CN116297841A - Railway track disease identification method based on optical fiber distributed vibration detection - Google Patents
Railway track disease identification method based on optical fiber distributed vibration detection Download PDFInfo
- Publication number
- CN116297841A CN116297841A CN202310249537.0A CN202310249537A CN116297841A CN 116297841 A CN116297841 A CN 116297841A CN 202310249537 A CN202310249537 A CN 202310249537A CN 116297841 A CN116297841 A CN 116297841A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- signal
- signals
- feature vector
- disease
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 57
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 57
- 239000013307 optical fiber Substances 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000001514 detection method Methods 0.000 title claims abstract description 19
- 239000013598 vector Substances 0.000 claims abstract description 87
- 230000004927 fusion Effects 0.000 claims abstract description 48
- 230000006378 damage Effects 0.000 claims abstract description 30
- 208000027418 Wounds and injury Diseases 0.000 claims abstract description 25
- 208000014674 injury Diseases 0.000 claims abstract description 24
- 230000003044 adaptive effect Effects 0.000 claims abstract description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 33
- 238000012706 support-vector machine Methods 0.000 claims description 33
- 238000000605 extraction Methods 0.000 claims description 15
- 239000004973 liquid crystal related substance Substances 0.000 claims description 8
- 241000251468 Actinopterygii Species 0.000 claims description 7
- 238000003745 diagnosis Methods 0.000 claims description 7
- 230000005484 gravity Effects 0.000 claims description 7
- 239000000835 fiber Substances 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 238000004804 winding Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 2
- 230000006835 compression Effects 0.000 claims description 2
- 238000007906 compression Methods 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims description 2
- 230000010354 integration Effects 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000011155 quantitative monitoring Methods 0.000 claims description 2
- 208000034656 Contusions Diseases 0.000 claims 1
- 208000034526 bruise Diseases 0.000 claims 1
- 238000012545 processing Methods 0.000 claims 1
- 238000012544 monitoring process Methods 0.000 abstract description 13
- 208000035874 Excoriation Diseases 0.000 abstract description 10
- 238000005299 abrasion Methods 0.000 abstract description 10
- 230000003287 optical effect Effects 0.000 abstract description 5
- 238000004891 communication Methods 0.000 abstract description 4
- 238000012423 maintenance Methods 0.000 abstract description 3
- 230000007774 longterm Effects 0.000 abstract description 2
- 230000007547 defect Effects 0.000 description 6
- 230000003595 spectral effect Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000011084 recovery Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000009545 invasion Effects 0.000 description 2
- 241001669679 Eleotris Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000008846 dynamic interplay Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000253 optical time-domain reflectometry Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000009418 renovation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/04—Analysing solids
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
- B61K9/08—Measuring installations for surveying permanent way
- B61K9/10—Measuring installations for surveying permanent way for detecting cracks in rails or welds thereof
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H9/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by using radiation-sensitive means, e.g. optical means
- G01H9/004—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by using radiation-sensitive means, e.g. optical means using fibre optic sensors
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4445—Classification of defects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2291/00—Indexing codes associated with group G01N29/00
- G01N2291/02—Indexing codes associated with the analysed material
- G01N2291/023—Solids
- G01N2291/0234—Metals, e.g. steel
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Abstract
The invention relates to the field of signal source monitoring, in particular to a railway track disease identification method based on optical fiber distributed vibration detection. According to the invention, the existing communication optical cable beside the railway is connected to obtain the acoustic wave sensing signal of the communication optical cable, and the original vibration signal S (p, q) is obtained by demodulating the acoustic wave sensing signal by adopting a differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising; then, extracting the characteristics of the expanded demodulation signals, and forming a multidimensional fusion characteristic vector through vector splicing; pre-weighting the feature vector based on the data sample set; and finally, classifying and identifying the feature vector by using the trained SVM model, recording the existence position of the disease, and outputting the disease position alarm of the optical fiber length. The invention realizes the identification and classification of typical diseases of the rails such as scale injury, wave abrasion injury, overhead injury and the like on the railway, positions and alarms the positions of the diseases in real time, and provides a long-term online monitoring means for the rail safety operation and maintenance monitoring of railway transportation.
Description
Technical Field
The invention relates to the field of signal source monitoring, in particular to a railway track disease identification method based on optical fiber distributed vibration detection.
Background
At present, railways are one of the main forces of the transportation industry in national economy construction. As train speed and transport capacity increase, wheel-rail interactions are inevitably exacerbated. The serious wheel-rail dynamic interaction can cause serious abrasion of wheels and rails, particularly when a freight train runs on a large steep slope, a small-radius curve section, the wheels are damaged, the contact relationship of the wheels and the rails is very bad, the damage of a foundation bed is increased, and fatigue damage phenomena such as oblique crack contact fatigue damage (fish scale damage), rail top surface wave abrasion (wave abrasion) and gaps (hollow hanging damage) between a sleeper and lower railway ballasts caused by serious wave abrasion are more serious. The fatigue damage of the rails aggravates the corresponding renovation work, thereby increasing the operation cost of railway transportation and even seriously interfering with the normal order of railway transportation. It is necessary to find the defect of the rail and perform maintenance in time. However, current monitoring for track diseases is mainly achieved through ultrasonic detection, eddy current detection, acceleration measurement, video images, and the like. The above-mentioned sensing methods mostly have non-linearity, zero drift, and low tolerance of the strong electromagnetic environment around the transmission line, and the sensor is also very complex and expensive to install and maintain.
In view of the limitations of conventional electrical sensors, fiber optic sensors have been used both at home and abroad to safely monitor railroad tracks using various methods. In 2017, wang et al uses a railway vibration detection scheme based on phase sensitive OTDR to propose multidimensional comprehensive analysis to identify train signals and illegal intrusions during high-speed railway operation, successfully monitor the train operation signals, calculate the length and the operation speed of the train, and identify illegal disturbance along a railway, but do not monitor and identify railway diseases. In 2016, buggy et al have tested 7 FBG strain sensors mounted to several rail components to analyze the dynamic strain induced by the train passing the sensor locations. By analysis and data collection methods, tension changes in bolt-torque can be categorized and any changes in the condition of the railway track components can be revealed. However, the research is not applied to real-time monitoring in actual railways, but only shows one feasibility. In 2020, xin et al have used signals detected by ultra-weak FBG arrays in subway tunnels to identify earth intrusion events. The event signal is extracted by combining the root mean square of the spectral subtraction and the power spectral density. And then, carrying out feature extraction by adopting local feature scale decomposition and multi-scale permutation entropy, and improving the event recognition rate from the aspect of multi-scale analysis. Experimental verification shows that the scheme can identify four events of subway trains, truck passing, discrete impact invasion and continuous impact invasion. The scheme completes the identification of subway related events, but does not realize the monitoring of rail diseases, and the FBG layout cost is still much higher than that of the existing optical cable.
Disclosure of Invention
Aiming at the problems or defects existing in the last time, the invention provides a railway track disease identification method based on optical fiber distributed vibration detection, which aims at solving the problems of timely diagnosis and real-time accurate positioning of railway track diseases.
A railway track disease identification method based on optical fiber distributed vibration detection comprises the following steps:
In order to effectively solve and restore a large vibration signal acquired by a system, improve the signal-to-noise ratio of the acquired signal and ensure the accuracy of the subsequent signal characteristic extraction and classification identification, the invention demodulates an optical fiber distributed acoustic wave sensing signal to obtain an original vibration signal S (p, q) by adopting a differential compression-restoration dynamic range expansion algorithm based on wavelet adaptive denoising aiming at the characteristics of the wheel track vibration signal, so that the original signal is well restored in a time domain, the frequency domain peak value of the original signal is increased, and the signal-to-noise ratio is improved, thereby ensuring the accurate monitoring of the whole distributed optical fiber sensing system on the optical fiber line.
And 2, constructing an initial data set for the original vibration signals S (p, q) obtained in the step 1, and marking data samples of the normal rail signals and 3 disease signals (fish scale injury, wave abrasion injury and empty hanging injury). And carrying out time-frequency domain feature extraction on sample data in the initial data set by using gravity center frequency, standard deviation frequency, power spectrum entropy, sample entropy and normalized wavelet energy feature vectors, and constructing a multidimensional fusion feature vector by using a vector splicing method to obtain an initial multidimensional fusion feature vector data set.
And 3, selecting M (M is more than or equal to 100) groups of data of the normal rail signals and the 3 disease signals from the initial multidimensional fusion feature vector data set obtained in the step 2, constructing a multidimensional fusion feature vector data set, and pre-weighting the multidimensional fusion feature vector based on the average standard deviation ASD of the multidimensional fusion feature vectors of the normal rail signals and the 3 disease signals as the weight of each feature value of the multidimensional fusion feature vector, so as to obtain a pre-weighted multidimensional fusion feature vector sample set.
And 4, inputting the pre-weighted multidimensional fusion feature vector sample set obtained in the step 3 into a Support Vector Machine (SVM) model for classification training, using the trained SVM model to realize classification identification of normal rail signals and 3 disease signals, recording the positions and types of the diseases, outputting the positions and types of the diseases, alarming, and realizing real-time disease diagnosis and positioning output.
Further, in the step 1, a differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising is adopted to demodulate an optical fiber distributed acoustic wave sensing signal, and the specific steps include:
step 1-1, using an arctangent demodulation algorithm to demodulate the optical fiber distributed acoustic wave sensing signal, the step of solving the phase using an arctangent function results in the phase value being limited to [ -pi, pi]In the range of (2), a phase unwrapping is required to obtain a true phaseThe classical phase unwrapping algorithm is expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the original wrapped signal, < >>Is the unwind signal. The original wrapped signal can be expressed as + ->Where k (n) is an integer series of numbers, the above formula can be further expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the threshold for unwind. But this general algorithm also has problems. If phi>The unwind signal will remain in the wound state and thus the change trace cannot be retrieved correctly. This problem is unavoidable when the amplitude of the phase signal varies drastically, especially for vibration events on the fiber with large amplitude or high frequency.
Aiming at the phenomenon, the invention provides a differential compression-reduction algorithm based on wavelet self-adaptive denoising. First, an original phase-wrapped signal is subjected to first-order differential compression:
wherein% l k (n) is still an integer number sequence and thus a unwrapping algorithm can be used for it. Discrete noiseN noise (n) are incorporated herein. Because in the original signal the noise is negligible, but in the higher order differential signal the noise rises exponentially with the order of the differential, while the signal amplitude is compressed and the signal to noise ratio drops dramatically.
Step 1-2, handleSine signal regarded as N discrete samples +.>And (3) summing; wherein A is i Is the signal amplitude omega i Is the signal frequency, θ i Is the primary phase of the signal). For each sinusoidal signal +.>Having a derivative of order I
The unwind threshold at this point may be expressed as:
When l epsilon [1,5], phi < pi can be satisfied, and then the winding can be unwound to obtain:
when the signal is affected by the post-differentiation noise, this results in phi>Pi, integer number column is equal to l k (n) cannot be completely unwoundThe algorithm eliminates, leaving the misinterpreted winding term k' (n). Directly carrying out first-order integration on the unwrapped signal to obtain:
where K' (n) is a complex polynomial of degree i. Thus, the misunderstanding wrapping term k' (n) and noise delta can be removed by using a wavelet denoising algorithm in the differential stage l N noise (n) and thus achieve correct demodulation, so that the phase demodulation of the overall signal is more tolerant to noise. The original vibration signal with large range and high signal-to-noise ratio is correctly restored by the differential compression-restoration dynamic range expansion algorithm, so that the follow-up accurate extraction and classification recognition of the signal characteristics are realized.
Further, the specific steps of extracting the time-frequency domain features by the normalized wavelet energy feature vector in the step 2 are as follows:
and 2-1, calculating normalized wavelet energy eigenvectors of the signals. And carrying out wavelet decomposition on the collected and restored signals, wherein each decomposition layer keeps high-frequency wavelet coefficients and further decomposes low-frequency wavelet coefficients. After L-layer wavelet decomposition, the low-frequency wavelet coefficient cA can be obtained L And high frequency wavelet coefficient cD 1 ,cD 2 ,…cD L ;
Step 2-2, the high-frequency wavelet coefficient and the low-frequency wavelet coefficient are sequentially arranged from low frequency to high frequency, and the energy E of each wavelet coefficient is solved by the formula (7) j Form wavelet energy feature vector E= [ E ] 1 ,E 2 ,…,E L+1 ];
Wherein C is j (m) is the wavelet coefficient after decomposition, and L is the number of wavelet decomposition layers.
Step 2-3, normalizing the wavelet energy feature vector obtained in step 2-2 according to formula (8) to obtain normalized wavelet energy feature vector E'=[E′ 1 ,E′ 2 ,…,E′ L+1 ];
In the step 2-4, when the wavelet decomposition is actually performed, the decomposition layer number L of the wavelet decomposition and the wavelet basis function are determined by normalizing the average standard deviation ASD of the wavelet energy eigenvectors.
And (3) carrying out wavelet decomposition test on each 1 group of data of 4 signals (normal rail signals and 3 disease signals) in the initial data set, enumerating different wavelet decomposition layers and wavelet functions, respectively carrying out wavelet decomposition to obtain normalized wavelet energy characteristic vectors, and calculating the Average Standard Deviation (ASD) under the conditions of the different wavelet decomposition layers and the wavelet functions as the degree of distinguishing the energy vectors according to a formula (9) and a formula (10). Determining the wavelet decomposition layer number and wavelet function with the maximum discrimination so as to enable the four types of signals to have the maximum discrimination;
SD j is the standard deviation of the two-dimensional image,the mean value of the jth wavelet energy component of the four types of signals, a being the corresponding sample type.
Further, the step 3 pre-weights the multidimensional fusion feature vector, and the specific steps include:
and 3-1, screening out M (M is more than or equal to 100) group data of normal rail signals and 3 disease signals (fish scale injury, wave abrasion injury and overhead injury) from the multidimensional fusion feature vector data set constructed in the step 2, and constructing a multidimensional fusion feature vector data set.
Step 3-2, for each multidimensional fusion feature vector, each multidimensional fusion feature vector is formed by combining 4 single feature values (center of gravity frequency, frequency standard deviation, power spectrum entropy and sample entropy) and normalized wavelet energy feature vectors with the length of L+1, and the total length is L+5. Each characteristic value differs for four types of signal discrimination, and the signal discrimination corresponding to each characteristic value is calculated as the weight w of the characteristic value according to the formula (11) n Constructing the weight vector of the sample data set as [ w ] 1 ,w 2 ,…w N+5 ]Weighting all the obtained multidimensional fusion feature vectors to construct a multidimensional fusion feature vector [ w ] after being pre-weighted 1 c 1 ,w 2 c 2 ,…w N+5 c N+5 ]。
Wherein M is the number of groups of each signal data in the data set, c s And the s-th eigenvalue of the multidimensional fusion eigenvector, r is the corresponding sample group number, and a is the corresponding sample type.
Further, the step 4 specifically includes:
and 4-1, training the support vector machine by using the pre-weighted multidimensional fusion feature vector sample set obtained in the step 3 to obtain a trained support vector machine SVM.
And 4-2, carrying out engineering deployment on the trained support vector machine SVM, carrying out feature extraction and pre-weighting on the original vibration signals by using the methods of the step 2 and the step 3, and taking the pre-weighted multidimensional fusion feature vector as an input vector of the support vector machine to finish classification recognition.
And 4-3, recording the identification result of the support vector machine SVM, and outputting the position and type of the disease, thereby realizing real-time disease diagnosis and positioning output.
The invention provides a railway track defect identification method based on optical fiber distributed vibration detection, which is applied to railway track defect monitoring. The invention adopts the distributed optical fiber acoustic wave sensing system with expanded dynamic range to monitor and record the vibration information generated when the train runs on the railway by connecting the existing communication optical cable beside the railway, realizes the waveform restoration and the signal-to-noise ratio improvement of a large-amplitude vibration signal under the wheel track vibration effect based on a dynamic range expansion algorithm, solves the problem of high-fidelity restoration of the large-amplitude vibration signal by the system, extracts a single characteristic value and carries out wavelet decomposition on the restored vibration information, constructs a multidimensional fusion characteristic vector in a vector splicing mode, and completes the identification and classification of various track disease types.
In summary, the distributed optical fiber acoustic wave sensing system realized by using the differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising combines multidimensional feature extraction and SVM classification to realize effective extraction and identification of railway track diseases along the railway, solve the problems of effective extraction and positioning of the railway track diseases, and can realize timely diagnosis and real-time positioning of the railway track diseases.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the differential compression-reduction dynamic range expansion algorithm of the present invention;
FIG. 3 is a signal normalization wavelet energy feature solution flow chart of an embodiment of the present invention;
FIG. 4 is a graph comparing the signal recovery of an embodiment of the present invention with a conventional unwind algorithm;
FIG. 5 is a signal-to-noise ratio versus conventional convolution algorithm for an embodiment of the present invention;
FIG. 6 is a graph showing the frequency of center of gravity, frequency variance, power spectral entropy, and sample entropy for the example 4 signals;
FIG. 7 is a graph showing normalized wavelet energy feature vectors for the signals of example 4;
fig. 8 is a schematic diagram of recognition results of a multi-dimensional fusion feature vector SVM of a single train in an embodiment.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples.
The method for identifying the railway track diseases based on the optical fiber distributed vibration detection (shown in figure 1) comprises the following specific implementation steps:
and step 1, transmitting detection light pulses to a sensing optical fiber by adopting an optical fiber distributed sensing system, demodulating an original large-amplitude vibration signal by adopting a differential compression-reduction dynamic range expansion algorithm based on wavelet self-adaptive denoising, quantitatively monitoring vibration information along the sensing optical fiber in real time, and acquiring a demodulated original vibration signal S (4000,900).
And 2, constructing an initial data set for the demodulated original vibration signal S (4000,900) obtained in the step 1, and marking data samples of the normal rail signals and the 3 disease signals (fish scale injury, wave abrasion injury and empty hanging injury). And respectively carrying out feature extraction on the signals by using 4 single feature values of center-of-gravity frequency, standard deviation frequency, power spectrum entropy and sample entropy, and then carrying out wavelet decomposition test on 1 group of data of each of the 4 signals in the initial data set to determine that the number of decomposition layers L=5 for wavelet decomposition, wherein when the wavelet function db9 is used for wavelet decomposition, the 4 signals have the largest degree of distinction.
The signal is subjected to wavelet decomposition, and a normalized wavelet energy feature vector of the signal is calculated. And constructing a multidimensional fusion feature vector by using a vector splicing method to construct a single feature value of the 4 signals and the normalized wavelet energy feature vector, so as to obtain an initial data set.
And 3, screening 150 groups of data of the normal rail signals and 3 disease signals (fish scale injury, wave injury and air injury) from the initial data set constructed in the step 2, constructing a multidimensional fusion feature vector sample set for each group of data, obtaining a weight vector of the sample data set as [7.26,606.42,0.18,0.31,0.20,0.11,0.07,0.03,0.0062,0.0025] according to the average standard deviation of multidimensional fusion feature vectors of each disease type data as the weight, and pre-weighting the multidimensional fusion feature vectors to obtain a pre-weighted multidimensional fusion feature vector sample set.
And 4, training and testing the Support Vector Machine (SVM) by using the pre-weighted multidimensional fusion feature vector sample set obtained in the step 3, and selecting a kernel function of the SVM as a Gaussian kernel function, wherein a penalty coefficient c=131 and a kernel parameter g=3.85. The pre-weighted multidimensional fusion feature vector of the demodulated vibration signal obtained by actual acquisition is input into a support vector machine SVM model, so that the classification and identification of the normal rail signal and 3 disease signals are realized, the position and the type of the disease are recorded, the position and the type of the disease are output, and the real-time disease diagnosis and positioning output are realized.
In this embodiment, the data sample used is the vibration signal data of multiple trains on a heavy haul railway. FIG. 2 is a flow chart of a differential compression-reduction dynamic range expansion algorithm. Fig. 3 is a flowchart of a normalized wavelet energy feature vector algorithm, taking the number of decomposition layers n=3 and the wavelet function db2 as an example. Fig. 4 is a signal reduction condition comparison of the differential compression-reduction dynamic range expansion algorithm of the present embodiment and the conventional unwrapping algorithm, and fig. 5 is a signal-to-noise ratio comparison of the differential compression-reduction dynamic range expansion algorithm of the present embodiment and the conventional unwrapping algorithm.
In the embodiment, the original signal after the expansion of the differential compression-reduction dynamic range in the step 1 is well restored in both the time domain and the frequency domain, and the signal-to-noise ratio of the original signal is further improved by introducing a wavelet adaptive denoising algorithm. In fig. 4, the dashed line is the error recovery condition of the conventional unwrapping algorithm for the large-amplitude signal, and the solid line is the demodulation result of the embodiment, so that it can be seen from the graph that the dynamic range expansion algorithm provided by the present invention can effectively recover the original large-range signal, and the conventional unwrapping algorithm has the condition of unable correct recovery.
Fig. 5 is a comparison of the signal-to-noise ratio of the traditional convolution algorithm and the differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising in the present embodiment, where both methods correctly restore the original waveform of the signal, but the dynamic range expansion algorithm provided by the present invention has a relatively higher signal-to-noise ratio.
In method step 2 above, center of gravity frequency, frequency variance, spectral entropy, sample entropy, and normalized wavelet energyThe feature vectors are all time-frequency domain related feature values, and the effective restoration and signal-to-noise ratio improvement of the time-frequency domain differential compression-restoration dynamic range expansion algorithm based on wavelet self-adaptive denoising ensures the accuracy of original signal feature extraction, and further improves the accuracy of final classification recognition. Four exemplary signal extraction result pairs such as center of gravity frequency, frequency variance, spectral entropy and sample entropy are shown in FIG. 6, where I 1 、I 2 、I 3 And I 4 The characteristic values of the 4 signal samples are obviously different, and the distinction can be made. Using the number of decomposition layers n=5, the wavelet function db9 performs wavelet decomposition on 4 typical signal samples to find normalized wavelet energy feature vectors, and the differences between the 4 signal samples are apparent from the normalized wavelet energy feature vector pairs of the 4 signals, such as those shown in fig. 7.
And vector splicing is further carried out on the extracted characteristic value and the normalized wavelet energy characteristic vector, a multidimensional characteristic vector is constructed, pre-weighting is carried out, and the multidimensional characteristic vector is input into a trained support vector machine SVM for classification and identification. FIG. 8 is a schematic diagram showing the results of SVM classification recognition after multidimensional fusion feature extraction of single pass train signals, wherein I 1 、I 2 、I 3 And I 4 The signals are normal track signals, fish scale injury, wave abrasion injury and empty hanging injury respectively. As can be seen from the figure, the method can realize the track defect monitoring and early warning of about 9km along the heavy-load railway, the early warning result and the defect position of the actual track can be in one-to-one correspondence, and the effectiveness of the method is fully illustrated.
In summary, the invention acquires the acoustic wave sensing signal by connecting the existing communication optical cable beside the railway, and demodulates and acquires the original vibration signal S (p, q) by adopting a differential compression-reduction dynamic range expansion algorithm based on wavelet self-adaptive denoising; then, extracting the characteristics of the expanded demodulation signals, and forming a multidimensional fusion characteristic vector through vector splicing; pre-weighting the feature vector based on the data sample set; and finally, classifying and identifying the feature vector by using the trained SVM model, recording the existence position of the disease, and outputting the disease position alarm of the optical fiber length. The invention realizes the identification and classification of typical diseases of the rails such as scale injury, wave abrasion injury, overhead injury and the like on the railway, positions and alarms the positions of the diseases in real time, and provides a long-term online monitoring means for the rail safety operation and maintenance monitoring of railway transportation.
Claims (5)
1. The railway track disease identification method based on the optical fiber distributed vibration detection is characterized by comprising the following steps of:
step 1, transmitting detection light pulses to a sensing fiber through an optical fiber distributed acoustic sensor system DAS to realize real-time quantitative monitoring of vibration information of the sensing fiber along the line, and acquiring original vibration signals S (p, q), wherein p is the number of acquired pulses, and q is the number of acquisition sections of the acoustic vibration signals along the line; the original acoustic vibration signals S (p, q) are obtained by demodulating optical fiber distributed acoustic wave sensing signals through a differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising;
step 2, constructing an initial data set for the original vibration signals S (p, q) obtained in the step 1, and marking data samples of the normal rail signals and the 3 disease signals; carrying out time-frequency domain feature extraction on sample data in an initial data set by using gravity center frequency, standard deviation frequency, power spectrum entropy, sample entropy and normalized wavelet energy feature vectors respectively, and completing construction of a multidimensional fusion feature vector by using a vector splicing method to obtain an initial multidimensional fusion feature vector data set; the 3 disease signals are fish scale injury, bruise injury and empty hanging injury;
step 3, selecting M groups of data of the normal rail signals and the 3 disease signals from the initial multidimensional fusion feature vector data set obtained in the step 2, constructing a multidimensional fusion feature vector data set, and pre-weighting the multidimensional fusion feature vectors based on the average standard deviation ASD of the multidimensional fusion feature vectors of the normal rail signals and the 3 disease signals as the weight of each feature value of the multidimensional fusion feature vectors to obtain a pre-weighted multidimensional fusion feature vector sample set;
and 4, inputting the pre-weighted multidimensional fusion feature vector sample set obtained in the step 3 into a Support Vector Machine (SVM) model for classification training, using the trained SVM model to realize classification identification of normal rail signals and 3 disease signals, recording the positions and types of the diseases, alarming, and realizing real-time disease diagnosis and positioning output.
2. The method for identifying railway track diseases based on optical fiber distributed vibration detection as claimed in claim 1, wherein:
in the step 1, a differential compression-reduction dynamic range expansion algorithm based on wavelet adaptive denoising demodulates an optical fiber distributed acoustic wave sensing signal, and the specific steps include:
step 1-1, using an arctangent demodulation algorithm to demodulate the optical fiber distributed acoustic wave sensing signal, the step of solving the phase using an arctangent function results in the phase value being limited to [ -pi, pi]In the range of (2), a phase unwrapping is required to obtain a true phaseThe classical phase unwrapping algorithm is expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the original wrapped signal, < >>Is an unwind signal; the original wrapped signal may be represented asWhere k (n) is an integer series of numbers, the above formula can be further expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the unwind threshold;
performing an l-order differential compression on the original phase wrapped signal:
wherein% l k (N) is an integer number sequence, so that a deconvolution algorithm can be used to introduce a discrete noise N noise (n);
Step 1-2: handleSine signal regarded as N discrete samples +.>Sum of which A is i Is the signal amplitude omega i Is the signal frequency, θ i Is the primary phase of the signal; for each sinusoidal signal +.>Having a derivative of order I
The unwind threshold at this point may be expressed as:
when l epsilon [1,5], phi < pi can be satisfied, and then the winding can be unwound to obtain:
when the signal is affected by the post-differentiation noise, this results in phi>Pi, integer number column is equal to l k (n) cannot be completely eliminated by the unwind algorithm, leaving misunderstanding the winding term k' (n); directly carrying out first-order integration on the unwrapped signal to obtain:
where K '(n) is a complex polynomial of order l, so that the misinterpreted convolution term K' (n) and noise delta can be removed in the differential stage using wavelet denoising algorithms l N noise (n) and thus achieve proper demodulation.
3. The method for identifying railway track diseases based on optical fiber distributed vibration detection as claimed in claim 1, wherein:
the specific steps of the step 2 of carrying out time-frequency domain feature extraction by normalizing wavelet energy feature vectors are as follows:
step 2-1, calculating normalized wavelet energy eigenvectors of signals; wavelet decomposition is carried out on the collected and restored signals, each decomposition layer keeps high-frequency wavelet coefficients, the low-frequency wavelet coefficients are further decomposed, and after L-layer wavelet decomposition, the low-frequency wavelet coefficients cA can be obtained L And high frequency wavelet coefficient cD 1 ,cD 2 ,…cD L ;
Step 2-2, the high-frequency wavelet coefficient and the low-frequency wavelet coefficient are sequentially arranged from low frequency to high frequency, and the energy of each wavelet coefficient is solved by the formula (7)Quantity E j Form wavelet energy feature vector E= [ E ] 1 ,E 2 ,…,E L+1 ];
Wherein C is j (m) is the wavelet coefficient after decomposition, L is the number of wavelet decomposition layers;
step 2-3, carrying out normalization processing on the wavelet energy feature vector obtained in the step 2-2 according to a formula (8) to obtain a normalized wavelet energy feature vector E '= [ E ]' 1 ,E′ 2 ,…,E′ L+1 ];
Step 2-4, selecting 1 group of data of 4 signals, namely a normal rail signal and 3 disease signals in an initial data set, performing wavelet decomposition test, enumerating different wavelet decomposition layers and wavelet functions, performing wavelet decomposition respectively to obtain normalized wavelet energy feature vectors, calculating an average standard deviation ASD under the conditions of the different decomposition layers and the wavelet functions according to a formula (9) and a formula (10), as the degree of distinguishing of the energy vectors, and determining the wavelet decomposition layers and the wavelet functions with the maximum degree of distinguishing so as to enable the 4 types of signals to have the maximum degree of distinguishing;
4. The method for identifying railway track diseases based on optical fiber distributed vibration detection as claimed in claim 1, wherein:
the specific step of pre-weighting the multidimensional fusion feature vector in the step 3 comprises the following steps:
step 3-1, screening M groups of data of normal rail signals and 3 disease signals from the initial data set constructed in the step 2, wherein M is more than or equal to 100, and constructing a multidimensional fusion feature vector data set;
step 3-2, combining 4 single characteristic values of gravity center frequency, frequency standard deviation, power spectrum entropy and sample entropy and normalized wavelet energy characteristic vectors with the length of L+1 for each multidimensional fusion characteristic vector, wherein the total length is L+5; each characteristic value differs for four types of signal discrimination, and the signal discrimination corresponding to each characteristic value is calculated as the weight w of the characteristic value according to the formula (11) n Constructing the weight vector of the sample data set as [ w ] 1 ,w 2 ,…w N+5 ]Weighting all the obtained multidimensional fusion feature vectors to construct a multidimensional fusion feature vector [ w ] after being pre-weighted 1 c 1 ,w 2 c 2 ,…w N+ 5 c N+5 ];
Wherein M is the number of groups of each signal data in the data set, c s And the s-th eigenvalue of the multidimensional fusion eigenvector, r is the corresponding sample group number, and a is the corresponding sample type.
5. The method for identifying railway track diseases based on optical fiber distributed vibration detection according to claim 1, wherein the step 4 specifically comprises:
step 4-1: training a support vector machine by using the pre-weighted multidimensional fusion feature vector sample set obtained in the step 3 to obtain a trained support vector machine SVM;
step 4-2: carrying out engineering deployment on the trained support vector machine SVM, carrying out feature extraction and pre-weighting on the original vibration signals by using the methods of the step 2 and the step 3, taking the pre-weighted multidimensional fusion feature vector as an input vector of the support vector machine, and completing classification recognition;
step 4-3: recording the identification result of the support vector machine SVM, outputting the position and type of the disease, and realizing real-time disease diagnosis and positioning output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310249537.0A CN116297841A (en) | 2023-03-15 | 2023-03-15 | Railway track disease identification method based on optical fiber distributed vibration detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310249537.0A CN116297841A (en) | 2023-03-15 | 2023-03-15 | Railway track disease identification method based on optical fiber distributed vibration detection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116297841A true CN116297841A (en) | 2023-06-23 |
Family
ID=86833829
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310249537.0A Pending CN116297841A (en) | 2023-03-15 | 2023-03-15 | Railway track disease identification method based on optical fiber distributed vibration detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116297841A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116956226A (en) * | 2023-09-19 | 2023-10-27 | 之江实验室 | DAS dynamic range improving method and device based on self-supervision type signal fusion |
-
2023
- 2023-03-15 CN CN202310249537.0A patent/CN116297841A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116956226A (en) * | 2023-09-19 | 2023-10-27 | 之江实验室 | DAS dynamic range improving method and device based on self-supervision type signal fusion |
CN116956226B (en) * | 2023-09-19 | 2023-12-22 | 之江实验室 | DAS dynamic range improving method and device based on self-supervision type signal fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Fault detection method for railway wheel flat using an adaptive multiscale morphological filter | |
Lederman et al. | Track-monitoring from the dynamic response of an operational train | |
CN107406090B (en) | Abnormal vehicle dynamics detection | |
Zhang et al. | Adaptive fault feature extraction from wayside acoustic signals from train bearings | |
Wang et al. | Optimization of segmentation fragments in empirical wavelet transform and its applications to extracting industrial bearing fault features | |
CN106323442A (en) | Railway health monitoring method based on distributed optical fiber vibration sensing system | |
CN105929025B (en) | One kind is based on the continuous wheel tread of the time and space and rail deformation detection method | |
CN116297841A (en) | Railway track disease identification method based on optical fiber distributed vibration detection | |
Ye et al. | Shock detection of rotating machinery based on activated time-domain images and deep learning: An application to railway wheel flat detection | |
CN105022912A (en) | Rolling bearing fault prediction method based on wavelet principal component analysis | |
JP4431163B2 (en) | Abnormality detection system for moving body and abnormality detection method for moving body | |
Wan et al. | Anomaly detection of train wheels utilizing short-time Fourier transform and unsupervised learning algorithms | |
CN113281414B (en) | Method and device for identifying short-wave disease types of steel rails and electronic equipment | |
GB2383635A (en) | Chromatic analysis of measured acoustic signals from a system | |
Dai et al. | Acceleration-guided acoustic signal denoising framework based on learnable wavelet transform applied to slab track condition monitoring | |
Cerullo et al. | Acoustic signal processing to diagnose transiting electric trains | |
JP4044837B2 (en) | Abnormality detection system for moving body, and abnormality detection method for moving body | |
CN107782548A (en) | One kind is based on to track vehicle parts detecting system | |
CN104271428B (en) | Method for investigating Wheel Rail Contact | |
Alten et al. | Detecting and classifying rail corrugation based on axle bearing vibration | |
CN113343919A (en) | Method and device for detecting continuous equidistant rubbing damage of steel rail and computer equipment | |
US20200284671A1 (en) | Method for detecting an anomaly of a rolling equipment exploiting a deformation signal from a rail support | |
WO2005003698A1 (en) | Monitoring the performance of a system by measuring a time-variant acoustic signal from the system | |
Wang et al. | A fiber-sensor-based long-distance safety monitoring system for buried oil pipeline | |
CN112986388A (en) | Turnout switch blade defect detection method and system based on broadband excitation |
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 |