Intelligent sensing tow seat feedback system

Abstract

The application relates to the field of rail transit safety technology and discloses an intelligent sensing traction seat feedback system. A data acquisition module of the system acquires a traction seat sensor data set collected by a high-speed rail transit safety detection system; a data division module divides time intervals according to an acquisition sequence and a measurement value, and obtains each sub-data unit; an abnormality identification module identifies signal characteristics of the sub-data units, and marks the sub-data units satisfying abnormality detection conditions; an evaluation parameter generation module generates safety evaluation parameters in combination with physical parameters corresponding to the sub-data units of the reference points and the marked points; and a database construction module integrates the safety evaluation parameters of the sub-data units, and constructs a comprehensive safety database. Through ordered processing, abnormality identification and parameter integration of the traction seat sensor data, the system realizes systematic feedback on the safety state of the traction seat, and provides data support for safety management of the traction seat of the rail transit.

Description

Technical Field

[0001] This invention relates to the field of rail transit safety technology, specifically to an intelligent sensing traction seat feedback system. Background Technology

[0002] In the field of high-speed rail transit, the traction seat, as a key component connecting the train body and the traction device, directly affects the train's operational safety. With the rapid development of the rail transit industry and the continuous increase in train speeds, the loads and stresses borne by the traction seat are becoming more complex, and the requirements for its safety testing are also increasing.

[0003] Traditional methods for inspecting towing seats rely heavily on manual inspections or data collection from single sensors, which have many limitations. Manual inspections are not only costly in terms of manpower and resources, but are also susceptible to subjective factors such as the experience and sense of responsibility of the inspectors, making it difficult to achieve real-time and accurate monitoring of the towing seat's condition. Furthermore, the long intervals between manual inspections make it difficult to capture transient anomalies that occur within a short period, easily overlooking potential safety hazards.

[0004] Data acquisition using a single sensor suffers from insufficient data dimensionality. The operating status of the traction device is affected by multiple factors, including changes in various physical parameters such as vibration, temperature, pressure, and displacement. Data from a single sensor cannot comprehensively reflect the true operating status of the traction device. Furthermore, existing data processing methods lack systematic integration and analysis. The collected sensor data is often stored in a scattered manner without effective time interval segmentation and feature extraction, resulting in low data utilization. When abnormal signals occur, it is difficult to quickly trace the time of occurrence of the anomaly and the corresponding changes in physical parameters, posing challenges to the investigation of the cause of the anomaly and safety assessment.

[0005] While some safety detection systems can achieve initial data collection, they lack targeted optimization in the data processing stage. Data segmentation fails to consider the correlation between the collection order and the measured values, resulting in a chaotic division of sub-data units that hinders subsequent signal feature analysis. Anomaly identification relies heavily on simple threshold judgments, lacking sensitivity to complex signal feature changes and prone to misjudgments or omissions. Furthermore, the lack of systematic generation and integration of safety assessment parameters prevents the formation of a comprehensive traction seat safety status database, failing to meet the rail transit industry's needs for long-term monitoring and trend analysis of traction seat safety status. These problems make it difficult to detect and address safety hazards in traction seats in a timely manner, posing a potential threat to the safe operation of high-speed rail transit. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent sensing traction seat feedback system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides an intelligent sensing traction seat feedback system, the system comprising:

[0008] The data acquisition module is used to acquire the data set of the traction seat sensor collected by the high-speed rail transit safety detection system;

[0009] The data segmentation module, connected to the data acquisition module, is used to divide the time interval of the traction seat sensor data set into sub-data units according to the acquisition order and measurement values ​​of the traction seat sensor data set.

[0010] An anomaly identification module, connected to the data partitioning module, is used to identify the signal characteristics of each sub-data unit and mark the sub-data unit corresponding to the signal characteristics that meet the anomaly detection conditions.

[0011] The evaluation parameter generation module, connected to the anomaly identification module, is used to generate security evaluation parameters by combining the physical parameters corresponding to the sub-data units of the reference point and the marker point.

[0012] The database construction module, connected to the evaluation parameter generation module, is used to associate and integrate the security evaluation parameters of the sub-data units to construct a comprehensive security database.

[0013] Preferably, based on the acquisition order and measurement values ​​of the traction seat sensor data set, the time interval of the traction seat sensor data set is divided to obtain each sub-data unit, including:

[0014] Based on the acquisition sequence and measurement values, the cumulative time of each sub-data unit is recorded based on a time interval to determine the initial time interval when there is no gap between each sub-data unit; when the total measurement duration of each sub-data unit is less than the total duration of the time interval, the signal characteristics of adjacent sub-data units are compared to mark the sub-data groups with gaps; the prediction time interval of the sub-data group is determined according to the feature difference degree in the signal characteristics; the prediction time interval is inserted into the initial time interval corresponding to the sub-data group, and the cumulative time of each sub-data unit is recalculated to obtain the sub-time interval of each sub-data unit.

[0015] Preferably, comparing the signal characteristics of adjacent sub-data units to mark sub-data groups with intervals includes:

[0016] Frequency features are extracted from the signal features of adjacent sub-data units; peak points and valley points are obtained from each frequency feature, and the peak difference and valley difference are calculated. When the peak difference and valley difference are greater than a set threshold, the sub-data group is determined to have an interval and is marked; when both the peak difference and valley difference are less than the set threshold, the matching set of the peak point and valley point with the smallest distance among the frequency features is obtained; the total distance of each matching set is calculated, and when the total distance is greater than the fit threshold, the sub-data group is determined to have an interval and is marked.

[0017] Preferably, obtaining the matching set of peak points and valley points with the smallest distance among each of the frequency features includes:

[0018] For each peak point of each frequency feature, calculate its distance to each valley point of another frequency feature, and select the valley point with the smallest distance to generate a matching set; for each valley point of each frequency feature, calculate its distance to each peak point of another frequency feature, and select the peak point with the smallest distance to generate a matching set.

[0019] Preferably, comparing the signal characteristics of adjacent sub-data units to mark sub-data groups with intervals further includes:

[0020] Extract amplitude features from the signal features of adjacent sub-data units; obtain target regions where the values ​​of each amplitude feature are in the same value range and located in the changing part; compare the target regions of each amplitude feature with the target regions of another amplitude feature; if there are no target regions with the same value range, determine that there is an interval in the sub-data group and mark it.

[0021] Preferably, identifying the signal features of each of the sub-data units and marking the sub-data units corresponding to the signal features that satisfy the anomaly detection conditions includes:

[0022] According to the acquisition sequence, each of the sub-data units is selected sequentially, and the amplitude features included in the signal characteristics of the sub-data unit are identified; the mean amplitude of the sub-data unit is determined based on the amplitude features, and the difference between the mean amplitude of the current sub-data unit and the mean amplitude of the previous sub-data unit is calculated; when the difference is within the abnormal difference range, the current sub-data unit is determined to meet the abnormal detection conditions and is marked.

[0023] Preferably, the security assessment parameters generated by combining the physical parameters corresponding to the sub-data units of the reference point and the marker point include:

[0024] Based on the correspondence between physical parameters and speed, the predicted speed corresponding to the corresponding time interval is determined, and safety assessment parameters are generated based on the mapping relationship between the time interval and the predicted speed.

[0025] Preferably, the system further includes:

[0026] The data filling module is used to acquire scanning information of data collected by the inspection equipment of the high-speed rail transit safety inspection system at various location points; determine the blank time interval with missing data based on the scanning information, select the sub-regions adjacent to the blank time interval, and obtain the filling data for the blank time interval by combining the difference of their physical parameters.

[0027] Preferably, selecting adjacent sub-regions of the blank time interval and combining them with the differences in their physical parameters, the filling data for the blank time interval includes:

[0028] Select the sub-data unit corresponding to the sub-time interval adjacent to the blank time interval as the target unit; calculate the difference between the same physical parameters in each target unit, and obtain the mean value corresponding to the physical parameters whose difference is less than the difference threshold; determine the mean value as the filling value of the corresponding physical parameter in the blank time interval, and obtain the filling data of the blank time interval based on the filling value.

[0029] Preferably, the system further includes a security report generation module for performing:

[0030] In the vehicle operation environment of the high-speed rail transit safety detection system, the state model data of the traction seat is acquired; based on the static analysis method, the operation control flow diagram and the operation data flow diagram are constructed according to the state model data.

[0031] By integrating and connecting the operation control flow diagram and the operation data flow diagram, a comprehensive operation model is obtained; based on safety constraint checks, anomaly checks are performed on the comprehensive operation model to identify vulnerable risk points;

[0032] Based on the vulnerable risk points, state analysis is conducted according to the comprehensive operation model to obtain safety risk information;

[0033] Based on the static analysis algorithm, the initial test pool is obtained by generating seeds through the seed initialization method according to the runtime control flow graph, runtime data flow graph and model interface.

[0034] Based on the dynamic feedback mechanism, dynamic testing is performed according to the initial test pool, state model data, runtime data flow diagram and risk mode to obtain real-time risk information;

[0035] A security report is generated based on real-time risk information and security risk information.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] The data acquisition module directly connects to the high-speed rail transit safety monitoring system, enabling it to comprehensively acquire data from the traction seat sensors, covering various physical parameters during the traction seat's operation. This comprehensive data acquisition method overcomes the limitations of traditional single-sensor data, allowing data analysis to be built upon a richer foundation of raw data, thus facilitating a more complete understanding of the traction seat's operational status.

[0038] The data partitioning module divides the time intervals based on the acquisition order and measurement values ​​of the traction seat sensor data set, resulting in sub-data units. This partitioning method fully considers the temporal correlation and numerical characteristics of the data, making the originally large and messy raw data orderly and structured. The ordered sub-data units not only facilitate the processing and analysis of each module, but also clearly reflect the changes in the operating status of the traction seat at different time stages, providing convenience for tracing parameter fluctuations within a specific time period.

[0039] The anomaly detection module focuses on identifying the signal characteristics of each sub-data unit and marking those that meet the anomaly detection criteria. Through precise analysis of signal characteristics, this module can keenly capture abnormal signals from normal data and promptly identify potential problems with the traction seat. Compared with traditional simple threshold judgment, anomaly detection based on signal characteristics is more targeted and accurate, reducing the possibility of false positives or false negatives, and enabling anomalies to be detected in a timely manner.

[0040] The evaluation parameter generation module combines the physical parameters corresponding to the sub-data units of the benchmark reference point and the marker point to generate safety evaluation parameters. The benchmark reference point provides a reasonable reference standard for the evaluation, while the sub-data units of the marker point focus on key information related to anomalies. The safety evaluation parameters generated by combining the two can objectively reflect the safety status of the traction seat. These parameters are no longer isolated values, but effective indicators closely related to the actual operating status, providing a clear basis for understanding the safety level of the traction seat.

[0041] The database construction module integrates and correlates the safety assessment parameters of sub-data units to build a comprehensive safety database. This comprehensive safety database systematically integrates the scattered assessment parameters, forming a complete record of the traction seat's safety status data. This database not only stores current safety information but also accumulates historical data, facilitating long-term tracking and trend analysis of the traction seat's safety status. By reviewing historical data, the patterns of safety status changes under different operating conditions can be summarized, providing a data foundation for predicting potential risks and offering comprehensive information support for traction seat maintenance and other related work. Attached Figure Description

[0042] Figure 1 This is a timing diagram of the intelligent sensing traction seat feedback system described in this invention;

[0043] Figure 2 A flowchart for labeling interval-based subgroups of data based on frequency characteristics;

[0044] Figure 3 A flowchart illustrating the process of marking intervals in sub-data groups based on amplitude characteristics;

[0045] Figure 4 Flowchart of the calculation process for filling blank time intervals with data. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Please see Figure 1 This invention provides an intelligent sensing traction seat feedback system, which includes five core modules: data acquisition, data segmentation, anomaly identification, evaluation parameter generation, and database construction, as well as two extended modules: data filling and safety report generation, which work together to realize real-time monitoring and safety assessment of the entire process of high-speed rail transit traction seats.

[0048] The data acquisition module first acquires the traction seat sensor data set collected by the high-speed rail transit safety detection system. The data segmentation module records the cumulative time of sub-data units based on the acquisition sequence, determines the initial time interval without gaps, and if the total measurement time is insufficient, it compares the frequency characteristics (calculates the peak-to-valley difference and the total distance of the matching set) and amplitude characteristics (compares the numerical interval of the target area) of adjacent sub-data units, marks the interval sub-data groups, inserts the predicted time interval, and recalculates the sub-time intervals and corresponding sub-data units. The anomaly identification module extracts the amplitude characteristics of sub-data units according to the acquisition sequence, calculates the difference between the current and previous unit's average amplitude, and marks the sub-data units whose differences are within the abnormal range. The evaluation parameter generation module combines the physical parameters of the benchmark reference point and the marked sub-data units, determines the predicted speed based on the correspondence between physical parameters and speed, and generates safety evaluation parameters. The database construction module integrates the evaluation parameters of each sub-data unit to build a comprehensive safety database. At the same time, the data filling module supplements the blank time interval data, and the safety report generation module combines static analysis and dynamic testing to generate a safety report, comprehensively ensuring the safe operation of the traction seat.

[0049] Example 1: See Figure 2The data partitioning module's specific workflow involves dividing the traction seat sensor data set into multiple sub-data units and identifying time intervals within the data. The following detailed explanation of this module's operation mechanism is illustrated with a concrete example:

[0050] Assume the data acquisition module obtains data from the traction seat sensor containing continuously collected vibration signals at a sampling frequency of 1000Hz, a total duration of 10 seconds, and 10,000 data points. The data partitioning module first divides the data into several sub-data units according to time sequence. Each sub-data unit contains 100 data points, corresponding to a time interval of 0.1 seconds. After the initial partitioning, the module accumulates the measurement duration of each sub-data unit and finds that the total measurement duration is 9.8 seconds, which is less than the total duration of 10 seconds, indicating a 0.2-second data gap or interval.

[0051] To pinpoint the interval position, the module compares and analyzes the signal characteristics of adjacent sub-data units. Taking adjacent sub-data units A and B as examples, the module extracts their frequency domain characteristics, including the amplitude and location of peak and valley points. The frequency domain characteristics of sub-data unit A show a peak amplitude of 5.2V and a valley amplitude of -4.8V; the peak amplitude of sub-data unit B is 5.5V and the valley amplitude is -5.1V. The module calculates the peak difference and valley difference between the two, which are 0.3V and 0.3V respectively. If the preset difference threshold is 0.4V, the difference does not exceed the threshold, and further analysis of the peak and valley matching relationship is required.

[0052] The module iterates through the peak points of sub-data unit A, calculates the distance between them and the valley points of sub-data unit B, and selects the valley point with the smallest distance to form a matching pair. For example, if the distance between the peak point P1 of sub-data unit A and the valley point V1 of sub-data unit B is 0.2, and the distance between V1 and V2 is 0.4, then V1 is selected as the matching pair. Simultaneously, the module iterates through the valley points of sub-data unit B, calculates the distance between them and the peak points of sub-data unit A, and selects the peak point with the smallest distance to form a matching pair. For example, if the distance between the valley point V1 of sub-data unit B and the peak point P1 of sub-data unit A is 0.2, and the distance between V1 and P2 is 0.3, then P1 is selected as the matching pair. The total distance of all matching pairs is 0.2 + 0.2 = 0.4. If the fit threshold is 0.5, the total distance does not exceed the threshold, and no gap is determined to exist. If the peak difference between another pair of adjacent sub-data units C and D is 0.5V, exceeding the threshold of 0.4V, then a gap is directly determined to exist between C and D. For sub-data unit pairs that do not exceed the difference threshold but whose total matching distance exceeds the fit threshold, the module also determines that there is a gap. For example, the peak difference between sub-data units E and F is 0.3V and the valley difference is 0.35V, both of which do not exceed the threshold, but the total matching distance is 0.6V, which exceeds the fit threshold of 0.5V. Therefore, it is determined that there is a gap between E and F.

[0053] For sub-data groups with intervals, the module predicts the time interval length based on the signal characteristic differences. For example, sub-data units C and D have a peak difference of 0.5V and a valley difference of 0.45V, with relatively large amplitude fluctuations, so the module predicts an interval length of 0.05 seconds. Sub-data units E and F have a large total matching distance but relatively small amplitude fluctuations, so the module predicts an interval length of 0.03 seconds. The module inserts the predicted time intervals into the initial time interval and recalculates the cumulative time for each sub-data unit. For example, a 0.05-second interval is inserted between sub-data units C and D, and a 0.03-second interval is inserted between sub-data units E and F, totaling 0.08 seconds. The remaining 0.12 seconds of missing time is supplemented by interval predictions from other sub-data groups.

[0054] When re-dividing the time interval, the module ensures that the inserted interval does not affect the continuity of existing sub-data units. For example, after inserting a 0.05-second interval between sub-data units C and D, the starting point of the time in sub-data unit D is adjusted from 0.6 seconds to 0.65 seconds, and the time stamp of the sub-data unit is correspondingly postponed. After the division is completed, the time intervals of each sub-data unit are strictly aligned, with no overlap or omissions.

[0055] The entire data partitioning process employs a sliding window mechanism to dynamically adjust the analysis range. For example, for data with varying sampling rates, the module automatically adjusts the length of sub-data units to ensure that each unit contains enough data points to extract stable features. Peak and valley detection utilizes a local extremum search algorithm to avoid noise interference. For instance, in data with high-frequency noise, the module performs smoothing processing before detecting extrema, reducing false positives.

[0056] Example 2: See Figure 3 The data partitioning module compares and analyzes the amplitude characteristics of adjacent sub-data units, and uses target region matching based on amplitude characteristics to identify data intervals. The workflow of this module is explained in detail below with a specific example:

[0057] Assume the data acquisition module obtains vibration acceleration signals from the traction seat sensor, with a sampling frequency of 500Hz, a total duration of 8 seconds, and 4000 data points. The data partitioning module has initially divided this data into 40 sub-data units, each containing 100 data points, corresponding to a 0.2-second time interval. When accumulating the measurement duration of each sub-data unit, it was found that the total measurement duration is 7.6 seconds, with a 0.4-second data gap, requiring further analysis to determine the interval location.

[0058] The module first selects adjacent sub-data units G and H for amplitude characteristic analysis. The time-domain waveforms of these two sub-data units show that unit G exhibits amplitude fluctuations of 3.2-3.5 m / s² within the 0.1-0.15 second interval, while unit H exhibits amplitude fluctuations of 3.3-3.6 m / s² within the 0.05-0.1 second interval. The module defines these two intervals as target regions and sets the numerical range to 3.0-3.8 m / s² for matching. Since the numerical ranges of the two target regions overlap and both are located within the waveform variation region, the module does not determine the existence of an interval at this stage.

[0059] Next, the adjacent sub-data units J and K were analyzed. Unit J exhibited amplitude fluctuations of 4.1-4.3 m / s² in the 0.05-0.1 second range, while unit K showed amplitude fluctuations of 2.8-3.0 m / s² in the 0.15-0.2 second range. The module was set to a numerical interval width of 0.5 m / s². It was found that the target region numerical interval for unit J was 3.9-4.5 m / s², and for unit K it was 2.6-3.4 m / s², with no overlap. The module further checked whether both target regions were located within the waveform variation region, and after confirmation, it was determined that there was an interval between J and K.

[0060] For another pair of adjacent sub-data units L and M, unit L exhibits amplitude fluctuations of 5.1-5.3 m / s² in the 0.1-0.12 second interval, while unit M exhibits amplitude fluctuations of 5.2-5.4 m / s² in the 0.08-0.11 second interval. Although the numerical intervals overlap, the module detects that the target region of unit L is located at the rising edge of the waveform, while the target region of unit M is located at the falling edge of the waveform, and they do not belong to the same trend. Therefore, an interval is still determined to exist.

[0061] After marking the sub-data groups with gaps, the module makes a comprehensive judgment based on the results of frequency feature analysis. For example, for sub-data groups J and K, amplitude feature analysis shows the presence of gaps, and frequency feature analysis also shows that the peak difference exceeds the threshold, thus doubly verifying the existence of gaps. However, for sub-data groups G and H, although amplitude feature analysis did not find obvious gaps, frequency feature analysis showed that the total matching distance exceeded the threshold, and the module ultimately determined that gaps existed.

[0062] After determining the interval position, the module predicts the interval duration based on the amplitude change trend of adjacent sub-data units. For sub-data groups with large amplitude differences, such as J and K, the predicted interval duration is longer; for sub-data groups with small amplitude differences, such as G and H, the predicted interval duration is shorter. During the prediction process, the module references the interval durations corresponding to similar amplitude characteristics in historical data to improve prediction accuracy.

[0063] When re-dividing the time intervals, the module maintains the integrity of the original data's waveform characteristics. For example, after inserting a prediction interval between sub-data units J and K, the time stamps of all sub-data units are adjusted, but the order of data points within each unit remains unchanged. Simultaneously, the module records the position and prediction duration of each insertion interval for analysis purposes.

[0064] The entire analysis process employs a multi-feature fusion strategy, considering both the numerical range matching of amplitude features and the judgment of waveform change trends. For complex waveforms, the module sets multiple target regions for analysis, avoiding the limitations of single-region judgment. Regarding numerical range settings, the module dynamically adjusts the range width based on the overall amplitude distribution of the data to adapt to data characteristics across different amplitude ranges.

[0065] Using this amplitude-based target region analysis method, the data segmentation module can effectively identify abnormal intervals in the traction seat sensor data. In the example 8 seconds of data, the module accurately located the distribution of the missing interval of 0.4 seconds, providing a reliable time reference for data processing. This analysis method is particularly suitable for vibration signals with obvious amplitude characteristics, and can effectively distinguish between normal fluctuations and abnormal changes caused by missing data.

[0066] Example 3: The collaborative working mechanism of the anomaly detection module and the safety assessment parameter generation module, focusing on the anomaly detection method based on amplitude feature analysis and the process of generating safety assessment parameters. The workflow of this module is explained in detail below through specific examples:

[0067] Assume the traction sensor data processed by the system comes from a vibration monitoring system of a high-speed train during operation, with a sampling frequency of 800Hz. The data partitioning module has divided the raw data into several sub-data units, each containing 160 data points, corresponding to a time interval of 0.2 seconds. The anomaly identification module processes these sub-data units in the order of acquisition, and is currently analyzing adjacent sub-data units numbered N-1 and N.

[0068] The anomaly detection module first extracts the amplitude characteristics of sub-data unit N and calculates its mean amplitude. The mean amplitude is calculated using a weighted average method, assigning different weights to the amplitude values ​​at different time points within the data unit. The specific calculation formula is as follows:

[0069]

[0070] in: This represents the weighted average amplitude of the current sub-data unit. This indicates the number of data points contained in the sub-data unit (m=160 in this example). This represents the amplitude value of the k-th data point. This represents the weight coefficient of the k-th data point. The weight coefficients are generated using a Gaussian distribution function, with the highest weight at the center point and decreasing towards both ends to highlight the characteristics of the middle part of the data unit. The module calculates the mean amplitude of sub-data unit N. The average amplitude of the previous sub-data unit N-1 The difference between the two is 0.8 m / s². The system's preset abnormal difference range is 0.5-1.2 m / s², and the current difference falls within this range, so the module marks sub-data unit N as abnormal.

[0071] When marking abnormal sub-data units, the module records information such as the anomaly type (amplitude mutation), the time of occurrence, and the average amplitude of the preceding and following units. If multiple consecutive sub-data units are marked as abnormal, the module will handle this specially, classifying it as a continuous abnormal event and recording it separately from single-point anomalies.

[0072] The evaluation parameter generation module receives the output from the anomaly identification module and combines it with the physical parameters of the benchmark reference point to generate safety evaluation parameters. The benchmark reference point is selected from a sub-data unit during the stable operation phase, and its physical parameters include vibration acceleration and displacement. The module establishes the correspondence between physical parameters and train speed, and determines the predicted speed corresponding to the current time interval by querying a preset parameter-speed mapping table.

[0073] Taking sub-data unit N as an example, its vibration acceleration when marked as abnormal is 4.6 m / s², and its displacement is 0.15 mm. Querying the parameter-velocity mapping table, the predicted velocity corresponding to vibration acceleration of 4.5-5.0 m / s² and displacement of 0.1-0.2 mm is 280-300 km / h. The module takes the median value of 290 km / h as the predicted velocity for this time interval and generates safety assessment parameters including timestamp, predicted velocity, and anomaly type. For sub-data units not marked as abnormal, the module also generates safety assessment parameters, but their anomaly type is marked as "normal." These parameters, together with the anomaly parameters, constitute the complete time-series assessment results, providing input for the comprehensive safety database.

[0074] During parameter generation, the module considers the combined effects of multiple physical parameters. For example, when vibration acceleration and displacement point to different velocity ranges, the module uses a weighted decision method to determine the final predicted velocity. The weight allocation is based on the sensitivity of each physical parameter to velocity changes, with vibration acceleration typically having a higher weight than displacement.

[0075] The safety assessment parameters also include a confidence index, reflecting the reliability of the prediction results. Confidence is calculated based on the degree of matching between the physical parameters and reference values ​​in the mapping table; a higher matching degree results in a higher confidence score. For boundary value cases, the module generates a lower confidence score to alert analysts. The entire processing flow is time-sensitive; the module maintains a time window buffer to store the analysis results of the most recent few sub-data units. This allows the module to identify trend characteristics of amplitude changes, rather than being limited to single-point comparisons. For example, when a continuously rising trend in the mean amplitude is detected, the module may generate a warning parameter even if a single difference does not reach the anomaly threshold. After parameter generation, the module performs a self-check to verify whether the generated predicted velocity maintains reasonable continuity with the velocity at adjacent time points. For abrupt prediction results, the module re-verifies the physical parameter measurements and mapping relationships, making corrections if necessary. This verification mechanism effectively avoids assessment bias caused by single-point data anomalies.

[0076] Based on amplitude feature analysis and multi-parameter mapping, the system can transform raw sensor data into assessment parameters with clear safety implications. Each safety assessment parameter includes key fields such as time information, prediction velocity, anomaly markers, and confidence levels, forming a structured description of the safety status and providing standardized input for comprehensive safety analysis and risk warning. After receiving these parameters, the database construction module organizes and stores them according to time series and builds multi-dimensional indexes to support efficient querying and analysis.

[0077] Example 4: See Figure 4 The data population module is responsible for handling missing data issues that occur in the high-speed rail transit safety detection system. The following example details the module's operation:

[0078] Suppose that during monitoring of a high-speed train operating section, the traction seat sensor system recorded 15 minutes of vibration data at a sampling frequency of 500Hz. Due to temporary signal interference, data was missing from 08:32:15.200 to 08:32:16.800 (duration 1.6 seconds), forming a blank time interval. The data filling module needs to fill this gap appropriately. The module determines the boundary positions of the blank time interval. The previous valid sub-data unit ends at 08:32:15.200, with the last three recorded vibration acceleration values ​​being 2.15 m / s², 2.18 m / s², and 2.20 m / s²; the next valid sub-data unit begins at 08:32:16.800, with the first three recorded vibration acceleration values ​​being 2.35 m / s², 2.32 m / s², and 2.30 m / s². The module uses these two sub-data units as target units for filling calculations.

[0079] Table 1: Comparison of physical parameters of adjacent sub-data units.

[0080] Parameter type Mean of previous sub-data units Sub-data unit mean Difference Difference threshold Vibration acceleration (m / s²) 2.17 2.33 0.16 0.20 Displacement (mm) 0.12 0.14 0.02 0.03 Temperature (°C) 42.5 42.7 0.2 0.5 Pressure (kPa) 101.2 101.5 0.3 0.4

[0081] The module calculates the differences in various physical parameters within the target unit and compares them with preset difference thresholds. As shown in the table, the difference in vibration acceleration is 0.16 m / s², less than the threshold of 0.20 m / s²; the difference in displacement is 0.02 mm, less than the threshold of 0.03 mm; the difference in temperature is 0.2℃, less than the threshold of 0.5℃; and the difference in pressure is 0.3 kPa, less than the threshold of 0.4 kPa. Since the differences in all parameters are less than their respective thresholds, the module uses the average of these parameters as the fill value.

[0082] For vibration acceleration, the module calculates a fill value of (2.17 + 2.33) / 2 = 2.25 m / s². Considering that a 1.6-second blank interval requires filling with 800 data points (500 Hz × 1.6 s), the module uses a linear interpolation method to generate a continuously changing data sequence. The initial value is 2.20 m / s² (the last valid value of the previous unit), the ending value is 2.35 m / s² (the first valid value of the subsequent unit), and the intermediate values ​​increase linearly. Displacement is filled using a similar method; after calculating an average value of 0.13 mm, a fill sequence is generated based on the displacement change trend of the preceding and following units. Temperature parameters, due to their slow change, are directly filled with a constant value of 42.6℃ throughout the blank interval. Pressure parameters, however, are filled with data that transitions smoothly by combining the rate of change before and after.

[0083] In special cases, when the difference in certain physical parameters exceeds a threshold, the module will adopt different filling strategies. For example, if the difference in vibration acceleration reaches 0.25 m / s², exceeding the threshold of 0.20 m / s², the module will abandon using the mean of this parameter and instead use the last value of the previous sub-data unit as the filling reference, combined with the typical rate of change of historical data to generate a filling sequence. After filling is completed, the module will perform a smoothness check on the generated transition data. The check includes numerical jump detection and physical rationality verification. For example, the change in adjacent data points of vibration acceleration should not exceed 0.05 m / s², and the change in displacement should conform to the motion law of the mechanical structure. When abnormal jumps are detected, the module will readjust the interpolation parameters to ensure the continuity of the filled data.

[0084] For longer blank intervals (e.g., exceeding 3 seconds), the module employs a segmented filling strategy. The long interval is divided into several smaller segments, each using different filling parameters to better simulate changes in the actual physical process. Simultaneously, markers are inserted to indicate which data were manually filled, facilitating identification during analysis. The module also maintains a filling record database, recording detailed information for each data filling, including the location of the blank interval, the filling method used, and the reference parameters used. These records are available for data analysis and for evaluating and improving the accuracy of the filling algorithm.

[0085] During the data population process, the module prioritizes ensuring the continuity of key parameters (such as vibration acceleration), while allowing relatively larger errors in the population of secondary parameters (such as temperature). It also considers the physical relationships between different parameters; for example, there is a certain proportional relationship between vibration acceleration and displacement, and this relationship will be maintained during the population process.

[0086] This imputation method, based on adjacent data feature analysis, effectively addresses missing data in monitoring data. In the example 1.6-second blank interval, the imputed data generated by the module smoothly connects with the preceding and following measured data, maintaining the continuity of the data curve and providing a complete data foundation for safety assessment. The labeling mechanism for the imputed data also ensures that analysts can clearly distinguish between measured values ​​and imputed values, avoiding misjudgments.

[0087] The entire filling process is fully automated, requiring no manual intervention. The module automatically selects the most suitable filling strategy based on the specific circumstances of data loss, ensuring data continuity while preserving the true characteristics of physical parameters to the greatest extent possible. This intelligent data filling capability significantly improves the reliability of the traction seat monitoring system, ensuring effective safety assessments can still be conducted even with partial data loss.

[0088] Example 5: The complete workflow of the safety report generation module. This module comprehensively evaluates the safety status of the traction seat system and generates a comprehensive safety report through a combination of static analysis and dynamic testing. The implementation process of this module is described in detail below through a specific operating scenario:

[0089] During high-speed train operation, the safety report generation module first acquires the state model data of the traction seat system. This data includes static information such as the geometric parameters, material properties, and connection methods of the traction seat structure, as well as dynamic parameters such as vibration characteristics and displacement obtained from real-time monitoring. The module processes this data using static analysis methods to construct the operation control flow diagram of the traction seat system. The control flow diagram details the working mode transition logic of the traction seat under various operating states, including the control flow under typical conditions such as normal driving, acceleration, deceleration, and curve passage. Simultaneously, the module constructs an operation data flow diagram to trace the transmission path of key physical parameters within the system. The data flow diagram clearly shows the complete process of vibration signal acquisition from sensors, through preprocessing, feature extraction, and anomaly detection, ultimately forming safety assessment parameters. Each node in the data flow diagram is labeled with the parameter type, value range, and processing method for easy analysis.

[0090] The module integrates the operational control flow diagram and operational data flow diagram to form a comprehensive operational model. This model not only reflects the static structural characteristics of the traction seat system but also embodies the data interaction relationships during dynamic operation. Based on this model, the module performs safety constraint checks to identify operational states that may violate safety rules. The checks include whether vibration amplitude exceeds limits, whether displacement is within allowable ranges, and whether the stress on each component is balanced. Through systematic checks, the module identifies multiple vulnerable points, including stress concentration areas on connecting bolts and wear-sensitive parts of the buffer device.

[0091] During the dynamic testing phase, the module uses a seed initialization method to generate test cases. Test seeds are derived from typical operating condition records in historical operational data, including traction seat status data under different speed levels and track conditions. The module performs mutation processing on these seeds to generate a set of test cases that include boundary conditions and abnormal situations. The test pool contains test cases simulating various scenarios such as extreme vibration, sudden load changes, and signal interference.

[0092] The module combines real-time acquired state model data and runtime data flow diagrams to perform dynamic testing on test cases in the test pool. During testing, the module monitors the response characteristics of the traction seat system under various abnormal operating conditions and records risk events such as parameter exceeding limits and control logic anomalies. Dynamic testing is conducted iteratively, adjusting the testing strategy based on the results of the previous round of testing, focusing on testing scenarios related to identified risk points.

[0093] Through static analysis and dynamic testing, the module obtains two types of risk information: structural risk points revealed by static analysis and operational risks discovered by dynamic testing. The module comprehensively analyzes this information to assess the severity and probability of occurrence of each risk point. Risk level classification considers multiple dimensions, including the degree of impact on driving safety, the detectability of the fault, and the time window for emergency response. The generated safety report adopts a structured format and includes main sections such as a system overview, analysis method description, risk list, and improvement recommendations. The risk list in the report details the location, characteristics, possible consequences, and handling recommendations for each risk point. The improvement recommendations section proposes specific measures for different types of risks, including hardware modification schemes, control parameter optimization, and detection frequency adjustment.

[0094] During report generation, the module employs visualization technology to aid in presentation. Different colors are used to mark risk levels in the 3D model of the traction seat structure, abnormal intervals are highlighted in the time-series data, and weak points are marked in the control flow chart. These visualization elements make the report more intuitive and easier to understand, allowing engineers to quickly grasp key information.

[0095] The module also features report version management, recording the complete process of each safety assessment. When the traction seat system is modified or its operating conditions change, historical reports can be traced to compare and analyze the evolution trend of the system's safety status. This longitudinal comparison helps to identify potential systemic risks. The entire safety report generation process is adaptive; the module dynamically adjusts its analysis focus based on the actual operating performance of the traction seat system. For parameters or parts that frequently exhibit anomalies, it automatically increases detection accuracy and testing intensity. Simultaneously, the module learns from the processing feedback of historical reports, optimizing the report's content organization and presentation.

[0096] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent sensing tow seat feedback system, characterized in that, include: The data acquisition module is used to acquire the data set of the traction seat sensor collected by the high-speed rail transit safety detection system; A data partitioning module, connected to the data acquisition module, is used to partition the time interval of the traction seat sensor data set according to the acquisition order and measurement values ​​to obtain each sub-data unit, including: Based on the acquisition sequence and measurement values, the cumulative time of each sub-data unit is recorded based on a time interval to determine the initial time interval when there is no gap between each sub-data unit; when the total measurement duration of each sub-data unit is less than the total duration of the time interval, the signal characteristics of adjacent sub-data units are compared to mark the sub-data groups with gaps; the prediction time interval of the sub-data group is determined according to the feature difference degree in the signal characteristics; the prediction time interval is inserted into the initial time interval corresponding to the sub-data group, and the cumulative time of each sub-data unit is recalculated to obtain the sub-time interval of each sub-data unit; An anomaly identification module, connected to the data partitioning module, is used to identify the signal characteristics of each sub-data unit and mark the sub-data unit corresponding to the signal characteristics that meet the anomaly detection conditions. The evaluation parameter generation module, connected to the anomaly identification module, is used to generate security evaluation parameters by combining the physical parameters corresponding to the sub-data units of the reference point and the marker point. The database construction module, connected to the evaluation parameter generation module, is used to associate and integrate the security evaluation parameters of the sub-data units to construct a comprehensive security database.

2. The intelligent sensing tow seat feedback system of claim 1, wherein, By comparing the signal characteristics of adjacent sub-data units, sub-data groups with intervals are identified as follows: Frequency features are extracted from the signal features of adjacent sub-data units; peak points and valley points are obtained from each frequency feature, and the peak difference and valley difference are calculated. When the peak difference and valley difference are greater than a set threshold, the sub-data group is determined to have an interval and is marked; when both the peak difference and valley difference are less than the set threshold, the matching set of the peak point and valley point with the smallest distance among the frequency features is obtained; the total distance of each matching set is calculated, and when the total distance is greater than the fit threshold, the sub-data group is determined to have an interval and is marked.

3. The intelligent sensing tow seat feedback system of claim 2, wherein, Obtaining the matching set of the peak points and valley points with the smallest distance among the frequency features includes: For each peak point of each frequency feature, calculate its distance to each valley point of another frequency feature, and select the valley point with the smallest distance to generate a matching set; for each valley point of each frequency feature, calculate its distance to each peak point of another frequency feature, and select the peak point with the smallest distance to generate a matching set.

4. The intelligent sensing tow seat feedback system of claim 3, wherein, By comparing the signal characteristics of adjacent sub-data units, marking sub-data groups with intervals further includes: Extract amplitude features from the signal features of adjacent sub-data units; obtain target regions where the values ​​of each amplitude feature are in the same value range and located in the changing part; compare the target regions of each amplitude feature with the target regions of another amplitude feature; if there are no target regions with the same value range, determine that there is an interval in the sub-data group and mark it.

5. The intelligent sensing seat feedback system of claim 1, wherein, Identifying the signal features of each of the sub-data units and marking the sub-data units corresponding to the signal features that satisfy the anomaly detection conditions includes: According to the acquisition sequence, each of the sub-data units is selected sequentially, and the amplitude features included in the signal characteristics of the sub-data unit are identified; the mean amplitude of the sub-data unit is determined based on the amplitude features, and the difference between the mean amplitude of the current sub-data unit and the mean amplitude of the previous sub-data unit is calculated; when the difference is within the abnormal difference range, the current sub-data unit is determined to meet the abnormal detection conditions and is marked.

6. The intelligent sensing traction seat feedback system according to claim 1, characterized in that, Combining the physical parameters corresponding to the sub-data units of the reference point and the marker point, the safety assessment parameters are generated as follows: Based on the correspondence between physical parameters and speed, the predicted speed corresponding to the corresponding time interval is determined, and safety assessment parameters are generated based on the mapping relationship between the time interval and the predicted speed.

7. The intelligent sensing traction seat feedback system according to claim 1, characterized in that, Also includes: The data filling module is used to acquire scanning information of data collected by the inspection equipment of the high-speed rail transit safety inspection system at various location points; determine the blank time interval with missing data based on the scanning information, select the sub-regions adjacent to the blank time interval, and obtain the filling data for the blank time interval by combining the difference of their physical parameters.

8. The intelligent sensing traction seat feedback system according to claim 7, characterized in that, By selecting adjacent sub-regions of the blank time interval and combining them with the differences in their physical parameters, the filling data for the blank time interval is obtained, including: Select the sub-data unit corresponding to the sub-time interval adjacent to the blank time interval as the target unit; calculate the difference between the same physical parameters in each target unit, and obtain the mean value corresponding to the physical parameters whose difference is less than the difference threshold; determine the mean value as the filling value of the corresponding physical parameter in the blank time interval, and obtain the filling data of the blank time interval based on the filling value.

9. The intelligent sensing traction seat feedback system according to claim 1, characterized in that, It also includes a security report generation module for performing: In the vehicle operation environment of the high-speed rail transit safety detection system, the state model data of the traction seat is acquired; based on the static analysis method, the operation control flow diagram and the operation data flow diagram are constructed according to the state model data. By integrating and connecting the operation control flow diagram and the operation data flow diagram, a comprehensive operation model is obtained. Based on safety constraint checks, anomaly checks are performed according to the comprehensive operation model to identify vulnerable risk points. Based on the vulnerable risk points, state analysis is conducted according to the comprehensive operation model to obtain safety risk information; Based on the static analysis algorithm, the initial test pool is obtained by generating seeds through the seed initialization method according to the runtime control flow graph, runtime data flow graph and model interface. Based on the dynamic feedback mechanism, dynamic testing is performed according to the initial test pool, state model data, runtime data flow diagram and risk mode to obtain real-time risk information; A security report is generated based on real-time risk information and security risk information.

Images