CN113553898A - Method for diagnosing loosening fault of escalator footing - Google Patents
Method for diagnosing loosening fault of escalator footing Download PDFInfo
- Publication number
- CN113553898A CN113553898A CN202110633800.7A CN202110633800A CN113553898A CN 113553898 A CN113553898 A CN 113553898A CN 202110633800 A CN202110633800 A CN 202110633800A CN 113553898 A CN113553898 A CN 113553898A
- Authority
- CN
- China
- Prior art keywords
- gradient
- footing
- fault
- gray
- escalator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 45
- 239000011159 matrix material Substances 0.000 claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 35
- 238000004458 analytical method Methods 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 14
- 238000003745 diagnosis Methods 0.000 claims description 12
- 238000010183 spectrum analysis Methods 0.000 claims description 11
- 238000001228 spectrum Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 9
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000008901 benefit Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000035772 mutation Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a method for diagnosing the loosening fault of a footing of an escalator, which comprises the following steps: collecting vibration signals of a main machine footing fixing bolt of the escalator in the state of loosening different turns as a training sample and a test sample; performing multi-scale decomposition on each group of training samples obtained in the step one by using empirical wavelet decomposition to obtain an empirical mode function and construct a gray-gradient co-occurrence matrix of the empirical mode function; extracting a plurality of textural features of the double spectrograms as fault feature vectors, and forming multi-dimensional fault feature vectors of the training samples; training the neural network model by using a plurality of groups of training samples, and establishing a footing fixing bolt loosening fault identification model; extracting multi-dimensional fault feature vectors of the test sample, inputting the trained neural network model, and identifying the loosening degree of the fixing bolt; the characteristic extraction of the loosening fault of the footing of the escalator can be realized; the bolt looseness fault of the footing is effectively identified, the bolt looseness degree is determined, and the identification rate of the bolt looseness fault is improved.
Description
Technical Field
The invention relates to the field of fault diagnosis, in particular to a method for diagnosing a looseness fault of a footing of an escalator.
Background
The escalator is an indispensable large-scale public transport device, and once a fault occurs, operation is necessarily influenced and even safety accidents are caused. The main machine footing is used as an important structural part of the escalator, and the loosening of a fixing bolt of the main machine footing can cause the abnormal operation of the escalator. The problem that the loosening fault characteristics of the footing bolt are difficult to extract is solved.
With the rapid development of society, escalators have become important transportation means in modern urban life. The escalator is widely applied to large buildings such as large shopping malls, railway stations, subway stations and the like, a large number of people and equipment are conveyed every day, and the safety of the escalator is closely related to the safety of citizens. Once the escalator breaks down, the operation is influenced if the escalator breaks down, and serious accidents can be caused if the escalator breaks down, so that the life and the economy of people are greatly damaged. The loosening phenomenon of the fixing bolt of the main machine footing of the escalator occurs frequently due to frequent overtime and overload operation. The main machine footing is used as a key part of the escalator, and the fixing bolt of the main machine footing is loosened to cause periodic impact during the operation of the escalator, so that the vibration of the escalator coupling system is aggravated, the stability of the operation of the escalator is influenced, and the operation safety of the escalator is endangered in serious cases. Therefore, it is necessary to monitor the state of the fixing bolt of the escalator footing and diagnose the fault.
Escalator footing vibration signals are typically nonlinear, non-stationary signals. For the characteristics of mechanical fault signal nonlinearity, non-stationarity, weak early fault characteristics and susceptibility to noise interference, domestic and foreign scholars have proposed a number of vibration signal fault diagnosis methods, such as short-time Fourier transform (STFT), Wavelet Transform (WT), Empirical Mode Decomposition (EMD), Variational Mode Decomposition (VMD), and the like. However, the window function of the short-time fourier transform cannot be adaptively adjusted according to the characteristic frequency of the signal itself, which affects the accuracy of fault diagnosis. Wavelet transformation (including continuous wavelet, discrete wavelet, dual-tree complex wavelet, etc.) requires manual selection of wavelet basis functions and the number of decomposition levels, and lacks adaptivity. Empirical Mode Decomposition (EMD) makes a major breakthrough in the aspect of extracting vibration signal fault information, but EMD has serious modal aliasing and end point effect and lacks necessary theoretical basis. In order to overcome the shortcomings of the EMD method, many improved EMD algorithms are proposed, such as Local Mean Decomposition (LMD), local feature-scale decomposition (LCD), Ensemble Empirical Mode Decomposition (EEMD), etc., but these methods also cannot completely solve the disadvantages of EMD. The Variable Mode Decomposition (VMD) overcomes the defects of self-adaptive decomposition methods such as EMD and LMD, converts signal decomposition into a variable problem, and determines the center frequency and bandwidth of a component signal by seeking the optimal solution of the variable problem, thereby realizing the effective separation of each component signal. The VMD has a perfect theoretical basis and can better inhibit modal aliasing, but before decomposition, the parameter combination and the decomposition times of penalty factors need to be determined, and different parameter combinations and decomposition times can influence the decomposition precision of signals, thereby bringing great difficulty to the accurate decomposition of the signals.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide the escalator footing looseness fault diagnosis method with sufficient theoretical support and high accuracy.
To achieve the above object, the present invention relates to: a method for diagnosing the loosening fault of the footing of the escalator comprises the following steps:
firstly, acquiring vibration signals of different circle number states of loosening of a fixing bolt of a footing of a main machine of an escalator, and acquiring multiple groups of signals as a training sample and a test sample in each state respectively;
step two, carrying out multi-scale decomposition on each group of training samples obtained in the step one by utilizing EWT (empirical wavelet decomposition) to obtain EMF (empirical mode function);
thirdly, constructing a gray matrix and a gradient matrix of each empirical mode component function through a two-dimensional contour map of bispectrum analysis, further constructing a gray-gradient co-occurrence matrix of each empirical mode component function, and performing normalization processing;
extracting a plurality of texture features of the double spectrograms through a gray-gradient co-occurrence matrix to be used as fault feature vectors;
step five, forming a multi-dimensional fault feature vector of the training sample by the fault feature vector obtained in the step four;
step six, training the neural network model by using a plurality of groups of training samples, and establishing a footing fixing bolt loosening fault recognition model;
and step seven, extracting the multi-dimensional fault feature vectors of the test samples in the step five, inputting the trained neural network model, identifying the loosening degree of the fixing bolt, and analyzing the identification accuracy of the proposed method.
Furthermore, the different number of turns is taken as 0 turn, 1 turn, 2 turns and 3 turns.
Further, the second step includes the following steps:
step 2.1: adaptively partitioning a signal spectrum: according to Meyer and Littlewood-Paley theory, a scale function of EWTAnd the definition of the wavelet function ψ (x) in the frequency domain are respectively:
in the formula, betan=x4(35-84x+70x2-20x3)
There are many functions that satisfy this property, among which β is more commonly usedn=x4(35-84x+70x2-20x3) 0 < gamma < 1 andis a parameter that ensures that the overlap area between two consecutive state intervals is minimal,the value of which is determined by the calculated boundary value, ωn+1And ωnAll are the segmentation boundary points, ω1…ωn+1All are segmentation boundary points, example: the signal is divided into n segments, and the range of the n segment is [ omega ]n,ωn+1])
The decomposition method of EWT is similar to that of the classical wavelet, and the detail coefficient and the approximate coefficient after decomposition are respectivelyBoth satisfy the following formula:
wherein, F-1(. represents an inverse Fourier transform, andandare defined by formula (1) and formula (2), respectively; omeganTo segment the boundaries, ω represents frequency, t represents time, τnRepresentative boundary bandwidth τn=γωnτ is the boundary bandwidth value (width of the transition phase interval), τ is τ according to the above equationnCalculated, each n corresponds to a τ; f (t) is the signal, f (τ) is a function of τ,representing the signal f (t) after fourier transformation.
Step 2.2: decomposing the signals by utilizing an orthogonal wavelet filter group to obtain modal component signals with tight support characteristics:
from equations (3) and (4), an Empirical Mode Function (EMF) calculation equation after EWT decomposition can be obtained:
after the signal f (t) is subjected to EWT decomposition, n empirical mode functions can be obtained through the calculation of the formula. f. of0(t),fn(t) represents an empirical mode decomposition function;
the footing vibration signal can obtain a group of EMF through EWT decomposition, and each EMF is subjected to double-spectrum analysis, so that the fault feature vector of each mode function can be extracted.
Further, the third step includes the following steps:
step 3.1: firstly, decomposing complex noise into a series of high-low frequency bands by adopting EWT (enhanced wavelet transform), and removing the influence of non-Gaussian noise in a signal; decomposing the escalator footing vibration signal by utilizing EWT to obtain a series of empirical mode components (EMFs);
calculating bispectrum estimates of vibration signals
The bispectrum is a complex-valued spectrum and has 2 frequency variables omega1And ω2Bispectrum at ω1And ω2The total number of the formed frequency plane is 12 symmetrical regions, only the bispectrum value in the main region needs to be calculated, and then (omega) can be obtained according to the symmetry1,ω2) All bispectral values on the plane; wherein, bk(ω1,ω2) The K-th data bispectrum estimated value is obtained; k belongs to (1, K), K is the total segment number of the observed data, fsRepresenting the sampling frequency, N0Is the number of data points, λ1、λ2Is the wavelength of the spectrum.
Step 3.2: in order to automatically identify whether the bolt is loosened and the loosening degree, feature information contained in the bispectrum coefficient needs to be extracted; the two-dimensional contour map of the bispectrum analysis contains the basic characteristic information of the vibration signal, the gradient map of the bispectrum analysis describes the edge and mutation information of the contour map, and meanwhile, the two-dimensional contour map and the gradient map of the bipectrum analysis are combined to extract the characteristics to generate a gray-gradient co-occurrence matrix; more accurate fault characteristics can be obtained;
step 3.3: in order to facilitate the extraction of the texture features of the gray-level-gradient co-occurrence matrix, normalization processing needs to be performed on the gray-level-gradient co-occurrence matrix, and the element value of the gray-level-gradient co-occurrence matrix after normalization processing isThen there are:
wherein c (x, y) is the element value of the gray-gradient co-occurrence matrix, the gray value is x, and the gradient value is y.
Further, in the fourth step: the plurality of texture features are fault feature vectors, and specifically include: small gradient dominance, non-uniformity of gray scale distribution, non-uniformity of gradient distribution, gray scale entropy, gradient entropy and mixed entropy;
the formula is as follows:
wherein the content of the first and second substances,the gray value is x, and the gradient value is y.
Furthermore, the neural network model in the seventh step is a Bi-LSTM neural network model, and the bidirectional LSTM (Bi-LSTM) is an extension of the conventional LSTM, which can effectively improve the model performance of the sequence classification problem, and the bidirectional LSTM trains two LSTMs on the input sequence, the first is the original sequence, and the second is the reverse copy of the input sequence. Bidirectional LSTM can provide additional context for the network model and lead to faster, even more adequate learning problems, which can more accurately describe the model after multiple iterations; the invention adopts bidirectional LSTM to diagnose the fault of the fixing bolt of the escalator footing.
In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects:
(1) the escalator footing looseness fault diagnosis method can effectively realize escalator footing looseness fault feature extraction through empirical wavelet decomposition (EWT) and bispectrum analysis.
(2) The diagnosis method for the loosening fault of the escalator footing can effectively identify the loosening fault of the footing bolt and determine the loosening degree of the bolt, and improves the identification rate of the loosening fault of the bolt.
Drawings
FIG. 1 is a flow chart of the logic processing of the preferred embodiment of the present invention;
FIG. 2 is a two-dimensional contour diagram of a two-dimensional spectrum analysis of a vibration signal of the foot when the fixing bolt is normally loose according to the preferred embodiment of the present invention;
FIG. 3 is a two-dimensional contour diagram of a dual-spectrum analysis of a vibration signal of the foot when the fixing bolt is loosened by 1 turn according to the preferred embodiment of the present invention;
FIG. 4 is a two-dimensional contour plot of a two-dimensional spectrum analysis of a footing vibration signal when the fixing bolt is loosened by 2 revolutions in accordance with a preferred embodiment of the present invention;
FIG. 5 is a two-dimensional contour diagram of a two-spectrum analysis of a footing vibration signal when the fixing bolt is loosened by 3 turns in accordance with the preferred embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
A method for diagnosing the loosening fault of the footing of the escalator comprises the following steps:
firstly, acquiring vibration signals of different circle number states of loosening of a fixing bolt of a footing of a main machine of an escalator, and acquiring multiple groups of signals as a training sample and a test sample in each state respectively;
step two, carrying out multi-scale decomposition on each group of training samples obtained in the step one by utilizing EWT (empirical wavelet decomposition) to obtain EMF (empirical mode function);
the method specifically comprises the following steps:
step 2.1: adaptively partitioning a signal spectrum: according to Meyer and Littlewood-Paley theory, a scale function of EWTAnd the definition of the wavelet function psi (x) in the frequency domain are respectively[19]:
In the formula, betan=x4(35-84x+70x2-20x3) 0 < gamma < 1 andis a parameter which ensures a minimum overlap between two successive state ranges, the value of which is determined by a calculated boundary value, ωn+1And ωnAll are the segmentation boundary points, ω1…ωn+1All are segmentation boundary points, example: the signal is divided into n segments, and the range of the n segment is [ omega ]n,ωn+1];
The decomposition method of EWT is similar to that of the classical wavelet, and the detail coefficient and the approximate coefficient after decomposition are respectivelyBoth satisfy the following formula:
wherein, F-1(. represents an inverse Fourier transform, andandare defined by formula (1) and formula (2), respectively; omeganTo segment the boundaries, ω represents frequency, t represents time, τnRepresenting the boundary zone taun=γωnτ is the boundary bandwidth value (width of the transition phase interval), τ is τ according to the above equationnCalculated, each n corresponds to a τ; f (t) is a signalF (τ) is a function of τ,representing the signal f (t) after fourier transformation;
step 2.2: decomposing the signals by utilizing an orthogonal wavelet filter group to obtain modal component signals with tight support characteristics:
from equations (3) and (4), an Empirical Mode Function (EMF) calculation equation after EWT decomposition can be obtained:
after the signal f (t) is subjected to EWT decomposition, n empirical mode functions can be obtained through the calculation of the formula; f. of0(t),fn(t) represents an empirical mode decomposition function;
the footing vibration signal can obtain a group of EMF through EWT decomposition, and each EMF is subjected to double-spectrum analysis, so that the fault feature vector of each mode function can be extracted.
Thirdly, constructing a gray matrix and a gradient matrix of each empirical mode component function through a two-dimensional contour map of bispectrum analysis, further constructing a gray-gradient co-occurrence matrix of each empirical mode component function, and performing normalization processing;
the method specifically comprises the following steps:
step 3.1: firstly, decomposing complex noise into a series of high-low frequency bands by adopting EWT (enhanced wavelet transform), and removing the influence of non-Gaussian noise in a signal; decomposing the escalator footing vibration signal by utilizing EWT to obtain a series of empirical mode components (EMFs);
calculating bispectrum estimates of vibration signals
The bispectrum is a complex-valued spectrum and has 2 frequency variables omega1And ω2Bispectrum at ω1And ω2The total number of the formed frequency plane is 12 symmetrical regions, only the bispectrum value in the main region needs to be calculated, and then (omega) can be obtained according to the symmetry1,ω2) All bispectral values on the plane; wherein, bk(ω1,ω2) The K-th data bispectrum estimated value is obtained; k belongs to (1, K), K is the total segment number of the observed data, fsRepresenting the sampling frequency, N0Is the number of data points, λ1、λ2Is the wavelength of the spectrum;
step 3.2: in order to automatically identify whether the bolt is loosened and the loosening degree, feature information contained in the bispectrum coefficient needs to be extracted; the two-dimensional contour map of the bispectrum analysis contains the basic characteristic information of the vibration signal, the gradient map of the bispectrum analysis describes the edge and mutation information of the contour map, and meanwhile, the two-dimensional contour map and the gradient map of the bipectrum analysis are combined to extract the characteristics to generate a gray-gradient co-occurrence matrix; more accurate fault characteristics can be obtained;
step 3.3: in order to facilitate the extraction of the texture features of the gray-level-gradient co-occurrence matrix, normalization processing needs to be performed on the gray-level-gradient co-occurrence matrix, and the element value of the gray-level-gradient co-occurrence matrix after normalization processing isThen there are:
wherein c (x, y) is the element value of the gray-gradient co-occurrence matrix, the gray value is x, and the gradient value is y.
The plurality of texture features are fault feature vectors, and specifically include: small gradient dominance, non-uniformity of gray scale distribution, non-uniformity of gradient distribution, gray scale entropy, gradient entropy and mixed entropy;
the formula is as follows:
wherein the content of the first and second substances,the gray value is x, and the gradient value is y.
Extracting a plurality of texture features of the double spectrograms through a gray-gradient co-occurrence matrix to be used as fault feature vectors;
step five, forming a multi-dimensional fault feature vector of the training sample by the fault feature vector obtained in the step four;
step six, training the neural network model by using a plurality of groups of training samples, and establishing a footing fixing bolt loosening fault recognition model;
and step seven, extracting the multi-dimensional fault feature vectors of the test samples in the step five, inputting the trained Bi-LSTM neural network model, identifying the loosening degree of the fixing bolt, and analyzing the identification accuracy of the proposed method.
The different number of turns is taken as 0 turn, 1 turn, 2 turns and 3 turns.
The actual diagnosis example, as shown in fig. 1, includes the following steps:
a) vibration signals of 4 states of looseness 1 ring, looseness 2 ring, looseness 3 ring, non-looseness and the like of a fixing bolt of a footing of a main machine of the escalator are collected, and multiple groups of signals are collected as training samples and test samples in each state respectively.
b) Decomposing each group of training samples by using EWT to form an Empirical Mode Function (EMF);
c) performing double-spectrum analysis on EMF of each group of training samples, and constructing a gray matrix and a gradient matrix of the EMF through a two-dimensional contour map (refer to fig. 2-5) of the double-spectrum analysis so as to construct a gray-gradient co-occurrence matrix of the EMF;
d) by EMFkThe gray level-gradient co-occurrence matrix extracts 6 texture features to form 18-dimensional fault feature vectors of the training samples.
The method comprises the following steps: performing double-spectrum analysis on three EMFs (2, 3 and 4), and extracting the above 6 texture features (small gradient dominance, non-uniformity of gray distribution, non-uniformity of gradient distribution, gray entropy, gradient entropy and mixed entropy) from each EMF by using a gray-gradient co-occurrence matrix to form 18-dimensional (3 x 6) fault feature vectors.
e) Bidirectional LSTM (Bi-LSTM) is an extension of traditional LSTM, and can effectively improve model performance of sequence classification problems. The bidirectional LSTM can provide additional context for a network model and lead to a faster and even more sufficient learning problem, the model can be described more accurately after repeated iteration, a Bi-LSTM network is trained by utilizing multiple groups of training data, and a footing fixing bolt loosening fault identification model is established;
f) and extracting 18-dimensional fault feature vectors of the test sample, inputting the trained Bi-LSTM network, identifying the loosening degree of the fixing bolt, and analyzing the identification accuracy of the proposed method.
The scheme adopts an escalator footing looseness fault feature extraction method based on empirical wavelet decomposition (EWT) and bispectrum analysis. Firstly, performing EWT decomposition on an original footing vibration acceleration signal to obtain a series of empirical mode component functions (EMF); then, for each empirical mode component function, calculating a double spectrogram by using a double spectrum analysis method, and extracting 6 texture features of the double spectrogram as fault feature vectors through a gray-gradient co-occurrence matrix; and finally, classifying and identifying four types of footing loosening fault signals with different degrees by using the extracted multi-scale fault feature vector and a bidirectional long-short time network (Bi-LSTM), and determining the types of the footing loosening faults.
The EWT signal processing is characterized in that:
(1) because the method is proposed based on wavelet theory, the method has sufficient theoretical support;
(2) modal aliasing and false modes can be avoided;
(3) the frequency domain has a plurality of segmentation methods, and can be self-organized and self-adaptive;
(4) the method has excellent effect in qualitative and quantitative processing, and can be suitable for nonlinear and uncertain signal processing.
The core of EWT signal processing is to combine the spectrum conversion theory of wavelet transform and decompose the signal into amplitude modulation and frequency modulation (am.fm) signals with characteristics in the frequency domain by means of empirical mode decomposition.
The vibration data collected in the actual environment has more or less interferences (the escalator is a large moving machine, and moving parts are more), and the interferences can be processed according to the characteristics of the wavelet algorithm, so that the analysis is more accurate.
Compared with a direct bipectrum analysis method (Bispecrum), the simulation experiment is carried out on various models of escalators such as Shenzhen subway, Nanjing subway and the like, and the overall accuracy is improved by about 21.38%; compared with the EMD-Bispec method, the overall accuracy is improved by about 10.01 percent; compared with the EEMD-Bispec method, the accuracy is improved by about 6.53 percent; the result shows that the method based on empirical wavelet decomposition and bispectrum analysis feature extraction can more effectively identify the looseness grade of the footing fixing bolt.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (6)
1. A method for diagnosing the loosening fault of the footing of the escalator is characterized by comprising the following steps:
firstly, acquiring vibration signals of different circle number states of loosening of a fixing bolt of a footing of a main machine of an escalator, and acquiring multiple groups of signals as a training sample and a test sample in each state respectively;
step two, performing multi-scale decomposition on each group of training samples obtained in the step one by utilizing EWT to obtain EMF;
thirdly, constructing a gray matrix and a gradient matrix of each empirical mode component function through a two-dimensional contour map of bispectrum analysis, further constructing a gray-gradient co-occurrence matrix of each empirical mode component function, and performing normalization processing;
extracting a plurality of texture features of the double spectrograms through a gray-gradient co-occurrence matrix to be used as fault feature vectors;
step five, forming a multi-dimensional fault feature vector of the training sample by the fault feature vector obtained in the step four;
step six, training the neural network model by using a plurality of groups of training samples, and establishing a footing fixing bolt loosening fault recognition model;
and step seven, extracting the multi-dimensional fault feature vectors of the test samples in the step five, inputting the trained neural network model, identifying the loosening degree of the fixing bolt, and analyzing the identification accuracy of the proposed method.
2. The escalator footing loose fault diagnosis method according to claim 1, wherein the different number of turns is taken as 0, 1, 2,3 loose turns.
3. The escalator footing looseness fault diagnosis method according to claim 1, wherein the second step includes the steps of:
step 2.1: adaptively partitioning a signal spectrum: according to Meyer and Littlewood-Paley theory, a scale function of EWTAnd the definition of the wavelet function ψ (x) in the frequency domain are respectively:
in the formula, betan=x4(35-84x+70x2-20x30 < gamma < 1 andis a parameter which ensures a minimum overlap between two successive state ranges, the value of which is determined by a calculated boundary value, ωn+1And ωnAll are the segmentation boundary points, ω1…ωn+1Are all the points of the segmentation boundary point,
the decomposition method of EWT is similar to that of the classical wavelet, and the detail coefficient and the approximate coefficient after decomposition are respectivelyBoth satisfy the following formula:
wherein, F-1(. represents an inverse Fourier transform, andandare defined by formula (1) and formula (2), respectively; omeganTo segment the boundaries, ω represents frequency, t represents time, τnRepresenting the boundary zone taun=γωnτ is the boundary bandwidth value, τ is τ according to the above equationnCalculated, each n corresponds to a τ; f (t) is the signal, f (τ) is a function of τ,representing the signal f (t) after fourier transformation;
step 2.2: decomposing the signals by utilizing an orthogonal wavelet filter group to obtain modal component signals with tight support characteristics:
from equations (3) and (4), an Empirical Mode Function (EMF) calculation equation after EWT decomposition can be obtained:
after the signal f (t) is subjected to EWT decomposition, n empirical mode functions can be obtained through the calculation of the formula; f. of0(t),fn(t) represents an empirical mode decomposition function;
the footing vibration signal can obtain a group of EMF through EWT decomposition, and each EMF is subjected to double-spectrum analysis, so that the fault feature vector of each mode function can be extracted.
4. The escalator footing looseness fault diagnosis method according to claim 1, wherein said step three includes the steps of:
step 3.1: firstly, decomposing complex noise into a series of high-low frequency bands by adopting EWT (enhanced wavelet transform), and removing the influence of non-Gaussian noise in a signal; decomposing the escalator footing vibration signal by utilizing EWT to obtain a series of empirical mode components (EMFs);
calculating bispectrum estimates of vibration signals
The bispectrum is a complex-valued spectrum and has 2 frequency variables omega1And ω2Bispectrum at ω1And ω2The total number of the formed frequency plane is 12 symmetrical regions, only the bispectrum value in the main region needs to be calculated, and then (omega) can be obtained according to the symmetry1,ω2) All bispectral values on the plane; wherein, bk(ω1,ω2) The K-th data bispectrum estimated value is obtained; k belongs to (1, K), K is the total segment number of the observed data, fsRepresenting the sampling frequency, N0Is the number of data points, λ1、λ2Is the wavelength of the spectrum;
step 3.2: in order to automatically identify whether the bolt is loosened and the loosening degree, feature information contained in the bispectrum coefficient needs to be extracted; the two-dimensional contour map of the bispectrum analysis contains the basic characteristic information of the vibration signal, the gradient map of the bispectrum analysis describes the edge and mutation information of the contour map, and meanwhile, the two-dimensional contour map and the gradient map of the bipectrum analysis are combined to extract the characteristics to generate a gray-gradient co-occurrence matrix; more accurate fault characteristics can be obtained;
step 3.3: in order to facilitate the extraction of the texture features of the gray-level-gradient co-occurrence matrix, normalization processing needs to be performed on the gray-level-gradient co-occurrence matrix, and the element value of the gray-level-gradient co-occurrence matrix after normalization processing isThen there are:
wherein c (x, y) is the element value of the gray-gradient co-occurrence matrix, the gray value is x, and the gradient value is y.
5. The escalator footing looseness fault diagnosis method according to claim 1, wherein in step four: the plurality of texture features are fault feature vectors, and specifically include: small gradient dominance, non-uniformity of gray scale distribution, non-uniformity of gradient distribution, gray scale entropy, gradient entropy and mixed entropy;
the formula is as follows:
6. The escalator footing looseness fault diagnosis method according to claim 1, wherein the neural network model in the seventh step is a Bi-LSTM neural network model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110633800.7A CN113553898A (en) | 2021-06-07 | 2021-06-07 | Method for diagnosing loosening fault of escalator footing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110633800.7A CN113553898A (en) | 2021-06-07 | 2021-06-07 | Method for diagnosing loosening fault of escalator footing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113553898A true CN113553898A (en) | 2021-10-26 |
Family
ID=78102024
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110633800.7A Pending CN113553898A (en) | 2021-06-07 | 2021-06-07 | Method for diagnosing loosening fault of escalator footing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113553898A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117105038A (en) * | 2023-10-17 | 2023-11-24 | 山西戴德测控技术股份有限公司 | Elevator operation monitoring method, device, equipment and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050171647A1 (en) * | 2004-02-02 | 2005-08-04 | Abb Inc. | High impedance fault detection |
CN108171742A (en) * | 2017-12-26 | 2018-06-15 | 太原理工大学 | A kind of method that the evaluation of carbon fibre composite surface topography three-dimensional is realized based on shade of gray feature |
CN110018447A (en) * | 2019-04-09 | 2019-07-16 | 电子科技大学 | Signal transmitter individual discrimination method based on double-spectrum analysis and convolutional neural networks |
WO2019205826A1 (en) * | 2018-04-28 | 2019-10-31 | 河北工业大学 | Rolling bearing fault diagnosis method based on wavelet packet energy spectrum and modulation signal bispectrum analysis |
CN110503025A (en) * | 2019-08-19 | 2019-11-26 | 重庆大学 | A kind of analog circuit Incipient Fault Diagnosis method based on semi-supervised coorinated training |
CN111753260A (en) * | 2020-06-29 | 2020-10-09 | 武汉科技大学 | Method for detecting loosening state of elevator fixing device, computer equipment and elevator |
CN111881933A (en) * | 2019-06-29 | 2020-11-03 | 浙江大学 | Hyperspectral image classification method and system |
-
2021
- 2021-06-07 CN CN202110633800.7A patent/CN113553898A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050171647A1 (en) * | 2004-02-02 | 2005-08-04 | Abb Inc. | High impedance fault detection |
CN108171742A (en) * | 2017-12-26 | 2018-06-15 | 太原理工大学 | A kind of method that the evaluation of carbon fibre composite surface topography three-dimensional is realized based on shade of gray feature |
WO2019205826A1 (en) * | 2018-04-28 | 2019-10-31 | 河北工业大学 | Rolling bearing fault diagnosis method based on wavelet packet energy spectrum and modulation signal bispectrum analysis |
CN110018447A (en) * | 2019-04-09 | 2019-07-16 | 电子科技大学 | Signal transmitter individual discrimination method based on double-spectrum analysis and convolutional neural networks |
CN111881933A (en) * | 2019-06-29 | 2020-11-03 | 浙江大学 | Hyperspectral image classification method and system |
CN110503025A (en) * | 2019-08-19 | 2019-11-26 | 重庆大学 | A kind of analog circuit Incipient Fault Diagnosis method based on semi-supervised coorinated training |
CN111753260A (en) * | 2020-06-29 | 2020-10-09 | 武汉科技大学 | Method for detecting loosening state of elevator fixing device, computer equipment and elevator |
Non-Patent Citations (2)
Title |
---|
赵博强: "高速列车轮对缺陷轨上感知算法研究", 《中国优秀硕士学位论文电子期刊.工程科技Ⅱ辑》 * |
陈宗祥,陈明星,焦民胜,葛芦生: "基于改进EMD和双谱分析的电机轴承故障诊断实现", 《电机与控制学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117105038A (en) * | 2023-10-17 | 2023-11-24 | 山西戴德测控技术股份有限公司 | Elevator operation monitoring method, device, equipment and storage medium |
CN117105038B (en) * | 2023-10-17 | 2024-01-05 | 山西戴德测控技术股份有限公司 | Elevator operation monitoring method, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sun et al. | Fault diagnosis of rolling bearing based on empirical mode decomposition and improved manhattan distance in symmetrized dot pattern image | |
Sun et al. | Bearing fault diagnosis based on EMD and improved Chebyshev distance in SDP image | |
Li et al. | An enhanced morphology gradient product filter for bearing fault detection | |
CN108052980B (en) | Image-based air quality grade detection method | |
Fard et al. | A new methodology for vehicle trajectory reconstruction based on wavelet analysis | |
CN107832525B (en) | Method for extracting bearing fault characteristic frequency by optimizing VMD (variable minimum vector machine) through information entropy and application of method | |
CN111751133B (en) | Intelligent fault diagnosis method of deep convolutional neural network model based on non-local mean embedding | |
Wodecki et al. | Local damage detection based on vibration data analysis in the presence of Gaussian and heavy-tailed impulsive noise | |
CN102117482A (en) | Non-local mean image denoising method combined with structure information | |
CN101727662A (en) | SAR image nonlocal mean value speckle filtering method | |
Naveed et al. | A statistical approach to signal denoising based on data-driven multiscale representation | |
Wang et al. | A computer vision based machine learning approach for fatigue crack initiation sites recognition | |
CN108171119B (en) | SAR image change detection method based on residual error network | |
CN111008585A (en) | Ship target detection method based on self-adaptive layered high-resolution SAR image | |
Guo et al. | An enhanced modulation signal bispectrum analysis for bearing fault detection based on non-Gaussian noise suppression | |
Chen et al. | Investigation on enhanced mathematical morphological operators for bearing fault feature extraction | |
CN111612640A (en) | Data-driven vehicle insurance fraud identification method | |
CN114112400A (en) | Mechanical bearing fault diagnosis method based on multi-angle information fusion | |
Zhang et al. | Adaptive filtering enhanced windowed correlated kurtosis for multiple faults diagnosis of locomotive bearings | |
CN113553898A (en) | Method for diagnosing loosening fault of escalator footing | |
Duan et al. | Impulsive feature extraction with improved singular spectrum decomposition and sparsity-closing morphological analysis | |
Belsak et al. | Wavelet analysis for gear crack identification | |
CN103325123A (en) | Image edge detection method based on self-adaptive neural fuzzy inference systems | |
JP2003132041A (en) | Signal processing method and device, signal processing program and recording medium | |
CN109271902B (en) | Infrared weak and small target detection method based on time domain empirical mode decomposition under complex background |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211026 |
|
RJ01 | Rejection of invention patent application after publication |