Wagon loading abnormity early warning method and system based on artificial intelligence

Abstract

The invention provides a railway wagon loading abnormity early warning method and system based on artificial intelligence, and the method comprises the steps: collecting the dynamic pressure distribution data of a compartment connection part and a plurality of groups of wheel supporting areas during the operation of a railway wagon; fusing the phase offset of the dynamic pressure distribution data at the compartment connecting part and the amplitude difference of the dynamic pressure distribution data of the multiple groups of wheel supporting areas, and generating space pressure gradient characteristics of each compartment of the railway wagon; inputting the space pressure gradient characteristics and the pressure distribution ground state of the railway wagon under the historical normal working condition into a pre-trained residual network, and outputting abnormal fluctuation components between the compartment connecting position and multiple groups of wheel supporting areas; and based on the abnormal fluctuation component, in combination with the current running speed and the track curvature parameter of the railway freight car, determining a loading abnormity type, and performing early warning according to an early warning mode of an early warning level to which the loading abnormity type belongs. According to the invention, the accuracy and timeliness of freight train operation safety early warning are improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a method and system for early warning of abnormal loading of railway freight cars based on artificial intelligence. Background Art

[0002] During the transportation of railway freight cars, due to problems such as uneven loading or overloading of goods, abnormal states such as partial load and overloading are likely to occur, seriously affecting the safety of train operation. Traditional detection means are difficult to achieve real-time monitoring under operating conditions.

[0003] Currently, there are monitoring methods based on track scales and vehicle-mounted weighing sensors. By installing weighing devices at specific positions on the track or deploying strain gauge sensors at the bottom of the carriage, static pressure data when the freight car passes is collected, and whether it is overloaded is judged by combining preset thresholds. Some improved schemes introduce machine learning algorithms to train historical weighing data to achieve preliminary classification of abnormal states.

[0004] This scheme relies on track scales or single-point weighing sensors at fixed positions, and can only obtain static load data at discrete time points. At the same time, the dynamic coupling force at the carriage connection and the cooperative force characteristics of multiple wheel support areas are not considered, resulting in insufficient recognition accuracy of the partial load state and a lack of fusion analysis with dynamic parameters such as running speed and track curvature, and a high false alarm rate. Summary of the Invention

[0005] This application provides a method and system for early warning of abnormal loading of railway freight cars based on artificial intelligence to solve the problems of low accuracy and poor timeliness of the early warning of the operation safety of freight trains in the prior art.

[0006] In a first aspect, this application provides a method for early warning of abnormal loading of railway freight cars based on artificial intelligence, including:

[0007] Collect the dynamic pressure distribution data at the carriage connection and multiple wheel support areas during the operation of the railway freight car;

[0008] Fuse the phase offset of the dynamic pressure distribution data at the carriage connection and the amplitude difference of the dynamic pressure distribution data of multiple wheel support areas to generate the spatial pressure gradient characteristics of each carriage of the railway freight car;

[0009] Input the spatial pressure gradient characteristics and the pressure distribution base state of the railway freight car under historical normal working conditions into a pre-trained residual network, and output the abnormal fluctuation component between the carriage connection and multiple wheel support areas;

[0010] Based on the abnormal fluctuation component, combined with the current running speed and track curvature parameters of the railway freight car, determine the type of abnormal loading, and issue an early warning according to the early warning method of the warning level to which the type of abnormal loading belongs.

[0011] Optionally, inputting the spatial pressure gradient feature and the pressure distribution ground state of the railway freight car under historical normal conditions into a pre-trained residual network, and outputting the abnormal fluctuation component between the carriage connection and multiple wheel support regions, including:

[0012] Inputting the spatial pressure gradient feature and the pressure distribution ground state of the railway freight car under historical normal conditions into a pre-trained residual network, and through the decomposition module in the pre-trained residual network, decomposing the spatial pressure gradient feature into a first balance parameter component parallel to the longitudinal axis direction of the carriage and a second balance parameter component perpendicular to the longitudinal axis direction of the carriage;

[0013] Through the matching module in the pre-trained residual network, hierarchically matching the corresponding components of the pressure distribution ground state to generate a dynamic coupling difference quantity;

[0014] Through the fusion module in the pre-trained residual network, weighted-fusing the dynamic coupling difference quantity to output the abnormal fluctuation component.

[0015] Optionally, the hierarchically matching the corresponding components of the pressure distribution ground state through the matching module in the pre-trained residual network to generate a dynamic coupling difference quantity includes:

[0016] In the residual network, hierarchically decomposing the pressure distribution ground state of the railway freight car under historical normal conditions through the matching module into a longitudinal ground state component corresponding to the first balance parameter component and a transverse ground state component corresponding to the second balance parameter component;

[0017] Hierarchically matching the first balance parameter component and the longitudinal ground state component to obtain an accumulated difference quantity, hierarchically matching the second balance parameter component and the transverse ground state component to generate a transverse force imbalance compensation coefficient;

[0018] Based on the accumulated difference quantity and the transverse force imbalance compensation coefficient, generating a dynamic coupling difference quantity.

[0019] Optionally, the hierarchically matching the first balance parameter component and the longitudinal ground state component to obtain an accumulated difference quantity, hierarchically matching the second balance parameter component and the transverse ground state component to generate a transverse force imbalance compensation coefficient includes:

[0020] Dividing the first balance parameter component into multiple equally spaced sections according to the carriage length, and each equally spaced section corresponds to a longitudinal ground state component reference value in the longitudinal ground state component;

[0021] For each equidistant section, calculate the instantaneous difference between the first balance parameter component corresponding to the equidistant section and the reference value of the longitudinal ground state component;

[0022] Recursively accumulate the instantaneous differences of all equidistant sections in the direction of the carriage movement to generate a cumulative difference;

[0023] Divide the second balance parameter component into a left rail contact component and a right rail contact component according to the wheel set position;

[0024] Calculate the left rail offset between the left rail contact component and the left rail reference value of the lateral ground state component, and the right rail offset between the right rail contact component and the right rail reference value of the lateral ground state component, respectively;

[0025] Calculate the lateral force imbalance compensation coefficient according to the reverse change characteristics of the left rail offset and the right rail offset of the same wheel set.

[0026] Optionally, generating a dynamic coupling difference based on the cumulative difference and the lateral force imbalance compensation coefficient includes:

[0027] Map the cumulative difference to the lateral plane of the carriage according to the corresponding wheel set position;

[0028] Adjust the distribution weight of the mapped cumulative difference in the left and right rail directions according to the lateral force imbalance compensation coefficient;

[0029] Integrate the adjusted cumulative differences corresponding to all wheel set positions to generate a dynamic coupling difference.

[0030] Optionally, determining the type of abnormal loading based on the abnormal fluctuation component, combined with the current running speed of the railway freight car and the track curvature parameter, includes:

[0031] Decompose the abnormal fluctuation component into a longitudinal eigenvector and a lateral eigenvector;

[0032] Based on the current running speed of the railway freight car and the track curvature parameter, correct the longitudinal eigenvector and the lateral eigenvector respectively;

[0033] Determine the overload level code corresponding to the overload level interval according to the frequency distribution of the corrected longitudinal eigenvector;

[0034] Determine the offloading direction code corresponding to the offloading direction interval according to the amplitude difference of the corrected lateral eigenvector;

[0035] Match the loading abnormal type corresponding to both the overload level code and the offloading direction code from a preset loading abnormal classification rule library.

[0036] Optionally, correcting the longitudinal feature vector and the transverse feature vector respectively based on the current running speed of the railway freight car and the track curvature parameter, includes:

[0037] Converting the current running speed of the railway freight car into a running speed correction coefficient consistent with the direction of the longitudinal axis of the carriage;

[0038] Decomposing the track curvature parameter of the railway freight car into a straight-line segment reference curvature and a curve-segment incremental curvature, and converting the curve-segment incremental curvature into a track curvature correction coefficient corresponding to the direction of the transverse feature vector;

[0039] Correcting the longitudinal feature vector according to the running speed correction coefficient to generate a corrected longitudinal feature vector;

[0040] Correcting the transverse feature vector according to the running speed correction coefficient and the track curvature correction coefficient to generate a corrected transverse feature vector.

[0041] In a second aspect, the present application provides an abnormal loading warning system for railway freight cars based on artificial intelligence, including:

[0042] An acquisition module, configured to acquire dynamic pressure distribution data at the carriage connection and multiple sets of wheel support areas during the running of the railway freight car;

[0043] A generation module, configured to fuse the phase offset of the dynamic pressure distribution data at the carriage connection and the amplitude difference of the dynamic pressure distribution data of the multiple sets of wheel support areas to generate a spatial pressure gradient feature of each carriage of the railway freight car;

[0044] An input module, configured to input the spatial pressure gradient feature and the pressure distribution base state of the railway freight car under historical normal conditions into a pre-trained residual network, and output an abnormal fluctuation component between the carriage connection and the multiple sets of wheel support areas;

[0045] A determination module, configured to determine the type of abnormal loading based on the abnormal fluctuation component, in combination with the current running speed and the track curvature parameter of the railway freight car, and issue a warning according to the warning method of the warning level to which the type of abnormal loading belongs.

[0046] In a third aspect, the present application provides a computing device, including a processor and a memory, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the methods for warning abnormal loading of railway freight cars based on artificial intelligence in the first aspect.

[0047] In a fourth aspect, the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method for warning of abnormal loading of railway freight cars according to any one of the first aspect is implemented.

[0048] In the present application, a method for warning of abnormal loading of railway freight cars based on artificial intelligence is provided. The method includes: collecting dynamic pressure distribution data of the carriage connection and multiple groups of wheel support areas during the operation of railway freight cars; fusing the phase offset of the dynamic pressure distribution data of the carriage connection and the amplitude difference of the dynamic pressure distribution data of the multiple groups of wheel support areas to generate spatial pressure gradient characteristics of each carriage of the railway freight car; inputting the spatial pressure gradient characteristics and the pressure distribution ground state of the railway freight car under historical normal conditions into a pre-trained residual network, and outputting an abnormal fluctuation component between the carriage connection and the multiple groups of wheel support areas; based on the abnormal fluctuation component, combining the current running speed and track curvature parameters of the railway freight car to determine the type of abnormal loading, and giving a warning according to the warning method corresponding to the warning level to which the type of abnormal loading belongs.

[0049] The technical solution provided by the present application has the following beneficial effects:

[0050] Through multi-region synchronous monitoring, the present application realizes the full-dimensional perception of the dynamic coupling force at the carriage connection and the contact pressure in the wheel support area, providing a complete data basis for abnormal detection; by fusing the phase offset and amplitude difference, a collaborative feature representation reflecting the imbalance of longitudinal force transmission and lateral force distribution of the car body is established, improving the timeliness of the safety warning for the operation of freight trains; through the hierarchical matching of the historical ground state and real-time features, the dynamic fluctuation component caused by abnormal loading is accurately separated; by dynamically adjusting the judgment threshold in combination with the running speed and track curvature, hierarchical and accurate warning under complex working conditions is realized.

[0051] Further, in the present application, the residual network decouples the spatial pressure gradient characteristics into longitudinal / transverse components through a decomposition module, generates a dynamic coupling difference amount through hierarchical difference comparison with the historical ground state by a matching module, and finally outputs an abnormal fluctuation component through weighted output by a fusion module.

[0052] Moreover, through physical dimension decoupling and hierarchical feature matching, this design improves the separation accuracy of partial load and overloading abnormalities, especially improving the detection sensitivity of lateral force imbalance under curved track conditions.

[0053] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0055] Figure 1 It is a flowchart of a method for early warning of abnormal loading of railway freight cars based on artificial intelligence provided by an embodiment of the present application;

[0056] Figure 2 It is a schematic structural diagram of a system for early warning of abnormal loading of railway freight cars based on artificial intelligence provided by an embodiment of the present application;

[0057] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0058] In order to enable those skilled in the art to better understand the solutions of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application.

[0059] In some processes described in the specification and claims of the present application and the above accompanying drawings, multiple operations that appear in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0060] Researchers found that the existing methods for monitoring the loading of railway freight cars are difficult to capture the multi-dimensional stress characteristics in the dynamic operation state in real time, and lack the analysis of the synergistic effect between the carriage connection area and the wheel-rail contact area, resulting in insufficient accuracy in detecting partial load and overloading. Based on this, the embodiments of the present application provide a method for early warning of abnormal loading of railway freight cars based on artificial intelligence. This method synchronously collects the dynamic pressure data of the carriage connection area and multiple sets of wheel support areas, constructs spatial pressure gradient features by fusing phase shift and amplitude difference, and uses a residual network to separate abnormal fluctuation components, and finally combines the operating environment parameters to achieve accurate classification and grading early warning of abnormal loading. The technical solution of the present application can be applied to the real-time safety monitoring scenario of heavy-haul railway freight cars under complex line conditions.

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0062] Figure 1 It is a flowchart of a method for early warning of abnormal loading of railway freight cars based on artificial intelligence provided by an embodiment of the present application. As Figure 1 shown, the method includes:

[0063] Step 101: Collect dynamic pressure distribution data at the carriage connection and multiple sets of wheel support areas during the operation of railway freight cars.

[0064] In this step, the carriage connection specifically refers to the mechanical connection part between adjacent carriages. The dynamic pressure distribution data at the carriage connection is used to characterize the dynamic coupling force between the car bodies. Multiple sets of wheel support areas refer to the distribution areas of the support points where each pair of wheels of the railway freight car contacts the track. The dynamic pressure distribution data of multiple sets of wheel support areas is used to characterize the contact pressure distribution between the wheels and the track. The dynamic pressure distribution data is sourced from a sensor network symmetrically arranged with the longitudinal axis of the carriage as the center and covers the support areas of at least two pairs of adjacent wheels, including time series data with time stamps, pressure values, and spatial position codes, and the sampling frequency ≥ 100Hz.

[0065] In the embodiments of the present application, first, a multi-axis force sensor array is deployed at the carriage connection, and dynamic pressure sensing technology is used to collect the interaction force data of the coupler and buffer device in real time; at the same time, a wheel-rail force sensor group is installed on the left and right rail contact surfaces of each set of wheels, and the wheel-rail contact pressure distribution is obtained through high-frequency sampling. After all sensor data is processed by anti-interference filtering, the data synchronization module unifies the time scale to ensure the spatio-temporal consistency of the data at each measurement point. The collected original pressure data is marked with spatial position attributes according to a preset coding rule to form a dynamic pressure data set with three-dimensional coordinate information, providing a standardized input for subsequent feature fusion.

[0066] For example, when a four-axis freight car on a certain freight dedicated line is running, the force sensor installed at the carriage connection detects that the dynamic pressure peak value at the No. 1 connection is 1.2 times a certain standard value (calculated by the calibration formula "pressure value = voltage reading × conversion coefficient" after directly measuring the original voltage signal through the pressure sensor), and at the same time, the left and right rail pressure difference detected at the position of the No. 2 wheel set reaches 80% of a certain safety threshold (this threshold is determined according to the axle material strength standard). All sensor data is synchronously collected at a fixed sampling interval and forms a complete data set containing time stamps and spatial coordinates after verification.

[0067] Step 102: Integrate the phase offset of the dynamic pressure distribution data at the carbody connection and the amplitude difference of the dynamic pressure distribution data in multiple wheel support areas to generate the spatial pressure gradient characteristics of each carbody of the railway freight car.

[0068] In this step, the phase offset represents the time delay difference of the pressure waveforms at the carbody connection (unit: millisecond), reflecting the dynamic force transmission delay between car bodies. The amplitude difference represents the difference between the pressure peaks of the left and right rails of the same group of wheels (unit: MPa), characterizing the degree of imbalance in the lateral load distribution. The spatial pressure gradient characteristics include a first balance parameter component parallel to the longitudinal axis of the carbody and a second balance parameter component perpendicular to the longitudinal axis of the carbody.

[0069] In the embodiment of the present application, cross-correlation analysis is performed on the time-series pressure data collected at the carbody connection to calculate the phase delay of the pressure waveforms at different measurement points, revealing the dynamic force transmission characteristics between car bodies; at the same time, the extreme values of the left and right rail pressure data in each wheel support area are extracted, and the real-time amplitude difference is calculated to reflect the lateral load distribution. The phase offset and the amplitude difference are input into the spatial mapping algorithm. By three-dimensional interpolation, the longitudinal force transmission distribution map and the lateral pressure imbalance map of the carbody are reconstructed. The feature fusion technology is used to convert the two types of distribution maps into a two-dimensional vector sequence including the longitudinal gradient intensity and the lateral imbalance degree, and finally the spatial pressure gradient characteristics with clear physical meanings are output.

[0070] For example, based on the collected data, the system calculates that there is a phase difference of a certain time unit between the pressure waveforms of the front and rear measurement points at the No. 1 connection (the time offset corresponding to the maximum correlation coefficient is obtained through the cross-correlation algorithm). At the same time, the extreme value difference between the left and right rails of the No. 2 wheel set is converted into a specific value in a certain measurement unit. After these parameters are input into the spatial feature construction module, a longitudinal distribution map showing that the maximum pressure gradient is located in the area of the 3rd wheel pair (the gradient intensity is calculated by the two-dimensional interpolation algorithm) and a lateral pressure nephogram with obvious right deviation (the imbalance degree is calculated according to the ratio formula of the left and right pressure differences "imbalance coefficient = (left rail pressure - right rail pressure) / average pressure") are generated.

[0071] Step 103: Input the spatial pressure gradient characteristics and the pressure distribution base state of the railway freight car under historical normal conditions into a pre-trained residual network, and output the abnormal fluctuation components between the carbody connection and multiple wheel support areas.

[0072] In this step, the pressure distribution base state represents a standard pressure mode library statistically obtained under historical normal conditions (stored by speed / curvature grading). The residual network includes a dedicated neural network with a three-level structure of decomposition-matching-fusion (256 nodes in the input layer). The abnormal fluctuation components are used to reflect the abnormal force characteristics deviating from the base state (including longitudinal feature vectors and lateral feature vectors).

[0073] In the embodiment of the present application, the real-time spatial pressure gradient feature is input into a pre-trained residual network. First, the network separates the input data into a longitudinal component parallel to the axis of the carriage and a transverse component perpendicular to the axis through a convolutional feature decomposition module; then, each component is compared layer by layer with the standard base-state data under the same working conditions in the historical database to calculate the feature deviation degree at each network level; finally, the deviation degree data of each level is integrated through a weighted fusion module to generate a composite anomaly index that simultaneously includes the longitudinal abnormal fluctuation intensity and the transverse abnormal area distribution, and the output is a standardized abnormal fluctuation code.

[0074] For example, the obtained feature data is compared with the normal pressure mode of trucks of the same model in a certain speed range in the historical database. The residual network first identifies that the longitudinal pressure fluctuation in the area of the No. 2 wheel set exceeds a certain percentage of the base state (calculated by the sum of squared differences at each point), and at the same time, the transverse offset at the No. 1 connection reaches a certain multiple of the base state value (the relative ratio after standardization). The network finally outputs an anomaly code, marking the main anomaly areas as the second wheel pair (the longitudinal anomaly level is determined by a preset fluctuation amplitude grading table) and the right side of the No. 1 connection (the transverse anomaly degree is divided according to the ratio of the offset to the safety threshold).

[0075] Step 104: Based on the abnormal fluctuation component, in combination with the current running speed and track curvature parameter of the railway freight car, determine the type of loading anomaly, and issue a warning according to the warning method corresponding to the warning level to which the loading anomaly type belongs.

[0076] In this step, the running speed represents the amplification factor of the influence of speed on the anomaly (formula: (v / v0)^2, v0 = 100 km / h). The track curvature parameter represents the adjustment coefficient of the lateral force for the curve (formula: 1 + 0.5 * |1 / R|, where R is the radius). The warning levels include four levels (Level I: minor; Level IV: urgent).

[0077] In the embodiment of the present application, the spectrum feature analysis of the abnormal fluctuation code output by the residual network is performed, the longitudinal abnormal main frequency component is extracted and the speed influence coefficient is calculated in combination with the real-time running speed, and the overload level is determined through a preset frequency-load relationship mapping table; at the same time, the spatial distribution pattern of the transverse abnormal area is analyzed, the curvature compensation coefficient is calculated by superimposing the current track curvature parameter, and the offloading direction is determined based on the abnormal difference ratio between the left and right tracks. Finally, the overload level and the offloading direction parameters are input into the hierarchical warning rule library, the corresponding warning level is matched, and an audible and visual alarm and data reporting are triggered to complete the closed-loop processing from anomaly detection to warning output.

[0078] For example, according to the anomaly code, the system makes a comprehensive judgment by combining the current vehicle speed (directly obtained through the on-vehicle speed sensor) and the track curvature (from the line digital map). The speed influence coefficient is calculated to be a certain value (calculated according to the formula "coefficient = (actual speed / reference speed) squared"), and the track curvature compensation coefficient is a certain value (determined according to the ratio of the reciprocal of the curve radius to the reference curvature). Finally, it is determined as a certain level of overloading combined with offloading in a certain direction (obtained by querying the pre-set anomaly level comparison table), triggering a warning signal of the corresponding level. The warning level is calculated and determined according to the anomaly index weighting formula "comprehensive index = longitudinal anomaly × speed coefficient + lateral anomaly × curvature coefficient".

[0079] Through the complete processes of multi-source sensor collaborative monitoring, dynamic feature fusion analysis, and intelligent warning decision-making, this technical solution realizes real-time and accurate monitoring of the loading state of railway freight cars. The method innovatively combines the collaborative force-bearing characteristics of the carriage connection and the wheel-rail contact area, and fully considers the influence of operating environment parameters, improving the recognition accuracy of offloading and overloading anomalies and the timeliness of warning under complex working conditions. The entire system operates stably and reliably, can effectively ensure the safe and efficient operation of railway freight transportation, and is particularly suitable for the real-time safety monitoring requirements of long and heavy trains under various line conditions.

[0080] To solve the problem of inaccurate extraction of dynamic features in the detection of abnormal loading of railway freight cars, in some embodiments, step 103: inputting the spatial pressure gradient feature and the pressure distribution base state of the railway freight car under historical normal working conditions into a pre-trained residual network, and outputting the abnormal fluctuation component between the carriage connection and multiple groups of wheel support areas, including:

[0081] Step 201: Input the spatial pressure gradient feature and the pressure distribution base state of the railway freight car under historical normal working conditions into a pre-trained residual network. Through the decomposition module in the pre-trained residual network, decompose the spatial pressure gradient feature into a first balance parameter component parallel to the longitudinal axis of the carriage and a second balance parameter component perpendicular to the longitudinal axis of the carriage.

[0082] In step 201, the first balance parameter component refers to the pressure change feature along the length direction of the carriage, reflecting the longitudinal force transmission state. The second balance parameter component refers to the pressure distribution feature perpendicular to the carriage axis, characterizing the lateral load balance situation.

[0083] In the embodiment of the present application, the residual network decomposition module adopts spatial convolution technology. First, perform direction filtering on the input spatial pressure gradient feature, extract the continuous pressure change pattern parallel to the carriage axis through the longitudinal convolution kernel to generate the first component, and at the same time use the lateral convolution kernel to capture the pressure difference pattern perpendicular to the axis to generate the second component. The two components respectively retain the mechanical information of different dimensions in the original feature map.

[0084] Step 202: Through the matching module in the pre-trained residual network, hierarchically match the corresponding components of the pressure distribution ground state to generate a dynamic coupling difference quantity.

[0085] In step 202, the dynamic coupling difference quantity refers to the matching difference between the real-time feature and the historical ground state at each network level, including a longitudinal difference spectrum and a transverse difference map.

[0086] In the embodiment of the present application, the matching module adopts a hierarchical correlation analysis method to perform multi-scale similarity comparison between the first component and the historical longitudinal ground state, and calculate the difference coefficient of each frequency band; at the same time, perform regional block-by-block matching between the second component and the historical transverse ground state to generate a spatial difference distribution. Through cross-level feature fusion technology, the two types of differences are uniformly quantified into a coupling difference index with physical significance.

[0087] Step 203: Through the fusion module in the pre-trained residual network, perform weighted fusion on the dynamic coupling difference quantity and output an abnormal fluctuation component.

[0088] In step 203, weighted fusion means applying dynamic weights according to the spatial distribution characteristics of the difference quantity to highlight key abnormal regions. The specific process of performing weighted fusion on the dynamic coupling difference quantity is as follows: First, generate a longitudinal weight coefficient according to the force transfer attenuation rate of the first balance parameter component in each detection section on the longitudinal axis of the carriage, and at the same time generate a transverse weight coefficient according to the amplitude difference ratio of the second balance parameter component in the left and right rail directions; then perform a dot product operation on the dynamic coupling difference quantity in the two dimensions of the carriage longitudinal and track transverse directions with the corresponding weight coefficients respectively, where the result of the dot product operation in the longitudinal dimension reflects the abnormal degree of the vehicle body structure force transfer, and the result of the dot product operation in the transverse dimension reflects the imbalance degree of the wheel-rail contact force; finally, dynamically proportionally mix the weighted results of the two dimensions according to the actual running speed of the carriage to generate an abnormal fluctuation component that simultaneously includes longitudinal force abnormality and transverse force imbalance characteristics, and the mixing ratio of the longitudinal and transverse characteristics in the abnormal fluctuation component increases the weight ratio of the longitudinal characteristics as the running speed increases.

[0089] In the embodiment of the present application, the fusion module first performs frequency domain energy analysis on the longitudinal difference spectrum to determine the weight coefficients of each frequency band; at the same time, performs regional detection on the transverse difference map to generate a spatial attention weight. An adaptive weighted algorithm is used to integrate the two types of difference quantities, and finally a fluctuation component that simultaneously includes the frequency domain abnormal intensity and the spatial abnormal distribution is output.

[0090] The following is a specific example:

[0091] During the operation of a four - axle freight car on a certain freight dedicated line, the system inputs the spatial pressure gradient features constructed from the dynamic pressure data at the No. 1 connection (the peak value is 1.2 times the standard value) and the left - right rail pressure difference data of the No. 2 wheel set (reaching 80% of the safety threshold) collected into the residual network. The decomposition module first extracts the pressure fluctuation features along the car body axis direction at the No. 1 connection through longitudinal convolution operations (the first balance parameter component, and the fluctuation amplitude is calculated as 1.15 times the change coefficient through "current pressure value / average value of the ground state"), and at the same time extracts the pressure distribution features perpendicular to the axis direction of the No. 2 wheel set through transverse convolution operations (the second balance parameter component, and the imbalance degree is calculated as 0.25 imbalance ratio according to "(left rail pressure - right rail pressure) / average ground state pressure"). The matching module hierarchically compares the first component with the longitudinal ground state under the historical straight - line working condition and finds the mid - frequency band difference (the difference coefficient is calculated as 0.18 through the mean square error of the feature maps layer by layer); at the same time, it regionally matches the second component with the transverse ground state under the historical curved - track working condition and identifies the abnormal pressure area on the right rail (the difference degree is calculated as 0.32 through the spatial correlation coefficient). The fusion module assigns a weight of 0.6 to the longitudinal difference (determined according to the proportion of the frequency band energy) and a weight of 0.4 to the transverse difference (determined according to the proportion of the area of the region), and finally outputs the abnormal fluctuation component to mark the existence of a complex abnormality that needs to be focused on in the area of the No. 2 wheel set.

[0092] In the embodiment of the present application, through multi - dimensional feature decoupling and intelligent matching, this solution realizes the accurate separation of abnormal features under complex working conditions, effectively avoids misjudgment caused by environmental interference in traditional methods, and enables the system to accurately identify the true dangerous state and give early warnings in a timely manner.

[0093] To further improve the accuracy of abnormal detection of railway freight cars, in some embodiments, step 202: The corresponding components of the pressure distribution ground state are hierarchically matched through the matching module in the pre - trained residual network to generate a dynamic coupling difference amount, including:

[0094] Step 301: In the residual network, the pressure distribution ground state of the railway freight car under historical normal working conditions is hierarchically decomposed through the matching module into a longitudinal ground state component corresponding to the first balance parameter component and a transverse ground state component corresponding to the second balance parameter component.

[0095] In step 301, the longitudinal ground state component refers to the reference pressure fluctuation mode along the axis direction of the car body under historical normal working conditions. The transverse ground state component refers to the reference pressure distribution mode perpendicular to the axis direction of the car body under historical normal working conditions.

[0096] In the embodiment of the present application, the matching module adopts a multi-scale decomposition technique to perform wavelet transform processing on the historical pressure distribution ground state, extracts the longitudinal pressure fluctuation characteristics in different frequency bands as longitudinal ground state components, and at the same time extracts the transverse pressure distribution characteristics of each region through spatial grid division as transverse ground state components to establish a complete reference feature library.

[0097] Step 302: Hierarchically match the first balance parameter component and the longitudinal ground state component to obtain an accumulated difference amount, and hierarchically match the second balance parameter component and the transverse ground state component to generate a transverse force imbalance compensation coefficient.

[0098] In step 302, the accumulated difference amount refers to the cumulative difference degree between the real-time longitudinal feature and the historical longitudinal ground state in each frequency band. The transverse force imbalance compensation coefficient refers to the correction parameter required when the real-time transverse feature deviates from the historical transverse ground state.

[0099] In the embodiment of the present application, wavelet packet transform is performed on the first balance parameter component, decomposed into the same frequency band as the longitudinal ground state component, and the accumulated value of the energy differences in each frequency band is calculated as the accumulated difference amount; at the same time, regional grid matching is performed on the second balance parameter component, and according to the deviation degree of the pressure difference between the left and right rails in each grid cell from the historical reference, the transverse force imbalance compensation coefficient of each region is calculated through a preset compensation formula.

[0100] Step 303: Generate a dynamic coupling difference amount based on the accumulated difference amount and the transverse force imbalance compensation coefficient.

[0101] In step 303, the dynamic coupling difference amount refers to a composite anomaly index after comprehensively considering longitudinal differences and transverse corrections.

[0102] In the embodiment of the present application, the accumulated difference amount is mapped to the longitudinal axis of the carriage according to the spatial position, and at the same time the transverse force imbalance compensation coefficient acts on the corresponding region. The two types of parameters are fused through a spatial superposition algorithm to generate a dynamic coupling difference amount matrix that includes both frequency domain anomaly characteristics and spatial distribution characteristics.

[0103] The following is a specific example:

[0104] In the operation monitoring of a four-axle freight car on a certain freight dedicated line, after the system inputs the extracted first balance parameter component (longitudinal fluctuation characteristics at the No. 1 connection point, variation coefficient 1.15) into the matching module, it first performs three-layer wavelet decomposition on the historical longitudinal ground state to obtain low-frequency, medium-frequency, and high-frequency reference components; by calculating the energy ratio of the current component to the reference component layer by layer (formula: "energy of the current layer / energy of the reference layer"), the difference coefficients of each layer are obtained as 0.9, 1.3, and 1.1 respectively. After cumulative summation, the longitudinal cumulative difference amount of 1.8 is obtained (0.9 + 1.3 + 1.1 - 1.5, where 1.5 is the normal fluctuation threshold). At the same time, for the second balance parameter component (lateral imbalance ratio of the No. 2 wheel set 0.25), the matching module compares it with the lateral ground state (reference imbalance ratio 0.15) at the same position under the historical curve condition, and calculates the lateral force imbalance compensation coefficient of 0.67 according to the formula "compensation coefficient = (current imbalance ratio - reference imbalance ratio) / reference imbalance ratio". Finally, the longitudinal cumulative difference amount is mapped to the No. 2 wheel set area according to the weight distribution of the carriage position (weight coefficient 0.7 is determined by the sensor density in this area), and after superimposing the lateral compensation coefficient, a dynamic coupling difference amount of 1.26 is generated (1.8×0.7 + 0.67×0.3).

[0105] In the embodiment of the present application, this solution realizes the accurate quantification and positioning of abnormal features through hierarchical ground state matching and multi-parameter fusion, effectively distinguishes normal working condition fluctuations from real abnormal signals, and provides a reliable quantitative basis for subsequent early warning decisions.

[0106] In order to further improve the accuracy and reliability of abnormal detection of railway freight cars, in some embodiments, step 302: hierarchically matching the first balance parameter component and the longitudinal ground state component to obtain a cumulative difference amount, and hierarchically matching the second balance parameter component and the lateral ground state component to generate a lateral force imbalance compensation coefficient, includes:

[0107] Step 401: Divide the first balance parameter component into multiple equally spaced sections according to the carriage length, and each equally spaced section corresponds to a longitudinal ground state component reference value in the longitudinal ground state component.

[0108] In step 401, the equally spaced section refers to a detection unit evenly divided along the carriage length direction. The longitudinal ground state component reference value refers to the reference pressure fluctuation amplitude of the corresponding section under historical normal working conditions.

[0109] In the embodiment of the present application, the system divides the longitudinal axis into detection sections of a fixed length according to the carriage structure characteristics, and the length of each section is determined according to the sensor distribution density, and calls the reference pressure fluctuation data of the corresponding section from the historical database as the reference value.

[0110] Step 402: For each equidistant section, calculate the instantaneous difference between the first balance parameter component corresponding to the equidistant section and the reference value of the longitudinal ground state component.

[0111] In step 402, the instantaneous difference refers to the instantaneous deviation amount between the real-time detected pressure fluctuation and the historical benchmark at the current moment.

[0112] In the embodiment of the present application, a point-to-point subtraction operation is performed on the real-time pressure data of each section and the corresponding historical benchmark data to calculate an instantaneous difference sequence reflecting the current degree of abnormality.

[0113] Step 403: Recursively accumulate the instantaneous differences of all equidistant sections in the direction of the carriage movement to generate a cumulative difference amount.

[0114] In step 403, the length direction of the carriage refers to the longitudinal axis direction of the carriage body structure (the fixed physical extension direction from the head to the tail of the carriage), which is static. The direction of the carriage movement refers to the moving direction of the freight car during actual operation (which may be deflected from the length direction of the carriage, such as when driving on a curve), which is dynamic.

[0115] In the embodiment of the present application, a sliding window integration algorithm is adopted to perform weighted accumulation on the instantaneous differences of each section in the direction of the carriage movement. The window size is adaptively adjusted according to the train speed, and finally a cumulative difference amount reflecting the overall abnormal trend is generated.

[0116] Step 404: Divide the second balance parameter component into a left rail contact component and a right rail contact component according to the position of the wheel set.

[0117] In step 404, the position of the wheel set is a logical calibration of the above support area, and each group of support areas is converted into a computable parametric position through coordinate coding. For example: the position of the first wheel set = the area 10 - 12 meters away from the head of the carriage, the position of the second wheel set = the area 20 - 22 meters away from the head of the carriage. Multiple groups of wheel support areas refer to the set of physical areas where all wheels in the railway freight car chassis structure contact the track, which is the spatial basis for the deployment of the sensor network. For example, a four-axle freight car contains 4 groups of independent support areas, and each group corresponds to the track contact surface of a wheel pair (left and right wheels). The left rail contact component and the right rail contact component respectively refer to the pressure distribution characteristics of a single wheel set on the left and right rails.

[0118] In the embodiment of the present application, according to the geometric position information of the wheel set, the lateral pressure data is separated into left and right rail independent components, and a corresponding relationship matrix is established.

[0119] Step 405: Calculate the left rail offset between the left rail contact component and the left rail reference value of the lateral ground state component, and the right rail offset between the right rail contact component and the right rail reference value of the lateral ground state component, respectively.

[0120] In step 405, the left rail offset and the right rail offset respectively reflect the deviation degrees of the left and right rail pressures from the historical benchmark.

[0121] In the embodiment of the present application, by calculating the relative change rates of the real-time left and right rail pressure values and the historical benchmark values, a quantified offset index is obtained.

[0122] Step 406: Calculate a lateral force imbalance compensation coefficient according to the reverse change characteristics of the left rail offset and the right rail offset of the same wheel set.

[0123] In step 406, the specific formation process of the reverse change characteristics is as follows: when the left rail offset of the same wheel set increases in the positive direction, its right rail offset will necessarily show a symmetric change of decreasing in the negative direction. This reverse change characteristic comes from the rigid connection characteristic of the railway freight car wheel pair structure - the left and right wheels are fixedly connected by the axle to form a motion coupling body. Under the action of the lateral force of the track, an increase in the left rail contact pressure will necessarily lead to an equal decrease in the right rail contact pressure; by real-time monitoring whether the algebraic sum of the left and right rail offsets of the same wheel set exceeds the elastic deformation threshold of the axle structure, when it is detected that the left rail offset increases by ΔP while the right rail offset synchronously decreases by ΔP, it is determined as a typical reverse change characteristic. At this time, a lateral force imbalance compensation coefficient is generated according to the ratio of ΔP to the elastic coefficient of the axle material, and this coefficient is proportional to the absolute value of the algebraic difference between the left and right rail offsets. The lateral force imbalance compensation coefficient refers to the correction parameter calculated according to the left and right rail offset characteristics.

[0124] In the embodiment of the present application, when it is detected that the left and right rail offsets of the same wheel set show a reverse change trend, a dynamic compensation coefficient is calculated according to the ratio of the absolute value of the difference between the two to the historical maximum allowable deviation.

[0125] The following is a specific example:

[0126] In the real-time monitoring of a four-axle freight car on a certain freight dedicated line, the system longitudinally divides the carriage into 6 equally spaced sections. Among them, in the 3rd section (corresponding to the position of the No. 2 wheel set), the first balance parameter component is detected to be 1.3 times the reference value (calculated by the ratio of the real-time reading of the pressure sensor in this section to the historical reference value). The instantaneous difference in this section is calculated to be 0.3 (1.3 - 1.0). The cumulative difference of 0.75 is obtained by accumulating the instantaneous differences of the first 3 sections along the traveling direction (0.2 in the 1st section, 0.25 in the 2nd section, and 0.3 in the 3rd section). At the same time, the system separates the second balance parameter component of the No. 2 wheel set according to the left and right rails. The left rail contact component is measured to be 1.2 times the reference value, and the right rail is 0.9 times. The left rail offset is calculated to be +20% ((1.2 - 1.0) / 1.0×100%), and the right rail offset is -10% ((0.9 - 1.0) / 1.0×100%). According to the reverse change characteristics of the left and right rail offsets (the total difference is 30%), the lateral force imbalance compensation coefficient of 1.2 (30% / 25%) is calculated according to "compensation coefficient = actual total difference / safety threshold" (the safety threshold is set to 25%).

[0127] In the embodiment of the present application, this solution realizes the precise positioning and quantitative evaluation of abnormal features through refined area division and dynamic parameter calculation, improves the accuracy and reliability of abnormal detection under complex working conditions, and provides a strong guarantee for the safe operation of the train.

[0128] In order to further improve the accuracy and reliability of abnormal detection of railway freight cars, in some embodiments, step 303: generating a dynamic coupling difference amount based on the cumulative difference amount and the lateral force imbalance compensation coefficient includes:

[0129] Step 501: Map the cumulative difference amount to the lateral plane of the carriage according to the corresponding wheel set position.

[0130] In step 501, the lateral plane of the carriage refers to the two-dimensional projection plane obtained by unfolding the carriage along the width direction, which is used to intuitively display the force distribution of each wheel set position.

[0131] In the embodiment of the present application, the system establishes a coordinate system according to the actual layout position of the wheel set at the bottom of the carriage, maps the cumulative difference amount to the lateral plane according to the spatial position, and forms a heat map reflecting the distribution of longitudinal anomalies in the width direction of the carriage.

[0132] Step 502: Adjust the distribution weights of the mapped cumulative difference amount in the left and right rail directions according to the lateral force imbalance compensation coefficient.

[0133] In step 502, the left-right rail direction refers to the direction of the line connecting the left rail to the right rail (parallel to the transverse extension direction of the track but bound to physical entities), with the actual rail contact surface as the reference. It is used to specifically calibrate the physical distribution direction of the wheel set pressure data and is the detection direction corresponding to the physical rail. The distribution weight refers to the process of dynamically correcting the longitudinal anomaly amount according to the degree of lateral force imbalance.

[0134] In the embodiment of the present application, a dynamic weighting algorithm based on a compensation coefficient is adopted. When the lateral force imbalance compensation coefficient of a certain side of the track is greater than 1, the weight of the cumulative difference amount on that side is correspondingly increased (weight adjustment coefficient = compensation coefficient × base weight), highlighting the abnormal concentration area.

[0135] Step 503: Integrate the adjusted cumulative difference amounts corresponding to all wheel set positions to generate a dynamic coupling difference amount.

[0136] In the embodiment of the present application, the adjusted difference amounts at each wheel set position are smoothed through a spatial interpolation algorithm to generate a dynamic coupling difference amount, and the peak area is extracted as the key object of concern.

[0137] The following is a specific example:

[0138] In the operation monitoring of a four-axle freight car on a certain freight dedicated line, the system maps the four wheel set cumulative difference amounts (No. 1: 0.9, No. 2: 1.8, No. 3: 1.2, No. 4: 1.0) calculated to the lateral plane of the car body according to the actual positions. For the lateral force imbalance compensation coefficient of 0.67 calculated for the No. 2 wheel set (calculated based on the deviation of the left-right rail pressure difference of this group from the historical benchmark), the system corrects according to the formula "adjusted difference amount = original difference amount × (1 + compensation coefficient × azimuth factor)" (azimuth factor: 0.6 for the left rail and 0.4 for the right rail). The adjusted difference amount on the left rail side of the No. 2 wheel set is 1.8×(1 + 0.67×0.6) = 2.52, and on the right rail side is 1.8×(1 + 0.67×0.4) = 2.28. After integrating all wheel set data, the system generates a final dynamic coupling difference amount of 1.85 through a weighted average algorithm (weights are allocated according to the number of sensors in each area).

[0139] In the embodiment of the present application, this solution realizes the accurate coupling analysis of longitudinal anomalies and lateral imbalances through the combination of spatial mapping and dynamic weighting, effectively improving the accuracy of anomaly positioning under complex working conditions and providing a reliable technical guarantee for the safe operation of railway freight cars.

[0140] To further improve the accuracy of classifying loading anomalies of railway freight cars, in some embodiments, step 104: Based on the abnormal fluctuation component, combined with the current running speed of the railway freight car and the track curvature parameter, determine the loading anomaly type, including:

[0141] Step 601: Decompose the abnormal fluctuation component into a longitudinal eigenvector and a transverse eigenvector.

[0142] In step 601, the longitudinal eigenvector refers to a set of parameters reflecting the abnormal fluctuation characteristics in the direction of the car body axis. The transverse eigenvector refers to a set of parameters reflecting the abnormal distribution characteristics perpendicular to the car body axis. In the example of the abnormal fluctuation component before decomposition, this component is a multi-dimensional vector group containing time series data, in the form of [timestamp, pressure value, spatial position code], where the pressure value includes both the dynamic coupling pressure at the measuring points of the car body connection (such as the peak coupler force of 120 kN ± 15%) and the contact pressures in four groups of wheel support areas (such as 85 MPa on the left rail of wheel 1, 78 MPa on the right rail of wheel 1... 92 MPa on the right rail of wheel 4); the longitudinal eigenvector obtained after decomposition is in the form of [timestamp, longitudinal force value, car body section code], and the content is a dynamic force sequence distributed along the length of the car body (such as +108 kN in section 1, +95 kN in section 2... +82 kN in section 6), and the transverse eigenvector is in the form of [timestamp, left-right rail pressure difference, wheel set number], and the content is the real-time unbalanced force at the position of each wheel set (such as the left - right rail pressure difference of +7 MPa in wheel set 1, -5 MPa in wheel set 2... +10 MPa in wheel set 4), where the longitudinal eigenvector reflects the abnormal force transmission of the car body structure, and the transverse eigenvector characterizes the imbalance in the distribution of wheel-rail contact forces.

[0143] In the embodiment of the present application, the principal component analysis method is used to decouple the direction of the abnormal fluctuation component, extract the spectral characteristics along the length of the car body as the longitudinal eigenvector, and extract the distribution characteristics of the left-right rail pressure difference as the transverse eigenvector.

[0144] Step 602: Based on the current running speed of the railway freight car and the track curvature parameters, respectively correct the longitudinal eigenvector and the transverse eigenvector.

[0145] In the embodiment of the present application, the longitudinal eigenvector is corrected by the speed square weighting method (correction coefficient = actual speed square / reference speed square), and the transverse eigenvector is corrected by the reciprocal of the curvature radius weighting method (correction coefficient = 1 + standard curvature radius / actual curvature radius).

[0146] Step 603: Determine the overload level code corresponding to the overload level interval according to the frequency distribution of the corrected longitudinal eigenvector.

[0147] In step 603, the overload level code refers to a quantitative level identifier divided according to the longitudinal abnormality degree.

[0148] In the embodiment of the present application, the main frequency component of the longitudinal feature vector is extracted through Fourier transform, and the specific overload level code is obtained by matching according to the preset "frequency - load" correspondence table.

[0149] Step 604: Determine the off - load direction code corresponding to the off - load direction interval according to the amplitude difference of the corrected lateral feature vector.

[0150] In step 604, the off - load direction code refers to the directional identifier determined according to the lateral abnormal distribution characteristics.

[0151] In the embodiment of the present application, the mean value of the left - right rail pressure difference of each wheel set in the lateral feature vector is calculated, and combined with the track curvature direction, the off - load direction code is generated through a direction discrimination algorithm.

[0152] Step 605: Match the loading anomaly type corresponding to both the overload level code and the off - load direction code from the preset loading anomaly classification rule library.

[0153] In step 605, the loading anomaly classification rule library refers to the set of predefined anomaly type discrimination criteria.

[0154] In the embodiment of the present application, a two - dimensional lookup table structure is used to store the combination relationship between the overload level code and the off - load direction code, and the final anomaly type identifier is determined through parallel matching.

[0155] The following is a specific example:

[0156] During the operation of a four - axle freight car on a certain freight dedicated line, the system decomposes the detected abnormal fluctuation components (including data that the pressure peak value at the 1st connection exceeds the standard value by 20% and the left - right rail pressure difference of the 2nd wheel set reaches the safety threshold by 80%). First, the longitudinal feature vector is extracted through spectrum analysis, and the measured main frequency is 1.3 times the reference value (calculated through fast Fourier transform); at the same time, the left - right rail pressure difference of the 2nd wheel set in the lateral feature vector is obtained as 1.25 times the standard value (calculated according to the formula "measured difference / reference difference"). Combining the current vehicle speed (measured by the speed sensor as 90% of the reference speed), the speed correction coefficient of 0.81 is calculated according to "correction coefficient = (actual speed / reference speed) squared", and the longitudinal feature after correction is 1.3×0.81≈1.05 times. Then, according to the track curvature radius (obtained from the line database as 80% of the reference radius), the curvature correction coefficient of 1.25 is calculated according to "curvature correction coefficient = reference radius / actual radius", and the lateral feature after correction is 1.25×1.25 = 1.56 times. According to the preset standard, 1.05 times longitudinally corresponds to the secondary overload code, and 1.56 times laterally corresponds to the right - hand off - load code. Finally, the anomaly type of "secondary overload combined with right - hand off - load" is matched from the rule library, and the system starts the corresponding - level warning program accordingly.

[0157] In the embodiments of the present application, this solution realizes the accurate discrimination of abnormal loading types through multi-dimensional feature decoupling and environmental parameter compensation, effectively overcomes the interference of speed and curve factors on the detection results, improves the accuracy of abnormal classification, and provides a reliable guarantee for the safe transportation of heavy-haul railways.

[0158] In order to further improve the accuracy of abnormal feature analysis of railway freight cars, in some embodiments, step 602: respectively correcting the longitudinal feature vector and the lateral feature vector based on the current running speed of the railway freight car and the track curvature parameter, includes:

[0159] Step 701: Convert the current running speed of the railway freight car into a running speed correction coefficient that is consistent with the direction of the longitudinal axis of the car body.

[0160] In step 701, the specific process of converting the current running speed of the railway freight car into a running speed correction coefficient is as follows: First, establish a speed-correction coefficient mapping function. This function takes the maximum design speed of the train (such as 120 km / h) as the reference value. When the real-time speed reaches the reference value, the correction coefficient is 1.0. When it is lower than the reference value, it decreases according to a quadratic function relationship (such as 60 km / h corresponding to 0.25). When it exceeds the reference value, it increases linearly (such as 150 km / h corresponding to 1.25). At the same time, introduce the real-time monitoring value of the longitudinal vibration frequency of the car body for dynamic adjustment. When the vibration frequency exceeds the safety threshold, apply an attenuation factor of 0.8 to the correction coefficient. Example: When the freight car is traveling at 100 km / h (reference speed 120 km / h) and the vibration frequency is normal, the speed correction coefficient calculated through the quadratic function is 0.69 ((100 / 120)2≈0.69). Being consistent with the direction of the longitudinal axis of the car body means that the acting direction of the physical quantity is completely collinear with the longitudinal axis of the car body (such as traction / braking force). In the present application, the "running speed correction coefficient" must be limited to "direction consistent" (because speed is an overall motion parameter of the car body). Being parallel to the longitudinal axis of the car body means that the acting direction of the physical quantity is parallel to the longitudinal axis of the car body but allows spatial offset (such as the longitudinal pressure distribution of multiple groups of wheels). In the present application, the "first balance parameter component" needs to be limited to "axis parallel" (because the pressure distribution may come from wheel groups at different spatial positions). The running speed correction coefficient is an adjustment parameter that reflects the influence degree of vehicle speed on the longitudinal force characteristics.

[0161] In the embodiments of the present application, obtain the real-time vehicle speed through a speed sensor, and map the current speed to the corresponding correction coefficient value according to a preset speed-correction coefficient look-up table (established based on the vehicle dynamics model). This coefficient has a non-linear positive correlation with the vehicle speed.

[0162] Step 702: Decompose the track curvature parameter of the railway wagon into a straight - segment reference curvature and a curve - segment incremental curvature, and convert the curve - segment incremental curvature into a track curvature correction coefficient corresponding to the direction of the lateral eigenvector.

[0163] In step 702, the track curvature correction coefficient refers to a compensation parameter reflecting the influence degree of the curve on the lateral force characteristics. The specific generation process is as follows: Decompose the track curvature parameter of the railway wagon into a straight - segment reference curvature and a curve - segment incremental curvature; Generate a basic correction amount based on the straight - segment reference curvature; Generate a dynamic adjustment coefficient according to the relationship between the curve - segment incremental curvature and the track bending radius; Multiply the basic correction amount by the dynamic adjustment coefficient to generate a track curvature correction coefficient associated with the direction of the second fluctuation eigenvector. The track bending radius refers to the curvature radius of the track center line in the curve segment (unit: meter), which is a basic parameter in railway line design. There is a reciprocal conversion relationship between it and the track curvature parameter (unit: 1 / meter): Track curvature parameter = 1 / track bending radius.

[0164] In the embodiment of the present application, extract the curvature data of the current section from the line database, decompose it into a straight - line reference value and a curve - segment incremental value, and calculate the correction coefficient corresponding to the incremental curvature through the curvature radius conversion formula. This coefficient is in a direct - proportion relationship with the curvature increment.

[0165] Step 703: Correct the longitudinal eigenvector according to the running speed correction coefficient to generate a corrected longitudinal eigenvector.

[0166] In step 703, the corrected longitudinal eigenvector refers to a standardized abnormal feature after eliminating the influence of speed.

[0167] In the embodiment of the present application, divide each component value of the original longitudinal eigenvector by the running speed correction coefficient to obtain standardized feature data independent of speed.

[0168] Step 704: Correct the lateral eigenvector according to the running speed correction coefficient and the track curvature correction coefficient to generate a corrected lateral eigenvector.

[0169] In step 704, the corrected lateral eigenvector refers to a standardized abnormal feature after eliminating the combined influence of speed and curve.

[0170] In the embodiment of the present application, the original lateral feature vector is first corrected for speed (divided by the speed correction coefficient), and then corrected for curvature (multiplied by the curvature correction coefficient), and finally the feature data reflecting the true load distribution is obtained. Specifically, the left-right rail pressure difference in the lateral feature vector is multiplied by the running speed correction coefficient to amplify the unbalanced force feature under high-speed conditions (such as the original pressure difference of wheel set 1 +7 MPa × 1.25 coefficient → +8.75 MPa); identify the incremental curvature direction of the track curve section, if it is a right turn, multiply the left rail pressure difference by (1 + track curvature correction coefficient), multiply the right rail pressure difference by (1 - track curvature correction coefficient), and reverse the operation for a left turn; apply the elastic deformation limit constraint of the axle material to the corrected pressure difference (such as ±15 MPa), and compress the excess part proportionally to the safe range; regenerate the lateral feature vector in the format of [timestamp, corrected pressure difference, wheel set number] for the pressure difference of each wheel set after processing.

[0171] The following is a specific example:

[0172] During the operation of a four-axle freight car on a certain freight dedicated line, the system detects that the current running speed is 90% of the reference speed (directly obtained through the on-vehicle speed measurement device), and calculates the speed correction coefficient 0.81 according to the speed correction formula "correction coefficient = (actual speed / reference speed) squared". At the same time, obtain the current track curvature parameters from the line database, where the reference curvature of the straight section is zero, and the radius corresponding to the incremental curvature of the curve section is 80% of the reference radius. Calculate the curvature correction coefficient of 1.25 according to the curvature correction formula "correction coefficient = reference radius / actual radius". Perform speed correction on the extracted longitudinal feature vector value of 1.15 (calculated by the ratio of the pressure peak value at the No. 1 connection to the reference value), and obtain the corrected longitudinal feature vector of 1.42 (1.15 / 0.81); perform composite correction on the lateral feature vector value of 0.25 (calculated by the ratio of the left-right rail pressure difference of the No. 2 wheel set to the reference value), first divide by the speed correction coefficient 0.81 to get 0.31, and then multiply by the curvature correction coefficient 1.25 to finally obtain the corrected lateral feature vector of 0.39.

[0173] In the embodiment of the present application, this solution effectively eliminates the interference of speed and curve factors on the detection results through dual environmental parameter compensation, making the abnormal feature analysis results more objective and accurate, and providing a reliable basis for subsequent abnormal classification.

[0174] Figure 2 The following is a schematic structural diagram of an abnormal loading warning system for railway freight cars based on artificial intelligence provided by the embodiment of the present application, as Figure 2 shown, the system includes:

[0175] An acquisition module 21, configured to acquire the dynamic pressure distribution data of the carriage connection and multiple groups of wheel support areas during the operation of the railway freight car.

[0176] A generating module 22, configured to fuse the phase offset of the dynamic pressure distribution data at the carriage connection and the amplitude difference of the dynamic pressure distribution data in the multiple groups of wheel support areas, and generate the spatial pressure gradient characteristics of each carriage of the railway freight car.

[0177] An input module 23, configured to input the spatial pressure gradient characteristics and the pressure distribution base state of the railway freight car under historical normal conditions into a pre-trained residual network, and output the abnormal fluctuation component between the carriage connection and the multiple groups of wheel support areas.

[0178] A determining module 24, configured to determine the type of abnormal loading based on the abnormal fluctuation component, in combination with the current running speed and the track curvature parameter of the railway freight car, and issue a warning according to the warning method of the warning level to which the type of abnormal loading belongs.

[0179] Figure 2 The described railway freight car abnormal loading warning system based on artificial intelligence can execute Figure 1 The described railway freight car abnormal loading warning method based on artificial intelligence in the illustrated embodiment, and its implementation principle and technical effects will not be elaborated. For the railway freight car abnormal loading warning system based on artificial intelligence in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0180] In a possible design, Figure 2 The railway freight car abnormal loading warning system based on artificial intelligence in the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, the computing device may include a storage component 31 and a processing component 32;

[0181] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0182] The processing component 32 is the above Figure 1 The railway freight car abnormal loading warning method based on artificial intelligence in the illustrated embodiment.

[0183] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-mentioned method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0184] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0185] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0186] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0187] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0188] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0189] The embodiments of the present application also provide a computer storage medium storing a computer program, which can implement the above-mentioned Figure 1 anomaly warning method for railway wagon loading based on artificial intelligence in the embodiments shown.

[0190] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0191] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

[0192] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0193] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An abnormal warning method for railway wagon loading based on artificial intelligence, characterized in that, Including: Collecting the dynamic pressure distribution data of the carriage connection part and the multi-group wheel support areas during the operation of railway freight cars; Fusing the phase offset of the dynamic pressure distribution data of the carriage connection part and the amplitude difference of the dynamic pressure distribution data of the multi-group wheel support areas to generate the spatial pressure gradient characteristics of each carriage of the railway freight car; Inputting the spatial pressure gradient characteristics and the pressure distribution ground state of the railway freight car under historical normal working conditions into a pre-trained residual network, and outputting the abnormal fluctuation component between the carriage connection part and the multi-group wheel support areas; Based on the abnormal fluctuation component, combining the current running speed and the track curvature parameters of the railway freight car, determining the type of abnormal loading, and giving an early warning according to the early warning method of the early warning level to which the abnormal loading type belongs.

2. The method according to claim 1, wherein The step of inputting the spatial pressure gradient characteristics and the pressure distribution ground state of the railway freight car under historical normal working conditions into a pre-trained residual network and outputting the abnormal fluctuation component between the carriage connection part and the multi-group wheel support areas includes: Inputting the spatial pressure gradient characteristics and the pressure distribution ground state of the railway freight car under historical normal working conditions into a pre-trained residual network, and through the decomposition module in the pre-trained residual network, decomposing the spatial pressure gradient characteristics into a first balance parameter component parallel to the longitudinal axis direction of the carriage and a second balance parameter component perpendicular to the longitudinal axis direction of the carriage; Through the matching module in the pre-trained residual network, hierarchically matching the corresponding components of the pressure distribution ground state to generate a dynamic coupling difference amount; Through the fusion module in the pre-trained residual network, performing weighted fusion on the dynamic coupling difference amount and outputting the abnormal fluctuation component.

3. The method according to claim 2, wherein The step of hierarchically matching the corresponding components of the pressure distribution ground state through the matching module in the pre-trained residual network to generate a dynamic coupling difference amount includes: In the residual network, hierarchically decomposing the pressure distribution ground state of the railway freight car under historical normal working conditions through the matching module into a longitudinal ground state component corresponding to the first balance parameter component and a transverse ground state component corresponding to the second balance parameter component; Performing hierarchical matching on the first balance parameter component and the longitudinal ground state component to obtain an accumulated difference amount, and performing hierarchical matching on the second balance parameter component and the transverse ground state component to generate a transverse force imbalance compensation coefficient; Based on the accumulated difference amount and the transverse force imbalance compensation coefficient, generating a dynamic coupling difference amount.

4. The method according to claim 3, characterized in that The step of performing hierarchical matching on the first balance parameter component and the longitudinal ground state component to obtain an accumulated difference amount, and performing hierarchical matching on the second balance parameter component and the transverse ground state component to generate a transverse force imbalance compensation coefficient includes: Dividing the first balance parameter component into multiple equally spaced sections according to the carriage length, and each equally spaced section corresponds to a longitudinal ground state component reference value in the longitudinal ground state component; For each equally spaced section, calculating the instantaneous difference between the first balance parameter component corresponding to the equally spaced section and the corresponding longitudinal ground state component reference value; Recursively accumulate the instantaneous differences of all equidistant sections in the traveling direction of the carriage to generate a cumulative difference quantity; Divide the second balance parameter component into a left rail contact component and a right rail contact component according to the wheel set position; Calculate the left rail offset between the left rail contact component and the left rail reference value of the lateral ground state component, and the right rail offset between the right rail contact component and the right rail reference value of the lateral ground state component respectively; Calculate the lateral force imbalance compensation coefficient according to the reverse change characteristics of the left rail offset and the right rail offset of the same wheel set; 5. The method according to claim 3, characterized in that, Generating a dynamic coupling difference quantity based on the cumulative difference quantity and the lateral force imbalance compensation coefficient, including: Map the cumulative difference quantity to the lateral plane of the carriage according to the corresponding wheel set position; Adjust the distribution weight of the mapped cumulative difference quantity in the left and right rail directions according to the lateral force imbalance compensation coefficient; Integrate the adjusted cumulative difference quantities corresponding to all wheel set positions to generate a dynamic coupling difference quantity; 6. The method according to claim 1, characterized in that, Determining the type of abnormal loading based on the abnormal fluctuation component, combined with the current running speed of the railway freight car and the track curvature parameter, including: Decompose the abnormal fluctuation component into a longitudinal eigenvector and a lateral eigenvector; Based on the current running speed of the railway freight car and the track curvature parameter, correct the longitudinal eigenvector and the lateral eigenvector respectively; Determine the overload level code corresponding to the overload level interval according to the frequency distribution of the corrected longitudinal eigenvector; Determine the offloading direction code corresponding to the offloading direction interval according to the amplitude difference of the corrected lateral eigenvector; Match the type of abnormal loading corresponding to both the overload level code and the offloading direction code from a preset abnormal loading classification rule library; 7. The method according to claim 6, characterized in that, The correcting the longitudinal eigenvector and the lateral eigenvector respectively based on the current running speed of the railway freight car and the track curvature parameter includes: Convert the current running speed of the railway freight car into a running speed correction coefficient consistent with the direction of the longitudinal axis of the carriage; Decompose the track curvature parameter of the railway freight car into a straight section reference curvature and a curve section incremental curvature, and convert the curve section incremental curvature into a track curvature correction coefficient corresponding to the direction of the lateral eigenvector; Correct the longitudinal eigenvector according to the running speed correction coefficient to generate a corrected longitudinal eigenvector; Correct the lateral eigenvector according to the running speed correction coefficient and the track curvature correction coefficient to generate a corrected lateral eigenvector; 8. An abnormal loading warning system for railway freight cars based on artificial intelligence, characterized in that, Including: A collection module for collecting dynamic pressure distribution data at the carriage connection and multiple wheel support areas during the operation of the railway freight car; A generation module for fusing the phase offset of the dynamic pressure distribution data at the carriage connection and the amplitude difference of the dynamic pressure distribution data at the multiple wheel support areas to generate the spatial pressure gradient characteristics of each carriage of the railway freight car; An input module for inputting the spatial pressure gradient characteristics and the pressure distribution ground state of the railway freight car under historical normal conditions into a pre-trained residual network, and outputting the abnormal fluctuation component between the carriage connection and the multiple wheel support areas; A determination module, configured to determine a type of abnormal loading based on the abnormal fluctuation component, in combination with the current running speed of the railway freight car and the track curvature parameter, and issue a warning in accordance with the warning method of the warning level to which the type of abnormal loading belongs.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an abnormal loading warning method for railway freight cars based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements an abnormal loading warning method for railway freight cars based on artificial intelligence as described in any one of claims 1 to 7.

Images