CN115290133A - Method and system for monitoring track structure at joint of light rail platform - Google Patents
Method and system for monitoring track structure at joint of light rail platform Download PDFInfo
- Publication number
- CN115290133A CN115290133A CN202210758978.9A CN202210758978A CN115290133A CN 115290133 A CN115290133 A CN 115290133A CN 202210758978 A CN202210758978 A CN 202210758978A CN 115290133 A CN115290133 A CN 115290133A
- Authority
- CN
- China
- Prior art keywords
- sensor
- signal
- displacement
- audio
- fusion
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- 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
- G06N3/08—Learning methods
Abstract
The invention discloses a method and a system for monitoring a track structure at a joint of a light rail platform, wherein the method collects various sensor signals and audio signals, and obtains primary signal fusion results of the sensors and primary signal fusion results of the audio characteristics by respectively carrying out feature extraction and primary signal fusion on the various sensors and the audio signals; and further performing secondary signal fusion on the two primary signal fusion results, thereby obtaining a monitoring result of the track structure at the joint of the light rail platform according to the signal fusion result. The invention realizes the intelligent monitoring of the rail structure at the joint of the light rail platform, thereby realizing the early discovery of the problems and avoiding the occurrence of serious accidents.
Description
Technical Field
The invention relates to the field of light rail platform monitoring, in particular to a method and a system for monitoring a track structure at a joint of a light rail platform.
Background
The light rail is a brand new traffic tool in modern society, has the advantages of high running speed, strong transportation capability and the like, and meanwhile, compared with the subway, the construction cost of the light rail is much lower, so that the comprehensive development of the light rail is an important support for the sustainable and stable development of the economy in China. At present, the construction of light rails has entered a climax, and many cities are actively constructing light rails. The research on the track monitoring of the light rail is less, and particularly the monitoring of the track structure at the connection position of the light rail platform is carried out.
In the prior art, most of the monitoring of the light rail platform or the rail structure is in the construction stage of a rail bridge and a platform, and the research on the structure monitoring of the light rail in the use stage is almost blank.
For the light rail elevated station platform, because the platform structure influences the acting force of the rail at the platform section of the light rail track, the stress of the rail at the platform section is different from the stress of the rail at other bridge sections, therefore, the stress of the rail structure at the platform connection part also comprises the influence of the elevated station structure on the rail structure, and the conditions of fatigue, deformation, even damage and the like are easy to occur to the rail structure at the light rail platform connection part. Therefore, it is necessary to monitor the track structure state at the connection of the light rail platform.
Disclosure of Invention
Technical problem to be solved
In order to solve the technical problems, the invention provides a method and a system for monitoring a track structure at a light rail platform joint.
(II) technical scheme
In order to solve the technical problems and achieve the purpose of the invention, the invention is realized by the following technical scheme:
in a first aspect, the invention discloses a method for monitoring a track structure at a joint of a light rail platform, which comprises the following steps:
arranging a plurality of sensors and audio acquisition equipment at the joint of the light rail platform, and acquiring the sensors and audio signals; the plurality of sensors comprise an orbit settlement meter, a strain sensor, a displacement sensor and an acceleration sensor;
respectively processing the sensor signals and the audio signals, wherein the processing comprises denoising and feature extraction of the sensor signals and the audio signals and feature fusion based on a neural network, namely primary signal fusion;
and performing secondary signal fusion on the two primary signal fusion results based on a D-S evidence theory.
And obtaining a monitoring result of the track structure at the joint of the light rail platform based on the signal fusion result, and performing classified output, wherein the classification result comprises the combination of structural damage, structural deformation, settlement, displacement and various abnormal conditions. And outputting the monitoring result to a remote server side, and sending warning information to the operation maintenance terminal.
Further, the sensors also include a pressure sensor, a speed sensor, and a temperature sensor.
Further, the rail settlement meter is arranged on the surface of the rail and used for detecting the settlement of the rail; the strain sensor is arranged at the lower side of the bridge deck pavement and used for detecting the strain of the bridge deck; the displacement sensor and the acceleration sensor are arranged on the bridge span, and the displacement sensor is used for measuring the displacement of the part of the viaduct bridge relative to the bridge pier, wherein the displacement comprises transverse displacement, longitudinal displacement and lateral rotation displacement; the acceleration sensor is used for measuring the vibration of the track structure.
Further, the audio signal denoising comprises improved empirical mode decomposition denoising, including processing the original signal, adding different white noise sequences into the original signal, and performing empirical mode decomposition based on an improved envelope fitting method;
the amplitude value of the added white noise is as follows:
where a is the amplitude of white noise, E is the amplitude of the original signal, and pa is the high frequency component in the original signal.
The improved envelope fitting method comprises the following steps:
calculating all local maximum and minimum points of the original signal, and calculating slopes between adjacent data points, and calculating a ratio of the slopes:
judging whether the data point is an extreme point according to the following conditions:
wherein G is a slope ratio threshold, the degree of curvature of a signal data point is obtained from the slope ratio, the greater the value of the degree of curvature, the more curvature, and if the slope ratio satisfies the threshold condition, the point is determined to be a false extreme point, and the point is denoted as D (t).
Preferably, G can be 1.5.
Further dividing the obtained false extreme points into false maximum points and false minimum points, and distinguishing the false maximum points and the false minimum points according to the concave-convex property of the original signal data point curve:
and respectively combining the maximum value point and the minimum value point based on the obtained false maximum value point and false minimum value point to form a new extreme value sequence.
Further, the sensor features include: average, energy value, maximum gradient value, average differential value, and variance value.
The audio feature parameters include: time domain characteristic parameters, waveform characteristics and frequency domain characteristic parameters; the time domain characteristic parameters comprise a mean value, a maximum value, a minimum value, a root mean square and a variance; the waveform characteristics comprise peak value, kurtosis, skewness and margin factors; the frequency domain characteristic parameters comprise frequency spectrum gravity center, 1/N octave and mean square frequency.
In a second aspect, the present invention further provides a system for monitoring a track structure at a connection of a light rail platform, including:
the system comprises a track settlement meter, a strain sensor, a displacement sensor, an acceleration sensor, a pressure sensor, a speed sensor, a temperature sensor and audio acquisition equipment;
and the sensor signal processing module is used for respectively processing the signals acquired by the track settlement gauge, the strain sensor, the displacement sensor, the acceleration sensor, the pressure sensor, the speed sensor and the temperature sensor, and comprises the step of filtering the sensor signals.
And performing feature extraction on the processed sensor data, and performing primary signal fusion on the features of the plurality of sensors based on a neural network model.
And the audio signal processing module is used for denoising the audio signal based on empirical mode decomposition, extracting characteristics including time domain characteristics, waveform characteristics and frequency domain characteristics, and performing primary signal fusion on the audio characteristics based on a neural network model.
And the data fusion module is used for performing secondary signal fusion based on the result of primary signal fusion respectively performed on the data of various sensors and the audio data in a secondary signal fusion mode.
And the output module is used for obtaining a monitoring result of the track structure at the joint of the light rail platform based on the signal fusion result and carrying out classified output, wherein the classified result comprises the combination of structural damage, structural deformation, settlement, displacement and various abnormal conditions. And outputting the monitoring result to a remote server side, and sending warning information to the operation maintenance terminal.
In a third aspect, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform the method for monitoring a track structure at a light rail station connection in the first aspect or any implementation form of the first aspect.
(III) advantageous effects
The beneficial effects of the invention are as follows:
(1) Based on various sensors and an intelligent data fusion algorithm combined with audio signals, the monitoring of the track structure at the connection position of the light rail platform is realized.
(2) The track is monitored in an all-dimensional mode based on various sensors including a track settlement meter, a strain sensor, a displacement sensor, an acceleration sensor, a pressure sensor, a speed sensor and a temperature sensor, the influences of load action, temperature change and the like on a measurement result are considered, and the measurement accuracy is improved.
(3) And based on the information fusion mode of the second-level fusion, the sensor signal and the audio signal are respectively subjected to the first-level signal fusion, and further based on the D-S evidence theory, the result of the first-level signal fusion is subjected to the second-level signal fusion, so that the multi-level signal fusion is realized.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a schematic flow chart illustrating a method for monitoring a track structure at a connection of a light rail station according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an improved empirical mode decomposition flow according to an embodiment of the present application;
fig. 3 is a schematic diagram of a signal fusion method according to an embodiment of the present application.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure in the specification. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be carried into practice or applied to various other specific embodiments, and various modifications and changes may be made in the details within the description and the drawings without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present disclosure, and the drawings only show the components related to the present disclosure rather than the number, shape and size of the components in actual implementation, and the type, amount and ratio of the components in actual implementation may be changed arbitrarily, and the layout of the components may be more complicated.
Referring to fig. 1, a method for monitoring a track structure at a connection point of a light rail platform includes the following steps:
s1: arranging a sensor and audio acquisition equipment and acquiring a sensor and an audio signal;
the sensor comprises a track settlement meter, a strain sensor, a displacement sensor and an acceleration sensor; the setting position comprises that the track settlement meter is arranged on the surface of the track and used for detecting the settlement of the track; the strain sensor is arranged at the lower side of the bridge deck pavement and used for detecting the strain of the bridge deck; the displacement sensor and the acceleration sensor are arranged on the bridge span, and the displacement sensor is used for measuring the displacement of the part of the viaduct bridge relative to the bridge pier, wherein the displacement comprises transverse displacement, longitudinal displacement and lateral rotation displacement; the acceleration sensor is used for measuring the vibration of the track structure.
The sensors of various types are uniformly arranged on the elevated station section and the front section and the rear section of the elevated station section of the light rail, the elevated station section extends forwards for a certain distance from the initial position of the platform to the final position of the platform and extends backwards for a certain distance, the extending distance is determined by actual needs, and optionally, the extending distance is 20 meters.
The audio acquisition equipment is used for acquiring track audio information, when the track structure at the platform joint is deformed or cracked and damaged, the gravity of the light rail vehicle entering the station can generate acting force on the damaged or deformed part, so that weak abnormal sound can be generated in the running process of the vehicle, and the audio acquisition equipment is different from audio signals in a normal state.
The knowledge of structural mechanics shows that the main causes of the deflection of the structure include load effect, temperature change, material expansion and contraction, and the like. The deflection caused by the load action can be calculated by the following formula:
wherein the content of the first and second substances,
、
、
bending moment, axial force and shearing force caused by actual load are respectively measured;
、
、
bending moment, axial force and shearing force caused by the dummy unit load respectively;
、
、
are respectively bending resistantStiffness, shear stiffness and tensile stiffness, k being the section coefficient.
Therefore, based on the influence of the track load, in order to improve the measurement accuracy, the invention combines the weight of different light rail vehicle types and the number of passengers to obtain measurement data based on different loads, and based on the measurement data, a pressure sensor is adopted to obtain the load weight, and a temperature sensor is adopted to obtain the temperature change.
The characteristics of the generated audio signals are different for different arrival speeds of the light rail train, so that the arrival speed of the light rail train needs to be considered in the aspect of processing the audio signals, and based on the arrival speed, the speed sensor is adopted to acquire the running speed of the light rail train.
S2: processing the sensor signal and the audio signal separately
S21: processing sensor signals
(1) And respectively processing signals acquired by the track settlement meter, the strain sensor, the displacement sensor, the acceleration sensor, the pressure sensor, the speed sensor and the temperature sensor, including filtering the sensor signals.
(2) And performing feature extraction on the processed sensor data, wherein the feature extraction specifically comprises an average value, an energy value, a maximum gradient value, an average differential value and a variance value.
The average differential value can comprehensively reflect the whole information process of the dynamic response process of the sensor, and is defined as:
(3) Performing primary signal fusion on the characteristics of the sensors based on a neural network model, and comprising the following steps of:
a. normalization process
And carrying out normalization processing on the characteristic data of different sensors to ensure the consistency of the characteristic data of each sensor.
b. Establishing a network and training a network
According to the invention, the LSTM recurrent neural network algorithm has the best fusion effect of performing data fusion on the characteristic data of the sensor according to the test of the inventor on various neural network models, so that the invention adopts the LSTM recurrent neural network algorithm to perform primary signal fusion on the sensor data, solves the problem of the traditional recurrent neural network on gradient and realizes the learning of long-term dependence information.
Further, the method also comprises the step of further improving the standard LSTM recurrent neural network model, so that the problem of gradient disappearance is avoided.
The forgetting door is used for judging whether the content of the memory unit is reserved or not, so that the memory content is screened, and the output is as follows:
wherein the content of the first and second substances,
to forget the output of the gate function, sig is the activation function,
to weight between the input and the forgetting gate,
as a weight between the history output and the forgetting gate,
in order to forget the biasing of the door,
for the input at the time t, the input is,
is the output at time t-1.
Alternatively, the forgetting gate activation function may be a RELU function.
The input gate is used for feature import and updating the state of the unit, when other parts of the neural network are to be written into the storage unit, the input gate needs to be passed first, and the output is:
wherein the content of the first and second substances,
is the output of the input gate or gates,
for the weight between the input and the input gate,
for the weight between the history output and the input gate,
biasing the input gate.
In addition, after obtaining the forgotten gate output and the input gate output, the current memory cell state can be obtained as follows:
wherein the content of the first and second substances,
is a state of a candidate cell, and,
is the weight between the input and the cell state;
is the weight between the history output and the cell state;
the cell state is biased.
The output gate is used for controlling the final output of the unit and determining whether external neurons can read out values from the storage unit, the unit output is obtained by multiplying the output of the output gate by the current unit state information and passing through an activation function, and therefore, the output gate and the unit output are as follows:
wherein, the first and the second end of the pipe are connected with each other,
in order to output the output of the gate,
is the weight between the input and output gates;
is the weight between the history output and the output gate;
in order to offset the output gate(s),
and is output by the unit at the time t.
And training the network model according to a network training method of a neural network algorithm. The model training adopts supervised learning, inputs marked training data, and realizes the training of the LSTM recurrent neural network through a large amount of data learning and iteration.
S22: audio signal processing
(1) Denoising process
The source of the platform noise of the light rail station is complex, and besides the sound of the light rail train, the platform noise also comprises great environmental noise and noise caused by people flow, so that the denoising processing of the collected audio signal is particularly important, and the denoising processing is carried out according to the characteristics of the audio signal.
The method comprises the steps of firstly obtaining station platform background noise when a light rail train does not enter a station, wherein the background noise is synthesized noise of environmental noise and people stream noise, and the background noise is mostly low-frequency noise known from the source of the noise, so that a first audio signal collected by audio collection equipment is subtracted from a background noise signal to obtain a second audio signal.
Further, audio denoising is carried out based on empirical mode decomposition, the second audio signal is subjected to denoising to obtain a third audio signal, and denoising processing of the non-stationary signal is achieved.
Common denoising methods include Fourier transform, wavelet denoising and the like, but the Fourier denoising can only be performed on stationary signals, the wavelet denoising has the problem of wavelet basis selection, wavelet basis functions cannot be selected in a self-adaptive manner, and therefore the denoising effect is poor due to the fact that the denoising method cannot be adapted to various regions of signals, the empirical mode decomposition decomposes the signals into a plurality of components, the basis functions and the decomposition layer number do not need to be set in advance, the signals are decomposed in a self-adaptive manner through a repeated iteration mode according to the characteristics of the signals, and therefore the denoising method is excellent in non-stationary signal processing.
Therefore, for the second audio signal after the background noise signal is removed, the method of the invention performs denoising in an empirical mode decomposition based mode, and can more effectively implement denoising treatment, specifically comprising the following steps:
the empirical mode decomposition generates intrinsic mode functions with the frequency from high to low and the size from small to large one by one, and the high-frequency intrinsic mode function influences the formation of the low-frequency intrinsic mode function. The specific steps of empirical mode decomposition of the signal Y (t) are as follows:
A. solving a maximum value point of the signal Y (t), fitting an envelope u (t) by utilizing cubic spline interpolation, solving a minimum value point of the Y (t), and fitting an envelope v (t) by utilizing the same method;
B. averaging u (t) and v (t) to obtain m (t):
and m (t) is the mean envelope of Y (t), and the mean envelope is subtracted from the solved signal Y (t) to obtain a new signal h (t).
C. And calculating a new signal h (t), judging whether the new signal h (t) meets the characteristics of the intrinsic mode functions, if so, recording the current h (t) as a first intrinsic mode function IMF1 (t), if not, recording Y (t) = h (t), repeating the steps until the characteristics of the intrinsic mode functions are met, and recording as IMF1 (t).
D. Calculate the first order residual signal r1 (t):
E. taking the residual signal as the signal to be processed, repeating the steps (1) to (4) to obtain a series of intrinsic mode functions
、
、
......
And a residual signal
、
......
. When the temperature is higher than the set temperature
Corresponding residual signal
And when the value is smaller than the preset threshold value, finishing the decomposition.
Further, an empirical mode decomposition algorithm can be improved, and the conventional classical empirical mode decomposition algorithm has the problem of noise aliasing, so that the denoising effect is not ideal, therefore, the conventional algorithm is further improved, and the improved algorithm flow is shown in fig. 2:
a. processing the original signal, and selecting the white noise to be added according to the frequency component of the original signal by adding different white noise sequences into the original signal, specifically, determining the amplitude of the white noise to be added according to the following method:
where a is the amplitude of white noise, E is the amplitude of the original signal, and pa is the high frequency component in the original signal.
b. Empirical mode decomposition is performed based on an improved envelope fitting method, so that each IMF component can be more fit to an original signal.
In the classical empirical mode decomposition method, only extreme points are considered during envelope fitting, so that the positions of the found extreme points may not be real extreme points, and therefore, the envelope fitting method is improved by the method, which specifically comprises the following steps:
calculating all local maximum and minimum points of the original signal, calculating slopes between adjacent data points, and calculating a ratio of the slopes:
judging whether the data point is an extreme point according to the following conditions:
wherein G is a slope ratio threshold, the degree of curvature of a signal data point is obtained from the slope ratio, the greater the value of the degree of curvature, the more curvature, and if the slope ratio satisfies the threshold condition, the point is determined to be a false extreme point, and the point is denoted as D (t).
Preferably, G can be 1.5.
Further dividing the obtained false extreme points into false maximum points and false minimum points, and distinguishing the false maximum points and the false minimum points according to the concave-convex property of the original signal data point curve:
and respectively combining the maximum value point and the minimum value point based on the obtained false maximum value point and false minimum value point to form a new extreme value sequence.
Based on the improvement on the envelope fitting method, the phenomena of overshoot and undershoot of the fitting envelope curve of the cubic spline curve are effectively solved.
(2) Feature extraction
And performing feature extraction based on the time domain feature parameters, the waveform features and the frequency domain feature parameters.
The time domain characteristic parameters comprise a mean value, a maximum value, a minimum value, a root mean square and a variance;
the waveform characteristics comprise peak value, kurtosis, skewness and margin factors;
the frequency domain characteristic parameters comprise a frequency spectrum center of gravity, a 1/N octave and a mean square frequency.
When a bridge structure is damaged or displaced, the characteristics of an audio signal in a frequency domain are changed more obviously than those of a signal in a time domain when a light rail train passes through, so that the method is indispensable to the analysis of the signal frequency domain. Because of the characteristics of the frequency domain, the invention has greater value for the above abnormity by testing different damage structures and displacements at different positions to obtain four parameters of the frequency spectrum gravity center, the 1/N octave, the mean square frequency and the frequency spectrum extensibility.
The spectral centroid reflects the approximate distribution of the power spectral density curve, which is smaller if the power spectral density curve is larger near the origin, and vice versa. The calculation formula of the center of gravity of the frequency spectrum is as follows:
wherein the content of the first and second substances,
，
fs is the sampling frequency and N is the number of sampling points.
The 1/N octave refers to dividing a frequency spectrum into a plurality of frequency bands and analyzing energy distribution of different frequency bands. The frequency range of each frequency band satisfies:
wherein the content of the first and second substances,
、
respectively, the upper and lower limit frequencies of the frequency band.
The mean square frequency represents the signal average energy, i.e., signal strength, and is expressed as follows:
(3) The method for performing first-level signal fusion on the audio features based on the neural network model comprises the following steps:
because the proportion of each audio feature to the fusion output result is different, the convergence rate of the neural network calculation can be accelerated by selecting the initial weight value of the neural network according to the correlation based on the correlation coefficient of each feature and the output result.
S3: performing two-stage signal fusion on sensor signal and audio signal
As shown in fig. 3, the data of various sensors and the audio data are respectively subjected to primary signal fusion based on a secondary signal fusion mode, and the two primary signal fusion results are further subjected to secondary signal fusion.
The global fusion of the two-stage signal fusion is carried out based on the D-S evidence theory, the neural network local fusion has the defect that the result is uncertain, and the D-S evidence theory is characterized in that the uncertainty problem can be solved. After the first-level signal fusion, local judgment can be obtained, then the output values of the neural networks in each area are normalized, and decision-level fusion is carried out by utilizing a D-S evidence theory.
Particularly, a fusion rule is determined based on the contradiction coefficient, when the contradiction coefficient is smaller than a set threshold value, the classical D-S evidence theory is adopted for calculation, and when the contradiction coefficient is larger than or equal to the set threshold value, the divergence entropy-based D-S evidence theory is adopted for calculation.
The contradiction coefficients are calculated as follows:
the contradiction coefficient indicates the degree of contradiction between focal elements.
、
Respectively representing the assignment of two groups of basic probabilities under the same identification frame corresponding to focal elements
、
，
Are respectively as
、
The basic trust distribution values distributed to the corresponding focal elements are intersected to form a matrix and expressed as a measure
。
The divergence entropy D-S evidence theory is calculated as follows:
based on divergence measurement based on conditional distribution, the divergence calculation method adopted by the invention is as follows:
wherein the content of the first and second substances,
、
for the two forms of probability distribution where the random variable X exists,
is the information entropy.
S4: outputting the monitoring result
And (4) obtaining a monitoring result of the track structure at the joint of the light rail platform based on the signal fusion result, and performing classified output, wherein the classification result comprises the combination of structural damage, structural deformation, settlement, displacement and various abnormal conditions. And outputting the monitoring result to a remote server side, and sending warning information to the operation maintenance terminal.
In the embodiment, the monitoring of the track structure at the connection position of the light rail platform is realized by combining the intelligent data fusion algorithm based on various sensors and audio signals, and the serious conditions of derailment, rollover, bridge fracture and the like of the light rail train caused by damage and deformation of the track structure at the connection position of the light rail platform are avoided.
The embodiment of the present invention further provides a system for monitoring a track structure at a connection of a light rail platform, which includes:
the system comprises a track settlement meter, a strain sensor, a displacement sensor, an acceleration sensor, a pressure sensor, a speed sensor, a temperature sensor and audio acquisition equipment;
and the sensor signal processing module is used for respectively processing the signals acquired by the track settlement gauge, the strain sensor, the displacement sensor, the acceleration sensor, the pressure sensor, the speed sensor and the temperature sensor, and comprises the step of filtering the sensor signals.
And performing feature extraction on the processed sensor data, and performing primary signal fusion on the features of the plurality of sensors based on a neural network model.
And the audio signal processing module is used for denoising the audio signal based on empirical mode decomposition, extracting characteristics including time domain characteristics, waveform characteristics and frequency domain characteristics, and performing primary signal fusion on the audio characteristics based on a neural network model.
And the data fusion module is used for performing secondary signal fusion based on the result of primary data fusion respectively performed on the data of the various sensors and the audio data in a secondary signal fusion mode.
And the output module is used for obtaining a monitoring result of the track structure at the joint of the light rail platform based on the signal fusion result and carrying out classified output, wherein the classified result comprises the combination of structural damage, structural deformation, settlement, displacement and various abnormal conditions. And outputting the monitoring result to a remote server side, and sending warning information to the operation maintenance terminal.
Embodiments of the present invention also provide a non-transitory computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause the one or more processors to perform the aforementioned method for monitoring a track structure at a light rail station junction.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solution of the present invention made by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims (10)
1. A method for monitoring a track structure at a connection part of a light rail platform is characterized by comprising the following steps:
arranging a plurality of sensors and audio acquisition equipment at the joint of the light rail platform, and acquiring the sensors and audio signals; the plurality of sensors comprise an orbit settlement meter, a strain sensor, a displacement sensor and an acceleration sensor;
respectively processing the sensor signals and the audio signals, wherein the processing comprises denoising and feature extraction of the sensor signals and the audio signals and feature fusion based on a neural network, namely primary signal fusion;
performing secondary signal fusion on the two primary signal fusion results;
obtaining a monitoring result of the track structure at the connection position of the light rail platform based on the signal fusion result, and performing classified output, wherein the classification result comprises structural damage, structural deformation, settlement, displacement and combination of the abnormal conditions; and outputting the monitoring result to a remote server side, and sending warning information to the operation maintenance terminal.
2. The method of claim 1, wherein the plurality of sensors further comprises a pressure sensor, a speed sensor, and a temperature sensor.
3. The method of claim 1, wherein the track settlement gauge is disposed on a surface of the track for detecting settlement of the track; the strain sensor is arranged at the lower side of the bridge deck pavement and used for detecting the strain of the bridge deck; the displacement sensor and the acceleration sensor are arranged on the bridge span, and the displacement sensor is used for measuring the displacement of the part of the viaduct bridge relative to the bridge pier, wherein the displacement comprises transverse displacement, longitudinal displacement and lateral rotation displacement; the acceleration sensor is used for measuring vibration of the track structure.
4. The method as claimed in claim 1, wherein the audio signal denoising comprises an improved empirical mode decomposition based denoising, which comprises processing the original signal, adding a different white noise sequence to the original signal, and performing an empirical mode decomposition based on an improved envelope fitting method.
5. The method of claim 4, wherein the white noise sequence has an amplitude value of:
where a is the amplitude of white noise, E is the amplitude of the original signal, and pa is the high frequency component in the original signal.
6. The method of claim 4, wherein the improved envelope fitting method comprises:
calculating all local maximum and minimum points of the original signal, and calculating slopes between adjacent data points, and calculating a ratio of the slopes:
judging whether the data point is an extreme point according to the following conditions:
wherein G is a ratio threshold of the slopes, the bending degree of the signal data point is obtained through the ratio of the slopes, the larger the value is, the more bending is shown, if the ratio of the slopes meets the threshold condition, the point is judged to be a false extreme point, and the point is marked as D (t);
further dividing the obtained false extreme points into false maximum points and false minimum points, and distinguishing the false maximum points and the false minimum points according to the concave-convex property of the original signal data point curve:
and respectively combining the maximum value point and the minimum value point based on the obtained false maximum value point and false minimum value point to form a new extreme value sequence.
7. The method of claim 1, wherein the sensor characteristics comprise: average, energy value, maximum gradient value, average differential value, and variance value.
8. The method of claim 1, wherein the audio parameters comprise: time domain characteristic parameters, waveform characteristics and frequency domain characteristic parameters; the time domain characteristic parameters comprise a mean value, a maximum value, a minimum value, a root mean square and a variance; the waveform characteristics comprise peak values, kurtosis, skewness and margin factors; the frequency domain characteristic parameters comprise frequency spectrum gravity center, 1/N octave and mean square frequency.
9. A system for monitoring the track structure at a light rail platform junction, the system comprising:
the system comprises a track settlement meter, a strain sensor, a displacement sensor, an acceleration sensor, a pressure sensor, a speed sensor, a temperature sensor and audio acquisition equipment;
the sensor signal processing module is used for respectively processing signals acquired by the track settlement gauge, the strain sensor, the displacement sensor, the acceleration sensor, the pressure sensor, the speed sensor and the temperature sensor, and comprises the step of filtering the sensor signals;
performing feature extraction on the processed sensor data, and performing primary signal fusion on the features of the plurality of sensors based on a neural network model;
the audio signal processing module is used for denoising the audio signal based on empirical mode decomposition, extracting characteristics including time domain characteristics, waveform characteristics and frequency domain characteristics, and performing primary signal fusion on the audio characteristics based on a neural network model;
the data fusion module is used for performing secondary signal fusion based on the result of primary signal fusion respectively performed on the data of various sensors and the audio data in a secondary signal fusion mode;
and the output module is used for obtaining a monitoring result of the track structure at the joint of the light rail platform based on the signal fusion result, classifying and outputting the monitoring result, wherein the classification result comprises the combination of structural damage, structural deformation, settlement, displacement and various abnormal conditions, outputting the monitoring result to a remote server side, and sending warning information to the operation and maintenance terminal.
10. A non-transitory computer-readable storage medium storing computer instructions which, when executed by one or more processors, cause the one or more processors to perform the method of light rail station joint track structure monitoring according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210758978.9A CN115290133A (en) | 2022-06-30 | 2022-06-30 | Method and system for monitoring track structure at joint of light rail platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210758978.9A CN115290133A (en) | 2022-06-30 | 2022-06-30 | Method and system for monitoring track structure at joint of light rail platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115290133A true CN115290133A (en) | 2022-11-04 |
Family
ID=83821870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210758978.9A Pending CN115290133A (en) | 2022-06-30 | 2022-06-30 | Method and system for monitoring track structure at joint of light rail platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115290133A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304581A (en) * | 2023-05-10 | 2023-06-23 | 佛山市钒音科技有限公司 | Intelligent electric control system for air conditioner |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1184931A (en) * | 1996-12-11 | 1998-06-17 | 唐秀家 | Method and apparatus for detecting and positioning leakage of fluid transferring pipeline |
| CN101900708A (en) * | 2010-08-18 | 2010-12-01 | 哈尔滨工业大学 | Vibration and audio signal-based high-speed train track defect detecting method |
| CN105719661A (en) * | 2016-01-29 | 2016-06-29 | 西安交通大学 | Automatic discrimination method for playing timbre of string instrument |
| CN110376002A (en) * | 2019-07-24 | 2019-10-25 | 西南交通大学 | A kind of high-speed train bogie method for diagnosing faults |
| CN112017639A (en) * | 2020-09-10 | 2020-12-01 | 歌尔科技有限公司 | Voice signal detection method, terminal device and storage medium |
| CN114386467A (en) * | 2022-01-29 | 2022-04-22 | 中国科学院声学研究所 | System and method for monitoring rail vehicle chassis fault abnormal sound based on acoustic characteristics |
-
2022
- 2022-06-30 CN CN202210758978.9A patent/CN115290133A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1184931A (en) * | 1996-12-11 | 1998-06-17 | 唐秀家 | Method and apparatus for detecting and positioning leakage of fluid transferring pipeline |
| CN101900708A (en) * | 2010-08-18 | 2010-12-01 | 哈尔滨工业大学 | Vibration and audio signal-based high-speed train track defect detecting method |
| CN105719661A (en) * | 2016-01-29 | 2016-06-29 | 西安交通大学 | Automatic discrimination method for playing timbre of string instrument |
| CN110376002A (en) * | 2019-07-24 | 2019-10-25 | 西南交通大学 | A kind of high-speed train bogie method for diagnosing faults |
| CN112017639A (en) * | 2020-09-10 | 2020-12-01 | 歌尔科技有限公司 | Voice signal detection method, terminal device and storage medium |
| CN114386467A (en) * | 2022-01-29 | 2022-04-22 | 中国科学院声学研究所 | System and method for monitoring rail vehicle chassis fault abnormal sound based on acoustic characteristics |
Non-Patent Citations (2)
| Title |
|---|
| 周国宪;伍星;刘韬;: "基于多传感器的神经网络和D-S证据理论在故障诊断中的应用", 测试技术学报 * |
| 谢春丽,夏虹,刘永阔: "多传感器数据融合技术在故障诊断中的应用", 传感器技术 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304581A (en) * | 2023-05-10 | 2023-06-23 | 佛山市钒音科技有限公司 | Intelligent electric control system for air conditioner |
| CN116304581B (en) * | 2023-05-10 | 2023-07-21 | 佛山市钒音科技有限公司 | Intelligent electric control system for air conditioner |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Fault detection method for railway wheel flat using an adaptive multiscale morphological filter | |
| Sarwar et al. | Deep autoencoder architecture for bridge damage assessment using responses from several vehicles | |
| CN103617684B (en) | Interference-type optical fiber circumference vibrating intruding recognizer | |
| Mousavi et al. | Damage detection and localization of a steel truss bridge model subjected to impact and white noise excitations using empirical wavelet transform neural network approach | |
| CN108170925A (en) | A kind of bridge damnification method for early warning based on arma modeling | |
| CN112834193B (en) | Operation bridge vibration and health state abnormity early warning method based on three-dimensional graph | |
| CN107885927B (en) | Railway bridge operation state early warning method | |
| CN105760934A (en) | Bridge abnormity monitoring restoration method based on wavelet and BP neural network | |
| CN102539098A (en) | Bridge dynamic load testing method based on neural network technology | |
| CN103994817A (en) | Vibration source identification method based on long-distance optical fiber frequent occurring events | |
| CN115879343B (en) | Ballastless track self-compacting concrete void two-stage identification method | |
| CN115290133A (en) | Method and system for monitoring track structure at joint of light rail platform | |
| Li et al. | Automatic uncoupling of massive dynamic strains induced by vehicle-and temperature-loads for monitoring of operating bridges | |
| CN108920766A (en) | A kind of Bridge Influence Line recognition methods indicated based on basic function with sparse regularization | |
| CN111353640B (en) | Method for constructing wind speed prediction model by combination method | |
| Liu et al. | Deep learning based identification and uncertainty analysis of metro train induced ground-borne vibration | |
| CN113204850A (en) | Temperature effect separation method in bridge deflection monitoring | |
| Chegeni et al. | New supervised learning classifiers for structural damage diagnosis using time series features from a new feature extraction technique | |
| Li et al. | Data-driven structural condition assessment for high-speed railway bridges using multi-band FIR filtering and clustering | |
| CN114595721A (en) | Vehicle part strain load prediction method based on road condition recognition | |
| CN103994816A (en) | Identification method based on optical fiber multiple events | |
| CN116359341A (en) | Method and device for monitoring damage condition of expansion joint based on noise | |
| Yessoufou et al. | Deep autoencoder model for direct monitoring of bridges subjected to a moving vehicle load under varying temperature conditions | |
| CN109711075B (en) | Steel girder bridge life and reliability analysis method based on sudden load nonlinear theory | |
| JP7452686B2 (en) | Vehicle detection device, method and program |
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 |