Rack rail railway track state monitoring method based on Internet of Things

Abstract

The invention discloses a rack rail railway track state monitoring method based on the Internet of Things, which comprises the following steps: acquiring transverse, longitudinal and vertical displacement data and environmental parameter signals of a track and a rack, and determining a deviation type label; obtaining a corrected position deviation trend sequence based on the deviation type label, judging a meshing stability risk level, and obtaining a risk assessment score; fusing real-time monitoring data streams according to the risk assessment scores, determining a priority ranking list of high-risk sections, and if the risk score of the first section in the priority ranking list is higher than a warning line, integrating supplementary signals of the sensor array to obtain an enhanced deviation detail report; and updating the whole track state model by enhancing the deviation detail report, judging whether meshing stability returns to normal or not, and obtaining a final monitoring output result. According to the invention, the operation safety and maintenance efficiency of the rack rail railway are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of rail transit condition monitoring technology, and in particular to a method for monitoring the condition of a rack railway track based on the Internet of Things. Background Technology

[0002] As a special form of railway transportation, rack and pinion railways play an irreplaceable role in mountainous or steep terrain, and their safety and stability are directly related to transportation efficiency and personnel safety. Especially in complex terrain environments, real-time monitoring of track conditions becomes a crucial link in ensuring operational safety. Due to its unique rack and pinion meshing structure, rack and pinion railways require extremely high track precision; even the slightest deviation can lead to significant risks. Therefore, researching how to achieve accurate track condition monitoring through advanced technologies is of paramount practical significance.

[0003] However, current methods for monitoring the condition of rack and pinion railway tracks mostly rely on manual inspections or traditional fixed equipment, which are ill-suited to dynamic changes in complex environments. These methods often suffer from limited coverage and slow response times, especially when faced with long-term track stress or natural environmental influences, making it difficult to detect potential hazards in a timely manner. More importantly, existing technologies are insufficient in capturing subtle positional changes between the track and rack, resulting in untimely and inaccurate warnings of safety hazards.

[0004] In this field, the core technical challenge lies in accurately obtaining the relative positional relationship between the track and the rack. The unique characteristic of rack-and-pinion railways is that the positional deviation between the rack and the track bearing surface directly affects the stability of gear meshing. This deviation is often influenced by a combination of factors, such as deformation caused by foundation settlement or temperature changes. Because this positional relationship involves subtle changes in three dimensions—lateral, longitudinal, and vertical—extremely high precision is required for monitoring; otherwise, it is impossible to determine whether the gear and rack can mesh properly. A deeper problem is that these deviations manifest inconsistently in different sections and time periods. For example, in some steep slope sections, vertical deviation may be exacerbated by gravity, thus affecting overall operational safety.

[0005] Therefore, how to achieve comprehensive and real-time monitoring of the positional deviation between the rack and pinion rail in complex environments has become a key issue in ensuring transportation safety. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention proposes an Internet of Things-based method for monitoring the track condition of a rack railway, ensuring the real-time performance and accuracy of the monitoring.

[0007] To achieve the above objectives, the present invention provides a method for monitoring the track condition of a rack railway based on the Internet of Things, comprising:

[0008] The lateral, longitudinal, and vertical displacement data of the track and rack, as well as environmental parameter signals, are collected to obtain an initial position deviation dataset. Based on the initial position deviation dataset, a convolutional neural network is applied to extract feature vectors, and abnormal patterns caused by foundation settlement or temperature changes are classified to determine the deviation type label.

[0009] If the deviation type label indicates that the vertical deviation exceeds a preset threshold, a time series subset is separated from the feature vector to obtain a corrected position deviation trend sequence. By comparing the corrected position deviation trend sequence with historical monitoring records, the meshing stability risk level is determined and a risk assessment score is obtained.

[0010] Based on the risk assessment score and real-time monitoring data stream, a dynamic deviation map is generated, and a priority ranking list of high-risk sections is determined. If the risk score of the first section in the priority ranking list is higher than the warning line, the supplementary signal from the sensor array is integrated to obtain an enhanced deviation detail report.

[0011] The overall track state model is updated by the enhanced deviation detail report to determine whether meshing stability has returned to normal and to obtain the final monitoring output result.

[0012] Preferably, obtaining the initial position deviation dataset includes:

[0013] The sensor array is used to collect real-time data on track displacement and rack displacement along the railway line, covering lateral, longitudinal and vertical data, while recording environmental parameters to form a multi-dimensional initial dataset.

[0014] For the initial dataset, a preprocessing method is used to denoise the collected horizontal, vertical, and longitudinal data to generate a cleaned displacement dataset.

[0015] Key change features of track displacement and rack displacement are extracted from the cleaned displacement dataset. Combined with environmental parameters, the position deviation is classified and predicted using a support vector machine model to obtain the initial position deviation dataset.

[0016] Preferably, determining the deviation type label includes:

[0017] Location deviation-related information is obtained from the initial location deviation dataset, and a data extraction method is used to preliminarily screen the abnormal patterns caused by foundation settlement and temperature changes to obtain an abnormal pattern set.

[0018] For the set of abnormal patterns, a convolutional neural network is applied to extract feature vectors for each set of abnormal patterns, generating a corresponding set of feature vectors.

[0019] Based on the feature vector set, the abnormal patterns are distinguished by a classification processing method. If a feature vector set matches a ground settlement-related pattern, it is marked as a settlement deviation type; if it matches a temperature change-related pattern, it is marked as a temperature deviation type. The deviation type label is then determined.

[0020] Preferably, after determining the deviation type label, the method further includes:

[0021] For the aforementioned deviation type label, a set of feature vectors related to the settlement deviation type is obtained, and a preset threshold is used for comparison. If the threshold is exceeded, the corresponding set of feature vectors is sorted by priority to obtain a settlement deviation priority sequence.

[0022] Based on the settlement deviation priority sequence, high-priority deviation data are extracted, and combined with the feature vector group of temperature deviation type, comprehensive deviation distribution information is generated through data extraction methods to determine the concentration interval of the deviation distribution.

[0023] For the concentrated interval of the deviation distribution, the corresponding abnormal pattern data is obtained, and the deviation type labels within the interval are grouped by the pattern analysis method to obtain the grouped deviation pattern set.

[0024] Based on the grouped set of deviation patterns, a classification process is used to perform a secondary verification of each group of deviation patterns. If a group of deviation patterns is consistent with the position deviation information in the initial position deviation dataset, the corresponding data is retained, and the final deviation classification result is determined.

[0025] Preferably, obtaining the corrected position deviation trend sequence includes:

[0026] If the deviation type label indicates that the vertical deviation exceeds the preset vertical deviation threshold, the corresponding time series subset is separated from the feature vector group, the time series subset is initially processed to obtain the processed time series data, and Kalman filtering is used to smooth the noise interference in the processed time series data to generate a smoothed time series dataset.

[0027] Based on the smoothed time series dataset, position deviation-related information is extracted and matched with a preset vertical deviation threshold using a data comparison method to determine the trend direction of the position deviation.

[0028] If the trend direction of the position deviation shows a continuous deviation, the key time point information is separated from the trend direction data, and the key time points are classified using a data grouping method to determine the core interval of the deviation change. For the core interval of the deviation change, the time series data within the interval is obtained, which is the corrected position deviation trend sequence.

[0029] Preferably, obtaining the risk assessment score includes:

[0030] Data segments for key time periods are obtained from the corrected position deviation trend sequence. The data segments are classified and organized using a data filtering method to obtain a set of classified deviation data. The set of classified deviation data is compared with historical monitoring data segment by segment. If the comparison results show that the difference between the deviation data and the historical monitoring data exceeds a preset difference threshold, it is marked as an abnormal data segment, and the distribution range of the abnormal data segment is determined.

[0031] Based on the distribution range of the abnormal data fragments, feature information related to the meshing state is extracted, and a distribution map of state changes is generated through a data mapping method to obtain the key intervals of state changes.

[0032] For the key intervals of the state change, acquire monitoring data segments within the intervals. If the fluctuation range of the data segments within the intervals exceeds the preset fluctuation range, they are classified as potential risk data, and the concentrated area of ​​potential risk data is determined.

[0033] Based on the concentrated areas of the potential risk data, the status judgment criteria within the areas are extracted, and the regional data are classified using a data grouping method to determine the preliminary risk level classification results. For the preliminary risk level classification results, the classified data categories are obtained, and the risk assessment scores are obtained by comparing them with preset risk assessment standards.

[0034] Preferably, a dynamic deviation map is generated by integrating the risk assessment score with real-time monitoring data streams to determine a priority ranking list of high-risk areas, including:

[0035] The data flow is continuously updated from real-time monitoring. The data flow is segmented for key time points to obtain a segmented monitoring data set. Deviation analysis is performed on the segmented monitoring data set. If the deviation between the monitoring data and the benchmark value exceeds a preset deviation threshold, it is marked as an anomaly and the distribution range of the anomaly is determined.

[0036] Based on the distribution range of the anomalies, a distribution graph of dynamic deviation is constructed. By identifying the clustering areas of anomalies in the graph, the location information of high-risk segments is determined. For the location information of the high-risk segments, the evaluation scores within the corresponding segments are extracted. The segments are initially sorted according to their scores to obtain a preliminary priority sequence.

[0037] Based on the initial priority sequence and combined with real-time updated monitoring data, a secondary deviation analysis is performed on high-risk segments. If the deviation continues to increase in the secondary analysis, the priority level of the segments is adjusted to determine the priority ranking list of the high-risk segments.

[0038] Preferably, obtaining the enhanced deviation detail report includes:

[0039] If the risk score of the segment ranked first in the priority sorting list exceeds the preset warning threshold, a supplementary signal acquisition task is performed through the sensor array to obtain an expanded signal dataset.

[0040] Based on the extended signal dataset, a pre-established filtering mechanism is used to denoise the acquired supplementary signals to determine the cleaned signal dataset. Then, based on the cleaned signal dataset, a hierarchical classification operation is performed on the signal segments related to the risk score to obtain the classified signal group set.

[0041] Based on the classified signal group set, a logical comparison method is used. If the signal fluctuation in a certain group exceeds the preset fluctuation range, it is marked as a high-interest category, and a high-interest signal group is obtained.

[0042] By grouping the high-interest signals, the signal feature data corresponding to the segment ranking is extracted. Combined with the analysis requirements of deviation details, a targeted deviation feature record is generated. Based on the deviation feature record, the real-time signal updates collected in the sensor array are integrated to construct the structured content of the enhanced report, determine the final deviation detail description, and obtain the enhanced deviation detail report.

[0043] Preferably, obtaining the final monitoring output result includes:

[0044] By combining real-time track status data with information from the enhanced deviation detail report, a preliminary data comparison is performed to obtain the basic assessment value of meshing stability. Based on the basic assessment value of meshing stability, a state model is used for in-depth analysis, and the changes are processed in layers to obtain the dynamic distribution of stability values.

[0045] By dynamically distributing stability values ​​and combining them with details of deviations, if the stability value of a certain area is lower than the preset stability threshold, targeted data collection is triggered to determine the detailed records of the abnormal area.

[0046] Based on the detailed records of the abnormal areas, a pre-established classification mechanism is used to structure and organize the report content to obtain a set of classified abnormal fragments.

[0047] By combining the categorized set of abnormal segments with the logic of model analysis, if the fluctuation of a certain abnormal segment exceeds the preset fluctuation range, it is marked as a key focus object, thus obtaining the key focus segment.

[0048] Based on the key monitoring segments and combined with the detailed update mechanism, a targeted monitoring and adjustment strategy is generated to determine the final monitoring results.

[0049] Compared with the prior art, the present invention has the following advantages and technical effects:

[0050] This invention collects multidimensional displacement and environmental data through a sensor array along the track, extracts features and classifies deviation types using a convolutional neural network, and combines this with Kalman filtering to correct noise interference, generating a precise position deviation trend sequence. After comparison with historical data, the risk level is assessed, and a dynamic deviation map and priority ranking list are generated. For high-risk sections, this invention integrates supplementary signals to update the track state model, ultimately determining the meshing stability. The core innovation of this invention lies in the multi-source data fusion and dynamic risk assessment mechanism, ensuring the real-time performance and accuracy of monitoring, effectively improving the operational safety and maintenance efficiency of rack and pinion railways, and providing reliable technical support for the stable operation of track systems in complex environments. Attached Figure Description

[0051] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0052] Figure 1 This is a flowchart of a method for monitoring the condition of a rack railway track based on the Internet of Things, according to an embodiment of the present invention. Detailed Implementation

[0053] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0054] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0055] This embodiment proposes a method for monitoring the track condition of a rack railway based on the Internet of Things (IoT), such as... Figure 1 ,include:

[0056] The lateral, longitudinal, and vertical displacement data of the track and rack, as well as environmental parameter signals, are collected to obtain an initial position deviation dataset. Based on the initial position deviation dataset, a convolutional neural network is applied to extract feature vectors, and abnormal patterns caused by foundation settlement or temperature changes are classified to determine the deviation type label.

[0057] If the deviation type label indicates that the vertical deviation exceeds a preset threshold, a time series subset is separated from the feature vector to obtain a corrected position deviation trend sequence. By comparing the corrected position deviation trend sequence with historical monitoring records, the meshing stability risk level is determined and a risk assessment score is obtained.

[0058] Based on the risk assessment score and real-time monitoring data stream, a dynamic deviation map is generated, and a priority ranking list of high-risk sections is determined. If the risk score of the first section in the priority ranking list is higher than the warning line, the supplementary signal from the sensor array is integrated to obtain an enhanced deviation detail report.

[0059] The overall track state model is updated by the enhanced deviation detail report to determine whether meshing stability has returned to normal and to obtain the final monitoring output result.

[0060] Further, obtaining the initial position deviation dataset includes:

[0061] The sensor array is used to collect real-time data on track displacement and rack displacement along the railway line, covering lateral, longitudinal and vertical data, while recording environmental parameters to form a multi-dimensional initial dataset.

[0062] For the initial dataset, a preprocessing method is used to denoise the collected horizontal, vertical, and longitudinal data to generate a cleaned displacement dataset.

[0063] Key change features of track displacement and rack displacement are extracted from the cleaned displacement dataset. Combined with environmental parameters, the position deviation is classified and predicted using a support vector machine model to obtain the initial position deviation dataset.

[0064] Specifically, in the scenario of displacement monitoring along railway lines, the deployment of sensor arrays is fundamental to the entire process. Sensors can be installed on key nodes of the track and rack to collect real-time displacement data in the lateral, longitudinal, and vertical directions, while simultaneously recording environmental parameters such as temperature and humidity. Assuming 100 sensor nodes are deployed along a section of railway, collecting data every 5 minutes, each node generates data on lateral displacement of 0.2 mm, longitudinal displacement of 0.1 mm, and vertical displacement of 0.3 mm, and records the current temperature as 25 degrees Celsius and humidity as 60%. This data forms the initial dataset, providing multi-dimensional information for subsequent analysis.

[0065] Noise removal is crucial in the data preprocessing stage. Raw data may be affected by vibration or electromagnetic interference, leading to some numerical anomalies. Noise can be removed through filtering methods, such as smoothing outliers in the lateral data that significantly deviate from the average. After processing, for example, adjusting the lateral displacement data of a node from 0.5 mm to 0.2 mm better reflects the actual trend. This cleaning process improves data quality and ensures the accuracy of subsequent analysis.

[0066] In the feature extraction and classification prediction stages, key change features, such as displacement rate or periodic fluctuations, are extracted from the cleaned dataset and combined with environmental parameters to classify the deviations using a support vector machine model. For example, assuming that the lateral displacement rate of the track increases to 0.3 mm / hour in a certain section when the temperature rises to 30 degrees Celsius, the model predicts that this deviation belongs to the moderate risk category. This result helps to accurately locate the problem area and reduce maintenance costs caused by misjudgments.

[0067] Further, determining the deviation type label includes:

[0068] Location deviation-related information is obtained from the initial location deviation dataset, and a data extraction method is used to preliminarily screen the abnormal patterns caused by foundation settlement and temperature changes to obtain an abnormal pattern set.

[0069] For the set of abnormal patterns, a convolutional neural network is applied to extract feature vectors for each set of abnormal patterns, generating a corresponding set of feature vectors.

[0070] Based on the feature vector set, the abnormal patterns are distinguished by a classification processing method. If a feature vector set matches a ground settlement-related pattern, it is marked as a settlement deviation type; if it matches a temperature change-related pattern, it is marked as a temperature deviation type. The deviation type label is then determined.

[0071] Specifically, in the field of positional deviation monitoring along rack railway lines, the processing of the initial dataset is a crucial starting point for the entire analysis process. For anomalies caused by ground settlement and temperature changes, preliminary screening can be performed using data extraction methods. For example, if data collected along a railway line shows that a certain area experienced a vertical displacement of 0.5 mm within a specific time period, while the temperature change was 5 degrees Celsius during the same period, this combination of anomalies is preliminarily classified as a potential anomaly pattern, forming an anomaly pattern set.

[0072] When using convolutional neural networks to extract feature vectors from a set of abnormal patterns, the focus can be on the fluctuation frequency and amplitude changes of displacement data. For example, if a set of data shows a fluctuation frequency of twice per hour with consistently high amplitude, this pattern can be extracted as a feature vector set, reflecting potential ground settlement or temperature-related characteristics.

[0073] When classifying feature vector groups, if the fluctuation pattern of a certain feature vector group highly matches the known foundation settlement feature database, such as a vertical displacement rate of 0.3 mm per hour, it is marked as a settlement deviation type. Conversely, if another set of data shows a positive correlation between displacement changes and temperature increases, such as a lateral displacement increase of 0.1 mm for every 1 degree Celsius increase in temperature, it is marked as a temperature deviation type. This distinction is helpful for subsequent targeted analysis.

[0074] Furthermore, after determining the deviation type label, the process also includes:

[0075] For the aforementioned deviation type label, a set of feature vectors related to the settlement deviation type is obtained, and a preset threshold is used for comparison. If the threshold is exceeded, the corresponding set of feature vectors is sorted by priority to obtain a settlement deviation priority sequence.

[0076] Based on the settlement deviation priority sequence, high-priority deviation data are extracted, and combined with the feature vector group of temperature deviation type, comprehensive deviation distribution information is generated through data extraction methods to determine the concentration interval of the deviation distribution.

[0077] For the concentrated interval of the deviation distribution, the corresponding abnormal pattern data is obtained, and the deviation type labels within the interval are grouped by the pattern analysis method to obtain the grouped deviation pattern set.

[0078] Based on the grouped set of deviation patterns, a classification process is used to perform a secondary verification of each group of deviation patterns. If a group of deviation patterns is consistent with the position deviation information in the initial position deviation dataset, the corresponding data is retained, and the final deviation classification result is determined.

[0079] Specifically, for feature vector groups of settlement deviation types, when setting a threshold for comparison, assuming the threshold is 0.4 mm / hour, if the displacement rate of a certain area reaches 0.6 mm / hour, it is listed as high priority, forming a settlement deviation priority sequence. This sorting method can quickly identify key areas of concern.

[0080] When generating comprehensive deviation distribution information, by combining high-priority settlement deviation data and feature vector groups of temperature deviation types, if it is assumed that settlement deviations are concentrated in the middle section of the railway within a certain interval, and temperature deviation data also shows that the temperature fluctuations in this area are large, then this interval is determined to be a region with concentrated deviation distribution. This comprehensive analysis can reveal the problem distribution more comprehensively.

[0081] When grouping deviation type labels within a concentrated deviation distribution range using pattern analysis, if a certain range contains both settlement and temperature deviations, it is divided into two pattern sets. This grouping refines the problem classification and facilitates subsequent processing.

[0082] When performing secondary verification on the grouped set of deviation patterns, if a certain group of settlement deviation patterns is consistent with the positional deviation information in the initial data, such as the vertical displacement data of a certain node consistently being 0.5 mm, then that group of data is retained as the final classification result. This secondary verification ensures the reliability of the data and provides an accurate basis for subsequent maintenance.

[0083] Further, the corrected position deviation trend sequence is obtained, including:

[0084] If the deviation type label indicates that the vertical deviation exceeds the preset vertical deviation threshold, the corresponding time series subset is separated from the feature vector group, the time series subset is initially processed to obtain the processed time series data, and Kalman filtering is used to smooth the noise interference in the processed time series data to generate a smoothed time series dataset.

[0085] Based on the smoothed time series dataset, position deviation-related information is extracted and matched with a preset vertical deviation threshold using a data comparison method to determine the trend direction of the position deviation.

[0086] If the trend direction of the position deviation shows a continuous deviation, the key time point information is separated from the trend direction data, and the key time points are classified using a data grouping method to determine the core interval of the deviation change. For the core interval of the deviation change, the time series data within the interval is obtained, which is the corrected position deviation trend sequence.

[0087] Specifically, in the field of positional deviation monitoring along rack railways, for scenarios where vertical deviation exceeds a preset vertical deviation threshold, the corresponding time series subset can be separated from the feature vector set. For example, if the vertical displacement data of a certain railway segment exceeds the threshold by 0.4 mm over a period of time, the data for that time period is extracted, and irrelevant interference items are removed using data filtering methods to obtain clear time series data, laying the foundation for subsequent analysis.

[0088] When using Kalman filtering to smooth noise in processed time-series data, we can assume that a set of data contains minute fluctuations caused by sensor jitter. After smoothing, these fluctuations are effectively reduced, and the resulting dataset better reflects the true displacement trend. This smoothing process helps improve the accuracy of data analysis.

[0089] When extracting positional deviation information from a smoothed time-series dataset, a data comparison method can be used to match the data with a preset threshold of 0.3 millimeters per hour. If the displacement rate of a certain data segment consistently exceeds this value, its trend direction is determined to be a continuous deviation. This judgment method provides a basis for subsequently locating the problem area.

[0090] For data showing a persistent trend deviation, when separating key time point information, the points in time when the displacement rate of a certain railway segment suddenly increases within a day can be marked. These time points can then be categorized by hour using data grouping methods to determine the core interval of the deviation change. This classification method helps to focus on the specific time period in which the problem occurred.

[0091] When generating a deviation distribution map within the core range of deviation changes, a data mapping method can be used to visualize the hourly displacement data within a certain range as a distribution map. For example, if the displacement data in a certain area is found to be concentrated in the northern section of the railway, then that area is marked as a region with concentrated deviation distribution. Generating such a distribution map facilitates intuitive identification of the problem location.

[0092] Further, obtaining the risk assessment score includes:

[0093] Data segments for key time periods are obtained from the corrected position deviation trend sequence. The data segments are classified and organized using a data filtering method to obtain a set of classified deviation data. The set of classified deviation data is compared with historical monitoring data segment by segment. If the comparison results show that the difference between the deviation data and the historical monitoring data exceeds a preset difference threshold, it is marked as an abnormal data segment, and the distribution range of the abnormal data segment is determined.

[0094] Based on the distribution range of the abnormal data fragments, feature information related to the meshing state is extracted, and a distribution map of state changes is generated through a data mapping method to obtain the key intervals of state changes.

[0095] For the key intervals of the state change, acquire monitoring data segments within the intervals. If the fluctuation range of the data segments within the intervals exceeds the preset fluctuation range, they are classified as potential risk data, and the concentrated area of ​​potential risk data is determined.

[0096] Based on the concentrated areas of the potential risk data, the status judgment criteria within the areas are extracted, and the regional data are classified using a data grouping method to determine the preliminary risk level classification results. For the preliminary risk level classification results, the classified data categories are obtained, and the risk assessment scores are obtained by comparing them with preset risk assessment standards.

[0097] Specifically, in the field of position deviation monitoring along rack railways, data segments for key time periods can be extracted from the corrected position deviation trend sequence using data filtering methods. For example, if the vertical displacement data of a certain railway line shows significant fluctuations in a particular hour of a day, then the data segment for that hour is extracted, categorized and organized according to displacement rate and direction, forming a preliminary deviation data set to provide a foundation for subsequent analysis.

[0098] For the categorized deviation data set, when comparing it segment by segment with historical monitoring data, a preset threshold can be set, such as a displacement rate difference of 0.5 mm per hour. If the displacement rate of a certain data segment exceeds this threshold compared with historical data, it is marked as an abnormal data segment, and its distribution range is further analyzed, such as determining whether the abnormal data is concentrated in a specific section of the railway. This marking and range determination helps to quickly locate potential problem areas.

[0099] After extracting the distribution range of abnormal data segments, a state change distribution map can be generated using data mapping methods based on the feature information related to the meshing state. For example, if the meshing state data of a certain railway segment is found to change frequently within a certain time interval, this interval is marked as the critical interval for state change. Generating such a distribution map facilitates a visual understanding of the concentrated periods and areas of state change.

[0100] For monitoring data segments within critical intervals of state change, if the fluctuation range of a certain data segment is found to exceed a preset range, such as a displacement rate fluctuation exceeding 0.4 mm per hour, it is classified as potential risk data, and the concentration area of ​​this data is further determined, such as whether it is concentrated at a certain end of the railway. This classification and regional determination provides an important basis for subsequent risk assessment.

[0101] When extracting criteria for assessing the state of potential risks within concentrated areas, data grouping methods can be used to classify the data by hourly intervals or by segments to initially determine risk levels. For example, if the displacement rate in a certain area consistently exceeds 0.3 mm per hour, it can be preliminarily classified as a higher risk level. This classification method helps to clarify the priority of risks.

[0102] When comparing the preliminary risk level classification with the preset risk assessment standards, an evaluation scoring system can be established, such as a maximum score of 100 points, with 5 points deducted for every 0.1 mm / hour increase in displacement rate. If the data for a certain area ultimately scores 70 points, it indicates that its risk level requires further attention. This scoring assessment method provides a quantitative reference for the final decision.

[0103] Throughout the entire process, from data segment classification and anomaly labeling to risk level grading, everything revolves around the core business of rack railway position deviation monitoring, ensuring the relevance and practicality of the analysis. This multi-step, multi-dimensional analysis method effectively improves the utilization efficiency of monitoring data and provides reliable support for railway maintenance.

[0104] Furthermore, based on the risk assessment scores and the real-time monitoring data stream, a dynamic deviation map is generated to determine a priority ranking list of high-risk areas, including:

[0105] The data flow is continuously updated from real-time monitoring. The data flow is segmented for key time points to obtain a segmented monitoring data set. Deviation analysis is performed on the segmented monitoring data set. If the deviation between the monitoring data and the benchmark value exceeds a preset deviation threshold, it is marked as an anomaly and the distribution range of the anomaly is determined.

[0106] Based on the distribution range of the anomalies, a distribution graph of dynamic deviation is constructed. By identifying the clustering areas of anomalies in the graph, the location information of high-risk segments is determined. For the location information of the high-risk segments, the evaluation scores within the corresponding segments are extracted. The segments are initially sorted according to their scores to obtain a preliminary priority sequence.

[0107] Based on the initial priority sequence and combined with real-time updated monitoring data, a secondary deviation analysis is performed on high-risk segments. If the deviation continues to increase in the secondary analysis, the priority level of the segments is adjusted to determine the priority ranking list of the high-risk segments.

[0108] Specifically, when performing deviation analysis on the segmented monitoring data set, a baseline value can be set, such as a normal range of 0.2 mm per hour for vertical displacement. If the average displacement of the data in a certain hour reaches 0.5 mm per hour, exceeding the preset threshold of 0.3 mm per hour, it is marked as an anomaly. Further determination is made as to whether these anomalies are concentrated in a few hours, forming a preliminary judgment on the distribution range.

[0109] When constructing a dynamic deviation distribution graph, the clustering areas of outliers can be visualized. For example, if the graph reveals a dense concentration of data points and significant deviations in a railway section between 2 AM and 4 AM, the corresponding railway section can be identified as a high-risk section, and its location information can be recorded to provide a basis for subsequent processing.

[0110] When extracting assessment scores from location information for high-risk sections, a simple scoring mechanism can be designed. Assuming a maximum score of 100 points, if the displacement deviation of a section is 0.4 mm per hour, 20 points are deducted, resulting in a score of 80. Multiple sections are then ranked according to their scores to form a preliminary priority sequence, facilitating resource allocation.

[0111] When performing secondary deviation analysis using real-time data, if the displacement deviation of a high-risk section increases from 0.4 mm / h to 0.6 mm / h, its priority level needs to be raised. The updated sorting list will place this section at the top to ensure it receives priority attention.

[0112] Furthermore, the enhanced deviation detail report is obtained, including:

[0113] If the risk score of the segment ranked first in the priority sorting list exceeds the preset warning threshold, a supplementary signal acquisition task is performed through the sensor array to obtain an expanded signal dataset.

[0114] Based on the extended signal dataset, a pre-established filtering mechanism is used to denoise the acquired supplementary signals to determine the cleaned signal dataset. Then, based on the cleaned signal dataset, a hierarchical classification operation is performed on the signal segments related to the risk score to obtain the classified signal group set.

[0115] Based on the classified signal group set, a logical comparison method is used. If the signal fluctuation in a certain group exceeds the preset fluctuation range, it is marked as a high-interest category, and a high-interest signal group is obtained.

[0116] By grouping the high-interest signals, the signal feature data corresponding to the segment ranking is extracted. Combined with the analysis requirements of deviation details, a targeted deviation feature record is generated. Based on the deviation feature record, the real-time signal updates collected in the sensor array are integrated to construct the structured content of the enhanced report, determine the final deviation detail description, and obtain the enhanced deviation detail report.

[0117] Specifically, for the section ranked first in the priority list, if its risk score exceeds a preset warning threshold, an activation command can be triggered. The purpose of this command is to collect supplementary signals through a sensor array to obtain more comprehensive data support. Suppose a railway section has a risk score of 85 points, while the preset warning threshold is 80 points, the system will automatically activate the sensor array to collect multi-dimensional signals such as vibration and displacement of that section within a specific time period, forming an expanded signal dataset.

[0118] For expanded signal datasets, noise reduction can be performed using pre-established filtering mechanisms. If the acquired signals contain environmental noise or equipment interference, the filtering mechanism will remove irrelevant information based on signal frequency characteristics, ensuring data accuracy. For example, if a acquired signal contains minor vibrations caused by wind, these will be identified and removed, resulting in a cleaned signal dataset that lays the foundation for subsequent analysis.

[0119] When classifying the cleaned signal dataset, the data can be categorized into groups such as displacement signals and vibration signals based on signal type and risk relevance. For example, if the fluctuation range of a certain displacement signal exceeds a preset value of 0.3 mm / hour, reaching 0.5 mm / hour, it is marked as a high-concern category. This classification method helps to focus on key problem areas and improves analysis efficiency.

[0120] For high-interest signal groups, signal feature data corresponding to the segment ranking can be extracted. Assuming the displacement signal of the top segment shows continuous anomalies, feature records containing time, location, and deviation amplitude can be generated based on the required deviation details. These records provide detailed information for subsequent analysis, helping to accurately pinpoint the problem.

[0121] When building enhanced reports, real-time signal updates from sensor arrays can be integrated, transforming feature records into structured content. For example, the report could detail the displacement deviation of a certain section within a specific hour as 0.4 millimeters per hour, and combine this with historical trends to generate a final detailed description of the deviation. This approach ensures both comprehensiveness and readability of the information.

[0122] Based on the detailed description of the final deviation, the risk score assessment criteria for the first segment can be updated. Assuming the original risk score was 85, it can be adjusted to 90 after analyzing the new data, generating an updated risk status record. This dynamic adjustment mechanism helps to reflect the true status of the segment in a timely manner, providing support for subsequent decision-making.

[0123] Throughout the process, each step is closely aligned with the core requirement of rack and pinion railway deviation monitoring, forming a complete analytical chain from data acquisition and feature extraction to report generation. This approach not only improves the accuracy of risk identification but also provides a reliable basis for resource allocation and emergency response.

[0124] Furthermore, obtaining the final monitoring output results includes:

[0125] By combining real-time track status data with information from the enhanced deviation detail report, a preliminary data comparison is performed to obtain the basic assessment value of meshing stability. Based on the basic assessment value of meshing stability, a state model is used for in-depth analysis, and the changes are processed in layers to obtain the dynamic distribution of stability values.

[0126] By dynamically distributing stability values ​​and combining them with details of deviations, if the stability value of a certain area is lower than the preset stability threshold, targeted data collection is triggered to determine the detailed records of the abnormal area.

[0127] Based on the detailed records of the abnormal areas, a pre-established classification mechanism is used to structure and organize the report content to obtain a set of classified abnormal fragments.

[0128] By combining the categorized set of abnormal segments with the logic of model analysis, if the fluctuation of a certain abnormal segment exceeds the preset fluctuation range, it is marked as a key focus object, thus obtaining the key focus segment.

[0129] Based on the key monitoring segments and combined with the detailed update mechanism, a targeted monitoring and adjustment strategy is generated to determine the final monitoring results.

[0130] Specifically, by combining real-time data on track status with enhanced report information, a basic assessment value for meshing stability can be obtained through multi-dimensional comparison.

[0131] Suppose that real-time data for a certain section shows a contact pressure value of 3.5 units at the track engagement point, while the historical average recorded in the reinforcement report is 3.2 units. By comparing these figures, a preliminary assessment can be made that there is a slight deviation in the current engagement stability, with a baseline assessment value set at 75 points, which is lower than the ideal value of 80 points. This comparison method helps to quickly identify potential problem areas.

[0132] When using a state model for in-depth analysis of the basic assessment values ​​for meshing stability, the data can be stratified by time period and segment location. Assuming that the assessment value for a certain segment fluctuates between 70 and 78 points over the past 24 hours, the state model will divide it into a short-term fluctuation layer and a long-term trend layer, obtaining a dynamic distribution of stability values. This distribution shows that a sudden drop in stability value to 72 points within a certain hour suggests a possible local anomaly, providing direction for subsequent analysis.

[0133] By combining the dynamic distribution of stability values ​​and detailed deviation information, if the stability value of a certain area falls below a preset threshold of 75 points, targeted data collection is triggered. For example, if a segment has a stability value of 71 points, the system will automatically activate a high-frequency sensor to record minute displacement changes in that area over the next 6 hours, generating a detailed record. This record includes time points and specific deviation values, providing a basis for anomaly localization.

[0134] When structuring the report content using a classification mechanism for detailed records of abnormal areas, the abnormal data can be categorized into two types: displacement anomalies and pressure anomalies. For example, if a certain section's records show a displacement deviation of 0.4 mm / hour and a pressure deviation of 0.3 units, the resulting classification forms a set of abnormal segments, facilitating subsequent focused analysis.

[0135] If the fluctuation of a certain abnormal segment exceeds the preset fluctuation range, such as a displacement deviation exceeding 0.35 mm per hour, it is marked as a key focus area. For example, if a segment's value is 0.42 mm per hour, the system marks it as a key focus segment and analyzes it from multiple perspectives. For instance, combining historical data reveals that the value has been rising continuously over the past week, or considering environmental factors, it suggests that recent rainfall may have played a role. This multi-faceted analysis ensures comprehensive monitoring.

[0136] For segments of particular interest, when generating monitoring adjustment strategies using a detailed update mechanism, the monitoring frequency for that segment can be increased to once per hour, and the sensor sensitivity can be adjusted. Assuming the final monitoring output shows that the deviation value for that segment is stable at 0.41 mm per hour, the adjusted strategy can capture changes more promptly.

[0137] By continuously tracking the final monitoring results and changes, when constructing a dynamic feedback dataset, it is possible to record whether the stability trend of a certain section changes from declining to stable over the past 48 hours. This trend judgment provides an important reference for subsequent maintenance plans, demonstrating the value of dynamic feedback.

[0138] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring the condition of a rack railway track based on the Internet of Things, characterized in that, include: The lateral, longitudinal, and vertical displacement data of the track and rack, as well as environmental parameter signals, are collected to obtain an initial position deviation dataset. Based on the initial position deviation dataset, a convolutional neural network is applied to extract feature vectors, and abnormal patterns caused by foundation settlement or temperature changes are classified to determine the deviation type label. If the deviation type label indicates that the vertical deviation exceeds a preset threshold, a time series subset is separated from the feature vector to obtain a corrected position deviation trend sequence. By comparing the corrected position deviation trend sequence with historical monitoring records, the meshing stability risk level is determined and a risk assessment score is obtained. Based on the risk assessment score and real-time monitoring data stream, a dynamic deviation map is generated, and a priority ranking list of high-risk sections is determined. If the risk score of the first section in the priority ranking list is higher than the warning line, the supplementary signal from the sensor array is integrated to obtain an enhanced deviation detail report. The overall track state model is updated by the enhanced deviation detail report to determine whether meshing stability has returned to normal and to obtain the final monitoring output result.

2. The method for monitoring the track condition of a rack railway based on the Internet of Things according to claim 1, characterized in that, Obtaining the initial position deviation dataset includes: The sensor array is used to collect real-time data on track displacement and rack displacement along the railway line, covering lateral, longitudinal and vertical data, while recording environmental parameters to form a multi-dimensional initial dataset. For the initial dataset, a preprocessing method is used to denoise the collected horizontal, vertical, and longitudinal data to generate a cleaned displacement dataset. Key change features of track displacement and rack displacement are extracted from the cleaned displacement dataset. Combined with environmental parameters, the position deviation is classified and predicted using a support vector machine model to obtain the initial position deviation dataset.

3. The method for monitoring the track condition of a rack railway based on the Internet of Things according to claim 2, characterized in that, Determining the deviation type label includes: Location deviation-related information is obtained from the initial location deviation dataset, and a data extraction method is used to preliminarily screen the abnormal patterns caused by foundation settlement and temperature changes to obtain an abnormal pattern set. For the set of abnormal patterns, a convolutional neural network is applied to extract feature vectors for each set of abnormal patterns, generating a corresponding set of feature vectors. Based on the feature vector set, the abnormal patterns are distinguished by a classification processing method. If a feature vector set matches a ground settlement-related pattern, it is marked as a settlement deviation type; if it matches a temperature change-related pattern, it is marked as a temperature deviation type. The deviation type label is then determined.

4. The method for monitoring the track condition of a rack railway based on the Internet of Things according to claim 3, characterized in that, After determining the deviation type label, the following is also included: For the aforementioned deviation type label, a set of feature vectors related to the settlement deviation type is obtained, and a preset threshold is used for comparison. If the threshold is exceeded, the corresponding set of feature vectors is sorted by priority to obtain a settlement deviation priority sequence. Based on the settlement deviation priority sequence, high-priority deviation data are extracted, and combined with the feature vector group of temperature deviation type, comprehensive deviation distribution information is generated through data extraction methods to determine the concentration interval of the deviation distribution. For the concentrated interval of the deviation distribution, the corresponding abnormal pattern data is obtained, and the deviation type labels within the interval are grouped by the pattern analysis method to obtain the grouped deviation pattern set. Based on the grouped set of deviation patterns, a classification process is used to perform a secondary verification of each group of deviation patterns. If a group of deviation patterns is consistent with the position deviation information in the initial position deviation dataset, the corresponding data is retained, and the final deviation classification result is determined.

5. The method for monitoring the condition of a rack railway track based on the Internet of Things according to claim 1, characterized in that, The corrected position deviation trend sequence is obtained by: If the deviation type label indicates that the vertical deviation exceeds the preset vertical deviation threshold, the corresponding time series subset is separated from the feature vector group, the time series subset is initially processed to obtain the processed time series data, and Kalman filtering is used to smooth the noise interference in the processed time series data to generate a smoothed time series dataset. Based on the smoothed time series dataset, position deviation-related information is extracted and matched with a preset vertical deviation threshold using a data comparison method to determine the trend direction of the position deviation. If the trend direction of the position deviation shows a continuous deviation, the key time point information is separated from the trend direction data, and the key time points are classified using a data grouping method to determine the core interval of the deviation change. For the core interval of the deviation change, the time series data within the interval is obtained, which is the corrected position deviation trend sequence.

6. The method for monitoring the condition of a rack railway track based on the Internet of Things according to claim 1, characterized in that, To obtain the risk assessment score, the following steps are taken: Data segments for key time periods are obtained from the corrected position deviation trend sequence. The data segments are classified and organized using a data filtering method to obtain a set of classified deviation data. The set of classified deviation data is compared with historical monitoring data segment by segment. If the comparison results show that the difference between the deviation data and the historical monitoring data exceeds a preset difference threshold, it is marked as an abnormal data segment, and the distribution range of the abnormal data segment is determined. Based on the distribution range of the abnormal data fragments, feature information related to the meshing state is extracted, and a distribution map of state changes is generated through a data mapping method to obtain the key intervals of state changes. For the key intervals of the state change, acquire monitoring data segments within the intervals. If the fluctuation range of the data segments within the intervals exceeds the preset fluctuation range, they are classified as potential risk data, and the concentrated area of ​​potential risk data is determined. Based on the concentrated areas of the potential risk data, the status judgment criteria within the areas are extracted, and the regional data are classified using a data grouping method to determine the preliminary risk level classification results. For the preliminary risk level classification results, the classified data categories are obtained, and the risk assessment scores are obtained by comparing them with preset risk assessment standards.

7. The method for monitoring the track condition of a rack railway based on the Internet of Things according to claim 1, characterized in that, Based on the risk assessment scores and real-time monitoring data streams, a dynamic deviation map is generated to determine a priority ranking list of high-risk areas, including: The data flow is continuously updated from real-time monitoring. The data flow is segmented for key time points to obtain a segmented monitoring data set. Deviation analysis is performed on the segmented monitoring data set. If the deviation between the monitoring data and the benchmark value exceeds a preset deviation threshold, it is marked as an anomaly and the distribution range of the anomaly is determined. Based on the distribution range of the anomalies, a distribution graph of dynamic deviation is constructed. By identifying the clustering areas of anomalies in the graph, the location information of high-risk segments is determined. For the location information of the high-risk segments, the evaluation scores within the corresponding segments are extracted. The segments are initially sorted according to their scores to obtain a preliminary priority sequence. Based on the initial priority sequence and combined with real-time updated monitoring data, a secondary deviation analysis is performed on high-risk segments. If the deviation continues to increase in the secondary analysis, the priority level of the segments is adjusted to determine the priority ranking list of the high-risk segments.

8. The method for monitoring the condition of a rack railway track based on the Internet of Things according to claim 1, characterized in that, The enhanced deviation details report includes: If the risk score of the segment ranked first in the priority sorting list exceeds the preset warning threshold, a supplementary signal acquisition task is performed through the sensor array to obtain an expanded signal dataset. Based on the extended signal dataset, a pre-established filtering mechanism is used to denoise the acquired supplementary signals to determine the cleaned signal dataset. Then, based on the cleaned signal dataset, a hierarchical classification operation is performed on the signal segments related to the risk score to obtain the classified signal group set. Based on the classified signal group set, a logical comparison method is used. If the signal fluctuation in a certain group exceeds the preset fluctuation range, it is marked as a high-interest category, and a high-interest signal group is obtained. By grouping the high-interest signals, the signal feature data corresponding to the segment ranking is extracted. Combined with the analysis requirements of deviation details, a targeted deviation feature record is generated. Based on the deviation feature record, the real-time signal updates collected in the sensor array are integrated to construct the structured content of the enhanced report, determine the final deviation detail description, and obtain the enhanced deviation detail report.

9. The method for monitoring the track condition of a rack railway based on the Internet of Things according to claim 1, characterized in that, The final monitoring output results are obtained, including: By combining real-time track status data with information from the enhanced deviation detail report, a preliminary data comparison is performed to obtain the basic assessment value of meshing stability. Based on the basic assessment value of meshing stability, a state model is used for in-depth analysis, and the changes are processed in layers to obtain the dynamic distribution of stability values. By dynamically distributing stability values ​​and combining them with details of deviations, if the stability value of a certain area is lower than the preset stability threshold, targeted data collection is triggered to determine the detailed records of the abnormal area. Based on the detailed records of the abnormal areas, a pre-established classification mechanism is used to structure and organize the report content to obtain a set of classified abnormal fragments. By combining the categorized set of abnormal segments with the logic of model analysis, if the fluctuation of a certain abnormal segment exceeds the preset fluctuation range, it is marked as a key focus object, thus obtaining the key focus segment. Based on the key monitoring segments and combined with the detailed update mechanism, a targeted monitoring and adjustment strategy is generated to determine the final monitoring results.

Images