsuccessfully, cannot modeling, can not diagnose; As there is accuracy τ=6 > 90% classification, in τ=1, o'clock observe, the class of mutually obscuring that does not reach accuracy 90% is merged into a class, and M=6 is set, π=π-X+1, τ=1, returns to second step; Wherein: π is identifying object, N is total fault category number to be sorted; X is the classification number of the class of merging.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310432071.4A CN103528836B (en) | 2013-09-22 | 2013-09-22 | The rotary machinery fault diagnosis method of word pattern is prohibited based on symbolic dynamics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310432071.4A CN103528836B (en) | 2013-09-22 | 2013-09-22 | The rotary machinery fault diagnosis method of word pattern is prohibited based on symbolic dynamics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103528836A true CN103528836A (en) | 2014-01-22 |
| CN103528836B CN103528836B (en) | 2016-04-20 |
Family
ID=49931028
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310432071.4A Expired - Fee Related CN103528836B (en) | 2013-09-22 | 2013-09-22 | The rotary machinery fault diagnosis method of word pattern is prohibited based on symbolic dynamics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103528836B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104863840A (en) * | 2015-03-16 | 2015-08-26 | 北京化工大学 | Reciprocating compressor intelligent diagnosis method based on EMD-PCA |
| CN104077231B (en) * | 2014-07-16 | 2017-02-08 | 国家电网公司 | Transformer maintenance optimization method based on symbol dynamics and LS-SVM |
| CN107991097A (en) * | 2017-11-16 | 2018-05-04 | 西北工业大学 | A kind of Method for Bearing Fault Diagnosis based on multiple dimensioned symbolic dynamics entropy |
| CN109034028A (en) * | 2018-07-16 | 2018-12-18 | 河北工业大学 | A kind of mechanical equipment fault feature extracting method based on symbolism and TF-IDF |
| CN109088770A (en) * | 2018-08-21 | 2018-12-25 | 西安交通大学 | A kind of Mechatronic Systems Internet modeling method based on self-adaptive symbol transfer entropy |
| CN112666918A (en) * | 2020-12-01 | 2021-04-16 | 沈阳化工大学 | Method for self-adaptive industrial process fault detection based on online compressed KECA |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH06273286A (en) * | 1993-03-18 | 1994-09-30 | Hitachi Ltd | Process state diagnosing system |
| US20030014692A1 (en) * | 2001-03-08 | 2003-01-16 | California Institute Of Technology | Exception analysis for multimissions |
| CN101201295A (en) * | 2006-12-13 | 2008-06-18 | 上海海事大学 | Method and device for predicting grey failure of rotating machinery wavelet |
| 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 |
-
2013
- 2013-09-22 CN CN201310432071.4A patent/CN103528836B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH06273286A (en) * | 1993-03-18 | 1994-09-30 | Hitachi Ltd | Process state diagnosing system |
| US20030014692A1 (en) * | 2001-03-08 | 2003-01-16 | California Institute Of Technology | Exception analysis for multimissions |
| CN101201295A (en) * | 2006-12-13 | 2008-06-18 | 上海海事大学 | Method and device for predicting grey failure of rotating machinery wavelet |
| 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 (3)
| Title |
|---|
| 王松岭等: "基于符号动力学信息熵与改进神经网络的风机故障诊断研究", 《华北电力大学学报》 * |
| 陈晓平等: "居于符号熵与支持向量机的滚动轴承故障诊断", 《中国机械工程》 * |
| 马东玲等: "基于符号化时间序列分析的电动机轴承故障诊断", 《微电机》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077231B (en) * | 2014-07-16 | 2017-02-08 | 国家电网公司 | Transformer maintenance optimization method based on symbol dynamics and LS-SVM |
| CN104863840A (en) * | 2015-03-16 | 2015-08-26 | 北京化工大学 | Reciprocating compressor intelligent diagnosis method based on EMD-PCA |
| CN107991097A (en) * | 2017-11-16 | 2018-05-04 | 西北工业大学 | A kind of Method for Bearing Fault Diagnosis based on multiple dimensioned symbolic dynamics entropy |
| CN109034028A (en) * | 2018-07-16 | 2018-12-18 | 河北工业大学 | A kind of mechanical equipment fault feature extracting method based on symbolism and TF-IDF |
| CN109034028B (en) * | 2018-07-16 | 2021-09-24 | 河北工业大学 | Symbolic and TF-IDF-based mechanical equipment fault feature extraction method |
| CN109088770A (en) * | 2018-08-21 | 2018-12-25 | 西安交通大学 | A kind of Mechatronic Systems Internet modeling method based on self-adaptive symbol transfer entropy |
| CN109088770B (en) * | 2018-08-21 | 2020-03-31 | 西安交通大学 | Electromechanical system interactive network modeling method based on self-adaptive symbol transfer entropy |
| CN112666918A (en) * | 2020-12-01 | 2021-04-16 | 沈阳化工大学 | Method for self-adaptive industrial process fault detection based on online compressed KECA |
| CN112666918B (en) * | 2020-12-01 | 2022-06-14 | 沈阳化工大学 | Method for self-adaptive industrial process fault detection based on online compressed KECA |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103528836B (en) | 2016-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Qiao et al. | Deep convolutional and LSTM recurrent neural networks for rolling bearing fault diagnosis under strong noises and variable loads | |
| CN103528836B (en) | The rotary machinery fault diagnosis method of word pattern is prohibited based on symbolic dynamics | |
| Wu et al. | Fault-attention generative probabilistic adversarial autoencoder for machine anomaly detection | |
| Yu et al. | Weak fault feature extraction of rolling bearings using local mean decomposition-based multilayer hybrid denoising | |
| Cui et al. | Improved fault size estimation method for rolling element bearings based on concatenation dictionary | |
| CN109187021B (en) | Multi-source Wind turbines Method for Bearing Fault Diagnosis based on entropy | |
| Fei et al. | Quantitative diagnosis of rotor vibration fault using process power spectrum entropy and support vector machine method | |
| Lin et al. | Hyper-spherical distance discrimination: A novel data description method for aero-engine rolling bearing fault detection | |
| Jiang et al. | Fault diagnosis of rotating machinery based on multisensor information fusion using SVM and time-domain features | |
| CN110617966A (en) | Bearing fault diagnosis method based on semi-supervised generation countermeasure network | |
| CN102156873B (en) | Chaos-based method for detecting and classifying early single-point faults of mechanical component | |
| Chen et al. | A sparse multivariate time series model-based fault detection method for gearboxes under variable speed condition | |
| Zheng et al. | Refined time-shift multiscale normalised dispersion entropy and its application to fault diagnosis of rolling bearing | |
| CN110060368A (en) | Mechanical method for detecting abnormality based on potential feature coding | |
| Prosvirin et al. | Global and local feature extraction using a convolutional autoencoder and neural networks for diagnosing centrifugal pump mechanical faults | |
| Gu et al. | Fault diagnosis of spindle device in hoist using variational mode decomposition and statistical features | |
| Xue et al. | Real-time diagnosis of an in-wheel motor of an electric vehicle based on dynamic Bayesian networks | |
| CN113256443B (en) | Nuclear power water pump guide bearing fault detection method, system, equipment and readable storage medium | |
| Chen et al. | Gear compound fault detection method based on improved multiscale permutation entropy and local mean decomposition | |
| Li et al. | Rolling bearing fault diagnosis method based on dispersion entropy and SVM | |
| Zhao et al. | Fault diagnosis of rolling bearings with noise signal based on modified kernel principal component analysis and DC‐ResNet | |
| CN111380680A (en) | Check valve fault feature extraction method based on improved permutation entropy | |
| Ding et al. | A novel weak feature extraction method for rotating machinery: link dispersion entropy | |
| Zhao et al. | An early fault diagnosis method of rolling element bearings based on med, dfa, and improved knn | |
| CN115099264A (en) | Ship part fault diagnosis method and device, computer and computer storage medium |
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: 20160420 Termination date: 20160922 |