Intelligent rail vehicle hinge device system monitoring method

Abstract

The intelligent rail vehicle hinge device system monitoring method comprises the following steps: step 1, carrying out data acquisition through a test; 2, preprocessing the data; and step 3, carrying out deep learning on the preprocessed data based on the adaptive residual capsule network to obtain parameters of a model used for monitoring, specifically, dividing the preprocessed data, then extracting prominent features in data signals by adopting adaptive residual, and finally outputting the prominent features through a capsule. Fourthly, the model with the determined parameters is installed on the intelligent rail vehicle for real-time monitoring. Real-time monitoring of the rotating angle of the intelligent rail vehicle and the pressure of the damping arm is achieved through the angle sensor and the pressure sensor of the hinge device, problems can be found in time, even part damage caused by long-term vibration is avoided, the safety of stable operation of the hinge device is improved, the deep learning algorithm is simple and practical, and the application range is wide. The method is suitable for engineering application.

Description

Technical Field

[0001] The present invention relates to a system monitoring method, and in particular to a system monitoring method for an articulated device of a smart rail vehicle. Background Art

[0002] In the field of urban transportation, with the development of social economy and the improvement of people's living standards, the number of motor vehicles has increased dramatically, which has led to the deterioration of the urban traffic environment, the reduction of road network traffic efficiency, and slow traffic on main and auxiliary roads. The design of intelligent rail vehicles combines the advantages of modern trams and buses. Intelligent rail vehicles use virtual trajectory following control technology to identify virtual track lines on the road surface through on-board sensors to achieve track-like travel. The articulation device of the intelligent rail vehicle enables the intelligent rail vehicle to have a smaller turning radius while maintaining the length of the train, solving the turning problem caused by the overlong body. However, the damping arm failure problem of the intelligent rail vehicle articulation device will occur after long-term service. If it is not repaired or replaced in time, it may cause excessive turning, which may cause serious traffic accidents at high speeds. In existing technologies, there is no monitoring technology for the articulation device of the intelligent rail vehicle, nor is there any consideration of real-time fault warning and diagnosis solutions for this component. Summary of the Invention

[0003] In response to the problem that the current smart rail vehicle articulation device system lacks monitoring technology, the present invention proposes a smart rail vehicle articulation device system monitoring method, which can realize real-time monitoring of the operating status of the smart rail vehicle articulation device during service, thereby improving the safety and stability of the articulation device.

[0004] The technical means adopted by the present invention to solve the above problems are: a monitoring method for the articulated device system of a smart rail vehicle, comprising the following steps: a first step, collecting data through experiments; a second step, preprocessing the data; a third step, performing deep learning on the preprocessed data based on an adaptive residual capsule network to obtain the parameters of the model used for monitoring, specifically: first dividing the preprocessed data, then using adaptive residuals to extract prominent features in the data signal, and finally outputting them through capsules; a fourth step, installing the model with determined parameters on the smart rail vehicle for real-time monitoring.

[0005] Furthermore, the tests in the first step include bench tests and actual vehicle tests. The bench test collects the pressure electrical signals of the articulated device sensors and converts the pressure electrical signals into digital pressure signals. The actual vehicle test is to complete multiple rounds of actual vehicle operation tests in actual environments under upper and lower load limit conditions after the articulated device is installed on the vehicle during the service period of the articulated device, collect the pressure electrical signals of the articulated device sensors, and convert the pressure electrical signals into digital pressure signals.

[0006] Furthermore, in the second step, when pre-processing the data, abnormal mutation signals are detected and removed to obtain the first hydraulic digital signal and the first vibration signal.

[0007] Furthermore, the specific method of preprocessing in the second step is: first, the digital pressure signal and angle signal are normalized to map the data to the range of 0 to 1; then, the normalized data is detected and removed for abnormal mutation values ​​to reduce the computational complexity and the impact of abnormal mutation values ​​on fault warning and diagnosis errors; finally, the data is overlapped and sampled, and data enhancement is performed to highlight data features.

[0008] Furthermore, in the third step, the adaptive residual calculates the optimal coefficient of each channel, thereby adaptively amplifying sensitive features and suppressing irrelevant features, and effectively maintaining the range of output values ​​through the Sigmoid function, which is expressed as:

[0009]

[0010] x i represents the characteristics of the i-th channel; σ i represents the adaptive coefficient of the i-th channel;

[0011] Then the obtained adaptive coefficient σ i Multiplied by the output feature α of another branch in the adaptive module:

[0012] y=σ i ×α.

[0013] In this way, the network can further retain effective information by identifying the optimal coefficients of each channel, adaptive methods and suppressing features according to the importance of features.

[0014] Furthermore, in the third step, the capsule network includes a primary capsule layer and a digital capsule layer. The digital capsule layer includes a dynamic routing algorithm. The main idea of ​​the capsule network is that the operation is divided into three stages. The first stage is completed in the primary capsule layer, and the second and third stages are completed in the digital capsule layer.

[0015] Furthermore, the first stage calculates the prediction vector U j|i , the calculation formula is as follows:

[0016] U j|i =W ij U i ,

[0017] U i is the output of the i-th neuron in the previous convolutional network; W ij is the output of the neuron U i Multiply the weight matrix.

[0018] Furthermore, the second stage calculates the prediction vector U obtained in the first stage. j|i Multiply by c ij Perform weighted summation and finally get an output vector S j , the calculation formula is as follows:

[0019] S j =∑ i c ij *U j|i

[0020] c ij It is a coupling coefficient determined by a dynamic routing algorithm, which allows the input neurons to autonomously select the best path to transmit to the next layer of neurons.

[0021] Furthermore, the coupling coefficient c ij By initializing the vector b ij Softmax normalization is performed, and the initial vector b ij By the prediction vector U j|i And the output vector V j Perform iterative updates, and the calculation formula is as follows:

[0022]

[0023] b ij =b ij +V j *U j|i .

[0024] Furthermore, the third stage converts the intermediate vector S obtained in the second stage into j The output vector V of the routing capsule layer is obtained by performing nonlinear transformation through the nonlinear activation vector compression function Squashing. j The advantage of this activation function is that it can keep the direction of the initial vector unchanged and compress the modulus of the input vector to between (0,1). The calculation formula is as follows:

[0025]

[0026] in, Represents the total input vector S j The reduction factor of Represents the total input vector S j The unit vector of .

[0027] The beneficial effects of the present invention are:

[0028] 1. This invention uses angle sensors and pressure sensors in the articulated joint to monitor the vehicle's rotation angle and damping arm pressure in real time. This allows for timely detection of problems and even prevents component damage caused by long-term vibration, improving the stability and safety of the articulated joint. It also addresses the issues of early warning of impending failure during service and subsequent fault diagnosis, thus avoiding the risk of component damage caused by untimely repair or replacement. Furthermore, the invention's deep learning algorithm is simple and practical, making it suitable for engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a schematic diagram of the structure of the hinge device system in Example 1;

[0030] Figure 2 This is a schematic diagram of monitoring the hinge device system in Example 1;

[0031] Figure 3 This is a monitoring flow chart for Example 1;

[0032] Figure 4 Schematic diagram of the adaptive residual capsule network in Example 1;

[0033] Figure 5 This is a schematic diagram of the construction principle of the reference database in Example 1;

[0034] Figure 6 This is a schematic diagram of the failure warning and diagnosis process of Example 1;

[0035] In the figure: 1. First damping arm, 2. Second damping arm, 3. First pressure sensor, 4. Second pressure sensor, 5. First angle sensor, 6. Second angle sensor, 7. Third angle sensor, 8. Articulated device, 9. Controller, 10. Data acquisition module, 11. Data analysis module, 12. Fault monitoring, early warning and diagnosis module, 13. Reference data test platform, 14. Host computer. DETAILED DESCRIPTION

[0036] The present invention is further described below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are schematic, not actual, representations. They should not be construed as limiting this patent. To better illustrate the embodiments of the present invention, some components in the accompanying drawings may be omitted, enlarged, or reduced in size, and do not represent the actual dimensions of the products. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted from the accompanying drawings.

[0037] Example 1

[0038] A method for monitoring the articulation system of a smart rail vehicle, such as Figure 1As shown, a controller 9 is provided inside the articulated device 8, and the controller 9 is connected to a first angle sensor 5, a second angle sensor 6 and a third angle sensor 7. A first damping arm 1 and a second damping arm 2 are provided outside the articulated device 8, and a first pressure sensor 3 and a second pressure sensor 4 are arranged on the first damping arm 1 and the second damping arm 2 respectively. When the smart rail vehicle turns, the first angle sensor 5, the second angle sensor 6 and the third angle sensor 7 will monitor the angle signal at this time, thereby limiting the turning radius of the smart rail vehicle through the damping of the first damping arm 1 and the second damping arm 2. The number of pressure sensors is at least two, and the number of angle sensors is at least three. Of course, there can be more, as long as the installation space requirements are met. As shown Figure 1 and Figure 2 As shown, the entire articulated device system also includes a data acquisition module 10 , a data analysis module 11 , a fault monitoring, early warning and diagnosis module 12 , a reference data testing platform 13 and a host computer 14 .

[0039] Among them, the first pressure sensor 3, the second pressure sensor 4 and the first angle sensor 5, the second angle sensor 6, and the third angle sensor 7 transmit the real-time measured pressure signals and angle signals to the data acquisition module 10, and then transmit the data to the data analysis module 11. The data analysis module 11 converts the signals to obtain the unprocessed original pressure signal and the original angle signal.

[0040] The data analysis module 11 transmits the converted signal to the fault monitoring, warning, and diagnosis module 12. This module first normalizes the original signal, eliminates outliers, and enhances the original data through overlapping sampling. The enhanced data set is then fed into a trained adaptive residual capsule network for fault warning and diagnosis. Finally, the real-time status of the articulated system is transmitted to the host computer 14 for display. Once the host computer 14 receives failure information, it issues a warning and performs fault diagnosis.

[0041] The reference data testing platform 13 is used to establish a reference database, which is then transmitted to the monitoring, warning and diagnosis module 12 to train the adaptive residual capsule network multiple times and save the optimal training model parameters. When the signals of the articulated device 8 during service are collected and transmitted to the monitoring, warning and diagnosis module 12, the trained model can be directly used for real-time status monitoring, warning and diagnosis. Figure 5 Schematic diagram of the construction principle of the reference database.

[0042] Specifically, as 1 and Figure 3As shown, the articulated joint system monitoring process of this embodiment includes the following steps: First, data acquisition is performed through testing, which includes bench testing and on-vehicle testing. During the bench testing, the data acquisition module 10 collects pressure signals from the first and second pressure sensors 3 and 4, and the data analysis module 11 converts the pressure signals into digital pressure signals using a specific relationship. The signals are then normalized and pre-processed, and data enhancement is performed before data storage. During the on-vehicle testing, after the articulated joint 8 is installed on the vehicle, multiple rounds of on-vehicle operation tests are conducted under both upper and lower load limits during the service life of the articulated joint 8. The data acquisition module 10 collects pressure signals from the first and second pressure sensors 3 and 4, as well as angle signals from the first, second, and third angle sensors 5, 6, and 7. The data analysis module 11 converts the pressure signals into digital pressure signals using a specific relationship. The digital pressure and angle signals are then normalized and pre-processed, and data enhancement is performed before data storage. Finally, the bench test data and on-vehicle test data are used to construct a reference database for training the adaptive residual capsule network.

[0043] The second step is to preprocess the data, detect and remove abnormal mutation signals, and obtain the first hydraulic digital signal and the first vibration signal. The specific process is as follows: (1) normalize the pressure digital signal and the angle signal, and map the data to the range of 0 to 1; (2) detect and remove abnormal mutation values ​​on the normalized data to reduce the computational complexity and the impact of abnormal mutation values ​​on fault warning and diagnosis; (3) overlap sampling the normalized data, enhance the data, and highlight the data features.

[0044] The third step, such as Figure 3 and Figure 4 As shown in the figure, based on the adaptive residual capsule network, the adaptive residual is used to calculate the optimal coefficient of each channel, adaptively amplify sensitive features and suppress irrelevant features. Specifically, it means: first divide the preprocessed data, then use the adaptive residual to extract the prominent features in the data signal, and finally output it through the capsule.

[0045] In this step, the adaptive model can obtain the optimal coefficient of each channel feature, thereby adaptively amplifying sensitive features and suppressing irrelevant features. In the adaptive model, the Sigmoid function can effectively maintain the range of output values. The adaptive output can be expressed as:

[0046]

[0047] Where: x i represents the characteristics of the i-th channel; σ i Represents the adaptive coefficient of the i-th channel. Then the adaptive coefficient σi Multiplied by the output feature a of another branch in the adaptation module:

[0048] y=σ i ×α.

[0049] In this way, the network can further retain effective information by identifying the optimal coefficients of each channel, adaptive methods and suppressing features according to the importance of features.

[0050] The capsule network includes the primary capsule layer and the digital capsule layer. The digital capsule layer includes a dynamic routing algorithm. The main idea of ​​the capsule network is divided into three stages. The first stage is completed in the primary capsule layer to calculate the prediction vector U j|i , the calculation formula is as follows:

[0051] U j|i =W ij U i ,

[0052] Among them, U i is the output of the i-th neuron in the previous convolutional network; W ij is the output of the neuron U i Multiply the weight matrix; U j|i is the prediction vector.

[0053] The second stage is to get the prediction vector U from the previous stage. j|i Multiply by c ij Perform weighted summation and finally get an output vector S j , which is calculated as follows:

[0054] S j =∑ i c ij *U j|i

[0055] c ij It is a coupling coefficient determined by a dynamic routing algorithm, which allows the input neurons to autonomously select the best path to transmit to the next layer of neurons.

[0056] In this stage, the core idea is the dynamic routing algorithm, and the core work of dynamic routing is to find the optimal coupling coefficient c ij , c ij The second operation stage of the capsule network plays a very critical role. It determines the connection between the current capsule layer and the previous capsule layer by adjusting the initial vector b. ij Softmax normalization is performed, and the initial vector is obtained by the prediction vector U j|i And the output vector V j Perform iterative updates, and the calculation formula is as follows:

[0057]

[0058] b ij =b ij +V j *U j|i

[0059] In each forward propagation, b ij Initialized to 0, the coupling coefficient c is calculated ij The initialization value of V is then calculated by the forward propagation of the network j . Each forward propagation updates b ij The value of c is then updated ij The value of S is further updated j The value of V j Through this step-by-step iterative update, a set of optimal coupling coefficients is found.

[0060] The third stage converts the intermediate vector S obtained in the second stage into j The output vector V of the routing capsule layer is obtained by performing nonlinear transformation through a new nonlinear activation vector compression function Squashing. j The advantage of this activation function is that it can keep the direction of the initial vector unchanged and compress the modulus of the input vector to between (0,1). Its calculation formula is as follows:

[0061]

[0062] in, Represents the total input vector S j The reduction factor of Represents the total input vector S j The unit vector of .

[0063] In the above process, the original data is first processed into multiple samples of the same length, and the processed samples are divided into training set and test set in a ratio of 8:2. The training set is first sent to the network model for model training, and then the test set is used for verification to determine the model parameters with the best fault warning and diagnosis effects.

[0064] The fourth step is as follows Figure 5As shown, the model with determined parameters is installed on the smart rail vehicle for real-time monitoring. In this embodiment, the different damping conditions of the damping arm of the articulated device are divided into: warning when the measured damping value deviates from the lower limit by 1%, warning when the measured damping value deviates from the lower limit by 2%, warning when the measured damping value deviates from the lower limit by less than 3%, warning when the measured damping value deviates from the lower limit by less than 4%, warning when the measured damping value deviates from the lower limit by more than 5%; the corresponding warning thresholds of the vibration amplitude are: threshold 1, threshold 2, threshold 3, threshold 4 and threshold 5. Failure forms include: sensor failure warning, warning when the measured damping value deviates from the lower limit by 1% to 5%. As Figure 6 As shown, the data analysis module 10 uses the pressure signals in the first pressure sensor 3 and the second pressure sensor 4 of the damping arm and the angle signals of the first angle sensor 5, the second angle sensor 6 and the third angle sensor 7 to monitor the working status of the articulated device 8 during its service. When the real-time signal of the articulated device 8 during its service is input into the network model, it will determine whether its working status is normal based on the signal characteristics of the input signal, and highlight the working status through the signal indicator light. If the tensor matrix values ​​output by the network model are all 0 or constant, the model outputs a sensor failure warning and synchronously displays its current status on the host computer 14, at which time the blue indicator light is on. A gradient threshold is set in the network model, and different thresholds represent different working states of the articulated device. If the RMS value of the tensor matrix output by the network model is greater than threshold 1 and less than threshold 2, the model outputs damping failure warning 1 (the measured damping value deviates from the lower limit by 1%), and synchronizes its current state on the host computer 14. At this time, the yellow indicator light is on and the flashing time interval is 15s; if the RMS value of the tensor matrix output by the network model is greater than threshold 2 and less than threshold 3, the model outputs damping failure warning 2 (the measured damping value deviates from the lower limit by 2%), and synchronizes its current state on the host computer 14 display. At this time, the yellow indicator light is on and the flashing time interval is 10s; if the RMS value of the tensor matrix output by the network model is greater than threshold 3 and less than threshold 4, the model outputs damping failure warning 3 (the measured damping value deviates from the lower limit by 3%), and the current status is displayed synchronously on the host computer 14. At this time, the yellow indicator light is on and the flashing time interval is 7s; if the RMS value of the tensor matrix output by the network model is greater than the threshold 4 and less than the threshold 5, the model outputs a damping failure warning 4 (the measured damping value deviates from the lower limit by 4%), and the current status is displayed synchronously on the host computer 14. At this time, the yellow indicator light is on and the flashing time interval is 5s; if the RMS value of the tensor matrix output by the network model is greater than the threshold 5, the model outputs a damping failure warning 5 (the measured damping value deviates from the lower limit by 5%), and the current status is displayed synchronously on the host computer 14. At this time, the red indicator light is on and the flashing time interval is 1s. The above failure warning situation can be carried out synchronously with the fault diagnosis.

[0065] The articulated device works by limiting the vehicle's turning radius and angle through a damping arm. The angle and pressure signals measured by the sensors act as feedback signals to each other. If the sensors or damping ratios fail, the vehicle's turning radius cannot be effectively controlled, potentially posing a safety hazard or even causing a serious traffic accident. This method uses the angle and pressure sensors in the articulated device to monitor the vehicle's turning angle and damping arm pressure in real time. This allows for timely detection of problems and even prevents component damage caused by long-term vibration, ensuring the vehicle's safety and reliability.

[0066] The above embodiments are only for the purpose of illustrating the present invention, and are not intended to limit the present invention. Those skilled in the art may make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of protection of the present invention, and the scope of protection of the present invention should be defined by the claims.

Claims

1. A method for monitoring an articulated joint system of a smart rail vehicle, characterized by: The method includes the following steps: the first step is to collect data through experiments; the second step is to preprocess the data; the third step is to perform deep learning on the preprocessed data based on the adaptive residual capsule network to obtain the parameters of the model used for monitoring, which specifically means: first divide the preprocessed data, then use adaptive residual to extract the prominent features in the data signal, and finally output it through the capsule; the fourth step is to install the model with determined parameters on the smart rail vehicle for real-time monitoring.

2. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 1, wherein: The tests in the first step include bench tests and actual vehicle tests. The bench test collects the pressure electrical signals of the articulated device sensors and converts them into digital pressure signals. The actual vehicle test is to install the articulated device on the vehicle and complete multiple rounds of actual vehicle operation tests under the upper and lower load limit conditions in actual environments during the service period of the articulated device, collect the pressure electrical signals of the articulated device sensors, and convert them into digital pressure signals.

3. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 1, wherein: In the second step, when preprocessing the data, abnormal mutation signals are detected and removed to obtain a first hydraulic digital signal and a first vibration signal.

4. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 3, wherein: The specific preprocessing method in the second step is: first, normalize the digital pressure signal and angle signal and map the data to the range of 0 to 1; then detect and remove abnormal mutation values ​​on the normalized data to reduce the computational complexity and the impact of abnormal mutation values ​​on incorrect judgment of fault warning and diagnosis; finally, overlap sample the data, enhance the data, and highlight the data features.

5. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 1, wherein: In the third step, the adaptive residual calculates the optimal coefficient of each channel, thereby adaptively amplifying sensitive features and suppressing irrelevant features, and effectively maintaining the range of output values ​​through the Sigmoid function, which is expressed as: x i represents the characteristics of the i-th channel; σ i represents the adaptive coefficient of the i-th channel; Then the obtained adaptive coefficient σ i Multiplied by the output feature α of another branch in the adaptive module: y=σ i ×a. In this way, the network can further retain effective information by identifying the optimal coefficients of each channel, adaptive methods and suppressing features according to the importance of features.

6. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 5, wherein: In the third step, the capsule network includes a primary capsule layer and a digital capsule layer. The digital capsule layer includes a dynamic routing algorithm. The main idea of ​​the capsule network is that the operation is divided into three stages. The first stage is completed in the primary capsule layer, and the second and third stages are completed in the digital capsule layer.

7. The method for monitoring the articulation system of a smart rail vehicle according to claim 6, characterized in that: The first stage calculates the prediction vector U j|i , the calculation formula is as follows: U j|i =W ij U i , U i is the output of the i-th neuron in the previous convolutional network; W ij is the output of the neuron U i Multiply the weight matrix.

8. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 7, wherein: The second stage is the prediction vector U obtained in the first stage j|i Multiply by c ij Perform weighted summation and finally get an output vector S j , the calculation formula is as follows: S j =∑ i c ij *U j|i c ij It is a coupling coefficient determined by a dynamic routing algorithm, which allows the input neurons to autonomously select the best path to transmit to the next layer of neurons.

9. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 8, wherein: Coupling coefficient c ij By initializing the vector b ij Softmax normalization is performed, and the initial vector b ij By the prediction vector U j|i And the output vector V j Perform iterative updates, and the calculation formula is as follows: b ij =b ij +V j *U j|i 。 10. The method for monitoring the articulated joint system of a smart rail vehicle according to claim 8, wherein: The third stage converts the intermediate vector S obtained in the second stage into j The output vector V of the routing capsule layer is obtained by performing nonlinear transformation through the nonlinear activation vector compression function Squashing. j , the calculation formula is as follows: in, Represents the total input vector S j The reduction factor of Represents the total input vector S j The unit vector of .

Images