Method and device for separating acceleration signals of high-speed railway vehicles

Abstract

The application discloses a method and device for separating a high-speed railway vehicle acceleration signal, and the method comprises the following steps: collecting a vehicle acceleration signal; normalizing the vehicle acceleration signal to obtain a transformed value of the vehicle acceleration signal; inputting the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network to separate source signals in the transformed value of the vehicle acceleration signal; and determining corresponding acceleration component signals according to the separated source signals. The application normalizes the vehicle acceleration signal, separates the vehicle acceleration signal through the nonlinear mixing and demixing model based on the radial basis function neural network, adopts blind source separation, and improves the accuracy of the vehicle acceleration signal separation.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method and apparatus for separating acceleration signals of high-speed railway vehicles. Background Technology

[0002] High-speed railways place higher demands on track conditions. Even minor track geometric irregularities can significantly impact the safety, smoothness, and comfort of high-speed train operation, as well as environmental noise. Utilizing high-speed train acceleration data allows for the timely extraction of track condition characteristics, ensuring the safe and reliable operation of high-speed trains and providing a theoretical basis for track condition adjustments.

[0003] To achieve signal separation, many researchers start with the instantaneous frequency of the signal. The instantaneous frequency is a quantitative physical description of the signal at a specific moment and is currently studied and applied in various communication and detection signals. From a physics perspective, signals are divided into single-component and multi-component signals. For single-component signals, the phase difference method, while relatively simple to calculate, has weak noise resistance and can produce phase ambiguity, making it only suitable for non-stationary signals with approximately linear phase changes. When the phase changes nonlinearly, phase modeling methods are more reliable for accurately estimating the instantaneous frequency. The zero-crossing method calculates the instantaneous frequency by counting the number of zeros the signal crosses, but it can only be used for single-frequency signals and is ineffective for multi-frequency signals. While the Teager energy operator method has low computational cost, it is not very suitable for low signal-to-noise ratio signals.

[0004] Vehicle dynamic response signals are a special type of nonlinear, non-stationary signal. Analyzing vehicle dynamic response data helps to better understand the various states of rail vehicle systems, especially providing greater assistance in diagnosing short-wave track defects. Existing techniques propose using wavelet analysis to analyze rail corrugation and separate the rail corrugation component contained in the vehicle acceleration signal. However, this requires knowledge of the source information, or noise needs to be added during the separation process, which can interfere with the separation results.

[0005] In summary, there is an urgent need for a method for separating acceleration signals from high-speed railway vehicles to address the problems existing in the aforementioned technologies. Summary of the Invention

[0006] This invention provides a method for separating acceleration signals of high-speed railway vehicles to improve the accuracy of vehicle acceleration signal separation. The method includes:

[0007] Collect vehicle acceleration signals;

[0008] The vehicle acceleration signal is normalized to obtain the transformed value of the vehicle acceleration signal;

[0009] The transformed value of the vehicle acceleration signal is input into a nonlinear mixing and demixing model based on a radial basis function neural network to separate the source signal from the transformed value of the vehicle acceleration signal.

[0010] The corresponding acceleration component signal is determined based on the separated source signal.

[0011] This invention also provides a device for separating acceleration signals of high-speed railway vehicles, to improve the accuracy of vehicle acceleration signal separation. The device includes:

[0012] The transformation module is used to acquire vehicle acceleration signals; it standardizes and normalizes the vehicle acceleration signals to obtain the transformed values ​​of the vehicle acceleration signals.

[0013] The separation module is used to input the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network, and to separate the source signal from the transformed value of the vehicle acceleration signal; and to determine the corresponding acceleration component signal based on the separated source signal.

[0014] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0015] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0016] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0017] In this embodiment of the invention, vehicle acceleration signals are acquired; the vehicle acceleration signals are normalized to obtain transformed values ​​of the vehicle acceleration signals; the transformed values ​​of the vehicle acceleration signals are input into a nonlinear mixing and demixing model based on a radial basis function neural network to separate the source signals in the transformed values ​​of the vehicle acceleration signals; the corresponding acceleration component signals are determined based on the separated source signals. Compared with the prior art, the vehicle acceleration signals are separated by using a nonlinear mixing and demixing model based on a radial basis function neural network after normalizing the vehicle acceleration signals, and blind source separation is adopted, which improves the accuracy of vehicle acceleration signal separation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0019] Figure 1 A flowchart illustrating the method for separating acceleration signals of high-speed railway vehicles provided by the present invention;

[0020] Figure 2 A flowchart illustrating the method for separating acceleration signals of high-speed railway vehicles provided by the present invention;

[0021] Figure 3 A schematic diagram of the post-nonlinear mixing and demixing model provided by the present invention;

[0022] Figure 4 A schematic diagram of the nonlinear mixing and demixing model based on radial basis function neural network provided by the present invention;

[0023] Figure 5 A schematic diagram of the continuous multi-wave signal separated from the vehicle acceleration signal provided by the present invention;

[0024] Figure 6 This is a schematic diagram of the structure of the device for separating acceleration signals of high-speed railway vehicles provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0026] Figure 1 This is a flowchart illustrating a method for separating acceleration signals from high-speed railway vehicles, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0027] Step 101: Collect vehicle acceleration signals.

[0028] It should be noted that vehicle acceleration signals include axle box acceleration signals, frame acceleration signals, and vehicle body acceleration signals.

[0029] In this embodiment of the invention, the sensor used for vehicle body acceleration has the function of measuring lateral, vertical, and longitudinal vibration acceleration, and the sensor used for axle box and frame acceleration has the function of measuring lateral and vertical vibration acceleration.

[0030] In this embodiment of the invention, an acceleration detection system is installed on a track inspection vehicle to utilize measured vehicle acceleration data.

[0031] Step 102: Standardize the vehicle acceleration signal to obtain the transformed value of the vehicle acceleration signal.

[0032] Step 103: Input the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network to separate the source signal from the transformed value of the vehicle acceleration signal.

[0033] Step 104: Determine the corresponding acceleration component signal based on the separated source signal.

[0034] In one possible implementation, this invention employs a multi-section vehicle dynamic detection system to collect real-time acceleration data of the vehicle body, frame, and axle boxes, aiding in the analysis of track smoothness. The system utilizes multi-channel distributed networked testing technology, using a computer to remotely control test equipment located at different locations to work synchronously. Data and synchronization information are transmitted via the network, offering advantages such as large data volume, geographically dispersed operation, high real-time performance and reliability, and long-distance collaborative operation. The system features online acquisition and processing of raw signals, storage of intermediate data and final results, online waveform display, network data transmission, output of over-limit reports, mileage correction, post-processing of stored data, and output of waveform data along with corresponding locations and speeds. It achieves data acquisition, raw data storage, data validity assessment, and waveform display.

[0035] The above scheme achieves vehicle acceleration signal separation by standardizing the normalized vehicle acceleration signal and then using a nonlinear mixing and demixing model based on a radial basis function neural network. Blind source separation is employed to improve the accuracy of vehicle acceleration signal separation. This enables auxiliary monitoring and timely maintenance of the railway's condition.

[0036] In step 102 of this embodiment of the invention, the vehicle acceleration signal is normalized to obtain the transformed value of the vehicle acceleration signal. The process flow is as follows: Figure 2 As shown, the details are as follows:

[0037] Step 201: Obtain the transmission relationship between vehicle acceleration signal and dynamic wheel load.

[0038] Step 202: Determine the dynamic wheel load data based on the transmission relationship between the vehicle acceleration signal and the dynamic wheel load.

[0039] Step 203: Take the logarithm of the dynamic wheel load data to obtain the transformed value of the vehicle acceleration signal.

[0040] The prerequisite for blind source separation is that at most one of the source signals is a Gaussian signal. When all the source signals (signals containing fault characteristic information) are Gaussian signals, their mixed signal is still a Gaussian signal, and they cannot be separated.

[0041] Fault characteristic signals contained in vehicle acceleration signals are generally impact signals, periodic signals, or their modulated signals, exhibiting a clear Gaussian distribution. Therefore, in order to perform blind source separation of the data, the source data needs to be transformed into signals that satisfy a non-Gaussian distribution before analysis.

[0042] In this embodiment of the invention, the specific formula for the transmission relationship between vehicle acceleration signal and dynamic wheel load is as follows:

[0043] P = k1P0 + k2Ma

[0044] Where P0 and M are constant values, a is the vehicle acceleration signal, P is the dynamic wheel load, and k1 and k2 are coefficients.

[0045] It follows a log-normal distribution, and its probability density function is:

[0046]

[0047] λ and ζ are respectively:

[0048] λ=lnP0-ζ 2 / 2

[0049] ζ 2 =ln[1+(σ p / P0) 2 ]

[0050] Where, σ p σ represents the standard deviation of wheel load. p / P0 is the wheel load deviation coefficient, and ζ uses the deviation coefficient σ. p / P0 indicates.

[0051] The vehicle acceleration signal, after the aforementioned transformation, is denoted as x. j (t) = lnP(t), and its inverse transform is:

[0052]

[0053] In this embodiment of the invention, after transforming the vehicle acceleration signal and taking its logarithm, the fault characteristic signal contained in the source signal no longer follows a Gaussian distribution, thus providing support for signal separation.

[0054] In step 103 of this embodiment of the invention, the transformed value of the vehicle acceleration signal is input into a nonlinear mixing and demixing model based on a radial basis function neural network to separate the source signal from the transformed value of the vehicle acceleration signal.

[0055] Assume there are Q source signals in the observed signal, denoted as s. i (t), i = 1, 2, 3, ..., Q. The acquired acceleration signal contains M sets of signals, each of which, after the aforementioned transformation, is represented by x. j (t), j = 1, 2, 3, ..., M, represents the signal. The amplitude of the i-th signal in the j-th group of signals is x. ij Then the received j-th group of signals is:

[0056]

[0057] Let s(t) = [s1(t), s2(t), ..., s Q (t)] T x(t) = [x1, x2, ..., x M ] T The transformed result of the received acceleration signal can then be expressed as:

[0058] x(t)=As(t)

[0059] Where, the non-singular mixture matrix A = [a1, a2, ..., a Q ], the i-th column vector a i Let represent the spatial vector of the i-th source signal, assuming that the spatial vectors of any two source signals are linearly independent. The blind source separation problem is to estimate the source signals and the mixing matrix when the mixing matrix and the source signals are unknown, but only the received signals and some assumptions about the independence between the sources are known.

[0060] Depending on the hybridization method, besides linear blind source separation, there are also convolutional blind source separation and nonlinear blind source separation. Compared to linear blind source separation, the nonlinear problem is more complex, but it better reflects the actual situation of vehicle dynamic response signals. Given the strong nonlinear characteristics of the track-vehicle system, the embodiments of this invention mainly consider the nonlinear blind source separation problem.

[0061] The separation of nonlinear blind source signals generally considers the post-nonlinear model, that is, the nonlinear aliasing still contains a certain amount of linear instantaneous aliasing:

[0062] x(t) = f(As(t))

[0063] Where f(·) is an unknown nonlinear function. The task of nonlinear blind source signal separation is to recover the source signal s(t) from the transformed value x(t) of the observed acceleration signal, such as... Figure 3 As shown.

[0064] Figure 3The left part represents a nonlinear mixing model of the signals, and the right part represents a nonlinear demixing neural network model. Assuming f(·) is invertible, its inverse function is expressed as f -1 (·). If there exists a function g(·) = f -1 (·), then there is an output signal.

[0065] y(t)=Bg(x(t))=Bf -1 (x(t))=Bf -1 (f(As(t)))=BAs(t)

[0066] If BA = I, then y(t) = s(t). B is the inverse matrix of A.

[0067] Clearly, without prior knowledge, the criterion of source signal independence alone cannot correctly separate the source signals. Therefore, it is also necessary to know the type of the nonlinear mixing function to achieve the desired result.

[0068] In this embodiment of the invention, the nonlinear mixing and demixing model based on radial basis function neural networks is as follows: Figure 4 As shown.

[0069] g(·,θ) is a nonlinear separation model, i.e., f -1 The parameter fitting function of (·) can be varied by the parameter θ to be estimated. If a suitable function can be found... Make g(·,θ) f -1 When (·) is a good approximation, it can be expressed as:

[0070] y(t)=Bg(x(t),θ)=Bf -1 (x(t))=Bf -1 (f(As(t)))=BAs(t)

[0071] The center of the RBF neural network is determined using the particle swarm optimization algorithm, and the variance is calculated. Using the mutual information of the output signal y as the objective function, the optimal parameter estimation is then determined. It can be expressed as

[0072]

[0073] Specifically, in this separation model, the mutual information of y (when the mutual information is minimized, the corresponding parameter is the separation parameter) can be further expressed as:

[0074]

[0075] The Jacobi determinant obtained by differentiating g(x,θ) with respect to x. Joint entropy H(y) t )for

[0076]

[0077] in, t = 1, 2, ..., n. E(·) represents the expectation. Since H(x) = -E[lgρ(x)], and ρ(x) is the joint probability density of x, which does not contain any parameters separated by the RBF neural network, it is ignored. The expression for the independence between the measured output signals is obtained:

[0078]

[0079] Then the gradient of I(y) with respect to the neural network parameters W is:

[0080]

[0081] The relationship between W and θ is θ=(W,ρ), where θ, W, and ρ all represent neural network parameters.

[0082] Therefore, the iterative formula for parameter W is:

[0083]

[0084] Where β is the iterative learning rate.

[0085] In one possible implementation, the termination condition for the above recursive iteration can be the overall change er of the variable at time t and time t+1, which determines the independence between the output signals, as the iteration termination condition.

[0086] In step 104 of this embodiment of the invention, the corresponding acceleration component signal is determined based on the separated source signal.

[0087] The separated source signal is inversely transformed to obtain the corresponding acceleration component signal.

[0088] The result s(t) calculated by the above scheme is obtained by taking the logarithm of the vehicle acceleration signal after transformation. Therefore, in order to recover the corresponding acceleration component signal, it is necessary to perform an inverse transformation on the data.

[0089] Recorded as as(t)=[as1(t), as2(t),…,as Q (t)] T

[0090]

[0091]

[0092] In this embodiment of the invention, the iteration ends when er is less than a preset threshold.

[0093] The above scheme recovers the vehicle acceleration signal components by performing an inverse transformation on the data.

[0094] In this embodiment of the invention, the continuous multi-wave signal separated from the vehicle body acceleration signal is as follows: Figure 5 As shown.

[0095] This invention also provides a device for separating acceleration signals of high-speed railway vehicles, as described in the following embodiments. This device... Figure 6 As shown, the device includes:

[0096] The transformation module 601 is used to acquire vehicle acceleration signals and standardize the vehicle acceleration signals to obtain transformed values ​​of the vehicle acceleration signals.

[0097] The separation module 602 is used to input the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network, and to separate the source signal from the transformed value of the vehicle acceleration signal; and to determine the corresponding acceleration component signal based on the separated source signal.

[0098] In this embodiment of the invention, the transformation module 601 is specifically used for:

[0099] Acquire the transmission relationship between vehicle acceleration signals and dynamic wheel loads;

[0100] Dynamic wheel load data is determined based on the transmission relationship between vehicle acceleration signals and dynamic wheel loads;

[0101] The logarithm of the dynamic wheel load data is taken to obtain the transformed value of the vehicle acceleration signal.

[0102] In this embodiment of the invention, the transformation module 601 is specifically used for:

[0103] The specific formula for the transmission relationship between vehicle acceleration signal and dynamic wheel load is as follows:

[0104] P = k1P0 + k2Ma

[0105] Where P0 and M are constant values, a is the vehicle acceleration signal, P is the dynamic wheel load, and k1 and k2 are coefficients.

[0106] In this embodiment of the invention, the separation module 602 is specifically used for:

[0107] The separated source signal is inversely transformed to obtain the corresponding acceleration component signal.

[0108] In this embodiment of the invention, the transformation module 601 is specifically used for:

[0109] The vehicle acceleration signals include axle box acceleration signals, frame acceleration signals, and vehicle body acceleration signals.

[0110] Since the principle behind this device is similar to that of the method for separating acceleration signals from high-speed railway vehicles, the implementation of this device can be found in the implementation of the method for separating acceleration signals from high-speed railway vehicles, and the repetitive parts will not be repeated.

[0111] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0112] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0113] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for separating acceleration signals of high-speed railway vehicles.

[0114] In this embodiment of the invention, vehicle acceleration signals are acquired; the vehicle acceleration signals are normalized to obtain transformed values ​​of the vehicle acceleration signals; the transformed values ​​of the vehicle acceleration signals are input into a nonlinear mixing and demixing model based on a radial basis function neural network to separate the source signals in the transformed values ​​of the vehicle acceleration signals; the corresponding acceleration component signals are determined based on the separated source signals. Compared with the prior art, the vehicle acceleration signals are separated by using a nonlinear mixing and demixing model based on a radial basis function neural network after normalizing the vehicle acceleration signals, and blind source separation is adopted, which improves the accuracy of vehicle acceleration signal separation.

[0115] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0117] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0118] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for high speed railway vehicle acceleration signal separation, characterized in that, The method comprises the following steps: collecting a vehicle acceleration signal; standard normalizing the vehicle acceleration signal to obtain a transformed value of the vehicle acceleration signal; inputting the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network to separate source signals in the transformed value of the vehicle acceleration signal; determining corresponding acceleration component signals according to the separated source signals; standard normalizing the vehicle acceleration signal to obtain a transformed value of the vehicle acceleration signal, comprising the following steps: obtaining a transfer relationship between the vehicle acceleration signal and a dynamic wheel load; determining dynamic wheel load data according to the transfer relationship between the vehicle acceleration signal and the dynamic wheel load; taking a logarithm of the dynamic wheel load data to obtain the transformed value of the vehicle acceleration signal; a specific formula of the transfer relationship between the vehicle acceleration signal and the dynamic wheel load is as follows: P=k1P0+k2Ma wherein P0 and M are constant values, a is the vehicle acceleration signal, P is the dynamic wheel load, and k1 and k2 are coefficients.

2. The method for high speed railway vehicle acceleration signal separation as claimed in claim 1, characterized in that, determining corresponding acceleration component signals according to the separated source signals, comprising the following step: performing inverse transformation on the separated source signals to obtain the corresponding acceleration component signals.

3. The method for high speed railway vehicle acceleration signal separation as claimed in claim 1, characterized in that, The vehicle acceleration signal comprises an axle box acceleration signal, a bogie acceleration signal and a car body acceleration signal.

4. A device for separating acceleration signals of a high-speed railway vehicle, characterized in that, The method comprises the following steps: a transformation module, configured to collect a vehicle acceleration signal; and standard normalize the vehicle acceleration signal to obtain a transformed value of the vehicle acceleration signal; a separation module, configured to input the transformed value of the vehicle acceleration signal into a nonlinear mixing and demixing model based on a radial basis function neural network to separate source signals in the transformed value of the vehicle acceleration signal; and determine corresponding acceleration component signals according to the separated source signals. The transformation module is specifically configured to: obtain a transfer relationship between the vehicle acceleration signal and a dynamic wheel load; determine dynamic wheel load data according to the transfer relationship between the vehicle acceleration signal and the dynamic wheel load; take a logarithm of the dynamic wheel load data to obtain the transformed value of the vehicle acceleration signal. The transformation module is specifically configured to: a specific formula of the transfer relationship between the vehicle acceleration signal and the dynamic wheel load is as follows: P=k1P0+k2Ma wherein P0 and M are constant values, a is the vehicle acceleration signal, P is the dynamic wheel load, and k1 and k2 are coefficients.

5. The device for separating acceleration signals of a high-speed railway vehicle according to claim 4, characterized in that, The separation module is specifically configured to: perform inverse transformation on the separated source signals to obtain the corresponding acceleration component signals.

6. The device for separating acceleration signals of a high-speed railway vehicle according to claim 4, characterized in that, The transformation module is specifically configured to: The vehicle acceleration signal comprises an axle box acceleration signal, a bogie acceleration signal and a car body acceleration signal.

7. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1 to 3.

9. A computer program product, characterised in that, The computer program product comprises a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1 to 3.

Images