A track smoothness influencing factor correlation analysis method

Abstract

The present application relates to the field of railway engineering, in particular to a track smoothness influence factor correlation analysis method, comprising: taking track gauge deviation, high-low irregularity, horizontal irregularity, track irregularity and curvature change rate corresponding to track smoothness as parameter components, using Gaussian mixture model to calculate the reference value and allowable threshold value between the parameter components, and determining the weight of each parameter component; identifying the deterioration characteristics of each parameter component under time series; using linear weighting form, taking the time of comprehensive deterioration degree update as the key node of each parameter component; verifying the correlation probability of each key node and maintenance record, setting the early warning parameter set, setting the change trend of each key node according to the number of continuous early warning in the early warning parameter set; based on the change trend of each key node, interactive analysis is carried out, the early warning range limit of each parameter component under parameter accumulation is determined; the accuracy and efficiency of track smoothness analysis are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of railway engineering, and in particular to a method for analyzing the correlation of factors affecting track smoothness. Background Art

[0002] Track irregularities refer to the deviation between the actual and designed geometry of a line during operation. Therefore, regular inspections and track maintenance are essential to ensure train safety and passenger comfort.

[0003] For example, Chinese patent publication number CN118862263A discloses a method for fine-tuning high-speed railway track irregularity intervals that integrates dynamic evaluation, including: 1) establishing a potential track geometric feature extraction architecture based on a deep Bayesian network; 2) establishing a mechanical module expression of an agent model based on the vehicle-track coupling dynamic equation; 3) establishing a vehicle-track coupling dynamic agent model based on a Fourier neural network; 4) establishing an active control function for vehicle response loss based on a track irregularity-vehicle response parameter intensive scale constraint algorithm; 5) designing a loss function that embeds adjustment cost and track smoothness quality optimization objectives; 6) embedding multiple track parameter constraint functions such as chord measurement, adjustment boundary, and unit adjustment amount; 7) designing and selecting track fine-tuning schemes under specific constraints; and 8) outputting track irregularity interval fine-tuning schemes based on Monte Carlo.

[0004] For example, Chinese patent publication number CN113919236A discloses a model generation method for predicting track irregularity and a method for predicting track irregularity, comprising: obtaining displacement, acceleration, and angular velocity during vehicle operation; calculating the displacement, acceleration, and angular velocity to obtain a sequence of actual track irregularity values, wherein the actual irregularity values ​​corresponding to different time points during vehicle operation are different; inputting the vertical acceleration in the acceleration into an initial neural network model for training, and adjusting the initial neural network model through a loss function loop; reducing the gap between the irregularity prediction value obtained after model training and the corresponding actual irregularity value, ensuring that the absolute value of the irregularity prediction value is less than or equal to the absolute value of the corresponding actual irregularity value, and ensuring that the difference between the peak and valley values ​​of the irregularity prediction value sequence and the corresponding actual irregularity value is less than a set difference; and thereby obtaining a target neural network model.

[0005] In the existing technology, track irregularities are represented by continuous curves to determine the adjustment method of the track irregularity interval in the vehicle acceleration scenario, and the measurement and processing method of track irregularities is constructed using displacement, acceleration, and angular velocity. However, the existing technology ignores the degradation and warning conditions represented by gauge deviation, height irregularity, horizontal irregularity, and track direction irregularity that affect track geometry measurement. As a result, when analyzing the time curve, it is impossible to adjust the warning interval set for each track according to the different measurement values ​​and warning conditions at each position on the track, resulting in false alarms in the acquired data and the lack of multi-parameter smoothness description. Summary of the Invention

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for correlation analysis of factors affecting track smoothness, including: S1, using gauge deviation, height irregularity, horizontal irregularity, track direction irregularity and curvature change rate corresponding to track smoothness as parameter components, using a Gaussian mixture model to calculate the baseline value and allowable threshold between the parameter components, and determine the weight of each parameter component.

[0007] S2, based on the values ​​of the parameter components in multiple time periods, identifies the degradation degree and distance value of each parameter component in the time series when degradation occurs, and quantifies the degradation characteristics under environmental damage.

[0008] S3, based on the degradation characteristics of each parameter component, uses the form of linear weighting to explain the comprehensive degradation degree of each parameter when degradation occurs, and uses the time of comprehensive degradation degree update as the key node of each parameter component.

[0009] S4 verifies the association probability between each key node and the maintenance record, and sets the warning parameter set according to the track section of each key node under the association probability and the length of the track section as the boundary condition. According to the number of consecutive warnings in the warning parameter set, the change trend of each key node is set.

[0010] S5, based on the change trend of each key node, interactive analysis is performed to determine the warning range limit of each parameter component under parameter accumulation.

[0011] The beneficial effects of the present invention are as follows: 1. The present invention uses five parameters such as gauge deviation and height irregularity as components, combines time and environmental information to perform multi-parameter association, and classifies the parameter components into statuses. The classified reference values, allowable thresholds and parameter weights are used to illustrate the reference status of the current parameter components, thereby improving the accuracy of the parameter reference values ​​and allowable thresholds, and providing a data basis for subsequent degradation analysis and evaluation.

[0012] 2. The present invention calculates the degradation degree and Mahalanobis distance value of the parameter components, and divides the critical points and extreme points close to the degradation state according to their values. Then, the degradation characteristics are output according to the degradation scenario. After emphasizing the characteristics of the distribution of degradation characteristics in different scenarios, the comprehensive degradation degree is determined by linear weighting. According to the comprehensive degradation degree, the key nodes that currently need maintenance and early warning processing are determined to avoid excessive maintenance and missed maintenance under track identification.

[0013] 3. The present invention verifies the association probability between key nodes and maintenance records, sets the warning parameter set according to the length of the track section; sets the key node change trend according to the number of consecutive warnings to illustrate the situation where warnings appear at each position of the key node in the continuous analysis scenario; then sets the warning range limit based on the contribution of the key node to the warning parameter set to improve the reliability of the key node under warning processing and enhance the adaptability of the warning range. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below with reference to the accompanying drawings and examples.

[0015] Figure 1 The present invention is a flowchart of a method for analyzing the correlation between factors affecting track smoothness.

[0016] Figure 2 The present invention is a flow chart of step S1 of a method for analyzing the correlation between factors affecting track smoothness.

[0017] Figure 3 The present invention is a flow chart of step S2 of a method for analyzing the correlation between factors affecting track smoothness.

[0018] Figure 4 The present invention is a flow chart of step S3 of a method for analyzing the correlation between factors affecting track smoothness.

[0019] Figure 5 The present invention is a flow chart of step S4 of a method for analyzing the correlation between factors affecting track smoothness.

[0020] Figure 6 The present invention is a flow chart of step S5 of the method for analyzing the correlation between factors affecting track smoothness. DETAILED DESCRIPTION

[0021] The following embodiments of the present invention are described in detail. The embodiments described below are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, the techniques or conditions described in the literature in the art or in the product specifications shall be followed.

[0022] See Figure 1, a method for correlation analysis of factors affecting track smoothness, comprising: S1, taking gauge deviation, height irregularity, horizontal irregularity, track direction irregularity and curvature change rate corresponding to track smoothness as parameter components, using a Gaussian mixture model to calculate the reference value and allowable threshold between the parameter components, and determining the weight of each parameter component.

[0023] S2, based on the values ​​of the parameter components in multiple time periods, identifies the degradation degree and distance value of each parameter component in the time series when degradation occurs, and quantifies the degradation characteristics under environmental damage.

[0024] S3, based on the degradation characteristics of each parameter component, uses the form of linear weighting to explain the comprehensive degradation degree of each parameter when degradation occurs, and uses the time of comprehensive degradation degree update as the key node of each parameter component.

[0025] S4 verifies the association probability between each key node and the maintenance record, and sets the warning parameter set according to the track section of each key node under the association probability and the length of the track section as the boundary condition. According to the number of consecutive warnings in the warning parameter set, the change trend of each key node is set.

[0026] S5, based on the change trend of each key node, interactive analysis is performed to determine the warning range limit of each parameter component under parameter accumulation.

[0027] When completing the analysis of factors affecting track smoothness, the problem essentially solved by the present invention is how to determine the abnormality type corresponding to the current track under the cumulative damage of the environment, as well as the deviation value and number of cyclic processing times corresponding to the abnormality type, to determine the multiple factors related to track smoothness.

[0028] For example, based on the parameter components collected after temperature cycle, weather environment, and driving vibration, it is determined whether any element in the current parameter components is abnormal, and the corresponding abnormal type is marked to obtain the warning range limit of each parameter component.

[0029] When analyzing track smoothness, a judgment matrix is ​​constructed based on its parameter components to form five sets of sample data, such as the judgment matrix for the five sets of parameters of track gauge deviation, height irregularity, horizontal irregularity, track direction irregularity and curvature change rate.

[0030] The architecture is extracted based on the locations of these parameters, including but not limited to setting a relative parameter matrix between the parameters, and setting a weight matrix in a normalized manner to explain the weight of each parameter.

[0031] Gauge deviation refers to the difference between the measured gauge 16mm below the inner rail top surface of the two rails and the standard value. The standard value of gauge deviation is 1435mm. Vertical irregularity refers to the amplitude of irregularity in the longitudinal vertical direction of the track, usually measured by the change in rail surface height. Horizontal irregularity refers to the height difference between the left and right rail surfaces of the track. Track irregularity refers to the amplitude of irregularity in the transverse direction of the track, that is, the lateral deviation of the track centerline in the horizontal plane, usually based on a chord length of 10m. Curvature change rate refers to the change in curvature per unit length of the curve segment, which is used to illustrate the curvature change of the current track.

[0032] like Figure 2 As shown, the implementation of step S1 includes: S11, when determining the input parameter component, performing multi-parameter association based on the time and environment obtained by the parameter component, and obtaining at least one state classification corresponding to the current parameter component.

[0033] S12, using the Gaussian mixture model to obtain the number of data included in each state classification, and obtaining the reference value and the allowable threshold value under the current state classification.

[0034] S13, using the reference value and allowable threshold of each state classification, set the weight of each parameter component.

[0035] When obtaining parameter components, it is first necessary to form a baseline state model based on the values ​​of the current parameter components in the initial healthy state. This model takes gauge deviation, height irregularity, horizontal irregularity, track irregularity and curvature change rate as input parameters. After normalizing the parameters, a Gaussian mixture model is used to describe the joint distribution of multiple parameters.

[0036] Baseline state model Expressed as: ;in, Represents the mixing coefficient of the kth Gaussian component, which indicates the weight or contribution ratio of the component in the overall component; Represents the mean vector of the k-th Gaussian component, describing the center position of the component; Represents the covariance matrix of the kth Gaussian component, describing the distribution shape and direction of the component. x represents a random variable, describing the step-by-step characteristics of the currently entered parameter component under the mixed distribution. K represents the number of Gaussian components, ranging from 1 to K. When setting the Gaussian component, the number of Gaussian components is based on the number of state classifications of the input parameter.

[0037] At this time, the corresponding data is input into the relevant state classification to illustrate the multiple states of the current track smoothness. For example, after setting the classification for the abnormal and normal parts of gauge deviation, height irregularity, horizontal irregularity, track irregularity and curvature change rate, the multi-parameter distribution form is used to describe the components of each parameter and the weights of the parameter components. At this time, the curvature change rate is used as a prerequisite for the previous four parameters to illustrate the changes in the track geometry, and as an auxiliary value for obtaining the parameter distribution in the Gaussian mixture model to describe the state of the overall track distribution.

[0038] After the final benchmark state model is implemented using a Gaussian mixture, the output data will include the benchmark values ​​of the mean vector and covariance matrix of the parameter components under multiple inputs, as well as the permissible threshold of the current benchmark state model. The benchmark values ​​of the mean vector and covariance matrix described here are obtained by averaging the values ​​under the corresponding state classification. The permissible threshold represents the value range of the corresponding state classification under multiple iterations. This value range directly adopts the permissible deviation limits for static geometric dimensions in the "High-Speed ​​Railway Maintenance Rules," such as a height deviation of ≤2mm, to illustrate the permissible thresholds under different state classifications. The corresponding benchmark value is then used as the benchmark value for the current state classification, and the permissible threshold is annotated to the corresponding state classification to illustrate the relevant values ​​of track irregularity when the track is irregular at a certain angle.

[0039] It should be noted that the initial weights of the currently input track gauge deviation, height irregularity, horizontal irregularity and track direction irregularity are set to 0.2, 0.3, 0.2 and 0.3 respectively; after obtaining the reference value and allowable threshold of the parameter component, the initial weight is weightedly summed with the reference value and the allowable threshold, and the value after weighted summation is regarded as the weight of the current output.

[0040] At this time, the weight is calculated by weighted summing the reference value, the upper and lower limits of the allowable threshold and the initial weight, and normalizing the summed weight so that the weights of the four adjusted values ​​can correspond.

[0041] As for the input curvature change rate, the value calculated with Gaussian mixture distribution is used as the label and as the data index of track gauge deviation, height irregularity, horizontal irregularity and track direction irregularity.

[0042] Preferably, when analyzing parameter components, a multi-parameter coupling analysis is performed on the data collected at the location of the curvature change rate and the four parameters of track gauge deviation, vertical irregularity, horizontal irregularity, and track irregularity. When using the baseline state model for calculation, the data coupled to the Gaussian mixture is used to determine whether the data collected on the current track has a sudden change in curvature and whether the curve segment transition is smooth. For example, the implementation of the curvature change rate also includes comparing the mean vector and covariance matrix solved under the curvature change rate baseline state with the portion calculated from data collected over multiple time periods. If they are consistent, an analysis is performed on the degradation of track gauge deviation, vertical irregularity, horizontal irregularity, and track irregularity. If they are inconsistent, the corresponding curvature change rate value is recorded, and the parameter component corresponding to the curvature change rate is considered a key node in the parameter component processing. Specifically, the relevant data for track gauge deviation, vertical irregularity, horizontal irregularity, and track irregularity under inconsistent curvature change rates are directly input into the key node to analyze the degree of degradation.

[0043] Preferably, the implementation of step S11 further includes: establishing a coordinate center of each parameter component according to the coordinate information of the currently acquired parameter component, and classifying the status of the local parameter components according to the parameter components contained under the coordinate center.

[0044] Compare the sizes of all state categories and take the state category with the largest amount of data as the initial center category; calculate the distance between the coordinate centers of the initial center category and other state categories. If the distance between the coordinate centers is greater than the preset distance, merge the corresponding state categories until all state categories are merged.

[0045] At this time, based on the spatial position of the obtained parameters, multiple spatial coordinate centers are selected, clustered using the coordinate centers, and then the state classifications are merged according to the distance between the coordinate centers to determine whether the Gaussian mixture model can contain enough data during processing to complete the smoothness analysis of the tracks covered by different curvature change rates.

[0046] As mentioned above, the preset distance is set based on the average distance between the coordinate centers when they are laid out plus three times the standard deviation. The average distance and standard deviation described here are calculated based on the distances between adjacent coordinate centers.

[0047] In one embodiment of the present invention, after quantifying the baseline values, allowable thresholds, and weights of multiple parameter components, it is necessary to conduct a centralized analysis of the gauge deviation, height irregularity, horizontal irregularity, track direction irregularity, and curvature change rate. These parameters represent the geometric measurement data of the track, which represent the inertial performance of the track under vehicle driving. If there are abnormal conditions that exceed the predicted values, it may indicate that there is a problem with the smoothness of the track. At this time, it is necessary to quantify the multiple factors that affect the smoothness, and conduct subsequent analysis based on the degradation form of the geometric measurement data in multiple time periods.

[0048] As for the description of degradation analysis, a relative degradation degree is set, and the degradation degree when degradation occurs is described according to the degradation degree of the relative degradation degree.

[0049] ;in, It represents the degradation degree of the jth parameter at time t. When its value is greater than 1, it means that the parameter exceeds the standard, indicating that there is a significant deviation in the track geometry parameters, which may lead to obvious track irregularities and other problems. represents the actual measured value of the jth parameter at time t, which is used to illustrate the deviation of the orbit at the corresponding time; The jth parameter represents the reference value obtained by the mean vector of the jth parameter. In this case, it represents the reference value relative to the jth parameter, usually the statistical mean under normal conditions, such as the average value under initial stable track operation. The jth parameter represents the number of the current parameter component. In this case, the four parameters of track gauge deviation, vertical irregularity, horizontal irregularity, and track irregularity can be numbered separately, and the value of j is set to 1 to 4 to complete the degradation analysis of multiple parameters. Indicates the allowable threshold of the jth parameter, which represents the threshold under the currently selected corresponding parameter.

[0050] As for the distance value obtained subsequently, it represents the comprehensive deviation degree between the current state and the baseline state under the joint distribution of quantitative multiple parameters, avoiding the simple superposition of the degradation degree of each parameter to explain the deviation caused by the current processing. At this time, the Mahalanobis distance is used to emphasize the difference between the current state and the baseline state.

[0051] ;in, Represents the distance value at time t, using Mahalanobis distance to illustrate the current multi-parameter correlation; represents a multi-parameter vector at time t, ; 、 、 and They represent the parameter vectors corresponding to gauge deviation, height irregularity, horizontal irregularity and track irregularity, respectively, and indicate the values ​​of the four parameters at time t; Indicates the reference value for the mean vector, where the reference values ​​of the four parameters are combined; represents the reference value of the covariance matrix, express The inverse matrix of , to prevent the variance of the original matrix from being too large or too small; and Both represent matrix transposition operations. After obtaining the corresponding degradation degree and distance values, the deviation and discrete state of the current track condition relative to the reference condition can be quantified based on these two values.

[0052] like Figure 3 As shown, the implementation method of step S2 also includes: S21, based on the degradation degree and distance value of each parameter component as input features, traversing the time series data, selecting the critical point of each parameter component, using the degradation degree of four parameters plus a Mahalanobis distance as input features, and explaining the variation of degradation under environmental damage according to its time trend analysis form. For example, periodic time series analysis is used to determine the environmental impact under the coordination of multiple parameters, and the parameter part of this part of the environmental impact is output as a degradation feature, so as to facilitate the subsequent identification of multiple factors affecting track smoothness.

[0053] It should be noted that the above-mentioned critical points are used to indicate that the gauge deviation, height irregularity, horizontal irregularity and track direction irregularity represented by the parameter components have data with degradation values ​​close to or exceeding 1, and the calculated distance value is greater than or equal to the distance value in the critical degradation state. The critical degradation state means that after taking the data of normal track operation in the historical data as input, the upper limit value of the confidence interval of the 95% confidence level composed of its data is used as the value judged to be close to degradation at this time, the part with degradation greater than the upper limit value and the part greater than 1 are taken as its critical points, and the Mahalanobis distance calculated according to the upper limit value of its confidence interval is used as the distance value in the critical degradation state. Data greater than or equal to these values ​​are regarded as multiple critical points of track degradation. The critical points will indicate the degradation situation under abnormal geometric parameter measurement, which is convenient for timely finding the factors causing track degradation in subsequent verification.

[0054] S22 introduces the extreme values ​​of each critical point as the time series progresses. This involves determining the extreme values ​​of track degradation that can be expressed by the corresponding input features as the time series changes when each critical point appears. Critical points are associated with these extreme values, and the time series is segmented into normal, progressive, and stable degradation states based on the degradation trend. The normal state represents the portion of the time series that does not contain a critical point, while the stable degradation state represents the portion where the absolute value of the rate of change of multiple consecutive points is less than 0.05% under the trend of change at the critical point. The remaining portion containing the critical point is considered to be in the progressive degradation state. In addition to these three states, the trend corresponding to the input features is also displayed in multiple time periods to illustrate the number of extreme values ​​corresponding to the critical points after segmentation. Alternatively, multiple analyses can be performed, with the current analysis data containing only the data per unit time corresponding to one time step. This data can be measured in days to determine the smoothness of the track during operation.

[0055] S23: When the critical point after scenario division corresponds to only one extreme point, it indicates that there may be sudden deterioration in a short period of time, and the degradation degree or Mahalanobis distance will quickly fall back or enter a stable state after reaching the critical value. This indicates that there is degradation of track geometry caused by a single factor such as train overload or temporary construction disturbance, and long-term environmental damage has occurred. It emphasizes short-term fluctuations and tends to display non-periodic situations. The corresponding data is output as a degradation feature using the amplitude value corresponding to the extreme point. The amplitude value represents the difference between the extreme point and other critical points under the current scenario division. This difference is expressed as a ratio to illustrate the growth rate of the extreme point in the corresponding scenario.

[0056] S24, when the critical point after scene division corresponds to multiple extreme points, it indicates that there is a periodic distribution, which may be affected by the periodic influence of freeze-thaw cycles and humidity fluctuations in environmental factors over a long period of time. The data obtained in this part may indicate the degradation of multiple parameters, which is consistent with the periodic influence of the cumulative nature of the environment. The data of the part where these extreme points are located needs to be used as the output degradation feature; the data of the continuous time length corresponding to the extreme points is used as the output degradation feature.

[0057] Preferably, when outputting degradation features, the implementation further includes obtaining the number of extreme value points of the output degradation features per unit time, and associating the time points at which degradation occurred based on the distribution of the number of extreme value points. In this case, the spatial and temporal locations of the degradation features are annotated to facilitate subsequent search for key nodes to be identified.

[0058] In one embodiment of the present invention, when performing linear weighting, it is necessary to use the weight of each parameter component combined with the degradation degree of each parameter classification to perform linear weighting, as shown below.

[0059] ;in, Indicates the comprehensive degradation degree at time t, which represents the comprehensive degradation value of any one of the parameters: gauge deviation, height irregularity, horizontal irregularity, and track irregularity at time t; n represents the amount of data used in the calculation of the comprehensive degradation degree. In this case, n can be set to 4 to represent the comprehensive degradation degree of the four parameters of gauge deviation, height irregularity, horizontal irregularity, and track irregularity at the current position, or the total amount of data collected using time series at the current position is used as its value range to calculate the comprehensive degradation degree relative to a track; represents the weight of the jth parameter, which is obtained by adjusting the reference value and the allowable threshold. The weight is used to illustrate the comprehensive degradation degree of any parameter among gauge deviation, height irregularity, horizontal irregularity and track irregularity at the corresponding moment according to the value of the weight at different moments, so as to emphasize the relative weight values ​​at different moments, illustrate the key nodes that need to be identified and processed in the time series, and emphasize adaptive adjustment to the environment.

[0060] Afterwards, a nonlinear correction method can be used to correct the value range and make the comprehensive degradation degree obtained by the four parameters able to be comprehensively summed.

[0061] ;in, Indicates the corrected comprehensive degradation degree at time t, which is used to map the value range to between 0 and 1 to prevent some calculated comprehensive degradation degree values ​​from being too large; represents the exponential constant, Indicates the curvature factor, which is set based on the value range of the curvature change rate corresponding to the current gauge deviation, height irregularity, horizontal irregularity and track irregularity. Its value range is usually 0.5-1. Based on the current linearly weighted data, the value range of its curvature change rate is mapped with the value range of the database flag, and the corresponding curvature factor is selected.

[0062] Afterwards, the recursive Bayesian update method can be used to check the time when the comprehensive degradation degree is updated to obtain the key nodes relative to the current processing. At this time, the comprehensive degradation degrees obtained in multiple time periods are queried one by one in the form of time steps. Based on the conditions that the baseline value in each time step and the probability value calculated by the corresponding Gaussian mixture meet the probability distribution, the time point when the baseline value and the allowable threshold obtained by the baseline state model change is obtained, and the corresponding comprehensive degradation degree value is checked based on the time position and spatial position of the change, and multiple periodic points under the periodic change of the comprehensive degradation degree are used as the key nodes of the output.

[0063] For example, by updating the sliding window, the baseline values ​​of the mean vector and covariance matrix are recalculated based on the data of the last 30 days to determine the data that has changed relative to the baseline state.

[0064] Recursive Bayesian updating ,in, Represents the posterior distribution, which means that all data from time step 1 to t are observed After that, the current parameters Probability distribution estimation of ; Represents the data from time step 1 to t, which includes all the data of track gauge deviation, height irregularity, horizontal irregularity and track direction irregularity when solving the reference state; Represents the data from time step 1 to t-1; Represents the data of the current time step t; represents the value of the mean vector under recursive Bayesian updating to illustrate the change of the baseline state at each time step; Represents the likelihood function, which means that given the current parameters Under the condition of probability; Represents the prior distribution, which means that when observing historical data from time step 1 to t-1 After that, the current parameters The probability distribution estimation of the current degradation characteristics is updated according to the form of these parameters, and the time when the change occurs is further verified.

[0065] like Figure 4 As shown, the implementation method of step S3 includes: S31, judging the currently acquired degradation characteristics, and when the baseline value and the allowable threshold corresponding to the degradation characteristics change, the time point of the change is used as the initial time point; here, multiple data showing degradation can be obtained, and the part of the corresponding data showing degradation transformation is identified, which will represent a change in the baseline condition of the track. When these baseline values ​​and allowable thresholds change, the use nature of the track will be seriously affected.

[0066] S32, use the comprehensive degradation degree obtained at the initial time point to perform feature solution, and obtain the periodic point of the initial time point with the spatial position and time position under the feature solution; as for the form of feature solution, the identified data is described in the form of a feature vector, that is, the degradation characteristics may contain some data on track gauge deviation, height unevenness, horizontal unevenness and track direction unevenness, as well as comprehensive degradation degree, with spatial position and time position as the theme, and the periodically updated parts therein are combined into a feature vector to illustrate the periodic points existing at the initial time point.

[0067] The periodic points can be decomposed in the form of time trend, and the data corresponding to the trend components and seasonal components are used as the periodic points at this time. For example, the STL decomposition calculation method is used to obtain the periodic points existing in the current input degradation characteristics. At this time, STL decomposition is used to decompose the current input data into an annual cycle. After determining the length of the current processing cycle, the parts of the trend components and seasonal components of the STL decomposition that have local extreme values ​​are regarded as periodic points, and the corresponding periodic points are output.

[0068] S33, using the available node information of each periodic point, output the periodic point according to the updateable time and number of times as the key node of each parameter component.

[0069] As for the available node information, it describes whether the data can be updated at the corresponding periodic point. After filtering out invalid nodes with missing data, the periodic points are sorted by time and then sorted again according to the number of updates. The sorted periodic points are used as the key nodes for output.

[0070] Preferably, when step S3 is implemented, its implementation method further includes: judging the value range of the comprehensive degradation degree corresponding to the current degradation characteristic, and when the comprehensive degradation degree exceeds the preset degradation value, considering the corresponding time point as the time for updating the comprehensive degradation degree.

[0071] At this time, the comprehensive value of the degradation degree is judged, and 0.6 and 0.8 are used as the preset degradation values ​​respectively. When it is greater than 0.6, the location and time of the corresponding degradation feature will be marked to indicate that there is an abnormal problem in the track section. After marking it in the form of a yellow warning, it will be processed according to the time point of the comprehensive degradation degree update to check whether the current track unevenness problem is a periodic problem. As for when it is greater than 0.8, a serious warning will be issued and marked in red to indicate the current tendency of track degradation, and finally its comprehensive processing will be completed.

[0072] For example, the ARIMA model is used to predict the value of the comprehensive degradation degree represented by its comprehensive degradation degree.

[0073] ;in, represents the comprehensive degradation degree at time t+1; Represents the constant term, which indicates the fixed part under multiple autoregressive analysis; Represents the autoregressive coefficient of the i-th order, where i ranges from 1 to p. Here, p represents the autoregressive order and the lag in the current regression calculation. represents the comprehensive degradation value of lag i; It represents the white noise term, which describes the unpredictable random fluctuation at time t. The value predicted by the ARIMA model is used as the current judgment of whether the degradation value exceeds the preset degradation value. The part that exceeds the value is also regarded as the key node that requires early warning to illustrate the components of the uneven part on the current track.

[0074] In one embodiment of the present invention, in step S4, the association probability between the position represented by each key node and the maintenance record can be calculated based on the ratio of the number of maintenance times of the corresponding key node to the total number of maintenance times, multiplying the ratio by the length of the track section corresponding to the key node, and dividing it by the total track section length during track maintenance to illustrate the association probability of each key node during maintenance; if the current key node appears for the first time, the comprehensive degradation degree during the current maintenance and the corresponding track gauge deviation, height irregularity, horizontal irregularity, track direction irregularity and curvature change rate are composed into a vector, and then the cosine similarity is calculated with the corresponding vector composed of the corresponding data in the maintenance record, and the value of the cosine similarity is regarded as its association probability.

[0075] The track sections described above are combined based on the locations of key nodes to form multiple train tracks, and the corresponding train tracks are used as the described track sections; or the length value of the actual sampling data of each track at the track where the key nodes are located is extracted. At this time, the length of the track section can be obtained from the database based on the current processing requirements to illustrate the train tracks that need maintenance and warnings in a larger range.

[0076] Since the key nodes are obtained based on multiple groups of analysis of time series, the analysis of track sections at this time tends to be based on spatial position analysis. The track sections formed by spatial position aggregation describe the relevant contents of their maintenance records to illustrate the parts that need to be analyzed at the current key nodes.

[0077] like Figure 5 As shown, the implementation method of step S4 also includes: S41, performing a constraint check on the track section where the key node is located, and observing whether the change trend of the key node in a single track section is single; this is used to describe the situation of maintenance warning in the track section where the key node is located. For example, the warning parameter set is used to issue a warning based on the comprehensive degradation degree. At this time, the warning parameter set is set to associate multiple key nodes with historical maintenance records to illustrate the relative data under historical maintenance conditions, and combined with the number of warnings contained in the currently identified data, to determine the probability of actual maintenance and warning situation of the corresponding key node.

[0078] As for the relationship between the number of warnings set and the changing trend of key nodes, by gradually analyzing the current space-time steps, counting the values ​​of the comprehensive degradation degree at each time step, and then marking the parts greater than 0.6 and 0.8, they are further divided into low-risk, medium-risk and high-risk parts according to the number of statistics at multiple time steps. For example, if there are 1-2 yellow warning signs at low risk, 3-5 yellow warnings or 1-2 red serious warnings at medium risk, and more than 6 yellow warnings and more than 3 red serious warnings at high risk, it means that the corresponding track section needs timely maintenance. These statistical values ​​are regarded as the changing trend of the current key nodes to illustrate which key nodes need timely maintenance under statistics in multiple time periods.

[0079] S42, if it is single, the number of key nodes in the current track section is accumulated, and each key node is marked with the association probability of the accumulated key nodes; when there is a single trend, it means that the current track section contains at least one key node, and when the trend of increase and decrease of the number of warnings is consistent, it means that the current key node needs to be centrally maintained according to the entire track section. At this time, the association between the current track section and the maintenance part in the historical data can be explained according to the sum of the association probabilities, thereby explaining the maintenance and processing method of the current track section; the order of maintenance and viewing of the key nodes can be set from large to small according to the sum of the association probabilities at this time.

[0080] If the trend is not uniform, then the key node with the most consecutive warnings is used as the main component, and the other key nodes are combined. The processing order of the key nodes is set based on the combined conditional probability. If the trend is not uniform, it indicates that some nodes in this track segment have an increasing number of warnings while others have a decreasing number of warnings. This means that multiple key nodes marked in the current track segment need to be processed in batches within a unit of time.

[0081] That is, taking the key node with the most warning times as the main body, the other key nodes are combined with this key node in pairs in sequence, and after obtaining the conditional probability based on the association probability of the two key nodes after the combination, the processing order of the current key node in the maintenance and warning scenarios is set from large to small.

[0082] In one embodiment of the present invention, when determining the limits of the warning range, data that make a significant contribution to degradation analysis and warning can be set based on the contribution of the corresponding features in each warning parameter set to the whole, and the value of this part of the data can be used as the limit value for the current warning range setting.

[0083] For example, the relative contribution of each warning parameter set can be quantified by taking the associated probability in each warning parameter set as the weight, the comprehensive degradation degree as the value, and the ratio of the weighted sum of the current warning parameter set to the weighted sum of all warning parameter sets; or the values ​​corresponding to track gauge deviation, height irregularity, horizontal irregularity, and track direction irregularity can be separated, and the ratio of the weighted sum of their values ​​and the weights set for the parameter components to the weighted sum of all warning parameter sets can be used to further illustrate the contribution of relevant parameters to the whole during the warning.

[0084] like Figure 6 As shown, the implementation method of step S5 also includes: S51, according to the contribution of the warning parameter set under the interactive analysis, starting from the warning parameter set with the maximum contribution, setting the initial range limit; the initial range node is set with the value range of the warning parameter set with the maximum contribution, so as to achieve the lower limit value of the parameter to be warned as the initial range limit, for example, the warning limit is set with the average value of the lower limit. At this time, the warning parameter sets with different contribution degrees will represent the track unevenness on different track sections. When the contribution degree is large, it means that the unevenness problem at the corresponding position is obvious, and the extracted parameter components will show a large value and a large value of the comprehensive degradation degree. At this time, the lower limit average value when the warning is reached is used to illustrate the value when the current track section needs to enter the warning. This value will include the directly measured values ​​of gauge deviation, height unevenness, horizontal unevenness and track direction unevenness to map the data in the warning parameter set to the corresponding range.

[0085] After that, it is also necessary to consider multiple warning parameter sets whose contributions are not the largest under the parameter component combination, and adjust the value range considered in the initial range limit.

[0086] S52, using the correlation coefficient between parameter components in other warning parameter sets, the correlation coefficient is based on the four coefficients of gauge deviation, height irregularity, horizontal irregularity and track direction irregularity, with normalized values, and using the Pearson correlation coefficient to calculate the correlation coefficients pairwise. At this time, the correlation coefficient will be determined based on the correlation between the current data and the historical data, and the data collected from the current data and historical data related to track warning and maintenance will be used for calculation; the range limit pair corresponding to the correlation coefficient, at this time the range limit pair represents a data pair composed of the data of the parameter components combined in pairs to reach the lower limit value of the warning and the correlation coefficient; the initial range limit is corrected, and the corrected initial range limit is used as the output warning range limit.

[0087] It should be noted that when the current judgment reaches the lower limit of the warning, by checking the part of the warning parameter set where the comprehensive degradation value reaches 0.6, these corresponding data are regarded as reaching the warning. After that, these data are mapped to the values ​​of the relevant parameter components, and then combined into the spatial nodes represented by the corresponding key nodes according to their lower limit values ​​to obtain the value of the currently marked warning.

[0088] As for the above-mentioned method of correcting the initial range limits, the warning range limits are set in the form of a moving average of the range limit pair and the initial range limit. For example, after introducing the values ​​of the initial range limits and the range limit pair at multiple adjacent key nodes, the moving average is used as the warning data for the corresponding orbital position; at the same time, the corresponding warning range limits can also be set in the form of a moving average using the value of the current initial range limit corresponding to the time dimension in the form of continuous time nodes and the value of the range limit pair extracted at the continuous time points; since the range limit pair is equivalent to describing two lower limit values, when correcting the initial range limit, the lower limit value corresponding to the initial range limit is used to obtain all range limit pairs related to the value in space and time for correction, so as to complete the warning processing method for track smoothness under cumulative degradation in multiple environments.

[0089] By combining the joint processing of warning parameter sets with different contribution levels, the geometric parameters obtained when identifying the geometric parameters of the current track can be more inclined to the values ​​of the relevant measurements of gauge deviation, height unevenness, horizontal unevenness, and track direction unevenness, so as to further obtain the changes in the track under environmental accumulation, as well as the parts that require warning processing, to improve the efficiency of track maintenance and processing.

[0090] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered by the scope of protection of the present invention.

Claims

1. A method for correlation analysis of factors affecting track smoothness, characterized in that: include: S1, using track gauge deviation, vertical irregularity, horizontal irregularity, track irregularity, and curvature change rate corresponding to track smoothness as parameter components, uses a Gaussian mixture model to calculate the baseline value and allowable threshold between the parameter components and determine the weight of each parameter component; S2, based on the values ​​of the parameter components in multiple time periods, identifies the degradation degree and distance value of each parameter component in the time series when degradation occurs, and quantifies the degradation characteristics under environmental damage; S3, based on the degradation characteristics of each parameter component, uses the form of linear weighting to explain the comprehensive degradation degree of each parameter when degradation occurs, and uses the time of comprehensive degradation degree update as the key node of each parameter component; S4: Verify the association probability between each key node and the maintenance record, and set the warning parameter set according to the track segment under the association probability of each key node, using the length of the track segment as the boundary condition. Set the change trend of each key node according to the number of consecutive warnings in the warning parameter set; The association probability between the position represented by each key node and the maintenance record is calculated based on the ratio of the number of maintenance times of the corresponding key node to the total number of maintenance times. This ratio is multiplied by the length of the track section corresponding to the key node and divided by the total length of the track section during the track maintenance to express the association probability of each key node during maintenance. If the current key node appears for the first time, the comprehensive degradation degree during the current maintenance is combined with the corresponding gauge deviation, height irregularity, horizontal irregularity, track irregularity, and curvature change rate to form a vector. The cosine similarity is calculated with the corresponding vector composed of the corresponding data in the maintenance record, and the value of the cosine similarity is regarded as its association probability. Track segments are grouped based on the locations of key nodes to form multiple train tracks, and the corresponding train tracks are used as the track segments to describe; or the length value of the actual sampled data of each track at the track where the key nodes are located is extracted; S5, based on the change trend of each key node, interactive analysis is performed to determine the warning range limit of each parameter component under parameter accumulation.

2. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of step S1 includes: S11, when determining the input parameter component, performing multi-parameter association based on the time and environment obtained by the parameter component, and obtaining at least one state classification corresponding to the current parameter component; S12, using the Gaussian mixture model to obtain the reference value and allowable threshold value of the current state classification based on the number of data included in each state classification; S13, using the reference value and allowable threshold of each state classification, set the weight of each parameter component.

3. The method for analyzing the correlation between factors affecting track smoothness according to claim 2, characterized in that: The implementation of step S11 further includes: According to the coordinate information of the currently acquired parameter components, the coordinate center of each parameter component is established, and the state of the local parameter components is classified according to the parameter components contained under the coordinate center; Compare the sizes of all state categories and take the state category with the largest amount of data as the initial center category; calculate the distance between the coordinate centers of the initial center category and other state categories. If the distance between the coordinate centers is greater than the preset distance, merge the corresponding state categories until all state categories are merged.

4. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of step S2 further includes: S21, based on the degradation degree and distance value of each parameter component as input features, traverse the time series data and select the critical point of each parameter component; S22, introduces the extreme value points of each critical point when the time series advances, associates the critical points with the extreme value points, and divides the time series into scenarios according to the changing trend of the degradation degree; S23, when the critical point after scene division corresponds to only one extreme point, the corresponding data is output as a degradation feature based on the amplitude value corresponding to the extreme point; S24: When the critical point after scene division corresponds to multiple extreme value points, the data of the continuous time length corresponding to the extreme value points is used as the output degradation feature.

5. The method for correlation analysis of factors affecting track smoothness according to claim 4, characterized in that: When outputting degradation characteristics, the implementation method also includes: The number of extreme value points of the output degradation feature per unit time is obtained, and the time point at which degradation occurs is associated with the distribution position of the number of extreme value points.

6. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of step S3 includes: S31, determining the currently acquired degradation feature, and when the reference value and the allowable threshold corresponding to the degradation feature change, using the time point of the change as the initial time point; S32, performing feature solution using the comprehensive degradation degree obtained at the initial time point, and obtaining the periodic point at the initial time point based on the spatial position and time position obtained from the feature solution; S33, using the available node information of each periodic point, output the periodic point according to the updateable time and number of times as the key node of each parameter component.

7. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of curvature change rate also includes: The mean vector and covariance matrix solved under the baseline state of the curvature change rate are compared with the parts calculated from the data collected in multiple time periods. If they are inconsistent, the corresponding curvature change rate value is recorded, and the parameter component corresponding to the curvature change rate is regarded as the key node in the parameter component processing.

8. The method for analyzing the correlation between factors affecting track smoothness according to claim 6, characterized in that: The implementation of step S3 further includes: The value range of the comprehensive degradation degree corresponding to the current degradation characteristic is determined. When the comprehensive degradation degree exceeds the preset degradation value, the corresponding time point is regarded as the time for updating the comprehensive degradation degree.

9. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of step S4 further includes: S41, performing constraint checks on the track sections where the key nodes are located to observe whether the change trend of the key nodes in a single track section is single; S42, if it is a single one, the number of key nodes in the current track section is accumulated, and each key node is marked according to the accumulated association probability of the key nodes; S43, if it is not a single one, then the key node with the most consecutive warning times is taken as the main body, and the other key nodes are combined, and the processing order of the key nodes is set according to the conditional probability after the combination.

10. The method for correlation analysis of factors affecting track smoothness according to claim 1, characterized in that: The implementation of step S5 further includes: S51, according to the contribution of the warning parameter set under the interactive analysis, starting from the warning parameter set with the greatest contribution, setting the initial range limit; S52, using the correlation coefficients between the parameter components in other warning parameter sets, and using the range limit pairs corresponding to the correlation coefficients, to correct the initial range limits, and using the corrected initial range limits as output warning range limits.

Images