WO2019201177A1 - 列车部件裂纹损伤监测方法和系统 - Google Patents

列车部件裂纹损伤监测方法和系统 Download PDF

Info

Publication number
WO2019201177A1
WO2019201177A1 PCT/CN2019/082492 CN2019082492W WO2019201177A1 WO 2019201177 A1 WO2019201177 A1 WO 2019201177A1 CN 2019082492 W CN2019082492 W CN 2019082492W WO 2019201177 A1 WO2019201177 A1 WO 2019201177A1
Authority
WO
WIPO (PCT)
Prior art keywords
damage
detection structure
lamb wave
signal
response signal
Prior art date
Application number
PCT/CN2019/082492
Other languages
English (en)
French (fr)
Inventor
蔡国强
王坚群
何明
Original Assignee
江苏必得科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏必得科技股份有限公司 filed Critical 江苏必得科技股份有限公司
Priority to EP19788319.2A priority Critical patent/EP3783356A4/en
Publication of WO2019201177A1 publication Critical patent/WO2019201177A1/zh

Links

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
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/46—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by spectral analysis, e.g. Fourier analysis or wavelet analysis
    • 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
    • G01N29/041—Analysing solids on the surface of the material, e.g. using Lamb, Rayleigh or shear waves
    • 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
    • G01N29/043—Analysing solids in the interior, e.g. by shear waves
    • 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
    • G01N29/06—Visualisation of the interior, e.g. acoustic microscopy
    • G01N29/0654—Imaging
    • G01N29/069—Defect imaging, localisation and sizing using, e.g. time of flight diffraction [TOFD], synthetic aperture focusing technique [SAFT], Amplituden-Laufzeit-Ortskurven [ALOK] technique
    • 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
    • G01N29/4454—Signal recognition, e.g. specific values or portions, signal events, signatures
    • 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
    • G01N29/4472—Mathematical theories or simulation
    • 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
    • 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
    • G01N2291/025—Change of phase or condition
    • G01N2291/0258—Structural degradation, e.g. fatigue of composites, ageing of oils
    • 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
    • G01N2291/04—Wave modes and trajectories
    • G01N2291/042—Wave modes
    • G01N2291/0427—Flexural waves, plate waves, e.g. Lamb waves, tuning fork, cantilever
    • 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
    • G01N2291/10—Number of transducers
    • G01N2291/106—Number of transducers one or more transducer arrays
    • 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
    • G01N2291/26—Scanned objects
    • G01N2291/269—Various geometry objects
    • G01N2291/2696—Wheels, Gears, Bearings

Definitions

  • the invention relates to the technical field of damage analysis, in particular to a method and a system for monitoring crack damage of train components.
  • the locomotive and EMU trains are important equipments for rail transportation.
  • the braking acceleration is large, and the running parts and connecting parts are subject to huge stress changes.
  • the steering frame is responsible for supporting the car body, running and steering functions. Its anti-fatigue performance is of great significance to the safe operation of the vehicle.
  • Most of the traditional bogie designs use a method of increasing the safety factor to ensure the strength margin. This method has the problem of rich static strength and insufficient or excessive fatigue strength. The fatigue life of the frame cannot be accurately predicted in the design stage.
  • the wheel is an important moving part of the rolling stock. It is easy to produce rim fatigue cracks during use. If it is not found in time, it will seriously endanger the driving safety of the rolling stock.
  • the locomotives put into use in China generally suffer from the damage of rim fatigue cracks.
  • This kind of damage of the wheel is a common phenomenon in history.
  • the train has a speed of up to 300km/h and the axle load of the heavy-duty truck reaches 30T.
  • the interaction between the wheel and the track also increases with the speed.
  • the linearity increases the law, and the contact stress also increases as the axle weight increases. Under such high speed and heavy load conditions, damage to the tread surface, peeling, etc. of the wheel may cause an impact load to act between the wheel and rail. These factors may cause an increase in the internal stress of the wheel rim.
  • the coupler is used as a steel casting, and there are some inevitable defects in the interior or surface, such as pores, slag inclusions and shrinkage.
  • a casting defect can be understood as a macroscopic crack, and the distribution of the cross-section stress is uneven due to the presence of macroscopic cracks, thereby causing an increase in the stress at the crack tip.
  • the greater the defect the more serious the stress concentration phenomenon, which makes the coupler crack expansion under lower load conditions and induces the occurrence of weld cracks. Under the action of dynamic load, a large amount of working stress will cause fatigue cracking of the casting.
  • a method for monitoring crack damage of a train component including:
  • the cellular sensor network device comprises a plurality of piezoelectric sensors, each of which serves as an excitation signal loading point and/or a response signal collection point;
  • damage data includes: crack length data of the detection structure
  • the damage data of the detection structure is transmitted to a central server for storage and subsequent processing.
  • the plurality of piezoelectric sensors are arranged in a honeycomb array comprising at least one regular hexagonal basic detection unit; each piezoelectric sensor is a node in the honeycomb array, loaded as an excitation signal Point and / or response signal acquisition points;
  • the basic detecting unit includes six piezoelectric sensors arranged in a regular hexagon shape;
  • a piezoelectric controller is embedded in the piezoelectric sensor, and the piezoelectric sensor functions as a Lamb wave excitation signal device and a Lamb wave signal receiver.
  • the method further includes:
  • Exciting signals are excited on the healthy detection structure at the first time interval at the excitation signal loading point, and Lamb waves are generated in the detection structure; and each response signal acquisition point collects a first Lamb wave response signal for the Lamb wave;
  • Exciting signals are excited on the detection structure to be detected at a second time interval at the excitation signal loading point, and Lamb waves are generated in the detection structure; and each response signal collection point collects a second Lamb wave response signal for the Lamb wave;
  • the second Lamb wave response signal is analyzed in the time domain and the frequency domain to extract feature information; the second Lamb wave response signal is used as a damage signal, and the first Lamb wave response signal is used as a reference signal, based on the damage signal, the reference signal, and The reference information and the feature information calculate a signal difference coefficient value SDC value corresponding to each of the response signal collection points;
  • the method further includes:
  • the central server predicts a growth rate of damage of the detection structure based on stored historical impairment data of the detection structure.
  • the increasing rate of the damage predicting the detection structure comprises:
  • the model parameters of the Bayesian probability prediction model are optimized by the Markov chain Monte Carlo method, and the growth rate of the damage of the detection structure is predicted.
  • a train component crack damage monitoring system including: a cellular sensor network device, a multi-channel data converter, an in-vehicle data processing device, and a central server.
  • the cellular sensor network device is disposed on a detection structure of a train component to be detected, and includes a plurality of piezoelectric sensors, each of which serves as an excitation signal loading point and/or a response signal collection point;
  • the multi-channel data converter is configured to transmit a response signal collected by the cellular sensor network device to the onboard data processing device;
  • the in-vehicle data processing apparatus is configured to pre-process a response signal collected by the cellular sensor network device to obtain damage data of the detection structure, and transmit the damage data of the detection structure to the central server;
  • the damage data includes: crack length data of the detection structure;
  • the central server is configured to store damage data of the detection structure and process the damage data to achieve remote monitoring of the detection structure of the train component to be detected.
  • the plurality of piezoelectric sensors are arranged in a honeycomb array comprising at least one regular hexagonal basic detection unit; each piezoelectric sensor is a node in the honeycomb array, loaded as an excitation signal Point and / or response signal acquisition points;
  • the basic detecting unit includes six piezoelectric sensors arranged in a regular hexagon shape;
  • a piezoelectric controller is embedded in the piezoelectric sensor, and the piezoelectric sensor functions as a Lamb wave excitation signal device and a Lamb wave signal receiver.
  • the onboard data processing device comprises:
  • the first signal excitation module is configured to use the probe to excite the excitation signal on the healthy detection structure at the first time interval at the excitation signal loading point, generate a Lamb wave in the detection structure, and collect the Lamb wave for each Lamb wave in each response signal collection point. a Lamb wave response signal;
  • the reference information acquisition module is configured to acquire the first Lamb wave response signal and establish a dispersion relationship of the Lamb wave with the propagation angle in the anisotropic composite material layer of the detection structure by Mindlin plate theory, and obtain a theoretical velocity distribution of the Lamb wave.
  • the baseline information As the baseline information;
  • a second signal excitation module is configured to use the probe to excite an excitation signal on the detection structure to be detected at a second time interval at the excitation signal loading point, and generate a Lamb wave in the detection structure, and each response signal collection point acquires a Lamb wave for the Lamb wave. a second Lamb wave response signal;
  • the damage information acquiring module is configured to analyze the second Lamb wave response signal in the time domain and the frequency domain to extract feature information, and use the second Lamb wave response signal as the damage signal, and use the first Lamb wave response signal as a reference signal. Calculating a signal difference coefficient value SDC value corresponding to each of the response signal collection points based on the damage signal, the reference signal, and the reference information and the feature information;
  • An imaging analysis module is configured to reconstruct a region where crack damage may exist in the detection structure according to the obtained SDC value and adopt a probability imaging principle; determine a crack direction based on the SDC value, and correct an SDC value in the crack direction for enhancing the crack direction
  • the crack damage image is reconstructed by the principle of probability imaging; the SDC profile is drawn, and the length of the crack is estimated based on the SDC profile.
  • the central server predicts an increase rate of damage of the detection structure based on stored historical impairment data of the detection structure.
  • the central server comprises:
  • a historical data acquisition module configured to acquire historical damage data of the detection structure; wherein the damage data includes: crack length data of the detection structure;
  • a prediction model establishing module configured to obtain a life distribution feature and a verification index parameter for the damage of the detection structure according to the historical damage data, and establish a Bayesian probability prediction model corresponding to the damage of the detection structure;
  • a damage growth prediction module configured to obtain a prior distribution of the verification indicator parameter according to the historical damage data analysis; and optimize a model parameter of the Bayesian probability prediction model by using a Markov chain Monte Carlo method, and predict the The growth rate of damage to the structure is detected.
  • the train component crack damage monitoring method and system can detect the train component crack damage based on the response signal, and store the damage data to predict and evaluate the development trend of the damage site based on the damage data, and provide reliability assessment and hidden danger. Early warning, monitoring the crack damage of train components, providing a more accurate and intuitive basis for train maintenance and replacement.
  • FIG. 1 is a flow chart of one embodiment of a method for monitoring crack damage of a train component in accordance with the present invention
  • FIG. 2 is a schematic view showing the arrangement of a sensor network device in one embodiment of a method for monitoring crack damage of a train component according to the present invention
  • FIG. 3 is a schematic structural view of an embodiment of a train component crack damage monitoring system according to the present invention.
  • FIG. 4 is a block diagram showing an embodiment of an in-vehicle data processing apparatus in a train component crack damage monitoring system according to the present invention
  • Figure 5 is a block diagram of an embodiment of a central server in a train component crack damage monitoring system in accordance with the present invention.
  • Embodiments of the invention may be applied to a computer system/central server that can operate with numerous other general purpose or special purpose computing system environments or configurations.
  • Examples of well-known computing systems, environments, and/or configurations suitable for use with computer systems/central servers include, but are not limited to, personal computer systems, central server computer systems, thin clients, thick clients, handheld or laptop devices Microprocessor-based systems, set-top boxes, programmable consumer electronics, networked personal computers, small computer systems, mainframe computer systems, and distributed cloud computing technology environments including any of the above, and the like.
  • the computer system/central server can be described in the general context of computer system executable instructions (such as program modules) being executed by a computer system.
  • program modules may include routines, programs, target programs, components, logic, data structures, and the like that perform particular tasks or implement particular abstract data types.
  • the computer system/central server can be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices that are linked through a communication network.
  • program modules may be located on a local or remote computing system storage medium including storage devices.
  • FIG. 1 is a flow chart of an embodiment of a method for monitoring crack damage of a train component according to the present invention, as shown in FIG.
  • Step 101 A cellular sensor network device is disposed on a detection structure of a train component to be detected; wherein the cellular sensor network device includes a plurality of piezoelectric sensors, each of which is used as an excitation signal loading point and/or a response signal acquisition point.
  • the train components can be key components such as high-speed trains, subways, etc., such as wheels, couplers, bogies, and the like.
  • the piezoelectric sensor can be embedded with a microcontroller as an excitation signal and a signal receiver, respectively, performing multiplex detection according to a certain timing, and storing the collected lamb wave signal in the sensor.
  • Step 102 Transmit a response signal collected by the cellular sensor network device to the in-vehicle data processing device.
  • the lambda signal data stored in the sensor is transmitted to the multi-channel data converter, and the multi-channel data converter transmits the lambda signal data to the in-vehicle data processing device for pre-processing to confirm the crack position and damage. Length, the damage data of the detected structure is obtained.
  • Step 103 Pre-process a response signal collected by the cellular sensor network device to obtain damage data of the detection structure.
  • the damage data includes: crack length data of the detection structure.
  • the data stored in the sensor is preprocessed, and the preprocessing process includes: 1. signal filtering: filtering out noise and false information, and compensating for temperature characteristics of the piezoelectric sensor; 2. using wavelet analysis, HHT Analysis, empirical mode decomposition and other methods to extract fault feature information, determine the damage degree of the key parts of the train; 3, through the piezoelectric element to receive the arrival signal and the delay time between them, use the elliptical positioning method to determine the damage location, near the damage near the sensor The positioning error will be eliminated by a regular hexagonal honeycomb arrangement.
  • the crack damage of the train component is detected based on the Lamb wave imaging, and the damage data of the detection structure can be obtained by the following steps;
  • an excitation signal is excited on the healthy detection structure at a first time interval at the excitation signal loading point to generate a Lamb wave in the detection structure.
  • Each of the response signal acquisition points acquires a first Lamb wave response signal for the Lamb wave.
  • the Lamb wave is an elastic guided wave propagating in a solid structure under free boundary conditions. It has a slow attenuation and a long propagation distance, and is very sensitive to small damage in the structure.
  • the charge amplifier can be used to amplify the excitation signal and then load it into the piezoelectric sensor to excite the Lamb wave in the detection structure.
  • the Lamb wave response signal of all excitation/sensing channels when the structure is healthy is collected as a reference signal for detecting the structure.
  • the first Lamb wave response signal is obtained and the dispersion relation of the Lamb wave with the propagation angle in the anisotropic composite laminate of the detection structure is established by the Mindlin plate theory, and the theoretical velocity distribution of the Lamb wave is obtained as the reference information.
  • the Mindlin plate theory is often referred to as the first-order shear deformation theory of the plate.
  • the Mindlin plate theory assumes that the plate displacement changes linearly in the plate thickness direction, but the plate thickness does not change, and it is assumed that the normal stress in the plate thickness direction is ignored, that is, the plane stress assumption.
  • the dispersion relation of Lamb wave in the anisotropic composite laminate with the propagation angle can be established by Mindlin plate theory, and the theoretical velocity distribution of Lamb wave is obtained, which provides the reference information for damage imaging.
  • the probe is used to excite the excitation signal on the detection structure to be detected at a second time interval, and a Lamb wave is generated in the detection structure.
  • Each of the response signal acquisition points acquires a second Lamb wave response signal for the Lamb wave.
  • the second Lamb wave response signal is analyzed in the time domain and the frequency domain to extract feature information.
  • the second Lamb wave response signal is used as a damage signal
  • the first Lamb wave response signal is used as a reference signal
  • the signal difference coefficient value SDC value corresponding to each response signal collection point is calculated based on the damage signal, the reference signal, the reference information, and the feature information.
  • the acquired second Lamb wave response signal is used as a damage signal, and then the first Lamb wave response signal is collected as a reference signal according to a healthy plate structure, and then the SDC values of all the excitation/sensing channels are calculated.
  • the wavelet transform can be used to analyze the Lamb wave signal excited and received by the piezoelectric sensing element in the time-frequency domain, extract the feature information, and measure the flight time and group velocity of the actual propagation of the Lamb wave in the monitored portion, and the reference Information is compared.
  • the energy of the damage scattering signal is superimposed and amplified by the focusing method, thereby improving the signal-to-noise ratio of the signal.
  • the time reversal method is used to adaptively focus the wave source, reconstruct the signal propagation fluctuation map, and display the damage location and region through signal focusing.
  • the region where the crack damage may exist in the detection structure is reconstructed.
  • the obtained SDC value according to the reflection and scattering of the Lamb wave monitoring signal by the crack damage, the reconstructed image information in the crack direction is strengthened by correcting the signal difference coefficient value SDC in the crack direction in the damaged region, thereby realizing the crack damage.
  • Image reconstruction using the principle of probability imaging to reconstruct the crack damage image.
  • the signal difference coefficient SDC can be used to characterize the statistical difference between the damage signal and the reference signal, and the magnitude of the SDC value reflects the degree of damage and the damage distance.
  • the reconstructed image information in the crack direction is strengthened by correcting the signal difference coefficient (SDC) in the crack direction in the damaged area, thereby realizing the crack damage.
  • SDC signal difference coefficient
  • the conventional probability imaging principle can be used to reconstruct the region where the crack damage may exist in the plate; the crack direction can be determined according to the difference of the SDC value calculated by the path of the piezoelectric sensor in the damaged region. .
  • the SDC value in the crack determination direction can be corrected to 1, and the crack damage image is reconstructed again using the conventional probability imaging principle.
  • the SDC values of the six excitation/sense channels are arranged into a regular hexagon, and the probability distribution maps corresponding to all the sensing paths are superimposed, thereby obtaining the damage distribution probability of any point in the detection area, and reconstructing the damage image of the crack. .
  • the crack direction is determined based on the SDC value, and the SDC value in the crack direction is corrected for enhancing the reconstructed image information in the crack direction, and the crack damage image is reconstructed by the principle of probability imaging.
  • the SDC profile is plotted and the length of the crack is estimated based on the SDC profile.
  • the SDC profile can be drawn using a variety of existing methods and the length of the crack can be estimated.
  • the energy of the damage scattering signal can be superimposed and amplified by the focusing method, thereby improving the signal-to-noise ratio of the signal; using the time reversal method to adaptively focus the wave source, reconstructing the signal propagation fluctuation pattern,
  • the signal focus shows the damage position and the area; according to the imaging result, the crack damage length is calculated according to the number and spacing of image pixel points exceeding the set threshold value on the monitoring path of the determined crack direction.
  • the energy of the damage scatter signal is superimposed and amplified by the focusing method, thereby improving the signal-to-noise ratio of the signal.
  • the adaptive focusing ability of the wave source is reconstructed by the time reversal method, and the signal propagation fluctuation map is reconstructed, and the damage position and region are displayed by signal focusing.
  • the crack size evaluation process according to the reflection and scattering of the Lamb wave monitoring signal by the crack damage, the reconstructed image information in the crack direction is strengthened by correcting the signal difference coefficient SDC value in the crack direction in the damaged area to achieve crack damage. The image was reconstructed and the length of the crack was estimated from the SDC profile at the receiving end.
  • a plurality of piezoelectric sensors are arranged in a honeycomb array including at least one regular hexagonal basic detecting unit.
  • Each piezoelectric sensor is a node in the honeycomb array as an excitation signal loading point and a response signal acquisition point.
  • the basic detecting unit includes six piezoelectric sensors arranged in a regular hexagon shape; the piezoelectric sensor is embedded with a microcontroller, and the piezoelectric sensor is used as a Lamb wave excitation signal device and a Lamb wave signal receiver.
  • the waveform generator is connected to the power amplifier via a wire, and the power amplifier is connected by a wire to an actuator in a monitoring path composed of a cellular sensor network device, the cellular sensor network device being disposed on the structure to be tested.
  • the sensor in the monitoring path is connected to the charge amplifier via a wire, and the charge amplifier is connected to the onboard data processing device via a wire.
  • a piezoelectric sensor with a certain number of embedded microcontrollers is used to form a regular hexagonal honeycomb array covering the surface of the part to be tested.
  • the number of arrangements of the cellular sensor network can be determined according to the actual situation of the structure to be monitored. Theoretically, six piezoelectric elements can form a monitoring unit.
  • the plurality of cellular sensor networks can be closely arranged according to the situation. The way of scanning is carried out.
  • the wavelet transform algorithm is used to analyze the second Lamb wave response signal in the time domain and the frequency domain, and the feature information is extracted to measure the time and speed of the actual propagation of the Lamb wave in the detection structure to be detected.
  • the Lamb wave signal excited and received by the piezoelectric sensing element is analyzed in the time-frequency domain by wavelet transform.
  • the first Lamb wave response signal is used as a base signal, and the second Lamb wave response signal is subtracted from the first Lamb wave response signal to obtain a damage scattering signal; the energy of the damage scattering signal is superimposed and amplified by a focusing method to improve damage scattering.
  • the second Lamb wave response signal including the defect information may be time-reversed, and the processed second Lamb wave response signal is loaded as a new wave source to the excitation signal loading point for transmission, and the detection structure to be detected is used.
  • the excitation signal is excited to realize the secondary focusing of the Lamb wave at the defect, and an amplitude focusing map is established to image and identify the damage location and the region.
  • a cellular sensor network device is first disposed on a portion to be tested, the monitoring region is divided into a plurality of basic monitoring units, and the nodes of the cellular sensor network device are used to excite and receive Lamb waves, and are reconstructed by focusing and time reversal.
  • the signal propagates a wave map that uses signal focus to show the location and area of the damage.
  • the signal difference coefficient value in the crack direction in the damaged region is corrected to reconstruct the reconstructed image information in the crack direction, and the length of the crack is evaluated by the SDC distribution map at the receiving end.
  • the Lamb wave imaging based train component damage data acquisition method provides a cellular sensor network device on the detection structure of the train component to be detected, and the piezoelectric sensor in the cellular sensor network device acts as an excitation signal loading point and/or Responding to the signal collection point; using the second Lamb wave response signal as the damage signal, using the first Lamb wave response signal as the reference signal, calculating the signal difference coefficient value SDC value corresponding to each response signal collection point, and reconstructing by using the probability imaging principle Crack damage image, and draw SDC distribution map, based on SDC distribution map to estimate the length of the crack, and can focus and time reversal of the response signal; can overcome the influence of environmental noise, measurement error, time delay and other factors, The energy calculation accuracy of the signal is high, and the false positive rate of damage judgment is low, which can reduce the "blind zone" in the detection, so that the damage monitoring range is larger, and the accuracy of the damage discrimination can be improved.
  • Step 104 Transmitting the damage data of the detection structure to a central server for storage and subsequent processing.
  • the central server can be used for various types of central servers, using the hadoop-spar piggybacking platform for data processing, and having a large data storage space to store the damage data of the detection structure.
  • the central server may include intelligent identification diagnostic software, using intelligent diagnosis methods and pattern recognition methods: artificial neural network, fuzzy logic, genetic algorithm, etc., which can be integrated into Lamb wave imaging and crack intelligent recognition technology, which can further improve the effective diagnosis. Sex and reliability.
  • the central server further predicts a growth rate of the damage of the detection structure based on historical damage data of the detection structure stored therein, and specifically includes the following steps:
  • Markov chain Monte Carlo method is used to optimize the model parameters of the Bayesian probability prediction model and predict the growth rate of the damage of the detection structure.
  • the basic theory of the Markov Chain Monte Carlo (MCMC) algorithm is the Markov process.
  • MCMC Markov Chain Monte Carlo
  • the fault feature information extracted by the in-vehicle data processing device is input as an input value to a central server including the smart identification diagnostic software, and the damage analysis data obtained under the real operating conditions is fitted and analyzed to obtain the most Excellent fitting function.
  • the life of a train component is subject to a statistically regular random variable, which is generally described by a distribution function of life (also called a cumulative distribution function).
  • the life span of train components is mostly subject to the probability distribution of continuous random variables, including logarithmic distribution, exponential distribution, normal distribution, and Weibull distribution.
  • the lognormal distribution is a relatively perfect distribution with non-negative properties. It is a probability distribution that can accurately describe the life of train components. It is suitable for the damage characteristics of train components accumulated with running mileage in the present invention.
  • the lifetime distribution of the damage to the detection structure is determined to be a lognormal distribution, wherein the density function of the lognormal distribution is:
  • ⁇ is the average value (mm) of the damage size
  • ⁇ is the standard deviation (mm) of the damage size.
  • the prior distribution of the verification index parameters is determined as the combination of the lognormal distribution f( ⁇ ) of the crack growth rate of the detection structure and the maximum likelihood estimate f(x
  • the crack data x of the detected structure in the damage data is obtained by the posterior distribution f( ⁇
  • Bayesian Decision Theory is based on incomplete intelligence, the subjective probability is estimated for the partially unknown state, then the Bayesian formula is used to correct the probability of occurrence, and finally the expected value and the modified probability are used to make the optimal.
  • the method of decision making can obtain a prior distribution.
  • the prior distribution is a combination of the lognormal distribution f( ⁇ ) of the crack growth rate of the observed variable train component and the maximum likelihood estimate f(x
  • x obtain the posterior distribution f( ⁇
  • Bayesian probability prediction model is described by the following formula:
  • ⁇ ldi represents the standard crack length damage (mm)
  • the normal distribution is used to improve the accuracy of the formula, that is, the accuracy of the prediction. At this point, all parameters are no longer fixed values and will be subject to their respective distributions.
  • Equation (6) is that the probability distribution obeyed by z 0 and z 1 in equation (5) can be disassembled into the product of two probability distributions. It should be noted that this formula is implicit in the calculation process and cannot be specifically expressed, but Does not affect the calculation.
  • Equation (7) indicates that z 0 obeys the log-positive distribution, ⁇ is the mean of z 0 , obtained from historical data, and ⁇ 2 is z 0 standard deviation, obeying the inverse gamma distribution.
  • ⁇ ) maximum likelihood estimation is first obtained by the historical damage data in reverse order, and then ⁇ -G(a,b) is determined through practical experience.
  • the Bayesian formula is used to obtain the verification index parameters.
  • the embodiment uses the winbugs software for the MCMC iteration, that is, the above process is completed, and the specific calculation of the above formula and the MCMC iteration can be performed by using various existing methods, since the maximum likelihood estimation is subject to many factors.
  • the interference cannot be expressed by a specific formula.
  • the purpose of using the MCMC method in this embodiment is to generate a large amount of data obeying the distribution law of historical data, which is used to represent the maximum likelihood estimate f(x
  • the method for predicting crack damage of train components collects historical damage data of key parts of high-speed trains (accumulated length of cracks under equal operating mileage); determines life distribution and verification index parameters of key parts of high-speed trains, and establishes a Bayesian statistical distribution model; Based on the historical data, the prior distribution of the parameters of the verification index is determined. Based on the MCMC method, the damage growth rate of the key parts of the high-speed train is predicted by WinBUGS.
  • the train component crack damage monitoring method can detect the train component crack damage based on the response signal, and store the damage data to predict and evaluate the development trend of the damage site based on the damage data, and provide reliability assessment and hidden danger warning. It realizes the monitoring of crack damage of train components and provides a more accurate and intuitive basis for train maintenance and replacement.
  • the train component crack damage monitoring system of the present invention includes a cellular sensor network device 30, a multi-channel data converter 31, an in-vehicle data processing device 32, and a center server 33.
  • the cellular sensor network device 30 is disposed on the detection structure of the train component to be detected and includes a plurality of piezoelectric sensors, each of which acts as an excitation signal loading point and/or a response signal acquisition point.
  • a plurality of piezoelectric sensors are arranged in a honeycomb array comprising at least one regular hexagonal basic detection unit. Each piezoelectric sensor is a node in the honeycomb array as an excitation signal loading point and a response signal acquisition point.
  • the basic detecting unit includes six piezoelectric sensors arranged in a regular hexagon.
  • the piezoelectric sensor has a built-in microcontroller, and the piezoelectric sensor acts as a Lamb wave excitation signal and a Lamb wave signal receiver.
  • the multi-channel data converter 31 is configured to transmit a response signal collected by the cellular sensor network device 30 to the onboard data processing device 32.
  • the multi-channel data converter 31 has a plurality of input and output channels through which all of the cellular sensor network devices 30 arranged on the list are networked for acquisition by all of the cellular sensor network devices 30.
  • the response signal is transmitted to the in-vehicle data processing device 32 for processing.
  • the in-vehicle data processing device 32 is configured to pre-process a response signal collected by the cellular sensor network device to obtain damage data of the detection structure, and transmit the damage data of the detection structure to the central server;
  • the damage data includes: crack length data of the detection structure.
  • the in-vehicle data processing device 32 of the present invention includes a first signal excitation module 321, a reference information acquisition module 322, a second signal excitation module 323, a damage information acquisition module 324, and imaging. Analysis module 325.
  • the first signal excitation module 321 excites the excitation signal on the healthy detection structure at the first time interval at the excitation signal loading point, generates a Lamb wave in the detection structure, and collects the first Lamb for the Lamb wave for each response signal collection point. Wave response signal.
  • the reference information acquisition module 322 obtains the first Lamb wave response signal and establishes the dispersion relationship of the Lamb wave in the anisotropic composite material layer of the detection structure by the Mindlin plate theory, and obtains the theoretical velocity distribution of the Lamb wave. Benchmark information.
  • the second signal excitation module 323 excites the excitation signal on the detection structure to be detected at the excitation signal loading point by using the probe at a second time interval, generates a Lamb wave in the detection structure, and collects a second Lamb wave for each Lamb wave in each of the response signal acquisition points. Lamb wave response signal.
  • the damage information obtaining module 324 analyzes the second Lamb wave response signal in the time domain and the frequency domain to extract feature information.
  • the damage information obtaining module 324 uses the second Lamb wave response signal as a damage signal, uses the first Lamb wave response signal as a reference signal, and calculates a signal corresponding to each response signal collection point based on the damage signal, the reference signal, the reference information, and the feature information. Difference coefficient value SDC value.
  • the imaging analysis module 325 reconstructs the regions in the detection structure where crack damage may exist based on the obtained SDC values and using the principle of probabilistic imaging.
  • the imaging analysis module 325 determines the crack direction based on the SDC value, corrects the SDC value in the crack direction, and is used to enhance the reconstructed image information in the crack direction, and reconstructs the crack damage image using the principle of probability imaging.
  • the imaging analysis module 325 plots the SDC profile and estimates the length of the crack based on the SDC profile.
  • the damage information obtaining module 324 analyzes the second Lamb wave response signal in the time domain and the frequency domain by using a wavelet transform algorithm, and extracts feature information for measuring the time and speed of the actual propagation of the Lamb wave in the detection structure to be detected. .
  • the damage information acquisition module 324 subtracts the second Lamb wave response signal from the first Lamb wave response signal by using the first Lamb wave response signal as a base signal to obtain a damage scatter signal.
  • the damage information acquisition module 324 superimposes the energy of the damage scatter signal by focusing to improve the signal-to-noise ratio of the damage scatter signal, and performs time reversal processing on the damage scatter signal.
  • the second signal excitation module 323 performs time reversal processing on the second Lamb wave response signal including the defect information, and loads the processed second Lamb wave response signal as a new wave source to the excitation signal loading point for transmission.
  • the excitation signal is excited on the detection structure to be detected, the secondary focus of the Lamb wave at the defect is realized, and the amplitude focus map is established to image and identify the damage location and the region.
  • the central server 33 is configured to store damage data of the detection structure and process the damage data to achieve remote monitoring of the detection structure of the train component to be detected.
  • the central server 33 includes a storage module 330, a historical data acquisition module 331, a prediction model establishment module 332, and a damage growth prediction module 334.
  • the storage module 330 is configured to store the damage data of the detection structure of the train component by the onboard data processing device 32, and the damage data includes: crack length data of the detection structure.
  • the historical data acquisition module 331 acquires historical damage data of the detection structure, and the damage data includes: detection of crack length data of the structure.
  • the prediction model establishing module 332 obtains the life distribution characteristics and the verification index parameters of the damage to the detection structure based on the historical damage data, and establishes a Bayesian probability prediction model corresponding to the damage of the detection structure.
  • the damage growth prediction module 333 obtains the prior distribution of the verification index parameters according to the historical damage data analysis, and optimizes the model parameters of the Bayesian probability prediction model by using the Markov chain Monte Carlo method, and predicts the growth rate of the damage of the detection structure.
  • the predictive model building module 332 determines that the lifetime distribution of the damage to the detected structure is a lognormal distribution, and the density function of the lognormal distribution is:
  • ⁇ is the average of the damage dimensions and ⁇ is the standard deviation of the damage dimensions.
  • the damage growth prediction module 333 determines the prior distribution of the verification index parameter as a combination of the lognormal distribution f( ⁇ ) of the crack growth rate of the detection structure and the maximum likelihood estimate f(x
  • the prediction model is combined with the crack data x of the detected structure in the historical damage data to obtain the posterior distribution f( ⁇
  • the central server 33 acquires historical damage data for the detection structure of the train component, obtains the life distribution feature and the verification index parameter for the damage of the detection structure according to the historical damage data, and establishes a Bayesian probability prediction model, and analyzes The prior distribution of the parameters of the verification index is obtained.
  • the model parameters of the Bayesian probability prediction model are optimized by the Markov chain Monte Carlo method, and the growth rate of the damage of the detection structure is predicted.
  • the Bayesian-MCMC method is applied and based on a large number of history. The data predicts the damage growth of train components. Subjective probability estimation can be applied to the damage of some unknown train components in the case of incomplete data. Then, the Bayesian formula is used to correct the probability of occurrence, and finally the expected value and the corrected probability are used to make the most. Excellent maintenance or update decision-making, providing a more accurate and intuitive basis for train maintenance and replacement, and high calculation accuracy.
  • the method and apparatus, apparatus of the present invention may be implemented in a number of ways.
  • the methods, apparatus, and apparatus of the present invention can be implemented in software, hardware, firmware, or any combination of software, hardware, and firmware.
  • the above-described sequence of steps for the method is for illustrative purposes only, and the steps of the method of the present invention are not limited to the order specifically described above unless otherwise specifically stated.
  • the invention may also be embodied as a program recorded in a recording medium, the program comprising machine readable instructions for implementing the method according to the invention.
  • the invention also covers a recording medium storing a program for performing the method according to the invention.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Mathematical Physics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

一种列车部件裂纹损伤监测方法和系统，该监测方法包括：在待检测列车部件的检测结构上设置蜂窝传感器网络装置（30）（101）；将对蜂窝传感器网络装置（30）采集的响应信号传输至车载数据处理装置（32）（102）；对蜂窝传感器网络装置（30）采集的响应信号进行预处理，以获得检测结构的损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据（103）；将检测结构的损伤数据传输至中心服务器（33），进行存储和后续处理（104）；该监测系统包括：蜂窝传感器网络装置（30）、多通道数据转换器（31）、车载数据处理装置（32）和中心服务器（33）；该列车部件裂纹损伤监测方法和系统可以基于响应信号对列车部件裂纹损伤进行检测，并对损伤数据进行存储以基于损伤数据对损伤部位的发展趋势进行预测评价，给出可靠性评估和隐患预警，实现对列车部件裂纹损伤的监测。

Description

列车部件裂纹损伤监测方法和系统 技术领域
本发明涉及损伤分析技术领域，尤其涉及一种列车部件裂纹损伤监测方法和系统。
背景技术
机车及动车组列车组是轨道交通运输的重要设备，工作时启动、制动加速度大，走行部，连挂部位皆承受着巨大的应力变化。在列车运行30万公里后，其走行部会出现较多的机械损伤。转向架构架作为机车车辆的关键承载部件，担负着支撑车体、运行和转向功能，其抗疲劳的性能对车辆安全运行有着格外重要意义。传统转向架设计大多采用增大安全系数来保障强度裕度的方法，这种方法存在着静强度富裕，局部的疲劳强度不足或者过量的问题，设计阶段不能准确预测构架疲劳寿命。
车轮是机车车辆的重要走行部件，在使用过程中易产生轮辋疲劳裂纹，若不及时发现会严重危及机车车辆的行车安全。当前我国投入使用的机车车辆普遍存在轮辋疲劳裂纹的伤损情况。车轮的这类损伤在历史中属于常见现象，但随着中国铁路的快速发展，列车时速高达300km/h、重载货车轴重达到30T，车轮与轨道之间的相互作用力也随速度的提高呈现线性递增规律，同时接触应力也随着轴重的提高而增加。在这种高速和重载条件下，车轮的踏面擦伤、剥离等损伤会导致冲击载荷作用在轮轨间，这些因素都会引起车轮轮辋内部应力的增加。
车钩作为铸钢件，内部或表面不可避免的存在一些缺陷，如气孔、夹渣和缩松等。这种铸造缺陷可以理解为宏观裂纹，由于宏观裂纹的存在，造成截面应力分布不均匀，从而引起裂纹尖端的应力增大。这种缺陷越大，应力集中现象也越严重，使得车钩在较低的载荷工况下就能造成裂纹的扩展并将诱发焊接裂纹的产生。车钩在动载荷作用下，大量的工作应力将导致铸件疲劳裂纹的发生。但在有铸造缺陷的部位，由于应力集中，受力截面上不大的工作应力同样可导致疲劳裂纹的发生，如果在这些应力较大区域例如钩耳孔、钩尾销孔或钩舌面存有铸造缺 陷，将加快疲劳裂纹的产生及扩展。
对于上述列车或轨道不同部位的结构损伤和相关的检测技术，国内外也有较多的研究和应用，但多为四点圆弧定位、三点双曲线定位，但不能避免环境噪声、测量误差、时间延迟计算方法误差等因素，尤其是无法解决边界损伤问题。
发明内容
在发明内容部分中引入了一系列简化形式的概念，这将在具体实施方式部分中进一步详细说明。本发明的发明内容部分并不意味着要试图限定出所要求保护的技术方案的关键特征和必要技术特征，更不意味着试图确定所要求保护的技术方案的保护范围。
根据本发明实施例的一个方面公开了一种列车部件裂纹损伤监测方法，包括：
在待检测列车部件的检测结构上设置蜂窝传感器网络装置；其中，所述蜂窝传感器网络装置包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点；
将对所述蜂窝传感器网络装置采集的响应信号传输至车载数据处理装置；
对所述蜂窝传感器网络装置采集的响应信号进行预处理，以获得所述检测结构的损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据；
将所述检测结构的损伤数据传输至中心服务器，进行存储和后续处理。
在本发明的一个实施例中，多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列；每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点和/或响应信号采集点；
所述基本检测单元包括呈正六边形排列的六个压电传感器；
所述压电传感器内嵌有微控制器，所述压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
在本发明的一个实施例中，还包括：
在激励信号加载点以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波；各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号；
获取第一Lamb波响应信号并建立Lamb波在检测结构的各向异性复合材料 层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息；
在激励信号加载点以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波；各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号；
对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息；将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个所述响应信号采集点相对应的信号差异系数值SDC值；
根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域；
基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像；绘制SDC分布图，基于SDC分布图评估出裂纹的长度，生成所述损伤数据。
在本发明的一个实施例中，还包括：
所述中心服务器基于存储的所述检测结构的历史损伤数据预测所述检测结构的损伤的增长速率。
在本发明的一个实施例中，所述预测所述检测结构的损伤的增长速率包括：
根据所述历史损伤数据获得对于所述检测结构的损伤的寿命分布特征和验证指标参数，并建立与所述检测结构的损伤相对应的贝叶斯概率预测模型；
根据所述历史损伤数据分析得出所述验证指标参数的先验分布；
采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测所述检测结构的损伤的增长速率。
本发明实施例的另一方面公开了一种列车部件裂纹损伤监测系统，包括：蜂窝传感器网络装置、多通道数据转换器、车载数据处理装置和中心服务器，
所述蜂窝传感器网络装置设置在待检测列车部件的检测结构上，且包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点；
所述多通道数据转换器配置为将所述蜂窝传感器网络装置采集的响应信号传输至所述车载数据处理装置；
所述车载数据处理装置配置为对所述蜂窝传感器网络装置采集的响应信号 进行预处理，以获得所述检测结构的损伤数据，并将所述检测结构的损伤数据传输至所述中心服务器；其中，损伤数据包括：所述检测结构的裂纹长度数据；
所述中心服务器配置为存储所述检测结构的损伤数据，并对所述损伤数据进行处理以实现对所述待检测列车部件的检测结构的远程监督。
在本发明的一个实施例中，多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列；每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点和/或响应信号采集点；
所述基本检测单元包括呈正六边形排列的六个压电传感器；
所述压电传感器内嵌有微控制器，所述压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
在本发明的一个实施例中，所述车载数据处理装置包括：
第一信号激发模块，用于在激励信号加载点利用探头以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号；
基准信息获取模块，用于获取第一Lamb波响应信号并通过Mindlin板理论建立Lamb波在检测结构的各向异性复合材料层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息；
第二信号激发模块，用于在激励信号加载点利用探头以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号；
损伤信息获取模块，用于对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息；将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个所述响应信号采集点相对应的信号差异系数值SDC值；
成像分析模块，用于根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域；基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像；绘制SDC分布图，基于SDC分布图评估出裂纹的长度。
在本发明的一个实施例中，所述中心服务器基于存储的所述检测结构的历史损伤数据预测所述检测结构的损伤的增长速率。
在本发明的一个实施例中，所述中心服务器包括：
历史数据获取模块，用于获取所述检测结构的历史损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据；
预测模型建立模块，用于根据所述历史损伤数据获得对于所述检测结构的损伤的寿命分布特征和验证指标参数，并建立与所述检测结构的损伤相对应的贝叶斯概率预测模型；
损伤增长预测模块，用于根据所述历史损伤数据分析得出所述验证指标参数的先验分布；采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测所述检测结构的损伤的增长速率。
根据本发明的列车部件裂纹损伤监测方法和系统可以基于响应信号对列车部件裂纹损伤进行检测，并损伤数据进行存储以基于损伤数据对损伤部位的发展趋势进行预测评价，给出可靠性评估和隐患预警，实现对列车部件裂纹损伤的监测，为列车的维修、更换提供更为准确的、直观化的依据。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其它的附图：
图1为根据本发明的列车部件裂纹损伤监测方法的一个实施例的流程图；
图2为根据本发明的列车部件裂纹损伤监测方法的一个实施例中的传感器网络装置的布置示意图；
图3为根据本发明的列车部件裂纹损监测系统的一个实施例中的结构示意图；
图4为根据本发明的列车部件裂纹损监测系统中的车载数据处理装置的一个实施例中的模块示意图；
图5为根据本发明的列车部件裂纹损监测系统中的中心服务器的一个实施例中的模块示意图。
具体实施方式
现在将参照附图来详细描述本发明的各种示例性实施例。应注意到：除非另外具体说明，否则在这些实施例中阐述的部件和步骤的相对布置、数字表达式和数值不限制本发明的范围。
同时，应当明白，为了便于描述，附图中所示出的各个部分的尺寸并不是按照实际的比例关系绘制的。
以下对至少一个示例性实施例的描述实际上仅仅是说明性的，决不作为对本发明及其应用或使用的任何限制。
对于相关领域普通技术人员已知的技术、方法和设备可能不作详细讨论，但在适当情况下，所述技术、方法和设备应当被视为说明书的一部分。
应注意到：相似的标号和字母在下面的附图中表示类似项，因此，一旦某一项在一个附图中被定义，则在随后的附图中不需要对其进行进一步讨论。
本发明实施例可以应用于计算机系统/中心服务器，其可与众多其它通用或专用计算系统环境或配置一起操作。适于与计算机系统/中心服务器一起使用的众所周知的计算系统、环境和/或配置的例子包括但不限于：个人计算机系统、中心服务器计算机系统、瘦客户机、厚客户机、手持或膝上设备、基于微处理器的系统、机顶盒、可编程消费电子产品、网络个人电脑、小型计算机系统﹑大型计算机系统和包括上述任何系统的分布式云计算技术环境，等等。
计算机系统/中心服务器可以在由计算机系统执行的计算机系统可执行指令(诸如程序模块)的一般语境下描述。通常，程序模块可以包括例程、程序、目标程序、组件、逻辑、数据结构等等，它们执行特定的任务或者实现特定的抽象数据类型。计算机系统/中心服务器可以在分布式云计算环境中实施，分布式云计算环境中，任务是由通过通信网络链接的远程处理设备执行的。在分布式云计算环境中，程序模块可以位于包括存储设备的本地或远程计算系统存储介质上。
下文中的“第一”、“第二”仅用于描述上相区别，并没有其它特殊的含义。
图1为根据本发明的列车部件裂纹损伤监测方法的一个实施例的流程图，如图1所示：
步骤101，在待检测列车部件的检测结构上设置蜂窝传感器网络装置；其中，所述蜂窝传感器网络装置包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点。列车部件可以为高速列车、地铁等的关键部件等，例如车轮、车钩、转向架等。所述压电传感器可内嵌微控制器，分别作为激发信 号器及信号接收器，按照一定时序进行多路检测，并将收集的lamb波信号储存在传感器。
步骤102，将对所述蜂窝传感器网络装置采集的响应信号传输至车载数据处理装置。在高速列车行至等公里时，将储存在传感器的lamb波信号数据传输至多通道数据转换器，多通道数据转换器将lamb波信号数据传输到车载数据处理装置进行预处理，确认裂纹位置及损伤长度，得到检测结构的损伤数据。
步骤103，对所述蜂窝传感器网络装置采集的响应信号进行预处理，以获得所述检测结构的损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据。
在一个实施例中，将储存在传感器的数据进行预处理，预处理过程包括：1、信号滤波：滤除噪声及虚假信息、对压电传感器的温度特性进项补偿；2、采用小波分析、HHT分析、经验模式分解等方法提取故障特征信息，确定列车关键部位损伤程度；3、通过压电原件接收到达信号及它们之间的延迟时间，利用椭圆定位的方法确定损伤位置，其中靠近传感器附近损伤定位的误差，将通过正六边形蜂窝状布置方式来消除。
在本实施例中，基于Lamb波成像对列车部件裂纹损伤进行检测，可以通过下述步骤获得所述检测结构的损伤数据；
首先，在激励信号加载点利用探头以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波。各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号。
Lamb波是在自由边界条件下，固体结构中传播的弹性导波，具有衰减慢传播距离远，且对结构中的微小损伤十分敏感。可以利用电荷放大器将激励信号放大后加载到压电传感器，从而在检测结构中激发出Lamb波。采集检测结构健康时的所有激励/传感通道的Lamb波响应信号，作为检测结构的基准信号。
接着，获取第一Lamb波响应信号并通过Mindlin板理论建立Lamb波在检测结构的各向异性复合材料层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息。Mindlin板理论通常被称为板的一阶剪切变形理论。Mindlin板理论假设在板厚度方向板位移线性变化，但是板厚度不变，并假设忽略板厚度方向的正应力，即平面应力假设。可以通过Mindlin板理论建立Lamb波在各向异性复合材料层板中随传播角度变化的频散关系，得到Lamb波的理论速度分布，为损伤成像提供基准信息。
接着，在激励信号加载点利用探头以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波。各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号。对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息。将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个响应信号采集点相对应的信号差异系数值SDC值。将采集到的第二Lamb波响应信号作为损伤信号，再根据在健康的板结构中采集第一Lamb波响应信号作为参考信号，然后计算所有激励/传感通道的SDC值。
可以利用小波变换对由压电传感元件激励和接收的Lamb波信号在时频域进行分析，提取特征信息，测量出Lamb波在监测的部位中实际传播的飞行时间和群速度，并与基准信息相比较。根据Lamb波信号传播自身的特性，通过聚焦的方法使损伤散射信号能量叠加放大，从而提高信号的信噪比。利用时间反转法对波源进行自适应聚焦能力，重建信号传播波动图，通过信号聚焦显示损伤位置和区域。
接着，根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域。根据得到的SDC值，根据裂纹损伤对Lamb波监测信号的反射及散射作用，通过校正损伤区域中裂纹方向上的信号差异系数值SDC，强化裂纹方向上的重构图像信息，实现对裂纹损伤的图像重构，采用概率成像原理重构裂纹损伤图像。
根据Lamb波传播的基本原理，当传播介质出现中断或不连续时，大多数的Lamb波信号会因受到阻碍而难以继续向前传播，即使裂纹宽度很窄，只要其长度大于Lamb波波长，就会造成在传播前进方向上的Lamb波出现十分明显的衰减。可以利用信号差异系数SDC来表征损伤信号和参考信号的统计特性差异，SDC值的大小反映了损伤程度和损伤距离。
根据裂纹损伤对Lamb波监测信号的反射及散射作用,通过校正损伤区域中裂纹方向上的信号差异系数值(Signal difference coefficient，SDC)，强化裂纹方向上的重构图像信息,实现对裂纹损伤的图像重构，并由接收端SDC分布图评估出裂纹的长度。
接着，根据得到的SDC值，可以利用传统概率成像原理，重构出板中裂纹损伤可能存在的区域；根据通过损伤区域内的压电传感器所在的路径计算其SDC 值差值，可以判定裂纹方向。可以将裂纹判定方向上的SDC值校正为1，采用传统概率成像原理再次重构裂纹损伤图像。把六个激励/传感通道的SDC值布置为一个正六边形，将所有传感路径对应的概率分布图进行叠加，从而得到检测区域内任意点的损伤分布概率，重构出裂纹的损伤图像。
接着，基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像。绘制SDC分布图，基于SDC分布图评估出裂纹的长度。可以采用现有的多种方法绘制SDC分布图，并评估出裂纹的长度。
可以根据Lamb波信号传播自身的特性，通过聚焦的方法使损伤散射信号能量叠加放大，从而提高信号的信噪比；利用时间反转法对波源的自适应聚焦能力，重建信号传播波动图，通过信号聚焦显示损伤位置和区域；根据成像结果，在判定的裂纹方向的监测路径上，按照超过设定阈值的图像像素点数量和间距计算裂纹损伤长度。
在基于Lamb的成像过程中，通过聚焦的方法使损伤散射信号能量叠加放大，从而提高信号的信噪比。在损伤定位过程中，利用时间反转法对波源的自适应聚焦能力，重建信号传播波动图，通过信号聚焦显示损伤位置和区域。在裂纹大小评估过程中，根据裂纹损伤对Lamb波监测信号的反射及散射作用，通过校正损伤区域中裂纹方向上的信号差异系数SDC值，强化裂纹方向上的重构图像信息，实现对裂纹损伤的图像重构，并由接收端SDC分布图评估出裂纹的长度。
如图2所示，多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列。每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点、响应信号采集点。基本检测单元包括呈正六边形排列的六个压电传感器；压电传感器内嵌有微控制器，压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
波形发生器通过导线与功率放大器连接，功率放大器通过导线与蜂窝传感器网络装置组成的监测路径中的激励器连接，该蜂窝传感器网络装置设置在待测结构上。监测路径中的传感器通过导线与电荷放大器连接，电荷放大器通过导线与车载数据处理装置连接。
根据检测区域大小，使用一定数量的内嵌微控制器的压电传感器组成正六边形蜂窝状阵列覆在待测部位表面。作为激励，蜂窝传感器网络的布置数量可以根 据待监测结构的实际情况进行确定，理论上六个压电元件就可以组成一个监测单元，结构较大时根据情况可通过紧密布置多个蜂窝传感器网络以扫查的方式进行。
采用小波变换算法对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息，用于测量出Lamb波在待检测的检测结构中的实际传播的时间和速度。在提取待测构件的特征信息过程中，利用小波变换对由压电传感元件激励和接收的Lamb波信号在时频域进行分析。
以第一Lamb波响应信号作为基信号，将第二Lamb波响应信号与第一Lamb波响应信号相减，得到损伤散射信号；通过聚焦的方法使损伤散射信号能量叠加放大，用以提高损伤散射信号的信噪比，并对损伤散射信号进行时间反转处理。可以对包含缺陷信息的第二Lamb波响应信号进行时间反转处理，并将处理后的第二Lamb波响应信号作为新的波源加载到激励信号加载点进行发射，用以在待检测的检测结构上激发激励信号，实现Lamb波在缺陷处的二次聚焦，建立幅值聚焦图，用以对损伤位置和区域进行成像识别。
在一个实施例中，首先在待测部位上布置蜂窝传感器网络装置，将监测区域分割成若干个基本监测单元，利用蜂窝传感器网络装置的节点激励和接收Lamb波，通过聚焦和时间反转法重建信号传播波动图，利用信号聚焦显示损伤位置和区域。根据Lamb波监测信号的反射及散射作用校正损伤区域中裂纹方向上的信号差异系数值强化裂纹方向上的重构图像信息，并由接收端SDC分布图评估出裂纹的长度。
上述实施例提供的基于Lamb波成像的列车部件损伤数据获取方法，在待检测列车部件的检测结构上设置蜂窝传感器网络装置，蜂窝传感器网络装置中的压电传感器都作为激励信号加载点和/或响应信号采集点；将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，计算与各个响应信号采集点相对应的信号差异系数值SDC值，采用概率成像原理重构裂纹损伤图像，并绘制SDC分布图，基于SDC分布图评估出裂纹的长度，并可以对响应信号进行聚焦和时间反转等处理；能够克服环境噪声、测量误差、时间延迟等因素的影响，对信号的能量计算准确度高，对损伤判断的误判率低，可以减少检测中的“盲区”，使得损伤监测范围更大，并能够提高损伤判别的精准性。
步骤104，将所述检测结构的损伤数据传输至中心服务器，进行存储和后续 处理。
所述中心服务器可以为各种类型的中心服务器，利用采用hadoop-spar搭载平台，进行数据处理，并且具有庞大的数据存储空间，以存储所述检测结构的损伤数据。所述中心服务器可以含有含智能辨识诊断软件，运用智能诊断方法和模式识别方法：比如人工神经网络、模糊逻辑、遗传算法等，其可以融入Lamb波成像与裂纹智能识别技术，能够进一步提高诊断有效性和可靠性。
在本发明一个实施例中，所述中心服务器还基于其内存储的所述检测结构的历史损伤数据预测所述检测结构的损伤的增长速率，具体包括下述步骤：
首先，根据所述历史损伤数据获得对于所述检测结构的损伤的寿命分布特征和验证指标参数，并建立与所述检测结构的损伤相对应的贝叶斯概率预测模型；
其次，根据所述历史损伤数据分析得出所述验证指标参数的先验分布；
最后，采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测所述检测结构的损伤的增长速率。
马尔可夫链蒙特卡罗(Markov Chain Monte Carlo,MCMC)算法的基础理论为马尔可夫过程，在MCMC算法中，为了在一个指定的分布上采样，根据马尔可夫过程，首先从任一状态出发，模拟马尔可夫过程，不断进行状态转移，最终收敛到平稳分布。
在一个实施例中，将车载数据处理装置提取的故障特征信息作为输入值，传入含智能辨识诊断软件的中心服务器，通过对获取的真实运营条件下的损伤增长数据进行拟合分析，得到最优的拟合函数。
列车部件的寿命服从一定统计规律的随机变量,一般用寿命的分布函数(也称累积分布函数)来描述。列车部件的寿命大多服从连续型随机变量的概率分布，包括对数分布、指数分布、正态分布、威布尔分布等。对数正态分布是一种比较完善的分布，具有非负性，是一种可以准确描述列车部件寿命的概率分布，适用于本发明中列车部件随运行里程累积产生的损伤特征。
确定对于检测结构的损伤的寿命分布为对数正态分布，其中，对数正态分布的密度函数为：
Figure PCTCN2019082492-appb-000001
其中，μ是损伤大小的平均值(mm)；σ是损伤大小的标准差(mm)。
将验证指标参数的先验分布确定为检测结构的裂纹增长速率的对数正态分布f(θ)与最大似然估计f(x|θ)的结合，运用贝叶斯概率预测模型并结合历史损伤数据中的检测结构的裂纹数据x，获得后验分布f(θ|x)，并进行迭代计算获得裂纹在单位里程中的增长速率。
由于贝叶斯决策(Bayesian Decision Theory)是在不完全情报下，对部分未知的状态用主观概率估计，然后用贝叶斯公式对发生概率进行修正，最后再利用期望值和修正概率做出最优决策的方法，根据如下贝叶斯公式，可以得到先验分布。
Figure PCTCN2019082492-appb-000002
在本实施例中先验分布为观察变量列车部件的裂纹增长速率的对数正态分布f(θ)与最大似然估计f(x|θ)的结合，运用贝叶斯决策模型结合实际数据x，获得后验分布f(θ|x)，通过大量迭代，最终得到裂纹增长速率均值(期望值)，这个均值也就是模型中高速列车关键部位裂纹增长趋于稳定后的单位里程增长速率。
例如，某城市发生了一起汽车撞人逃跑事件，该城市只有两种颜色的车，蓝色15％，绿色85％，事发时有一个人A在现场看见了，A指证是蓝车。但是根据专家在现场分析，那种条件能看正确的可能性是80％，那么，肇事的车是蓝车的概率采用下述的方法进行计算：
A：目击证人看到的车辆颜色为蓝色；B 1：肇事车辆是蓝色；B 2肇事车辆是绿色
P(A|B 1)＝80％；P(A)＝80％P(B 1)+20％P(B 2)
Figure PCTCN2019082492-appb-000003
。贝叶斯概率预测模型用以下公式描述：
σ 1di＝z 1+z 0T i，i＝1，...，n～N(z 1+z 0T i；σ 2)   (5)；
f(z 0，z 1，σ 2)＝f(z 0，z 1)*f(σ 2)   (6)；
log(z 0)～N 2(μ，σ 2)   (7)；
σ 2～IG(a，b)   (8)；
其中，δ ldi表示标准裂纹长度损伤量(mm)，z 1是裂纹初始的长度，如果列车部件不存在初始裂纹，z 0是裂纹在单位里程中的增长速率(mm/3000km、 mm/5000km或mm/10000km，根据实际不同测量数据值确定)，T是自上一次采集历史损伤数据后的积累运营公里数,τ＝σ-2，则τ－G(a,b)，其中a＝b＝0.01。
式(5)为裂纹长度与积累运营公里数的线性关系，其中，δ ldi服从均值为σ 1di＝z 1+z 0T i的正态分布，该公式用于计算累计裂纹长度，但考虑δ ldi受多方因素影响，引入贝叶斯概率公式，使用正态分布提高公式准确性，即预测的准确性，至此，所有参数都不再是定值，将服从各自分布。
式(6)为式(5)中z 0、z 1所服从的概率分布可拆解成两种概率分布的乘积，需要说明的是，此公式隐含于计算过程中，不能具体表达，但并不影响计算。
式(7)表示z 0服从对数正太分布，μ为z 0的均值，由历史数据获得，σ2为z 0标准差，服从逆gamma分布。以式(5)-(8)为顺序，在实际计算过程中是逆序通过历史损伤数据首先得到f(x|θ)最大似然估计，之后通过实际经验确定τ－G(a,b)，其中a＝b＝0.01，如式(7)先验分布已确定为对数正态分布f(θ)，最终，结合式(6)及通过历史损伤数据得到的f(x|θ)最大似然估计使用贝叶斯公式，从而得到验证指标参数。建立预测模型后，本实施例使用winbugs软件进行MCMC迭代，即完成上述过程，上述公式的具体计算以及进行MCMC迭代可以采用现有的多种方法具体进行执行，由于最大似然估计受太多因素干扰，不能用具体公式表示，本实施例使用MCMC方法的目的旨在产生大量服从历史数据分布规律的数据，该数据用于表示最大似然估计f(x|θ)，最终获得z 0的均值，即损伤增长速率。
上述实施例中的列车部件裂纹损伤预测方法，收集高速列车关键部位历史损伤数据(等运行里程下裂纹累计长度)；确定高速列车关键部位寿命分布和验证指标参数，建立贝叶斯统计分布模型；基于历史数据确定验证指标参数的先验分布，基于MCMC方法，利用WinBUGS预测高速列车关键部位损伤增长速率。
根据本发明的列车部件裂纹损伤监测方法可以基于响应信号对列车部件裂纹损伤进行检测，并损伤数据进行存储以基于损伤数据对损伤部位的发展趋势进行预测评价，给出可靠性评估和隐患预警，实现对列车部件裂纹损伤的监测，为列车的维修、更换提供更为准确的、直观化的依据。
图3为根据本发明的列车部件裂纹损监测系统的一个实施例中的结构示意图。如图3所示，本发明的列车部件裂纹损监测系统包括蜂窝传感器网络装置30、多通道数据转换器31、车载数据处理装置32和中心服务器33。
所述蜂窝传感器网络装置30设置在待检测列车部件的检测结构上，且包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点。多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列。每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点、响应信号采集点。基本检测单元包括呈正六边形排列的六个压电传感器。压电传感器内嵌有微控制器，压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
所述多通道数据转换器31配置为将所述蜂窝传感器网络装置30采集的响应信号传输至所述车载数据处理装置32。所述多通道数据转换器31具有多个输入输出通道，通过所述多通道数据转换器31将所述列出上布置的所有蜂窝传感器网络装置30联网，以将所有蜂窝传感器网络装置30采集的响应信号传输至所述车载数据处理装置32进行处理。
所述车载数据处理装置32配置为对所述蜂窝传感器网络装置采集的响应信号进行预处理，以获得所述检测结构的损伤数据，并将所述检测结构的损伤数据传输至所述中心服务器；其中，损伤数据包括：所述检测结构的裂纹长度数据。
在一个实施例中，如图4所示，本发明的所述车载数据处理装置32包括第一信号激发模块321、基准信息获取模块322、第二信号激发模块323、损伤信息获取模块324和成像分析模块325。
第一信号激发模块321在激励信号加载点利用探头以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号。
基准信息获取模块322获取第一Lamb波响应信号并通过Mindlin板理论建立Lamb波在检测结构的各向异性复合材料层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息。
第二信号激发模块323在激励信号加载点利用探头以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号。损伤信息获取模块324对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息。损伤信息获取模块324将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个响应信号采集点相对应的信号差异系数值SDC值。
成像分析模块325根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域。成像分析模块325基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像。成像分析模块325绘制SDC分布图，基于SDC分布图评估出裂纹的长度。
损伤信息获取模块324采用小波变换算法对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息,用于测量出Lamb波在待检测的检测结构中的实际传播的时间和速度。损伤信息获取模块324以第一Lamb波响应信号作为基信号，将第二Lamb波响应信号与第一Lamb波响应信号相减，得到损伤散射信号。损伤信息获取模块324通过聚焦的方法使损伤散射信号能量叠加放大，用以提高损伤散射信号的信噪比，并对损伤散射信号进行时间反转处理。
第二信号激发模块323对包含缺陷信息的第二Lamb波响应信号进行时间反转处理，并将处理后的第二Lamb波响应信号作为新的波源加载到激励信号加载点进行发射，用以在待检测的检测结构上激发激励信号，实现Lamb波在缺陷处的二次聚焦，建立幅值聚焦图，用以对损伤位置和区域进行成像识别。
所述中心服务器33配置为存储所述检测结构的损伤数据，并对所述损伤数据进行处理以实现对所述待检测列车部件的检测结构的远程监督。
在一个实施例中，如图5所示，所述中心服务器33包括：存储模块330、历史数据获取模块331、预测模型建立模块332和损伤增长预测模块334。
存储模块330用于存储所述车载数据处理装置32获得列车部件的检测结构的损伤数据，所述损伤数据包括：检测结构的裂纹长度数据。
历史数据获取模块331获取检测结构的历史损伤数据，损伤数据包括：检测结构的裂纹长度数据。
预测模型建立模块332根据历史损伤数据获得对于检测结构的损伤的寿命分布特征和验证指标参数，并建立与检测结构的损伤相对应的贝叶斯概率预测模型。
损伤增长预测模块333根据历史损伤数据分析得出验证指标参数的先验分布，采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测检测结构的损伤的增长速率。
预测模型建立模块332确定对于检测结构的损伤的寿命分布为对数正态分 布，对数正态分布的密度函数为：
Figure PCTCN2019082492-appb-000004
其中，μ是损伤尺寸的平均值，σ是损伤尺寸的标准差。
损伤增长预测模块333将验证指标参数的先验分布确定为检测结构的裂纹增长速率的对数正态分布f(θ)与最大似然估计f(x|θ)的结合，运用贝叶斯概率预测模型并结合历史损伤数据中的检测结构的裂纹数据x，获得后验分布f(θ|x)，并进行迭代计算获得裂纹在单位里程中的增长速率。
上述实施例提供的中心服务器33获取对于列车部件的检测结构的历史损伤数据，根据历史损伤数据获得对于检测结构的损伤的寿命分布特征和验证指标参数，并建立贝叶斯概率预测模型，分析得出验证指标参数的先验分布，采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测检测结构的损伤的增长速率；通过应用Bayesian-MCMC的方法并基于大量历史数据预测列车部件损伤增长，可以在数据不完全的情况下，对部分未知的列车部件损伤采用主观概率估计，然后用贝叶斯公式对发生概率进行修正，最后再利用期望值和修正概率做出最优维修或更新决策，为列车的维修、更换提供更为准确的、直观化的依据，并且计算准确度高。
可能以许多方式来实现本发明的方法和装置、设备。例如，可通过软件、硬件、固件或者软件、硬件、固件的任何组合来实现本发明的方法和装置、设备。用于方法的步骤的上述顺序仅是为了进行说明，本发明的方法的步骤不限于以上具体描述的顺序，除非以其它方式特别说明。此外，在一些实施例中，还可将本发明实施为记录在记录介质中的程序，这些程序包括用于实现根据本发明的方法的机器可读指令。因而，本发明还覆盖存储用于执行根据本发明的方法的程序的记录介质。
本发明的描述是为了示例和描述起见而给出的，而并不是无遗漏的或者将本发明限于所公开的形式。很多修改和变化对于本领域的普通技术人员而言是显然的。选择和描述实施例是为了更好说明本发明的原理和实际应用，并且使本领域的普通技术人员能够理解本发明从而设计适于特定用途的带有各种修改的各种实施例。

Claims (10)

  1. 一种列车部件裂纹损伤监测方法，其特征在于，包括：
    在待检测列车部件的检测结构上设置蜂窝传感器网络装置；其中，所述蜂窝传感器网络装置包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点；
    将对所述蜂窝传感器网络装置采集的响应信号传输至车载数据处理装置；
    对所述蜂窝传感器网络装置采集的响应信号进行预处理，以获得所述检测结构的损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据；
    将所述检测结构的损伤数据传输至中心服务器，进行存储和后续处理。
  2. 如权利要求1所述的方法，其特征在于，
    多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列；每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点和/或响应信号采集点；
    所述基本检测单元包括呈正六边形排列的六个压电传感器；
    所述压电传感器内嵌有微控制器，所述压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
  3. 如权利要求2所述的方法，其特征在于，还包括：
    在激励信号加载点以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波；各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号；
    获取第一Lamb波响应信号并建立Lamb波在检测结构的各向异性复合材料层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息；
    在激励信号加载点以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波；各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号；
    对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息；将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个所述响应信号采集点相对应的信号差异系数值SDC值；
    根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域；
    基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像；绘制SDC分布图，基于SDC分布图评估出裂纹的长度，生成所述损伤数据。
  4. 如权利要求1所述的方法，其特征在于，还包括：
    所述中心服务器基于存储的所述检测结构的历史损伤数据预测所述检测结构的损伤的增长速率。
  5. 如权利要求4所述的方法，其特征在于，所述预测所述检测结构的损伤的增长速率包括：
    根据所述历史损伤数据获得对于所述检测结构的损伤的寿命分布特征和验证指标参数，并建立与所述检测结构的损伤相对应的贝叶斯概率预测模型；
    根据所述历史损伤数据分析得出所述验证指标参数的先验分布；
    采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测所述检测结构的损伤的增长速率。
  6. 一种列车部件裂纹损伤监测系统，其特征在于，包括：蜂窝传感器网络装置、多通道数据转换器、车载数据处理装置和中心服务器，
    所述蜂窝传感器网络装置设置在待检测列车部件的检测结构上，且包括多个压电传感器，每个压电传感器都作为激励信号加载点和/或响应信号采集点；
    所述多通道数据转换器配置为将所述蜂窝传感器网络装置采集的响应信号传输至所述车载数据处理装置；
    所述车载数据处理装置配置为对所述蜂窝传感器网络装置采集的响应信号进行预处理，以获得所述检测结构的损伤数据，并将所述检测结构的损伤数据传输至所述中心服务器；其中，损伤数据包括：所述检测结构的裂纹长度数据；
    所述中心服务器配置为存储所述检测结构的损伤数据，并对所述损伤数据进行处理以实现对所述待检测列车部件的检测结构的远程监督。
  7. 根据权利要求6所述的系统，其特征在于，
    多个压电传感器排列成包括至少一个正六边形的基本检测单元的蜂窝状阵列；每个压电传感器都为蜂窝状阵列中的一个节点，作为激励信号加载点和/或响应信号采集点；
    所述基本检测单元包括呈正六边形排列的六个压电传感器；
    所述压电传感器内嵌有微控制器，所述压电传感器作为Lamb波激发信号器和Lamb波信号接收器。
  8. 根据权利要求7所述的系统，其特征在于，所述车载数据处理装置包括：
    第一信号激发模块，用于在激励信号加载点利用探头以第一时间间隔在健康的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第一Lamb波响应信号；
    基准信息获取模块，用于获取第一Lamb波响应信号并通过Mindlin板理论建立Lamb波在检测结构的各向异性复合材料层板中随传播角度变化的频散关系，获得Lamb波的理论速度分布，作为基准信息；
    第二信号激发模块，用于在激励信号加载点利用探头以第二时间间隔在待检测的检测结构上激发激励信号，在检测结构中产生Lamb波，各个响应信号采集点采集对于此Lamb波的第二Lamb波响应信号；
    损伤信息获取模块，用于对第二Lamb波响应信号在时域和频域上进行分析，提取特征信息；将第二Lamb波响应信号作为损伤信号，将第一Lamb波响应信号作为参考信号，基于损伤信号、参考信号以及基准信息、特征信息计算与各个所述响应信号采集点相对应的信号差异系数值SDC值；
    成像分析模块，用于根据获得的SDC值并采用概率成像原理，重构出检测结构中裂纹损伤可能存在的区域；基于SDC值判定裂纹方向，校正裂纹方向上的SDC值，用于强化裂纹方向上的重构图像信息，采用概率成像原理重构裂纹损伤图像；绘制SDC分布图，基于SDC分布图评估出裂纹的长度。
  9. 根据权利要求6所述的系统，其特征在于，所述中心服务器基于存储的所述检测结构的历史损伤数据预测所述检测结构的损伤的增长速率。
  10. 根据权利要求9所述的系统，其特征在于，所述中心服务器包括：
    历史数据获取模块，用于获取所述检测结构的历史损伤数据；其中，损伤数据包括：所述检测结构的裂纹长度数据；
    预测模型建立模块，用于根据所述历史损伤数据获得对于所述检测结构的损伤的寿命分布特征和验证指标参数，并建立与所述检测结构的损伤相对应的贝叶斯概率预测模型；
    损伤增长预测模块，用于根据所述历史损伤数据分析得出所述验证指标参数 的先验分布；采用马尔科夫链蒙特卡洛方法优化贝叶斯概率预测模型的模型参数，并预测所述检测结构的损伤的增长速率。
PCT/CN2019/082492 2018-04-17 2019-04-12 列车部件裂纹损伤监测方法和系统 WO2019201177A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP19788319.2A EP3783356A4 (en) 2018-04-17 2019-04-12 METHOD AND SYSTEM FOR MONITORING CRACK DAMAGE AFFECTING A TRAIN COMPONENT

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810351360.4A CN110389169A (zh) 2018-04-17 2018-04-17 列车部件裂纹损伤监测方法和系统
CN201810351360.4 2018-04-17

Publications (1)

Publication Number Publication Date
WO2019201177A1 true WO2019201177A1 (zh) 2019-10-24

Family

ID=68240439

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/082492 WO2019201177A1 (zh) 2018-04-17 2019-04-12 列车部件裂纹损伤监测方法和系统

Country Status (3)

Country Link
EP (1) EP3783356A4 (zh)
CN (1) CN110389169A (zh)
WO (1) WO2019201177A1 (zh)

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111077222A (zh) * 2020-03-02 2020-04-28 四川陆通检测科技有限公司 一种轨道板离缝的无损检测方法
CN111210409A (zh) * 2019-12-30 2020-05-29 浙江大学 一种基于条件生成对抗网络的结构损伤识别方法
CN111812210A (zh) * 2020-07-08 2020-10-23 丁刚 一种三维编织复合材料损伤源定位方法及装置
CN112268799A (zh) * 2020-10-16 2021-01-26 中国直升机设计研究所 一种复合材料结构静强度和疲劳强度一体化试验验证方法
CN112580264A (zh) * 2020-12-25 2021-03-30 中国人民解放军国防科技大学 基于bp神经网络算法的损伤点尺寸分布预测方法及系统
CN112949190A (zh) * 2021-03-08 2021-06-11 西南石油大学 基于r6-fad和bp-mcs的含裂纹缺陷管道的模糊随机可靠度评估方法
CN113155465A (zh) * 2021-04-22 2021-07-23 上海工程技术大学 一种便携式地铁牵引电机轴承状态检测装置
CN113237951A (zh) * 2021-05-11 2021-08-10 重庆大学 一种基于形状上下文动态时间规整的金属板疲劳损伤超声导波检测方法
CN113465732A (zh) * 2021-08-13 2021-10-01 重庆大学 基于emd-dnn的振动台结构位移响应预测方法及装置
CN113465733A (zh) * 2021-08-13 2021-10-01 重庆大学 基于eemd-dnn的振动台结构位移响应预测方法及装置
CN113484336A (zh) * 2021-07-06 2021-10-08 杭州电子科技大学 基于表面波和bp神经网络的亚表面裂纹尺寸测量方法
CN113791034A (zh) * 2021-09-30 2021-12-14 合肥德泰轨交数据有限公司 一种用于钢轨探伤的样本采集分类监管系统
CN113963174A (zh) * 2021-10-26 2022-01-21 国家高速列车青岛技术创新中心 多视角强度域与频域融合的转向架标识图像特征提取方法
CN114264727A (zh) * 2021-11-20 2022-04-01 重庆大学 一种基于运营列车动态响应的轨道—桥梁系统损伤识别的方法
CN114460175A (zh) * 2022-02-28 2022-05-10 西北工业大学 一种薄壁结构损伤检测方法
CN114595514A (zh) * 2022-02-24 2022-06-07 河北建筑工程学院 轨道安全智能计算方法
CN114878696A (zh) * 2022-07-06 2022-08-09 太原理工大学 一种弧形复合材料层合板分层损伤的识别方法
CN115593464A (zh) * 2022-10-31 2023-01-13 华东交通大学（Cn） 一种高速列车车轮伤损评估方法、存储介质及计算机
CN115753998A (zh) * 2022-12-07 2023-03-07 大连理工大学 用于服役环境下飞行器复合材料结构损伤的监测方法及监测系统
CN116343966A (zh) * 2023-03-27 2023-06-27 山东大学 基于延迟因子的概率乘累加结构损伤成像定位方法及系统
CN117538412A (zh) * 2023-11-09 2024-02-09 韶关学院 一种用于钢轨裂纹损伤的超声波检测方法

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112782236B (zh) * 2019-11-11 2022-05-06 株洲中车时代电气股份有限公司 变流器柜体的材料状态监测方法、系统、装置及存储介质
CN111323483A (zh) * 2020-03-20 2020-06-23 嘉兴博传科技有限公司 火车车钩系统损伤监测传感器网络的布置方法
CN111504666B (zh) * 2020-05-25 2022-05-06 中车青岛四方机车车辆股份有限公司 一种确定转向架损伤系数的方法及装置
CN112903953B (zh) * 2021-01-21 2022-03-15 北京航空航天大学 一种金属板结构损伤类型识别系统和方法
CA3215000A1 (en) * 2021-04-07 2022-10-13 Nathan BOLANDER Systems and methods for hybrid prognostics
CN113533513A (zh) * 2021-06-29 2021-10-22 北京交通大学 一种钢轨损伤实时监测方法及其监测装置
CN115754008A (zh) * 2022-09-28 2023-03-07 哈尔滨工业大学（威海） 结构损伤联合监测方法、系统、计算机设备和存储介质
CN116106430B (zh) * 2023-04-12 2023-06-27 中南大学 基于声发射技术的熔铸用耐火材料开裂诊断方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006177705A (ja) * 2004-12-21 2006-07-06 Mitsubishi Heavy Ind Ltd タンク検査装置及びタンク検査方法
CN103134857A (zh) * 2013-02-21 2013-06-05 南京邮电大学 利用Lamb波反射场的工程结构裂纹损伤监测评估方法
CN104279424A (zh) * 2013-07-02 2015-01-14 香港理工大学 检测及显示管道结构破损情况的方法和系统
CN104502457A (zh) * 2014-12-23 2015-04-08 南京邮电大学 一种改进Lamb波工程结构裂纹损伤监测与评估层析成像方法
CN105051527A (zh) * 2013-01-04 2015-11-11 西门子公司 疲劳破坏预测和结构完整性评定的无损检测的嵌入式缺陷的概率建模和尺寸测定
CN107870204A (zh) * 2016-09-22 2018-04-03 淮阴师范学院 一种准确性高的复合材料结构损伤监测方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103226129A (zh) * 2013-02-27 2013-07-31 常州兆能电子科技有限公司 一种基于平面电磁传感器的板材损伤探测装置及方法
CN106596726A (zh) * 2016-11-30 2017-04-26 南京邮电大学 一种十字正交扫描Lamb波工程结构裂纹损伤监测的方法
CN107133400B (zh) * 2017-05-03 2019-12-31 厦门大学 一种飞机结构疲劳可靠度贝叶斯组合预测方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006177705A (ja) * 2004-12-21 2006-07-06 Mitsubishi Heavy Ind Ltd タンク検査装置及びタンク検査方法
CN105051527A (zh) * 2013-01-04 2015-11-11 西门子公司 疲劳破坏预测和结构完整性评定的无损检测的嵌入式缺陷的概率建模和尺寸测定
CN103134857A (zh) * 2013-02-21 2013-06-05 南京邮电大学 利用Lamb波反射场的工程结构裂纹损伤监测评估方法
CN104279424A (zh) * 2013-07-02 2015-01-14 香港理工大学 检测及显示管道结构破损情况的方法和系统
CN104502457A (zh) * 2014-12-23 2015-04-08 南京邮电大学 一种改进Lamb波工程结构裂纹损伤监测与评估层析成像方法
CN107870204A (zh) * 2016-09-22 2018-04-03 淮阴师范学院 一种准确性高的复合材料结构损伤监测方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3783356A4 *

Cited By (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111210409B (zh) * 2019-12-30 2022-08-23 浙江大学 一种基于条件生成对抗网络的结构损伤识别方法
CN111210409A (zh) * 2019-12-30 2020-05-29 浙江大学 一种基于条件生成对抗网络的结构损伤识别方法
CN111077222A (zh) * 2020-03-02 2020-04-28 四川陆通检测科技有限公司 一种轨道板离缝的无损检测方法
CN111812210A (zh) * 2020-07-08 2020-10-23 丁刚 一种三维编织复合材料损伤源定位方法及装置
CN112268799A (zh) * 2020-10-16 2021-01-26 中国直升机设计研究所 一种复合材料结构静强度和疲劳强度一体化试验验证方法
CN112268799B (zh) * 2020-10-16 2022-09-09 中国直升机设计研究所 一种复合材料结构静强度和疲劳强度一体化试验验证方法
CN112580264A (zh) * 2020-12-25 2021-03-30 中国人民解放军国防科技大学 基于bp神经网络算法的损伤点尺寸分布预测方法及系统
CN112580264B (zh) * 2020-12-25 2023-07-07 中国人民解放军国防科技大学 基于bp神经网络算法的损伤点尺寸分布预测方法及系统
CN112949190A (zh) * 2021-03-08 2021-06-11 西南石油大学 基于r6-fad和bp-mcs的含裂纹缺陷管道的模糊随机可靠度评估方法
CN113155465A (zh) * 2021-04-22 2021-07-23 上海工程技术大学 一种便携式地铁牵引电机轴承状态检测装置
CN113237951A (zh) * 2021-05-11 2021-08-10 重庆大学 一种基于形状上下文动态时间规整的金属板疲劳损伤超声导波检测方法
CN113484336A (zh) * 2021-07-06 2021-10-08 杭州电子科技大学 基于表面波和bp神经网络的亚表面裂纹尺寸测量方法
CN113465732A (zh) * 2021-08-13 2021-10-01 重庆大学 基于emd-dnn的振动台结构位移响应预测方法及装置
CN113465733A (zh) * 2021-08-13 2021-10-01 重庆大学 基于eemd-dnn的振动台结构位移响应预测方法及装置
CN113791034B (zh) * 2021-09-30 2022-09-06 合肥德泰轨交数据有限公司 一种用于钢轨探伤的样本采集分类监管系统
CN113791034A (zh) * 2021-09-30 2021-12-14 合肥德泰轨交数据有限公司 一种用于钢轨探伤的样本采集分类监管系统
CN113963174A (zh) * 2021-10-26 2022-01-21 国家高速列车青岛技术创新中心 多视角强度域与频域融合的转向架标识图像特征提取方法
CN114264727A (zh) * 2021-11-20 2022-04-01 重庆大学 一种基于运营列车动态响应的轨道—桥梁系统损伤识别的方法
CN114264727B (zh) * 2021-11-20 2023-11-24 重庆大学 一种基于运营列车动态响应的轨道—桥梁系统损伤识别的方法
CN114595514A (zh) * 2022-02-24 2022-06-07 河北建筑工程学院 轨道安全智能计算方法
CN114460175A (zh) * 2022-02-28 2022-05-10 西北工业大学 一种薄壁结构损伤检测方法
CN114460175B (zh) * 2022-02-28 2024-03-15 西北工业大学 一种薄壁结构损伤检测方法
CN114878696A (zh) * 2022-07-06 2022-08-09 太原理工大学 一种弧形复合材料层合板分层损伤的识别方法
CN114878696B (zh) * 2022-07-06 2022-10-04 太原理工大学 一种弧形复合材料层合板分层损伤的识别方法
CN115593464A (zh) * 2022-10-31 2023-01-13 华东交通大学（Cn） 一种高速列车车轮伤损评估方法、存储介质及计算机
CN115593464B (zh) * 2022-10-31 2023-09-08 华东交通大学 一种高速列车车轮伤损评估方法、存储介质及计算机
CN115753998A (zh) * 2022-12-07 2023-03-07 大连理工大学 用于服役环境下飞行器复合材料结构损伤的监测方法及监测系统
CN116343966B (zh) * 2023-03-27 2023-11-17 山东大学 基于延迟因子的概率乘累加结构损伤成像定位方法及系统
CN116343966A (zh) * 2023-03-27 2023-06-27 山东大学 基于延迟因子的概率乘累加结构损伤成像定位方法及系统
US12062167B1 (en) 2023-03-27 2024-08-13 Shandong University Probability multiply-sum structural damage imaging positioning method and system based on delay factor
CN117538412A (zh) * 2023-11-09 2024-02-09 韶关学院 一种用于钢轨裂纹损伤的超声波检测方法
CN117538412B (zh) * 2023-11-09 2024-05-14 韶关学院 一种用于钢轨裂纹损伤的超声波检测方法

Also Published As

Publication number Publication date
EP3783356A4 (en) 2022-01-05
EP3783356A1 (en) 2021-02-24
CN110389169A (zh) 2019-10-29

Similar Documents

Publication Publication Date Title
WO2019201177A1 (zh) 列车部件裂纹损伤监测方法和系统
WO2019201176A1 (zh) 列车部件裂纹损伤预测方法和装置
Alemi et al. Condition monitoring approaches for the detection of railway wheel defects
CN110389170B (zh) 基于Lamb波成像的列车部件裂纹损伤检测方法和系统
Wang et al. Real-time defect detection of high-speed train wheels by using Bayesian forecasting and dynamic model
Falamarzi et al. A review on existing sensors and devices for inspecting railway infrastructure
US6951132B2 (en) Rail and train monitoring system and method
Roveri et al. Real-time monitoring of railway infrastructures using fibre Bragg grating sensors
JP7374104B2 (ja) 鉄道車両の車軸の監視
CN111351664A (zh) 一种基于lstm模型的轴承温度预测和报警诊断方法
CN101900708A (zh) 一种基于振动和声频信号的高速列车轨道伤损探测方法
CN105923014B (zh) 一种基于证据推理规则的轨道高低不平顺幅值估计方法
CN110606105B (zh) 一种无人驾驶地铁车辆走行部安全监测系统
CN111591319B (zh) 一种用于高速铁路的轨道状况动态监测方法
CN115758289B (zh) 一种基于多任务学习神经网络的钢轨波磨识别方法
CN105510438A (zh) 基于运营车辆的钢轨核伤检测系统及方法
Kundu et al. A review on condition monitoring technologies for railway rolling stock
Ye et al. Multislice Time-Frequency image Entropy as a feature for railway wheel fault diagnosis
Shaikh et al. State-of-the-art wayside condition monitoring systems for railway wheels: A comprehensive review
Wang et al. A practical structural health monitoring system for high-speed train car-body
CN107782548A (zh) 一种基于对轨道交通工具零部件检测系统
Chellaswamy et al. Optimized vehicle acceleration measurement for rail track condition monitoring
Brizuela et al. NDE system for railway wheel inspection in a standard FPGA
CN109211556B (zh) 一种轨道交通工具零部件检测系统
Vincent et al. A Cognitive Rail Track Breakage Detection System Using Artificial Neural Network

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19788319

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019788319

Country of ref document: EP

Effective date: 20201117