Method and device for fitting track geometric irregularities to vehicle vertical accelerations

Abstract

The application discloses a track geometric irregularity and vehicle vertical acceleration fitting method and device, and the method comprises the following steps: acquiring track irregularity data of a vehicle, and acquiring a wavelength corresponding to a periodic change feature from the track irregularity data; adopting a variational mode decomposition method VMD to decompose the track irregularity data, and extracting a specified modal component from a plurality of groups of modal components IMFs obtained through the decomposition; and inputting the specified modal component into a vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data of the specified modal component. The application can improve the prediction accuracy and prediction speed of the vehicle vertical dynamic response data.

Description

Technical Field

[0001] This invention relates to the field of railway engineering technology, and in particular to a method and apparatus for fitting track geometric irregularities with vehicle vertical acceleration. Background Technology

[0002] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section.

[0003] Track irregularities are a major cause of train vibration and increased wheel-rail forces, especially vertical track irregularities, which directly affect the stability and safety of train operation. In terms of safe operation, poorly maintained vertical track irregularities can lead to excessive vibration and impact forces during train operation. This not only accelerates the wear and tear on track and vehicle components but may also cause derailment or other safety accidents in extreme cases. Therefore, accurately predicting the vertical dynamic response of the vehicle is crucial for the timely detection of potential track safety risks. Meanwhile, the comfort of high-speed trains largely depends on the smoothness of their operation. Vertical track irregularities cause train vibration, affecting passenger comfort. From a data analysis perspective, constructing a fitting relationship between the track and the vehicle using deep learning models provides guidance for assessing track service conditions and designing vehicle suspension systems.

[0004] Traditional deep learning models, limited by their learning capacity and model complexity, are prone to overfitting when faced with exceptionally complex and massive datasets, impacting their generalization and prediction performance. In complex task prediction, early single deep neural networks, constrained by their network depth and model structure, were unable to learn complex latent features effectively and suffered from vanishing or exploding gradients, resulting in poor fitting performance. Summary of the Invention

[0005] This invention provides a method for fitting track geometric irregularities to vehicle vertical acceleration, thereby improving the prediction accuracy and speed of vehicle vertical dynamic response data and optimizing the fitting effect between track irregularity data and vehicle vertical dynamic response. The method includes:

[0006] Acquire track irregularity data of the vehicle, and extract the wavelengths corresponding to the periodic variation characteristics from the track irregularity data;

[0007] The track irregularity data is decomposed using VMD (Variational Mode Decomposition), and the specified mode components are extracted from the multiple sets of mode components (IMFs) obtained by decomposition.

[0008] The specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training the Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model introduces an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

[0009] This invention also provides a device for fitting track geometry irregularities and vehicle vertical acceleration, used to improve the prediction accuracy and prediction speed of vehicle vertical dynamic response data, and optimize the fitting effect between track irregularity data and vehicle vertical dynamic response. The device includes:

[0010] The track irregularity data acquisition module is used to acquire track irregularity data of the vehicle and extract the wavelengths corresponding to periodic changes in the track irregularity data.

[0011] The track irregularity data decomposition module is used to decompose track irregularity data using VMD and extract the specified modal components from the multiple sets of modal components (IMFs) obtained from the decomposition.

[0012] The vertical dynamic response data prediction module is used to input specified modal components into the vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training an Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model incorporates an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

[0013] 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 fitting track geometry irregularities with vehicle vertical acceleration.

[0014] 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 fitting track geometric irregularities to vehicle vertical acceleration.

[0015] 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 fitting track geometry irregularities to vehicle vertical acceleration.

[0016] In this embodiment of the invention, track irregularity data of a vehicle is acquired, and wavelengths corresponding to periodic variations in the track irregularity data are obtained. Variational Mode Decomposition (VMD) is used to decompose the track irregularity data, and specified modal components are extracted from the multiple sets of Intrinsic Modal Components (IMFs) obtained from the decomposition. The specified modal components are input into a vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training an Informer model based on historical track irregularity data and corresponding historical actual vehicle vertical dynamic response data. The Informer model incorporates an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variations in the track irregularity data to a pre-existing position encoder. The periodic variations include wheel circumference, track slab length, and simply supported beam length. In the above process, this embodiment of the invention uses VMD to decompose track irregularity data, which can effectively decompose complex data, extract key features, reduce noise interference, and provide a foundation for the prediction accuracy of vehicle vertical dynamic response data. By incorporating the wavelengths corresponding to the periodic variations in track irregularity data into a pre-existing position encoder, an optimized Informer model is obtained. This enhances the positional information of the periodic variations, improves the fitting effect of the Informer model, and enables efficient fitting of track irregularity data with vehicle vertical dynamic response, thereby improving the prediction accuracy and speed of vehicle vertical dynamic response data. Attached Figure Description

[0017] 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:

[0018] Figure 1 This is a flowchart illustrating a method for fitting track geometry irregularities to vehicle vertical acceleration in an embodiment of the present invention.

[0019] Figure 2 This is a flowchart illustrating the prediction of vehicle vertical dynamic response data in an embodiment of the present invention;

[0020] Figure 3This is a flowchart of the sparse self-attention mechanism in an embodiment of the present invention;

[0021] Figure 4 This is a flowchart of the improved Informer encoder distillation mechanism in an embodiment of the present invention;

[0022] Figure 5 This is a flowchart of the VMD-improved Informer model in an embodiment of the present invention;

[0023] Figure 6 This is a diagram showing the fitting results of track geometry irregularities and vehicle vertical acceleration based on VMD-improved Informer in an embodiment of the present invention.

[0024] Figure 7 This is a schematic diagram of a device for fitting track geometry irregularities and vehicle vertical acceleration in an embodiment of the present invention;

[0025] Figure 8 This is a schematic diagram of a specific track geometry irregularity and vehicle vertical acceleration fitting device in an embodiment of the present invention;

[0026] Figure 9 This is a schematic diagram of a specific track geometry irregularity and vehicle vertical acceleration fitting device in an embodiment of the present invention. Detailed Implementation

[0027] 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.

[0028] Figure 1 This is a flowchart of a method for fitting track geometry irregularities and vehicle vertical acceleration according to an embodiment of the present invention. The method includes:

[0029] Step 101: Obtain the track irregularity data of the vehicle, and extract the wavelength corresponding to the periodic variation characteristics from the track irregularity data;

[0030] Step 102: The variational mode decomposition method (VMD) is used to decompose the track irregularity data, and the specified mode components are extracted from the multiple sets of mode components (IMFs) obtained by decomposition.

[0031] Step 103: Input the specified modal components into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components; the vehicle vertical dynamic response prediction model is obtained by training the Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data; the Informer model introduces an optimized position encoder; the optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder; the periodic variation features include wheel circumference, track slab length, and simply supported beam length.

[0032] Each step is explained in detail below.

[0033] In step 101, track irregularity data of the vehicle is acquired, and the wavelength corresponding to the periodic variation characteristics is obtained from the track irregularity data.

[0034] In one embodiment, after obtaining the vehicle's track irregularity data, the method further includes: performing data removal processing on the vehicle's track irregularity data to obtain processed track irregularity data.

[0035] The wavelengths corresponding to periodic variations in orbital irregularities are extracted from the data, including:

[0036] The wavelengths corresponding to the periodic variation characteristics are obtained from the processed track irregularity data;

[0037] VMD is used to decompose the track irregularity data, including:

[0038] VMD is used to decompose the processed track irregularity data.

[0039] In step 102, the variational mode decomposition method (VMD) is used to decompose the track irregularity data, and the specified mode components are extracted from the multiple sets of mode components (IMFs) obtained by decomposition.

[0040] In a specific embodiment, variational mode decomposition (VMD) is used to decompose the track geometry irregularity data into multiple sets of intrinsically modal components (IMFs). The vertical channels of the track geometry irregularity include left high-low channels, right high-low channels, horizontal channels, and triangular pit channels. VMD analysis was used to demonstrate the track vertical irregularity components and its ability to effectively separate noise components and useful signals in the time series. Specific modal components are extracted from the multiple sets of IMFs, each with a specific frequency bandwidth. These components can serve as powerful features for further analysis and prediction of the time series, helping to improve the model's ability to capture the inherent dynamics of the data. Especially for nonlinear and non-stationary signals related to track irregularities and vehicle vertical dynamic response, it can process complex signals into sequence data that is easier for deep learning models to learn and fit, benefiting subsequent prediction tasks of deep learning models. Different modes can more clearly and accurately characterize the changes in different wavelengths. The specified modal components are those most strongly correlated with the vehicle's vertical dynamic response. VMD decomposition can effectively decompose complex signals, extract key and relevant features, reduce noise interference, and provide a foundation for more accurate predictions.

[0041] In a specific embodiment, it is assumed that the track irregularity data f is decomposed into k modal components, ensuring that the decomposed sequence is a modal component with a center frequency and a finite bandwidth, while the sum of the estimated bandwidths of each mode is minimized, and the constraint condition is that the sum of all modes is equal to the original signal.

[0042] To solve the aforementioned constrained optimization problem, the constrained variational problem is transformed into an unconstrained variational problem. Leveraging the advantages of quadratic penalty terms and the Lagrange multiplier method, an augmented Lagrangian function is introduced. Since the performance of VMD is jointly influenced by the number of modes K and the penalty coefficient α, an excessively large value of K can lead to mode fragmentation and over-decomposition; an excessively small value of K can result in multiple modes being mixed together or missing effective modes, leading to under-decomposition. Furthermore, as the value of K increases, the computation time also increases significantly. Considering both program execution efficiency and decomposition effect, after experimental comparison, the optimal decomposition results were found to be: for the left and right vertical channels, horizontal channels, and triangular pit channels, a mode number K=8 and a penalty coefficient α=2000 were selected; for the vehicle vertical acceleration channel, a mode number K=6 and a penalty coefficient α=3000 were selected.

[0043] In step 103, the specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training the Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model introduces an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

[0044] In one embodiment, the optimized position encoder includes a wheel circumference periodic position encoder, a track slab periodic position encoder, and a simply supported beam periodic position encoder.

[0045] like Figure 2 As shown, in one embodiment, the specified modal components are input into the vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components, including:

[0046] Step 201: Determine the wheel position information based on the wheel circumference periodic position encoder; determine the track slab position information based on the track slab periodic position encoder; determine the simply supported beam position information based on the simply supported beam periodic position encoder.

[0047] Step 202: Input the wheel position information, track slab position information, simply supported beam position information, and specified modal components into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components.

[0048] In one embodiment, the wheel position information is determined according to the following formula:

[0049] R-SE (pos，r) =Conv1D(pos|r,d) model );

[0050] Among them, R-SE {pos，r) This represents the wheel position information, r represents the wheel circumference, ConvlD() represents one-dimensional convolution, and d model Represents dimensional space;

[0051] The track slab position information is determined using the following formula:

[0052]

[0053] in, Indicates track slab position information, L a The length of the track slab is represented by d, Conv1D() represents one-dimensional convolution, and d represents the length of the track slab.model Represents dimensional space;

[0054] The location information of a simply supported beam is determined using the following formula:

[0055]

[0056] in, Indicates track slab position information, L b Represents the length of a simply supported beam, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space.

[0057] In one embodiment, it further includes:

[0058] The performance of the vehicle response prediction model is evaluated based on the mean absolute error, root mean square error, and mean absolute percentage error between the predicted vertical dynamic response data and the actual vertical dynamic response data of the vehicle.

[0059] In a specific embodiment, the core function of the sliding window in predicting the vehicle's vertical dynamic response from track geometry irregularities is to maintain the continuity of the predicted vertical dynamic response. Given that the Informer model's encoder and decoder need to process input sequences of varying lengths and positions, the sliding window needs to be designed to overlap to ensure the continuity of the vertical irregularity sequence. This design aims to improve prediction accuracy while adapting to the requirements of the model structure, making the prediction more coherent and effectively supporting the predictive analysis of the vehicle's dynamic response.

[0060] Using a sliding window to segment the input data for the Informer model's encoder and decoder involves dividing time series data into overlapping segments. To ensure that the vertically uneven and continuous output of the model is maintained, the length of the sliding window needs to be set.

[0061] For the encoder input, the data is divided into fixed-size windows, maintaining overlap between consecutive windows. Each window becomes an input sequence for the encoder, and the length of this input sequence is set. The positional order of the vertical dynamic response data is preserved. For the decoder, the decoder input is the encoder post-sequence up to the end of the entire window length. The decoder input is the sequence from the next point of the decoder input to the point after the cutoff position, but represents a window shifted one position backward from the decoder input.

[0062] In a specific embodiment, the differences between the predicted and actual vertical dynamic response results are measured using MAE (Mean Absolute Error), MSE (Mean Squared Error), and MAPE (Mean Absolute Percentage Error). Their functions and corresponding formulas are as follows:

[0063] MAE (Mean Absolute Error): Measures the average absolute difference between predicted and actual values, and is insensitive to outliers. The formula is expressed as:

[0064]

[0065] MSE (Mean Squared Error): Measures the average squared difference between predicted and actual values, and is more sensitive to large errors. Formula:

[0066]

[0067] MAPE (Mean Absolute Percentage Error): Measures the percentage error of the predicted value relative to the actual value. It provides a more intuitive assessment of percentage error.

[0068]

[0069] Where N is the number of samples, Y i These are actual observed values. That is the corresponding predicted value.

[0070] MAE is not sensitive to outliers and reflects the prediction of outliers; MSE is more sensitive to large errors and reflects the prediction accuracy; while MAPE provides a percentage error measure, making the relative magnitude of the error clearer.

[0071] In a specific embodiment, when selecting specified modal components to input into the vehicle vertical dynamic response prediction model, in channels with small distribution differences, modal components covering all wavelengths of the vehicle's vertical acceleration are selected, while modal components not covered by the vehicle's vertical dynamic response are discarded; in channels with large distribution differences, the modal component with the highest energy proportion is selected. By selecting channels with similar distribution characteristics to the vehicle's vertical acceleration, it helps to retain key information related to the target acceleration, ensuring better model performance on key features. At the same time, selecting modal components with high energy proportions means paying more attention to the frequency components that dominate the overall signal, which helps to reduce noise interference to the model and improve the significance of the signal.

[0072] The above analysis shows that the left and right high / low channels and the horizontal channel exhibit high consistency in time-frequency characteristics and energy distribution with the IMF modal components of the vehicle's vertical acceleration after VMD decomposition, while the triangular pit channel shows significant differences from the vehicle's vertical dynamic response. Furthermore, the wavelength range covered by IMF1-IMF7 in the left and right high / low channels and the horizontal channel includes all signal wavelength components of the vehicle's vertical dynamic response, while IMF8, with its low proportion in the vehicle's vertical acceleration components, is considered noise data and not selected as input for fitting. Additionally, the IMF1-IMF5 in the triangular pit channel covers over 98% of the energy distribution. Although it does not include the entire vehicle's vertical acceleration wavelength range, due to its significant distribution differences, IMF6-IMF8 are considered noise data and not selected as input for the track geometry vertical direction.

[0073] In a specific embodiment, optimizing the Informer model includes the following steps:

[0074] 1. For example Figure 3 As shown, based on the sparse self-attention mechanism, the Informer model expression is determined, where Q is the query matrix, containing vectors of queries (i.e., the positions to be evaluated), and K is the key matrix, containing vectors of keys (i.e., the positions compared with the query). T The transpose of the key matrix. k :Dimension of the key vector. In this formula, d k Used to scale the attention score to prevent the calculated inner product from becoming too large and causing gradient vanishing. V: value matrix, which contains the value vector, which is multiplied by the score in the attention mechanism to produce the output.

[0075] 2. For example Figure 4 As shown, a distillation mechanism is used for the encoder in the Informer model.

[0076] 3. An optimized position encoder has been introduced into the Informer model. Figure 5 To improve the Informer model based on VMD, Figure 6 The result is a fitting of track geometry irregularities and vehicle vertical acceleration.

[0077] The Informer model needs to add contextual information, i.e., the location information of relevant detection data, to the input features through positional encoding. The optimization steps can be divided into several parts:

[0078] The first part consists of the IMF components extracted after VMD decomposition and correlation selection based on left-right height, horizontality, and triangular pits, in d. model Spatial projection, d modelThe dimensions are defined when expanding the dimensions of the original data.

[0079] The second part is the position encoder mode based on the traditional Transformer model. Each position is encoded using a linear variation of a sine and cosine function to obtain the relevant position of the detection data at a certain moment in the entire input sequence. The specific calculation method is as follows:

[0080]

[0081]

[0082] Where pos and i represent the position and dimension of the current data in the input sequence, respectively, and PE is the sine and cosine position encoder of the Transformer.

[0083] The third part focuses on the dynamic detection position encoding information of the track. Addressing the periodic variations in track irregularities, the study investigates the periodic variations of wheels, track slabs, and simply supported beams at different wavelengths, including a wheelset circumference period of 2.67m, a track slab length period of 6.45m, and a simply supported beam length period of 32m. Furthermore, Section 2.3 demonstrates the existence of these periodic wavelengths in the track geometry sequence. At the mechanistic and data analysis levels, the use of periodic position encoders relative to wheel circumference, track slab length, and simply supported beam length has been proven and supported.

[0084] Wheel circumference period position encoder:

[0085] R-SE (pos，r) =Conv1D(pos|r,d) model )

[0086] Track slab periodic position encoder:

[0087]

[0088] Simply supported beam periodic position encoder:

[0089]

[0090] Among them, R-SE (pos，r) , These represent encoders for wheel position, track slab position, and simply supported beam position, respectively. r represents the wheel circumference, and L... a L represents the length of the track slab. b This represents the length of the simply supported beam. It is then mapped to d through a first convolution. model Dimensional space.

[0091] In summary, the input vector of the Informer model can be obtained as follows:

[0092]

[0093] Where i∈{1,...,L} x} represents the sequence length, N represents the detection data for different vertical irregularity channels, and K represents the number of VMD decomposition IMFs per channel. For the spatial projection of IMF decomposition, PE is the sine and cosine position encoder of Transformer, R-SE is the optimized position encoder, where p is the number of different track position encoders, including wheelset circumference period, track slab length period, and simply supported beam length period position encoders, Lx is the sequence length of track geometric irregularities, and Lp is the periodic wavelength of different position encoders.

[0094] This invention also provides a device for fitting track geometry irregularities to vehicle vertical acceleration, as described in the following embodiments. Since the principle by which this device solves the problem is similar to the method for fitting track geometry irregularities to vehicle vertical acceleration, the implementation of this device can refer to the implementation of the method for fitting track geometry irregularities to vehicle vertical acceleration; repeated details will not be elaborated further. Figure 7 As shown, the device includes:

[0095] The track irregularity data acquisition module 701 is used to acquire track irregularity data of the vehicle and extract the wavelengths corresponding to periodic variation characteristics from the track irregularity data.

[0096] The track irregularity data decomposition module 702 is used to decompose track irregularity data using the variational mode decomposition method (VMD) and extract the specified mode components from the multiple sets of mode components (IMFs) obtained from the decomposition.

[0097] The vertical dynamic response data prediction module 703 is used to input specified modal components into the vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training an Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model incorporates an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

[0098] like Figure 8 As shown, in one embodiment, it also includes a data preprocessing module 801, specifically used to: perform data removal processing on the vehicle track irregularity data to obtain processed track irregularity data;

[0099] The wavelengths corresponding to periodic variations in orbital irregularities are extracted from the data, including:

[0100] The wavelengths corresponding to the periodic variation characteristics are obtained from the processed track irregularity data;

[0101] VMD is used to decompose the track irregularity data, including:

[0102] VMD is used to decompose the processed track irregularity data.

[0103] In one embodiment, the optimized position encoder includes a wheel circumference periodic position encoder, a track slab periodic position encoder, and a simply supported beam periodic position encoder.

[0104] In one embodiment, the vertical dynamic response data prediction module 703 is specifically used for:

[0105] The wheel position information is determined based on the wheel circumference periodic position encoder; the track slab position information is determined based on the track slab periodic position encoder; and the simply supported beam position information is determined based on the simply supported beam periodic position encoder.

[0106] The wheel position information, track slab position information, simply supported beam position information, and specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components.

[0107] In one embodiment, the wheel position information is determined according to the following formula:

[0108] R-SE (pos,r) =Conv1D(pos|r,d) model );

[0109] Among them, R-SE (pos,r) This represents the wheel position information, r represents the wheel circumference, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space;

[0110] The track slab position information is determined using the following formula:

[0111]

[0112] in, Indicates track slab position information, L a The length of the track slab is represented by d, Conv1D() represents one-dimensional convolution, and d represents the length of the track slab. model Represents dimensional space;

[0113] The location information of a simply supported beam is determined using the following formula:

[0114]

[0115] in, Indicates track slab position information, L b Represents the length of a simply supported beam, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space.

[0116] like Figure 9 As shown, in one embodiment, it further includes a vehicle response prediction model performance evaluation module 901, specifically used for:

[0117] The performance of the vehicle response prediction model is evaluated based on the mean absolute error, root mean square error, and mean absolute percentage error between the predicted vertical dynamic response data and the actual vertical dynamic response data of the vehicle.

[0118] 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-mentioned method for fitting track geometric irregularities with vehicle vertical acceleration.

[0119] 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 fitting track geometric irregularities to vehicle vertical acceleration.

[0120] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-mentioned method for fitting track geometric irregularities to vehicle vertical acceleration.

[0121] In this embodiment of the invention, track irregularity data of a vehicle is acquired, and wavelengths corresponding to periodic variations in the track irregularity data are obtained. Variational Mode Decomposition (VMD) is used to decompose the track irregularity data, and specified modal components are extracted from the multiple sets of Intrinsic Modal Components (IMFs) obtained from the decomposition. The specified modal components are input into a vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training an Informer model based on historical track irregularity data and corresponding historical actual vehicle vertical dynamic response data. The Informer model incorporates an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variations in the track irregularity data to a pre-existing position encoder. The periodic variations include wheel circumference, track slab length, and simply supported beam length. In the above process, this embodiment of the invention uses VMD to decompose track irregularity data, which can effectively decompose complex data, extract key features, reduce noise interference, and provide a foundation for the prediction accuracy of vehicle vertical dynamic response data. By incorporating the wavelengths corresponding to the periodic variations in track irregularity data into a pre-existing position encoder, an optimized Informer model is obtained. This enhances the positional information of the periodic variations, improves the fitting effect of the Informer model, and enables efficient fitting of track irregularity data with vehicle vertical dynamic response, thereby improving the prediction accuracy and speed of vehicle vertical dynamic response data.

[0122] 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.

[0123] 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 1A device that provides the functions specified in one or more boxes.

[0124] 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.

[0125] 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.

[0126] 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 fitting track geometric irregularities to vehicle vertical acceleration, characterized in that, include: Acquire track irregularity data of the vehicle, and extract the wavelengths corresponding to the periodic variation characteristics from the track irregularity data; Variational Mode Decomposition (VMD) is used to decompose track irregularity data, and specified mode components are extracted from multiple sets of mode components (IMFs) obtained by decomposition. The specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training the Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model introduces an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

2. The method as described in claim 1, characterized in that, After obtaining the vehicle's track irregularity data, the process also includes: performing data removal processing on the vehicle's track irregularity data to obtain processed track irregularity data. The wavelengths corresponding to periodic variations in orbital irregularities are extracted from the data, including: The wavelengths corresponding to the periodic variation characteristics are obtained from the processed track irregularity data; VMD is used to decompose the track irregularity data, including: VMD is used to decompose the processed track irregularity data.

3. The method as described in claim 1, characterized in that, The optimized position encoders include wheel circumference periodic position encoders, track slab periodic position encoders, and simply supported beam periodic position encoders.

4. The method as described in claim 3, characterized in that, The specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data for the specified modal components, including: The wheel position information is determined based on the wheel circumference periodic position encoder; the track slab position information is determined based on the track slab periodic position encoder; and the simply supported beam position information is determined based on the simply supported beam periodic position encoder. The wheel position information, track slab position information, simply supported beam position information, and specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components.

5. The method as described in claim 4, characterized in that, The wheel position information is determined using the following formula: R-SE (pos,r) =Conv1D(pos|r,d model ); Among them, R-SE (pos,r) This represents the wheel position information, r represents the wheel circumference, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space; The track slab position information is determined using the following formula: in, Indicates track slab position information, L a The length of the track slab is represented by d, Conv1D() represents one-dimensional convolution, and d represents the length of the track slab. model Represents dimensional space; The location information of a simply supported beam is determined using the following formula: in, Indicates track slab position information, L b Represents the length of a simply supported beam, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space.

6. The method as described in claim 1, characterized in that, Also includes: The performance of the vehicle response prediction model is evaluated based on the mean absolute error, root mean square error, and mean absolute percentage error between the predicted vertical dynamic response data and the actual vertical dynamic response data of the vehicle.

7. A device for fitting track geometry irregularities to vehicle vertical acceleration, characterized in that, include: The track irregularity data acquisition module is used to acquire track irregularity data of the vehicle and extract the wavelengths corresponding to the periodic variation characteristics from the track irregularity data. The track irregularity data decomposition module is used to decompose track irregularity data using VMD and extract the specified modal components from the multiple sets of modal components (IMFs) obtained from the decomposition. The vertical dynamic response data prediction module is used to input specified modal components into the vehicle vertical dynamic response prediction model to obtain vehicle vertical dynamic response prediction data for the specified modal components. The vehicle vertical dynamic response prediction model is obtained by training an Informer model based on historical track irregularity data and corresponding actual historical vehicle vertical dynamic response data. The Informer model incorporates an optimized position encoder. The optimized position encoder is obtained by adding the wavelengths corresponding to the periodic variation features in the track irregularity data to a pre-existing position encoder. The periodic variation features include wheel circumference, track slab length, and simply supported beam length.

8. The apparatus as claimed in claim 7, characterized in that, It also includes a data preprocessing module, which is specifically used to: perform data removal processing on the track irregularity data of the vehicle to obtain processed track irregularity data; The wavelengths corresponding to periodic variations in orbital irregularities are extracted from the data, including: The wavelengths corresponding to the periodic variation characteristics are obtained from the processed track irregularity data; VMD is used to decompose the track irregularity data, including: VMD is used to decompose the processed track irregularity data.

9. The apparatus as claimed in claim 7, characterized in that, The optimized position encoders include wheel circumference periodic position encoders, track slab periodic position encoders, and simply supported beam periodic position encoders.

10. The apparatus as claimed in claim 9, characterized in that, The vertical dynamic response data prediction module is specifically used for: The wheel position information is determined based on the wheel circumference periodic position encoder; the track slab position information is determined based on the track slab periodic position encoder; and the simply supported beam position information is determined based on the simply supported beam periodic position encoder. The wheel position information, track slab position information, simply supported beam position information, and specified modal components are input into the vehicle vertical dynamic response prediction model to obtain the vehicle vertical dynamic response prediction data of the specified modal components.

11. The apparatus as claimed in claim 10, characterized in that, The wheel position information is determined using the following formula: R-SE (pos,r) =Conv1D(pos|r,d model ); Among them, R-SE (pos,r) This represents the wheel position information, r represents the wheel circumference, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space; The track slab position information is determined using the following formula: in, Indicates track slab position information, L a The length of the track slab is represented by d, Conv1D() represents one-dimensional convolution, and d represents the length of the track slab. model Represents dimensional space; The location information of a simply supported beam is determined using the following formula: in, Indicates track slab position information, L b Represents the length of a simply supported beam, Conv1D() represents one-dimensional convolution, and d model Represents dimensional space.

12. The apparatus as claimed in claim 7, characterized in that, It also includes a vehicle response prediction model performance evaluation module, specifically used for: The performance of the vehicle response prediction model is evaluated based on the mean absolute error, root mean square error, and mean absolute percentage error between the predicted vertical dynamic response data and the actual vertical dynamic response data of the vehicle.

13. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

15. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images