CN115728389A - Rail transit vehicle component quality detection device and method - Google Patents
Rail transit vehicle component quality detection device and method Download PDFInfo
- Publication number
- CN115728389A CN115728389A CN202310040803.9A CN202310040803A CN115728389A CN 115728389 A CN115728389 A CN 115728389A CN 202310040803 A CN202310040803 A CN 202310040803A CN 115728389 A CN115728389 A CN 115728389A
- Authority
- CN
- China
- Prior art keywords
- rail transit
- transit vehicle
- vehicle component
- probe
- kalman filtering
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention discloses a rail transit vehicle component quality detection device and method. The quality detection device comprises data processing equipment, a base, an upper support, a first inspection robot, a second inspection robot and a clamp; the first inspection robot is transversely movably arranged on the base; the first inspection robot and the second inspection robot respectively comprise a robot body, and the first inspection robot is provided with an upper detection probe; a detection probe under the second inspection robot; the upper detection probe and the lower detection probe respectively comprise a connecting plate, a probe horizontal plane rotation driving motor, a probe vertical plane rotation driving motor, a mounting plate, an ultrasonic detection probe and a camera; the ultrasonic detection probe is connected with the signal generator; the data processing equipment comprises an ultrasonic detection risk factor calculation module, a prediction module and a risk point judgment module; therefore, the method improves the sensitivity of detecting the defects of the rail transit vehicle members, can effectively detect the defects and further improves the product detection efficiency.
Description
Technical Field
The invention relates to a device and a method for detecting the quality of a rail transit vehicle component, and belongs to the technical field of detection.
Background
The rail transit is a type of transportation means or transportation system in which an operating vehicle needs to travel on a specific rail, and the most typical rail transit is a railway system consisting of a traditional train and a standard railway.
At present, the quality detection mode of rail transit vehicle components has the advantages of large detection blind area, poor near surface resolution, low signal-to-noise ratio and low sensitivity, increases the difficulty of defect identification and resolution, is not beneficial to defect detection, influences defect judgment, and influences detection efficiency on certain specification products.
Disclosure of Invention
The invention aims to provide a rail transit vehicle component quality detection device and method, which are used for improving the sensitivity of rail transit vehicle component defect detection and effectively detecting defects so as to improve the product detection efficiency.
In order to achieve the technical purpose, the invention adopts the following technical scheme:
a rail transit vehicle component quality detection device is used for detecting defects of a rail transit vehicle component to be detected and comprises data processing equipment, a base, an upper support, a first inspection robot, a second inspection robot and a clamp; wherein:
the upper support is arranged above the base through a support upright post;
the first inspection robot is transversely movably arranged on the base; the second inspection robot is transversely movably arranged on the upper support, and meanwhile, a clamp for clamping a rail transit vehicle component to be inspected is also arranged on the upper support;
the first inspection robot and the second inspection robot respectively comprise a robot body, and an upper detection probe is arranged on the surface of the robot body of the first inspection robot, which faces to the rail transit vehicle component to be detected; a lower detection probe is arranged on the surface of the robot body of the second inspection robot, which faces the rail transit vehicle component to be detected;
the upper detection probe and the lower detection probe respectively comprise a connecting plate, a probe horizontal plane rotation driving motor, a probe vertical plane rotation driving motor, a mounting plate, an ultrasonic detection probe and a camera; the probe horizontal plane rotation driving motor is connected with the robot body through a connecting plate, the power output end of the probe horizontal plane rotation driving motor is connected with the fixed part of the probe vertical plane rotation driving motor, the power output end of the probe vertical plane rotation driving motor is connected with the mounting plate, and the ultrasonic detection probe and the camera are respectively mounted on the mounting plate;
the ultrasonic detection probe and the camera are respectively connected with the data processing equipment, and the ultrasonic detection probe is connected with the signal generator; the camera is used for carrying out photographing detection on the rail transit vehicle component to be detected so as to obtain an image detection risk factor of the rail transit vehicle component to be detected, and synchronously uploading the image detection risk factor to the data processing equipment;
the data processing equipment comprises an ultrasonic detection risk factor calculation module, a prediction module and a risk point judgment module;
under the cooperative work of the probe horizontal plane rotation driving motor and the probe vertical plane rotation driving motor, the ultrasonic detection probe of the upper detection probe is contacted with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe of the lower detection probe is contacted with the lower surface of the rail transit vehicle component to be detected;
during detection, under the control of data processing equipment, a signal generator is started, sends a first excitation wave signal to the upper surface of a rail transit vehicle component to be detected through an ultrasonic detection probe of an upper detection probe, acquires a corresponding first received wave signal through a second probe, and feeds the first received wave signal back to an ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module calculates a corresponding time domain mirror image signal according to the received first received wave signal; under the control of data processing equipment, the time domain mirror image signal is used as a second excitation wave signal and is sent to a rail transit vehicle component to be detected through a second probe, and at the moment, a corresponding second receiving wave signal can be collected through the first probe and fed back to the ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module can obtain the ultrasonic detection risk factor of the rail transit vehicle component to be detected by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal, and synchronously uploads the ultrasonic detection risk factor to the prediction module;
the prediction module is constructed based on a neural network algorithm and can process according to the received ultrasonic detection risk factors and image detection risk factors to obtain fusion estimation risk factors and synchronously upload the fusion estimation risk factors to the risk point judgment module;
a defect risk threshold is preset in the risk point judgment module; and judging each defect risk point of the rail transit vehicle component to be detected by comparing the fusion estimation risk factor with the defect risk threshold.
Preferably, a first guide rail is arranged on the base along the transverse direction, and a second guide rail is arranged on the upper support along the transverse direction;
a first self-driven walking wheel is mounted below the robot body of the first inspection robot and can be assembled in a first guide rail;
the second is patrolled and examined the below of robot body of robot and is installed the second and from driving the walking wheel, the second is from driving the walking wheel and can assemble in the second guide rail.
Preferably, the clamp comprises two clamping plates; the two clamping plates can clamp/loosen the rail transit vehicle component to be detected under the driving of the clamping driving mechanism;
the clamping driving mechanism comprises a clamping driving motor, a screw, a guide pillar and a gear transmission mechanism;
the fixed part of the clamping driving motor is arranged on the upper support; the screw rod and the guide pillar are arranged in parallel and are positioned and supported by the upper support, and the power output end of the clamping driving motor is connected with the screw rod through a gear transmission mechanism;
the two clamping plates are arranged in parallel; each clamping plate is in threaded fit connection with the screw; and simultaneously, each clamping plate is connected with the guide post in a guiding way.
Preferably, the prediction model comprises a component prediction model, an adaptive weight fusion prediction model and a fusion layer, wherein the component prediction model comprises an RNN network model and a first kalman filtering model; the self-adaptive weight fusion prediction model comprises an LSTM network model and a second kalman filtering model;
in the reverse propagation process of the RNN model, performing first Kalman filtering estimation on a predicted value predicted in the forward propagation process through a first Kalman filtering model, and calculating the obtained first Kalman filtering estimation value in the reverse propagation process of the RNN model; the obtained first Kalman filtering estimation value enters an LSTM network model and a fusion layer;
in the backward propagation process of the LSTM network model, performing second Kalman filtering estimation on a predicted value obtained by prediction in the forward propagation process through a second Kalman filtering model, and calculating the obtained second Kalman filtering estimation value in the backward propagation process of the LSTM network model; the obtained second Kalman filtering estimation value enters a fusion layer;
performing self-adaptive updating on the risk factor and the weight of the fusion layer through the first kalman filtering estimation value and the second kalman filtering estimation value to obtain an updated fusion estimation risk factor;
preferably, the RNN network model includes a first input layer, a first hidden layer, and a first output layer, which are connected in sequence, and the first input layer is connected to the camera detection unit and the ultrasonic detection risk factor calculation module; the input end of the first kalman filtering model is connected with the first input layer and the first output layer respectively, and the output end of the first kalman filtering model is connected with the first hidden layer, the fusion layer and the adaptive weight fusion prediction model respectively.
Preferably, the LSTM network model includes a second input layer, a second fully-connected long-short term memory network, and a second output layer, which are connected in sequence, where the second input layer is connected with the output end of the first kalman filter model; the input end of the second kalman filtering model is connected with the second input layer and the second output layer respectively; the output end of the second kalman filtering model is connected with the second fully-connected long-short term memory network and the fusion layer respectively.
Preferably, the second fully-connected long-short term memory network comprises a fully-connected excitation-function-free layer, a 3-level connected long-short term memory network and a discarding layer which are connected in sequence.
Preferably, a cross entropy loss function is used as a loss function in the LSTM network model to calculate a loss value; calculating a loss value in the RNN model by using a Huber error loss function as a loss function; activating the RNN model by using a tanh activation function or a softmax activation function; and the LSTM network model is activated by adopting a sigmoid activation function or a tanh activation function.
The invention also provides a rail transit vehicle component quality detection method, which is realized based on the rail transit vehicle component quality detection device and comprises the following steps:
step one, mounting a workpiece
Installing the rail transit vehicle component to be detected above the base and below the upper support through a clamp;
step two, starting the first and second inspection robots
Synchronously starting the first and second inspection robots, so that the detection probes of the first inspection robot and the second inspection robot are distributed on the upper and lower sides of the rail transit vehicle component to be inspected and are opposite to each other;
step three, starting a probe horizontal plane rotation driving motor and a probe vertical plane rotation driving motor
Synchronously starting a probe horizontal plane rotation driving motor and a probe vertical plane rotation driving motor, so that an ultrasonic detection probe of a first inspection robot is in contact with the lower surface of a rail transit vehicle component to be detected, an ultrasonic detection probe of a second inspection robot is in contact with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe of the first inspection robot is opposite to the ultrasonic detection probe of the second inspection robot;
step four, acquiring and synchronously uploading risk factors for ultrasonic detection
Starting a signal generator, sending a first excitation wave signal to a rail transit vehicle component to be detected through an ultrasonic detection probe of a first inspection robot, and acquiring a corresponding first receiving wave signal through an ultrasonic detection probe of a second inspection robot;
then calculating a time domain image signal of the first received wave signal; then the time domain mirror image signal is used as a second excitation wave signal and is sent to a rail transit vehicle component to be detected through an ultrasonic detection probe of a second inspection robot, and at the moment, a corresponding second receiving wave signal can be acquired through the ultrasonic detection probe of the first inspection robot; the ultrasonic detection risk factor of the rail transit vehicle component to be detected can be obtained by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal;
step five, acquiring and synchronously uploading image information
The method comprises the steps that a camera is adopted to carry out photographing detection on a rail transit vehicle component to be detected, image detection risk factors of the rail transit vehicle component to be detected are obtained, and the obtained image detection risk factors are synchronously transmitted to data processing equipment;
processing the received ultrasonic detection risk factors and image detection risk factors by adopting a prediction model constructed based on a neural network algorithm in the data processing equipment to obtain fusion estimation risk factors; and comparing the obtained fusion estimation risk factor with a defect risk threshold preset in the data processing equipment, thereby judging each defect risk point of the rail transit vehicle component to be detected.
Preferably, in the sixth step, the prediction model comprises a component prediction model, an adaptive weight fusion prediction model and a fusion layer, wherein the component prediction model comprises an RNN network model and a first kalman filtering model; the self-adaptive weight fusion prediction model comprises an LSTM network model and a second kalman filtering model;
in the reverse propagation process of the RNN model, performing first Kalman filtering estimation on a predicted value predicted in the forward propagation process through a first Kalman filtering model, and calculating the obtained first Kalman filtering estimation value in the reverse propagation process of the RNN model; the obtained first Kalman filtering estimation value enters an LSTM network model and a fusion layer;
in the backward propagation process of the LSTM network model, performing second kalman filtering estimation on a predicted value obtained by prediction in the forward propagation process through a second kalman filtering model, and calculating the obtained second kalman filtering estimation value in the backward propagation process of the LSTM network model; the obtained second kalman filtering estimation value enters the fusion layer;
and performing self-adaptive updating on the risk factor and the weight of the fusion layer through the first Kalman filtering estimation value and the second Kalman filtering estimation value to obtain an updated fusion estimation risk factor.
Based on the technical purpose, compared with the prior art, the invention has the following advantages:
1. the invention provides a rail transit vehicle component quality detection device, and particularly discloses a detection means for ultrasonic detection risk factors, which can further improve the sensitivity of ultrasonic flaw detection, so that the quality of rail transit vehicle component quality detection is further improved.
2. In the reverse propagation process, the RNN model performs first Kalman filtering estimation on a predicted value predicted in the forward propagation process, and the obtained first Kalman filtering estimation value enters the reverse propagation process for calculation. And in the backward propagation process, the LSTM network model performs second Kalman filtering estimation on the predicted value obtained by prediction in the forward propagation process, and the obtained second Kalman filtering estimation value enters the backward propagation process for calculation. The fusion layer is adaptively updated through the first Kalman filtering estimation value and the second Kalman filtering estimation value to obtain updated fusion estimation risk factors, so that the method has high robustness and high precision of the obtained fusion estimation risk factors.
Drawings
FIG. 1 is a schematic structural diagram of a rail transit vehicle component quality detection device according to the present invention;
FIG. 2 is a schematic structural view of the upper/lower inspection probe of FIG. 1;
FIG. 3 is a control flow chart of the rail transit vehicle component quality detection device according to the present invention;
in fig. 1 to 2: 1-a base; 2-a first guide rail; 3-a first inspection robot; 4-supporting the upright post; 5, an upper support; 6-a second guide rail; 7-clamping plate; 8-guide column; 9-screw rod; 10-upper detection probe; 11-a clamping drive motor; 12-a second inspection robot; 13-lower detection probe; 14-probe horizontal plane rotation driving motor; 15-driving motor for vertical rotation of probe; 16-a mounting plate; 17-an ultrasonic detection probe; 18-a camera; 19. a connecting plate.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. The relative arrangement of the components and steps, expressions and numerical values set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise. Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate. In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting. Thus, other examples of the exemplary embodiments may have different values.
For ease of description, spatially relative terms such as "over 8230 \ 8230;,"' over 8230;, \8230; upper surface "," above ", etc. may be used herein to describe the spatial relationship of one device or feature to another device or feature as shown in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in the figures is turned over, devices described as "above" or "on" other devices or configurations would then be oriented "below" or "under" the other devices or configurations. Thus, the exemplary terms "at 8230; \8230; 'above" may include both orientations "at 8230; \8230;' above 8230; 'at 8230;' below 8230;" above ". The device may also be positioned in other different ways (rotated 90 degrees or at other orientations).
As shown in fig. 1 and 2, the rail transit vehicle component quality detection device of the invention is used for detecting defects of a rail transit vehicle component to be detected, and comprises a data processing device, a base 1, an upper support 5, a first inspection robot 3, a second inspection robot 12 and a clamp; wherein:
the upper support 5 is arranged above the base 1 through a support upright post 4; in this embodiment, two support columns 4 are symmetrically disposed above the base 1, and the upper support 5 is supported by the two support columns 4. In addition, the upper support 5 in this embodiment is a rectangular frame structure, and the base 1 is a flat plate structure.
The first inspection robot 3 is transversely movably arranged on the base 1; the second inspection robot 12 is transversely movably arranged on the upper support 5, and meanwhile, a clamp for clamping a rail transit vehicle component to be detected is also arranged on the upper support 5; specifically, in this embodiment, two first guide rails 2 parallel to each other are arranged on the base 1 along the transverse direction, and two second guide rails 6 parallel to each other are also arranged on the upper support 5 along the transverse direction. In addition, the clamp of the invention comprises two clamping plates 7; the two clamping plates 7 can clamp/loosen the rail transit vehicle component to be detected under the driving of the clamping driving mechanism; the clamping driving mechanism comprises a clamping driving motor 11, a screw 9, a guide post 8 and a gear transmission mechanism; the fixed part of the clamping driving motor 11 is arranged on the upper support 5; the screw rod 9 and the guide post 8 are arranged in parallel and are positioned and supported by the upper support 5, and the power output end of the clamping driving motor 11 is connected with the screw rod 9 through a gear transmission mechanism; the two clamping plates 7 are arranged in parallel; each clamping plate 7 is in threaded fit connection with the screw 9; and simultaneously, each clamping plate 7 is connected with the guide post 8 in a guiding way.
The first and second inspection robots 12 each comprise a robot body, and the robot body of the first inspection robot 3 is provided with an upper detection probe 10 on the surface facing the rail transit vehicle member to be inspected; the robot body of the second inspection robot 12 is provided with a lower detection probe 13 facing the surface of the rail transit vehicle member to be inspected; a first self-driven walking wheel is arranged below the robot body of the first inspection robot 3, the first self-driven walking wheels can be assembled in the first guide rail 2, and the number of the first self-driven walking wheels on each side is two, namely four first self-driven walking wheels are arranged below the robot body of the first inspection robot 3; the second is patrolled and examined the below of the robot body of robot 12 and is installed the second and from driving the walking wheel, the second is from driving the walking wheel and can assemble in second guide rail 6, and the second of every side is from driving the walking wheel and have two, and four second are installed from driving the walking wheel in the below of the robot body of robot 12 is patrolled and examined to the second promptly.
The upper detection probe 10 and the lower detection probe 13 respectively comprise a connecting plate 19, a probe horizontal plane rotation driving motor 14, a probe vertical plane rotation driving motor 15, a mounting plate 16, an ultrasonic detection probe 17 and a camera 18; the probe horizontal plane rotation driving motor 14 is connected with the robot body through a connecting plate 19, the power output end of the probe horizontal plane rotation driving motor 14 is connected with the fixed part of the probe vertical plane rotation driving motor 15, the power output end of the probe vertical plane rotation driving motor 15 is connected with the mounting plate 16, and the ultrasonic detection probe 17 and the camera 18 are respectively mounted on the mounting plate 16;
the ultrasonic detection probe 17 and the camera 18 are respectively connected with the data processing equipment, and the ultrasonic detection probe 17 is connected with the signal generator; the camera 18 is used for photographing and detecting the rail transit vehicle component to be detected so as to obtain an image detection risk factor of the rail transit vehicle component to be detected, and synchronously uploading the image detection risk factor to the data processing equipment;
the data processing equipment comprises an ultrasonic detection risk factor calculation module, a prediction module and a risk point judgment module;
under the cooperative work of the probe horizontal plane rotation driving motor 14 and the probe vertical plane rotation driving motor 15, the ultrasonic detection probe 17 of the upper detection probe 10 is contacted with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe 17 of the lower detection probe 13 is contacted with the lower surface of the rail transit vehicle component to be detected;
during detection, under the control of data processing equipment, a signal generator is started, sends a first excitation wave signal to the upper surface of a rail transit vehicle component to be detected through an ultrasonic detection probe 17 of an upper detection probe 10, acquires a corresponding first receiving wave signal through a second probe, and feeds the first receiving wave signal back to an ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module calculates a corresponding time domain mirror image signal according to the received first received wave signal; under the control of data processing equipment, the time domain mirror image signal is used as a second excitation wave signal and is sent to a rail transit vehicle component to be detected through a second probe, and at the moment, a corresponding second receiving wave signal can be collected through the first probe and fed back to the ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module can obtain the ultrasonic detection risk factor of the rail transit vehicle component to be detected by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal, and synchronously uploads the ultrasonic detection risk factor to the prediction module;
the prediction module is constructed based on a neural network algorithm and can process according to the received ultrasonic detection risk factors and image detection risk factors to obtain fusion estimation risk factors and synchronously upload the fusion estimation risk factors to the risk point judgment module;
a defect risk threshold is preset in the risk point judgment module; and judging each defect risk point of the rail transit vehicle component to be detected by comparing the fusion estimation risk factor with the defect risk threshold.
Based on the rail transit vehicle component quality detection device, the invention provides a rail transit vehicle component quality detection method, which specifically comprises the following steps:
step one, mounting a workpiece
Installing the rail transit vehicle component to be detected above the base 1 and below the upper support 5 through a clamp;
step two, starting the first and second inspection robots 12
Synchronously starting the first and second inspection robots 12, so that the detection probes of the first inspection robot 3 and the second inspection robot 12 are distributed on the upper and lower sides of the rail transit vehicle component to be inspected and are opposite to each other;
step three, starting the probe horizontal plane rotation driving motor 14 and the probe vertical plane rotation driving motor 15
Synchronously starting a probe horizontal plane rotation driving motor 14 and a probe vertical plane rotation driving motor 15, so that the ultrasonic detection probe 17 of the first inspection robot 3 is in contact with the lower surface of the rail transit vehicle component to be detected, the ultrasonic detection probe 17 of the second inspection robot 12 is in contact with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe 17 of the first inspection robot 3 is opposite to the ultrasonic detection probe 17 of the second inspection robot 12;
step four, acquiring and synchronously uploading risk factors for ultrasonic detection
Starting a signal generator, sending a first excitation wave signal to a rail transit vehicle component to be detected through an ultrasonic detection probe 17 of the first inspection robot 3, and acquiring a corresponding first receiving wave signal through an ultrasonic detection probe 17 of the second inspection robot 12;
then calculating a time domain image signal of the first received wave signal; then, the time domain mirror image signal is used as a second excitation wave signal and is sent to the rail transit vehicle component to be detected through the ultrasonic detection probe 17 of the second inspection robot 12, and at the moment, a corresponding second received wave signal can be acquired through the ultrasonic detection probe 17 of the first inspection robot 3; the ultrasonic detection risk factor of the rail transit vehicle member to be detected can be obtained by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal;
step five, acquiring image information and synchronously uploading
The method comprises the steps that a camera 18 is adopted to carry out photographing detection on a rail transit vehicle component to be detected, so that an image detection risk factor of the rail transit vehicle component to be detected is obtained, and the obtained image detection risk factor is synchronously transmitted to data processing equipment;
processing the received ultrasonic detection risk factors and image detection risk factors by adopting a prediction model constructed based on a neural network algorithm in the data processing equipment to obtain fusion estimation risk factors; and comparing the obtained fusion estimation risk factor with a defect risk threshold preset in the data processing equipment so as to judge each defect risk point of the rail transit vehicle component to be detected.
In the invention, the prediction model comprises a component prediction model, a self-adaptive weight fusion prediction model and a fusion layer, wherein the component prediction model comprises an RNN network model and a first kalman filtering model. The adaptive weight fusion prediction model includes an LSTM network model and a second kalman filter model.
The RNN network model comprises a first input layer, a first hidden layer and a first output layer which are sequentially connected, wherein the first input layer is connected with the image pickup detection unit, the ultrasonic flaw detection unit and the impact detection unit (optionally). The input end of the first Kalman filtering model is connected with a first input layer and a first output layer respectively, and the output end of the first Kalman filtering model is connected with a first hidden layer, a fusion layer and an adaptive weight fusion prediction model respectively.
The neurons of the first hidden layer of the RNN network model are provided with a feedback mechanism to realize the transmission of front and back information, so that the RNN has the capability of processing sequence data. And after the RNN model finishes training, predicting the output at the next moment.
RNN network model:
an input layer:
hiding the layer:
an output layer:
wherein the content of the first and second substances,
representing input at a first input level
The data of the time of day is,
，
is shown in
At the first moment
The first of the rail transit detected by the individual detecting unit
The risk factor of the individual probe points,
，
the number of the detection units is shown,
，
which represents a camera-shooting detection unit,
an ultrasonic inspection unit is shown,
it is indicated that the impact detection unit,
，
the number of the detection points representing the rail transit,
representing RNN networks
The hidden state of the moment of time,
the RNN network hidden layer activation function is represented, the RNN network hidden layer activation function is selected from a tanh activation function,
representing RNN network hidden layers
The weight matrix of the input data at a time,
representing RNN network hidden layers
The weight matrix of the time instant output data,
representing an RNN network hidden layer offset matrix,
representing RNN network output layers
The estimated value of the output of the time of day,
the RNN output layer activation function is expressed, the RNN output layer activation function is selected from a softmax activation function,
representing RNN network output layers
The weight matrix of the input data at a time,
representing the RNN network output layer offset matrix.
RNN network model in
In the forward propagation process of the time, a layer of unidirectional circulation neural network is formed along the time axis direction; and then, a layer of circulating neural network is taken as a unit along the network level direction, and the layer-by-layer superposition is carried out to form the deep circulating neural network. In the backward propagation process, performing first Kalman filtering estimation on a predicted value obtained by prediction in the forward propagation process to obtain a first Kalman filtering estimation value, and obtaining the first Kalman filtering estimation value through lossAfter loss calculation is carried out on the function, maximum likelihood estimation of RNN network model parameters is obtained by adopting a gradient descent method, and the obtained maximum likelihood estimation is used as a hidden state in the forward propagation process at the next moment.
The neural network model adopts a gradient descent method to obtain maximum likelihood estimation of the model parameters, and the back propagation is a process of solving the partial derivative (namely parameter gradient) of the model parameters of the loss function and updating the model parameters by using the partial derivative.
Kalman filtering:
consider a discrete system as shown below:
wherein the content of the first and second substances,
which indicates the current time of day,
it is indicated that the last time of day,
is the state vector of the system at the current time,
is the state vector of the system at the last moment,
is a matrix of the system and is,
in order to control the input of the electronic device,
in order to input the matrix, the input matrix is,
in order to be the noise of the system,
is the output of the system at the current moment,
to be the output matrix, the output matrix is,
noise is measured for the system.
wherein the content of the first and second substances,
representing a mathematical expectation.
The following covariance matrix is defined:
wherein the content of the first and second substances,
which represents the operation of the covariance,
to represent
The covariance matrix of (a) is determined,
to represent
Covariance matrix of (2), superscript
Representing a matrix transpose operation.
Suppose that at the current time,:
wherein, the first and the second end of the pipe are connected with each other,
an estimate value that is indicative of the current time of day,
a predicted value representing the current time is shown,
is a kalman gain matrix, which is to be solved,
is the output of the system at the current moment,
is an output matrix.
Simultaneous equations (1) and (4) can be obtained, the following equation relationships:
wherein the content of the first and second substances,
is the deviation between the true value and the estimated value,
is an identity matrix.
The following covariance matrix is defined:
wherein the content of the first and second substances,
is composed of
The covariance matrix of (2).
Simultaneous (3), (5) and (6) gives:
wherein, the first and the second end of the pipe are connected with each other,
is the deviation between the true value and the estimated value,
，
is the deviation between the true value and the predicted value,
，
is the covariance of the deviation between the true value and the predicted value,
。
for is to
Tracing, namely:
wherein the content of the first and second substances,
representing trace operations.
To formula (8) about
Partial differentiation of (d) can give:
wherein the content of the first and second substances,
is a partial differential operation.
According to the kalman filter criterion, let equation (9) be zero, and obtain:
wherein, the upper label
Representing a matrix inversion operation.
By combining the formulas (7) and (10), the following compounds can be obtained:
since at the current moment, the following are satisfied:
wherein, the first and the second end of the pipe are connected with each other,
is composed of
An estimate of the time of day.
Simultaneous (1) and (12) gives:
wherein, the first and the second end of the pipe are connected with each other,
。
then the user can either, for example,
wherein the content of the first and second substances,
。
as can be seen from the above derivation process, the kalman estimation actually consists of two processes: prediction and correction, in the prediction phase, the filter uses the estimate of the last state to make a prediction of the current state. In the correction phase, the filter uses the observed value for the current state to correct the predicted value obtained in the prediction phase to obtain a new estimate that is closer to the true value.
Predicted value equation at current moment:
updating the equation of the prediction covariance matrix:
calculation equation of kalman gain matrix:
the optimal estimation value equation at the current moment:
updating the equation of the estimated covariance matrix:
therefore, the Kalman filter algorithm only needs to give an initial state
And initial estimate covariance matrix
Can be based on the measured value at the current time
Obtaining the optimal estimated value of the system state
. Firstly, a predicted value of the current moment is obtained based on an estimated value of the previous moment, and the accuracy of prediction is described through a prediction covariance matrix. At the same time, a Kalman gain matrix is calculated based on the matrix
. Then, according to
And
and obtaining the optimal estimation value of the current moment. Finally, the estimated covariance matrix is updated by (19) and for the next time instant (A), (B)
) Prepare for recursion. And the Kalman filtering adopts a recursion method, continuously predicts and updates according to the measured values at different moments, and obtains the optimal estimated value of the system state. In the invention, the RNN network model is adopted to carry out first Kalman filtering estimation on a predicted value predicted in the forward propagation process through a first Kalman filtering model in the backward propagation process, and the obtained first Kalman filtering estimation value enters the backward propagation process of the RNN network model to be calculated. The obtained first Kalman filtering estimation value entersLSTM network model and fusion layer.
Therefore, the filtering process of the first kalman filter model is:
predicted value equation at current moment:
updating the equation of the prediction covariance matrix:
calculation equation of kalman gain matrix:
the optimal estimation value equation at the current moment is as follows:
updating the equation of the estimated covariance matrix:
wherein the content of the first and second substances,
representing the first kalman filtered prediction value,
，
is shown in
At the first moment
The first of the rail transit detected by the detecting unit
The first kalman filter prediction value of each probe,
，
the number of the detection units is shown,
，
which represents a camera-shooting detection unit,
an ultrasonic flaw detection unit is shown,
it is indicated that the impact detection unit,
，
the number of the detection points representing the rail transit,
representing the first kalman filter estimate,
system matrix for first kalman filtering， 
For the first kalman filtered input matrix,
is the control input to the first kalman filter,
the covariance matrix is predicted for the first kalman filter,
a covariance matrix is estimated for the first kalman filter,
for the first kalman filtered system noise covariance matrix,
for the first kalman filtered gain matrix,
for the first kalman filtered output matrix,
for the first kalman filtered measurement noise covariance matrix,
is an identity matrix.
The LSTM network model comprises a second input layer, a second full-connection long-short term memory network and a second output layer which are connected in sequence, wherein the second full-connection long-short term memory network comprises a full-connection non-excitation function layer, a 3-level connected long-short term memory network and a discarding layer which are connected in sequence. And the second kalman filtering model is respectively connected with the first input layer, the second input layer and the second output layer.
The second full-connection long-short term memory network comprises a forgetting gate, an input gate and an output gate, wherein:
forget the door:
an input gate:
and (3) updating the cell state:
an output gate:
wherein, the first and the second end of the pipe are connected with each other,
the forgetting of the door is shown,
is an input to the LSTM network and,
，
in the state of being hidden, the first electrode is in a hidden state,
in order to update the state of the cell,
is the output of the LSTM network.
and calculating a loss value in the LSTM network model by using a cross entropy loss function as a loss function.
Wherein the content of the first and second substances,
a label representing a sample, a positive class of 1, a negative class of 0,
is a sample
The probability of a prediction being a positive class,
representing the total number of samples.
The LSTM network model is in
In the forward propagation process of the time, a layer of unidirectional circulation neural network is formed along the direction of a time axis; then, a layer of circulating neural network is taken as a unit along the network level direction, and the layer-by-layer superposition is carried out to form a deep circulating neural network; in the backward propagation process, performing second Kalman filtering estimation on the predicted value obtained by prediction in the forward propagation process, performing loss calculation on the obtained second Kalman filtering estimation value through a loss function, obtaining the maximum likelihood estimation of the LSTM network model parameter by adopting a gradient descent method, and taking the obtained maximum likelihood estimation as the maximum likelihood estimation before the next momentTowards a hidden state in the propagation process.
As can be seen from equations (15) - (19), the filtering process of the second kalman filtering model is as follows:
predicted value equation at current moment:
updating the equation of the prediction covariance matrix:
calculation equation of kalman gain matrix:
the optimal estimation value equation at the current moment:
updating the equation of the estimated covariance matrix:
wherein, the first and the second end of the pipe are connected with each other,
is shown in
At the first moment
The first of the rail transit detected by the individual detecting unit
A second kalman filter predicted weights of the probe points,
，
the number of the detection units is shown,
，
which represents a camera-shooting detection unit,
an ultrasonic inspection unit is shown,
it is shown that the impact detection unit,
，
the number of the detection points representing the rail transit,
is shown in
At the first moment
The first of the rail transit detected by the detecting unit
A second kalman filter of the probe points estimates the weights,
system matrix for second kalman filtering， 
For the second kalman filtered input matrix,
is the control input to the second kalman filter,
the covariance matrix is predicted for the second kalman filter,
a covariance matrix is estimated for the second kalman filter,
for the second kalman filtered system noise covariance matrix,
for the second kalman filtered gain matrix,
for the output matrix of the second kalman filter,
for the second kalman filtered measurement noise covariance matrix,
is an identity matrix.
In the invention, the LSTM network model is adopted to carry out second Kalman filtering estimation on the predicted value obtained by prediction in the forward propagation process through the second Kalman filtering model in the backward propagation process, and the obtained second Kalman filtering estimation value enters the backward propagation process of the LSTM network model to be calculated. And the obtained second kalman filter estimation value enters the fusion layer.
The fusion layer is respectively connected with the first kalman filtering model and the second kalman filtering model. And performing self-adaptive updating on the risk factor and the weight of the fusion layer through the first Kalman filtering estimation value and the second Kalman filtering estimation value to obtain an updated fusion estimation risk factor.
Wherein the content of the first and second substances,
is shown in
The first time of rail transit
The fusion risk factors of the individual probe points,
is shown in
At the first moment
The first of the rail transit detected by the individual detecting unit
The first kalman filter predictor of each probe point,
is shown in
At the first moment
The first of the rail transit detected by the detecting unit
A second kalman filter of the prediction weights for each probe point,
indicating the number of detection units.
The risk point judgment unit is used for judging a risk point according to the fusion estimation risk factor, if the fusion estimation risk factor is within a set risk point threshold range, the risk point of the rail transit vehicle component to be detected is judged to be safe, and a safety signal of the risk point is output through the output unit. And if the fusion estimation risk factor is not in the set risk point threshold range, judging that the risk point of the rail transit vehicle component to be detected has risk, and outputting the risk point alarm signal of the rail transit vehicle component to be detected through an output unit.
According to the invention, the image detection risk factor, the ultrasonic detection risk factor and the impact detection risk factor (optional) are subjected to cyclic network neural calculation through the RNN network model and the first kalman filter, so that the robustness of the system is improved, and the prediction accuracy of the image detection risk factor, the ultrasonic detection risk factor and the impact detection risk factor is improved. The fusion weight of the fusion layer is updated adaptively through the LSTM network model and the second kalman filter, and the fusion layer obtains updated fusion estimation risk factors according to the updated fusion weight and the first kalman filter estimation.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention, and such modifications and adaptations are intended to be within the scope of the invention.
Claims (10)
1. A rail transit vehicle component quality detection device is used for detecting defects of a rail transit vehicle component to be detected and is characterized by comprising data processing equipment, a base, an upper support, a first inspection robot, a second inspection robot and a clamp; wherein:
the upper support is arranged above the base through a support upright post;
the first inspection robot is transversely movably arranged on the base; the second inspection robot is transversely movably arranged on the upper support, and meanwhile, a clamp for clamping a rail transit vehicle component to be inspected is also arranged on the upper support;
the first inspection robot and the second inspection robot respectively comprise a robot body, and an upper detection probe is arranged on the surface of the robot body of the first inspection robot, which faces to the rail transit vehicle component to be detected; a lower detection probe is arranged on the surface of the robot body of the second inspection robot, which faces the rail transit vehicle component to be detected;
the upper detection probe and the lower detection probe respectively comprise a connecting plate, a probe horizontal plane rotation driving motor, a probe vertical plane rotation driving motor, a mounting plate, an ultrasonic detection probe and a camera; the probe horizontal plane rotation driving motor is connected with the robot body through a connecting plate, the power output end of the probe horizontal plane rotation driving motor is connected with the fixed part of the probe vertical plane rotation driving motor, the power output end of the probe vertical plane rotation driving motor is connected with the mounting plate, and the ultrasonic detection probe and the camera are respectively mounted on the mounting plate;
the ultrasonic detection probe and the camera are respectively connected with the data processing equipment, and the ultrasonic detection probe is connected with the signal generator; the camera is used for carrying out photographing detection on the rail transit vehicle component to be detected so as to obtain an image detection risk factor of the rail transit vehicle component to be detected, and synchronously uploading the image detection risk factor to the data processing equipment;
the data processing equipment comprises an ultrasonic detection risk factor calculation module, a prediction module and a risk point judgment module;
under the cooperative work of the probe horizontal plane rotation driving motor and the probe vertical plane rotation driving motor, the ultrasonic detection probe of the upper detection probe is contacted with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe of the lower detection probe is contacted with the lower surface of the rail transit vehicle component to be detected;
during detection, under the control of data processing equipment, a signal generator is started, sends a first excitation wave signal to the upper surface of a rail transit vehicle component to be detected through an ultrasonic detection probe of an upper detection probe, acquires a corresponding first received wave signal through a second probe, and feeds the first received wave signal back to an ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module calculates a corresponding time domain mirror image signal according to the received first received wave signal; under the control of data processing equipment, the time domain mirror image signal is used as a second excitation wave signal and is sent to a rail transit vehicle component to be detected through a second probe, and at the moment, a corresponding second receiving wave signal can be collected through the first probe and fed back to the ultrasonic detection risk factor calculation module; the ultrasonic detection risk factor calculation module can obtain the ultrasonic detection risk factor of the rail transit vehicle component to be detected by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal, and synchronously uploads the ultrasonic detection risk factor to the prediction module;
the prediction module is constructed based on a neural network algorithm, and can process according to the received ultrasonic detection risk factors and image detection risk factors to obtain fusion estimation risk factors and synchronously upload the fusion estimation risk factors to the risk point judgment module;
a defect risk threshold is preset in the risk point judgment module; and judging each defect risk point of the rail transit vehicle component to be detected by comparing the fusion estimation risk factor with the defect risk threshold.
2. The rail transit vehicle component quality detection device of claim 1, wherein a first guide rail is disposed laterally on the base, and a second guide rail is disposed laterally on the upper mount;
a first self-driven walking wheel is mounted below a robot body of the first inspection robot and can be assembled in a first guide rail;
the second is patrolled and examined the robot's robot body's below and is installed the second and from driving the walking wheel, the second can assemble in the second guide rail from driving the walking wheel.
3. The rail transit vehicle component quality detection device of claim 1, wherein the clamp includes two clamp plates; the two clamping plates can clamp/loosen the rail transit vehicle component to be detected under the driving of the clamping driving mechanism;
the clamping driving mechanism comprises a clamping driving motor, a screw, a guide pillar and a gear transmission mechanism;
the fixed part of the clamping driving motor is arranged on the upper support; the screw rod and the guide pillar are arranged in parallel and are positioned and supported by the upper support, and the power output end of the clamping driving motor is connected with the screw rod through a gear transmission mechanism;
the two clamping plates are arranged in parallel; each clamping plate is in threaded fit connection with the screw; and simultaneously, each clamping plate is connected with the guide post in a guiding way.
4. The rail transit vehicle component quality detection apparatus of claim 1, wherein the prediction model comprises a component prediction model, an adaptive weight fusion prediction model, and a fusion layer, the component prediction model comprising an RNN network model and a first kalman filter model; the self-adaptive weight fusion prediction model comprises an LSTM network model and a second kalman filtering model;
in the reverse propagation process of the RNN model, performing first Kalman filtering estimation on a predicted value predicted in the forward propagation process through a first Kalman filtering model, and calculating the obtained first Kalman filtering estimation value in the reverse propagation process of the RNN model; the obtained first Kalman filtering estimation value enters an LSTM network model and a fusion layer;
in the backward propagation process of the LSTM network model, performing second Kalman filtering estimation on a predicted value obtained by prediction in the forward propagation process through a second Kalman filtering model, and calculating the obtained second Kalman filtering estimation value in the backward propagation process of the LSTM network model; the obtained second kalman filtering estimation value enters the fusion layer;
and performing self-adaptive updating on the risk factor and the weight of the fusion layer through the first Kalman filtering estimation value and the second Kalman filtering estimation value to obtain an updated fusion estimation risk factor.
5. The rail transit vehicle component quality detection device of claim 4, wherein the RNN model comprises a first input layer, a first hidden layer and a first output layer which are sequentially connected, and the first input layer is connected with the image pickup detection unit and the ultrasonic detection risk factor calculation module; the input end of the first kalman filtering model is connected with the first input layer and the first output layer respectively, and the output end of the first kalman filtering model is connected with the first hidden layer, the fusion layer and the adaptive weight fusion prediction model respectively.
6. The rail transit vehicle component quality detection device of claim 5, wherein the LSTM network model comprises a second input layer, a second fully-connected long-short term memory network and a second output layer which are connected in sequence, and the second input layer is connected with an output end of the first kalman filtering model; the input end of the second kalman filtering model is connected with the second input layer and the second output layer respectively; the output end of the second kalman filtering model is connected with the second fully-connected long-short term memory network and the fusion layer respectively.
7. The rail transit vehicle component quality detection device of claim 6, wherein the second fully-connected long-short term memory network comprises a fully-connected non-excitation function layer, a 3-level connected long-short term memory network and a discarding layer which are connected in sequence.
8. The rail transit vehicle component quality detection apparatus according to claim 7, wherein a cross entropy loss function is used in the LSTM network model as a loss function to calculate a loss value; calculating a loss value in the RNN model by using a Huber error loss function as a loss function; activating the RNN model by using a tanh activation function or a softmax activation function; and the LSTM network model is activated by adopting a sigmoid activation function or a tanh activation function.
9. A rail transit vehicle component quality detection method is realized based on the rail transit vehicle component quality detection device of claim 1, and is characterized by comprising the following steps:
step one, installing a workpiece
Installing the rail transit vehicle component to be detected above the base and below the upper support through a clamp;
step two, starting the first and second inspection robots
Synchronously starting the first and second inspection robots, so that the detection probes of the first inspection robot and the second inspection robot are distributed on the upper and lower sides of the rail transit vehicle component to be detected and are opposite to each other;
step three, starting a probe horizontal plane rotation driving motor and a probe vertical plane rotation driving motor
Synchronously starting a probe horizontal plane rotation driving motor and a probe vertical plane rotation driving motor, so that an ultrasonic detection probe of a first inspection robot is in contact with the lower surface of a rail transit vehicle component to be detected, an ultrasonic detection probe of a second inspection robot is in contact with the upper surface of the rail transit vehicle component to be detected, and the ultrasonic detection probe of the first inspection robot is opposite to the ultrasonic detection probe of the second inspection robot;
step four, acquiring and synchronously uploading risk factors for ultrasonic detection
Starting a signal generator, sending a first excitation wave signal to a rail transit vehicle component to be detected through an ultrasonic detection probe of a first inspection robot, and acquiring a corresponding first receiving wave signal through an ultrasonic detection probe of a second inspection robot;
then calculating a time domain image signal of the first received wave signal; then the time domain mirror image signal is used as a second excitation wave signal and is sent to a rail transit vehicle component to be detected through an ultrasonic detection probe of a second inspection robot, and at the moment, a corresponding second receiving wave signal can be acquired through the ultrasonic detection probe of the first inspection robot; the ultrasonic detection risk factor of the rail transit vehicle component to be detected can be obtained by calculating the correlation coefficient between the first excitation wave signal and the second receiving wave signal;
step five, acquiring and synchronously uploading image information
The method comprises the steps that a camera is adopted to carry out photographing detection on a rail transit vehicle component to be detected, image detection risk factors of the rail transit vehicle component to be detected are obtained, and the obtained image detection risk factors are synchronously transmitted to data processing equipment;
processing the received ultrasonic detection risk factors and image detection risk factors by adopting a prediction model constructed based on a neural network algorithm in the data processing equipment to obtain fusion estimation risk factors; and comparing the obtained fusion estimation risk factor with a defect risk threshold preset in the data processing equipment so as to judge each defect risk point of the rail transit vehicle component to be detected.
10. The rail transit vehicle component quality detection method according to claim 9, wherein in the sixth step, the prediction model comprises a component prediction model, an adaptive weight fusion prediction model and a fusion layer, and the component prediction model comprises an RNN network model and a first kalman filtering model; the self-adaptive weight fusion prediction model comprises an LSTM network model and a second kalman filtering model;
in the reverse propagation process of the RNN model, performing first Kalman filtering estimation on a predicted value predicted in the forward propagation process through a first Kalman filtering model, and calculating the obtained first Kalman filtering estimation value in the reverse propagation process of the RNN model; the obtained first Kalman filtering estimation value enters an LSTM network model and a fusion layer;
in the backward propagation process of the LSTM network model, performing second Kalman filtering estimation on a predicted value obtained by prediction in the forward propagation process through a second Kalman filtering model, and calculating the obtained second Kalman filtering estimation value in the backward propagation process of the LSTM network model; the obtained second Kalman filtering estimation value enters a fusion layer;
and performing self-adaptive updating on the risk factor and the weight of the fusion layer through the first Kalman filtering estimation value and the second Kalman filtering estimation value to obtain an updated fusion estimation risk factor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310040803.9A CN115728389B (en) | 2023-01-12 | 2023-01-12 | Rail transit vehicle component quality detection device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310040803.9A CN115728389B (en) | 2023-01-12 | 2023-01-12 | Rail transit vehicle component quality detection device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115728389A true CN115728389A (en) | 2023-03-03 |
| CN115728389B CN115728389B (en) | 2023-05-16 |
Family
ID=85302051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310040803.9A Active CN115728389B (en) | 2023-01-12 | 2023-01-12 | Rail transit vehicle component quality detection device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115728389B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5042305A (en) * | 1988-07-30 | 1991-08-27 | Hitachi Construction Machinery Co., Ltd. | Ultrasonic flaw detecting system |
| JPH0510928A (en) * | 1991-06-28 | 1993-01-19 | Hitachi Constr Mach Co Ltd | Ultrasonic wave image inspection apparatus |
| US20110030479A1 (en) * | 2008-03-31 | 2011-02-10 | Junichi Murai | Ultrasonic flaw detection method and ultrasonic flaw detection equipment |
| CN104634866A (en) * | 2015-02-16 | 2015-05-20 | 爱德森（厦门）电子有限公司 | Device and method for evaluating directions and depths of metal cracks by using ultrasonic electromagnetic principle |
| JP2015206607A (en) * | 2014-04-17 | 2015-11-19 | 三菱電機株式会社 | Measuring apparatus |
| JP2017223510A (en) * | 2016-06-14 | 2017-12-21 | 日立Ｇｅニュークリア・エナジー株式会社 | Submerged visualization device and submerged visualization method |
| CN107735680A (en) * | 2016-04-25 | 2018-02-23 | 非破坏检查株式会社 | The stripping inspection method and stripping check device of layered product |
| CN108593767A (en) * | 2018-01-24 | 2018-09-28 | 天津大学 | A kind of method for building up of shoal buried pipes supersonic sounding echo model |
| CN111521679A (en) * | 2020-05-06 | 2020-08-11 | 京冶华诚(天津)钢结构有限公司 | Flaw detection method for steel structure |
| CN114715216A (en) * | 2022-03-26 | 2022-07-08 | 北京交通运输职业学院 | Intelligent maintenance system for rail transit and control method thereof |
| CN115540790A (en) * | 2022-10-30 | 2022-12-30 | 苏州爱思尔提科技有限公司 | High-precision ultrasonic thickness measurement method and device |
| CN115577860A (en) * | 2022-11-21 | 2023-01-06 | 南京地铁运营咨询科技发展有限公司 | Intelligent maintenance method and system for rail transit based on adaptive control |
-
2023
- 2023-01-12 CN CN202310040803.9A patent/CN115728389B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5042305A (en) * | 1988-07-30 | 1991-08-27 | Hitachi Construction Machinery Co., Ltd. | Ultrasonic flaw detecting system |
| JPH0510928A (en) * | 1991-06-28 | 1993-01-19 | Hitachi Constr Mach Co Ltd | Ultrasonic wave image inspection apparatus |
| US20110030479A1 (en) * | 2008-03-31 | 2011-02-10 | Junichi Murai | Ultrasonic flaw detection method and ultrasonic flaw detection equipment |
| JP2015206607A (en) * | 2014-04-17 | 2015-11-19 | 三菱電機株式会社 | Measuring apparatus |
| CN104634866A (en) * | 2015-02-16 | 2015-05-20 | 爱德森（厦门）电子有限公司 | Device and method for evaluating directions and depths of metal cracks by using ultrasonic electromagnetic principle |
| CN107735680A (en) * | 2016-04-25 | 2018-02-23 | 非破坏检查株式会社 | The stripping inspection method and stripping check device of layered product |
| JP2017223510A (en) * | 2016-06-14 | 2017-12-21 | 日立Ｇｅニュークリア・エナジー株式会社 | Submerged visualization device and submerged visualization method |
| CN108593767A (en) * | 2018-01-24 | 2018-09-28 | 天津大学 | A kind of method for building up of shoal buried pipes supersonic sounding echo model |
| CN111521679A (en) * | 2020-05-06 | 2020-08-11 | 京冶华诚(天津)钢结构有限公司 | Flaw detection method for steel structure |
| CN114715216A (en) * | 2022-03-26 | 2022-07-08 | 北京交通运输职业学院 | Intelligent maintenance system for rail transit and control method thereof |
| CN115540790A (en) * | 2022-10-30 | 2022-12-30 | 苏州爱思尔提科技有限公司 | High-precision ultrasonic thickness measurement method and device |
| CN115577860A (en) * | 2022-11-21 | 2023-01-06 | 南京地铁运营咨询科技发展有限公司 | Intelligent maintenance method and system for rail transit based on adaptive control |
Non-Patent Citations (1)
| Title |
|---|
| 郑阳;何存富;吴斌;周进节;: "采用超声导波阵列技术研究板类结构大面积检测", 声学学报 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115728389B (en) | 2023-05-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jin et al. | Environmental boundary tracking and estimation using multiple autonomous vehicles | |
| CN113987834A (en) | CAN-LSTM-based railway train bearing residual life prediction method | |
| US11428658B2 (en) | Gas sensing device and method for operating a gas sensing device | |
| CN108267123A (en) | A kind of double-current vehicle-mounted pedestrian vehicle Forecasting Methodology based on bounding box and range prediction | |
| CN104899866A (en) | Intelligent infrared small target detection method | |
| EP1540046A1 (en) | Method and apparatus for real time monitoring of industrial electrolytes | |
| CN109477810B (en) | Electrochemical method for determining the sensitivity of a gas sensor by means of a pulse sequence | |
| CN107703496A (en) | Tracking before a kind of motor-driven Faint target detection of interactive multimode Bernoulli Jacob filtering | |
| CN107831485A (en) | The method of the multiple water body optical signature parameters of the more visual field laser radar detections of boat-carrying | |
| CN111862165A (en) | Target tracking method for updating Kalman filter based on deep reinforcement learning | |
| CN115577860B (en) | Intelligent maintenance method and system for rail transit based on adaptive control | |
| CN111030889A (en) | Network traffic prediction method based on GRU model | |
| CN113203464A (en) | Sensor fault detection method of dynamic vehicle-mounted weighing system | |
| CN112162545A (en) | Automobile fault diagnosis method and system | |
| CN115728389A (en) | Rail transit vehicle component quality detection device and method | |
| Li | Research on fault detection algorithm of pantograph based on edge computing image processing | |
| CN112070808A (en) | Trajectory prediction method, apparatus and computer-readable storage medium | |
| CN109191788B (en) | Driver fatigue driving judgment method, storage medium, and electronic device | |
| CN112836859A (en) | Intelligent fusion and analysis method for river mouth area pollutant monitoring data | |
| CN109835333B (en) | Control system and control method for keeping vehicle running in middle of lane | |
| CN113705865B (en) | Automobile stability factor prediction method based on deep neural network | |
| CN111561929B (en) | Time delay and noise reduction method for vehicle-mounted MEMS inertial sensor | |
| CN115047853A (en) | Micro fault detection method based on recursion standard variable residual error and kernel principal component analysis | |
| US6009185A (en) | Neural network based contact state estimator | |
| CN114091160A (en) | Intelligent detection method and system for bridge structure damage |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |