Urban rail transit equipment fault prediction method based on big data analysis

Abstract

The invention discloses an urban rail transit equipment fault prediction method based on big data analysis, and relates to the technical field of industrial big data analysis, and the method comprises the steps: building a health state index, a degradation trend index and a risk characteristic index according to the equipment type based on a standardized operation data set, and forming an index sequence; based on the index sequence, performing fault prediction calculation in a big data analysis environment to obtain a fault occurrence probability, a health state score and a residual life evaluation result; the fault occurrence probability, the health state score, the residual life evaluation result and the risk level are summarized, and fault prediction information and risk early warning information of operation and maintenance management are generated; a multi-dimensional dynamic index sequence fusing the health state, the deterioration trend and the risk characteristics is constructed, an intelligent operation and maintenance decision with clear priority and reasonable time sequence is supported and formed, and the safety, reliability and resource utilization efficiency of operation and maintenance of urban rail transit equipment are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of industrial big data analysis technology, and in particular to a method for predicting faults in urban rail transit equipment based on big data analysis. Background Technology

[0002] In recent years, with the rapid development of the Internet of Things, sensor networks, and big data analytics, intelligent operation and maintenance systems based on multi-source sensor data have gradually become a research hotspot. Existing technologies generally use threshold alarms or simple statistical models to detect anomalies in single parameters (such as temperature or vibration) and combine them with regular maintenance strategies. Although some advanced systems have introduced machine learning methods such as support vector machines (SVM) or shallow neural networks to identify fault modes of specific equipment, their data processing capabilities are still limited by traditional architectures. When faced with massive, high-dimensional, and heterogeneous operational data, it is difficult to fully realize the potential of big data analytics.

[0003] While current technologies have improved the timeliness of fault identification to some extent, they still have significant limitations in multi-dimensional heterogeneous data fusion, health status evolution modeling, and risk quantification assessment. Specifically, existing solutions typically do not fully integrate multi-modal operating parameters such as voltage, current, temperature, vibration, and acoustics, and lack a systematic expression of equipment degradation trends and risk characteristics in their indicator system construction. This results in insufficient accuracy in predicting future failure probabilities, remaining lifespan, and health scores using big data analytics, thus affecting the optimal allocation of operation and maintenance resources. Especially in high-concurrency, high-dynamic urban rail transit scenarios, traditional methods struggle to leverage big data analytics to achieve dynamic mapping of equipment risk levels and intelligent prioritization of handling, hindering the substantial transformation of operation and maintenance management from "passive response" to "proactive prevention." Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a fault prediction method for urban rail transit equipment based on big data analysis to solve the problems of low fault prediction accuracy and delayed operation and maintenance decision-making caused by the lack of dynamic quantification of risk levels.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for predicting faults in urban rail transit equipment based on big data analysis. The method includes: collecting voltage, current, temperature, vibration, and acoustic operation data, combining them with historical fault logs and maintenance records, preprocessing them according to time windows to generate a standardized operation dataset; constructing health status indicators, degradation trend indicators, and risk characteristic indicators based on the standardized operation dataset according to equipment type, forming an indicator sequence; performing fault prediction calculations based on the indicator sequence in a big data analysis environment to obtain fault occurrence probability, health status score, and remaining life assessment results; performing uncertainty assessment on the fault occurrence probability, health status score, and historical statistical distribution information to generate risk quantification values, and mapping these risk quantification values ​​to risk levels; determining the priority and timing of equipment operation and maintenance based on risk levels and risk change trends, generating operation and maintenance management decision information; and summarizing the fault occurrence probability, health status score, remaining life assessment results, and risk levels to generate fault prediction information and risk warning information for operation and maintenance management.

[0007] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for generating the standardized operation dataset are as follows: Collect voltage, current, temperature, vibration and acoustic operating data generated by urban rail transit equipment during operation, and combine them with historical fault logs and maintenance records to generate multimodal operating parameters; Based on the multimodal operating parameters organized according to time windows, abnormal sampling points are identified and noise suppression is performed to generate preprocessed multimodal operating parameters; dimensional normalization is performed on different types of operating parameters in the preprocessed multimodal operating parameters to form a standardized operating dataset.

[0008] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for forming the index sequence are as follows: The standardized operational dataset is grouped according to device identifier and device type, and divided into multiple continuous time windows. For each time window, the baseline deviation of each operational data point relative to historical healthy samples of the same device type is calculated, and a health status index is generated through monotonic mapping relationship and weighted fusion. By jointly characterizing the health status indicators and the rate of change of baseline deviation over time in adjacent time windows, a deterioration trend indicator is generated; risk characteristic indicators are generated based on the fluctuation amplitude of the standardized operational dataset within the time window, the cross-modal correlation, and the frequency of abnormal sampling points. Health status indicators, deterioration trend indicators, and risk characteristic indicators within the same time window are correlated to form a time window-level indicator combination, which is then arranged in chronological order to form an indicator sequence.

[0009] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for obtaining the fault occurrence probability, health status score, and remaining life assessment results are as follows. Based on the indicator sequence organized by device identifier, and using a sliding time window to segment the continuous time series, multiple time window level indicator combinations within each sliding time window are spliced ​​together in chronological order to form the prediction input sequence; In a big data analytics environment, the predicted input sequence is jointly analyzed to extract the equipment operating status features; based on the equipment operating status features, fault prediction calculation is performed to obtain the probability of fault occurrence, and a status score is calculated to obtain a health status score. Based on historical fault logs, fault events are statistically analyzed to form a fault occurrence time distribution. The equipment operating status characteristics are matched with the equipment operating status characteristics corresponding to the historical degradation stages. Based on the fault occurrence time distribution, lifetime estimation is performed to obtain the remaining lifetime assessment results.

[0010] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the big data analysis environment refers to a data processing and operation environment formed on the basis of continuous collection and unified organization of multi-source operation data, historical fault logs and maintenance records of urban rail transit equipment. The big data analysis environment can perform parallel statistical analysis, prediction calculation and historical distribution update on the indicator sequence organized by time window.

[0011] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for generating the risk quantification value are as follows: Based on the probability of failure and health status score, the data is organized according to equipment identification and prediction time to form prediction results data; Collect historical statistical distribution information and establish historical distributions of failure probability and health status to form historical distribution data; compare and analyze the prediction results data with the historical distribution data, and calculate the degree of deviation of failure probability and health status score from the historical distribution to generate deviation amount; The uncertainty characterization of the prediction results is calculated based on the deviation, and the uncertainty characterization is jointly calculated with the failure probability and health status score to generate a risk quantification value.

[0012] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for mapping risk quantification values ​​to risk levels are as follows: Based on historical statistical distribution information, the relative position of the risk quantification value in the overall distribution is characterized to obtain the distribution position characterization; Based on the distribution location representation, the relative position range of the risk quantification value in the historical statistical distribution information is determined, and the corresponding risk level is assigned according to the position range.

[0013] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for generating operation and maintenance management decision information are as follows: The risk levels and their corresponding quantitative values ​​are compiled to form a time series of risk levels; Based on the risk level time series analysis, the direction and magnitude of risk level changes over time are analyzed to output the risk change trend; based on the risk level within the current time window and combined with the risk change trend, the urgency of equipment operation and maintenance is comprehensively assessed to form the equipment operation and maintenance priority; Based on the priority of operation and maintenance (O&M) handling and the rate of change corresponding to the trend of risk change, the order of equipment handling time is sorted to form the equipment handling sequence; based on the O&M handling priority and the equipment handling sequence, O&M management decision information is generated.

[0014] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for summarizing the fault occurrence probability, health status score, remaining life assessment results, and risk level are as follows. Based on the probability of failure, health status score, remaining life assessment results and risk level, the data is compiled according to equipment identification and future time window to form a summary data; The aggregated data is aligned and correlated to form a correspondence between failure probability, health status score, remaining life assessment results and risk level, and generate predictive element correlation data.

[0015] As a preferred embodiment of the urban rail transit equipment fault prediction method based on big data analysis described in this invention, the specific steps for generating fault prediction information and risk warning information for operation and maintenance management are as follows: Based on the associated data of predictive elements, and using the future time window as the time index, the results of the failure probability and remaining life assessment are organized and summarized in sequence, and the failure prediction information is formed by centrally expressing the possibility of failure and the change of remaining life within the future time window. The device identifier, future time window, and risk level are extracted from the associated data of the predictive elements, and combined with the changes in the risk level, risk warning records are generated. The risk warning records are then summarized to form risk warning information for operation and maintenance management.

[0016] The beneficial effects of this invention are as follows: by constructing a multi-dimensional dynamic indicator sequence that integrates health status, deterioration trend and risk characteristics, and combining historical statistical distribution to assess the uncertainty of prediction results, a risk quantification value that can be mapped to risk level is generated. Under the framework of big data analysis, this invention effectively enhances the credibility, comparability and decision guidance value of fault prediction results, supports the formation of intelligent operation and maintenance decisions with clear priorities and reasonable timing, and thus greatly improves the safety, reliability and resource utilization efficiency of urban rail transit equipment operation and maintenance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for predicting faults in urban rail transit equipment based on big data analysis.

[0019] Figure 2 A flowchart for generating a standardized running dataset.

[0020] Figure 3 A flowchart for constructing the indicator sequence.

[0021] Figure 4 The flowchart for fault prediction calculation. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4As one embodiment of the present invention, this embodiment provides a method for predicting faults in urban rail transit equipment based on big data analysis, including the following steps: S1. Collect voltage, current, temperature, vibration and acoustic operation data, and combine them with historical fault logs and maintenance records. Perform preprocessing according to time windows to generate a standardized operation dataset.

[0026] S1.1. Collect voltage, current, temperature, vibration and acoustic operating data generated by urban rail transit equipment during operation, and combine them with historical fault logs and maintenance records to generate multimodal operating parameters.

[0027] Specifically, when urban rail transit equipment is in normal operation, voltage and current are acquired from the power supply circuit corresponding to the urban rail transit equipment, temperature operation data is acquired from the urban rail transit equipment, vibration operation data is acquired from the mechanical structure parts of the urban rail transit equipment, and acoustic signals generated during the operation of the urban rail transit equipment are collected. The acquired acoustic operation data reflects the operating sound characteristics of the urban rail transit equipment. After completing the acquisition of voltage, current, temperature, vibration and acoustic operation data, and combining historical fault logs and maintenance records, the corresponding acquisition time and equipment identification are recorded for each operation data to form multimodal operation parameters.

[0028] S1.2. Based on the multimodal operating parameters, organize them according to the time window, identify abnormal sampling points and perform noise suppression processing to generate preprocessed multimodal operating parameters; perform dimensional normalization processing on different types of operating parameters in the preprocessed multimodal operating parameters to form a standardized operating dataset.

[0029] Specifically, based on the acquisition time information recorded in the multimodal operating parameters, the multimodal operating parameters are organized sequentially according to time windows, so that the voltage operating parameters, current operating parameters, temperature operating parameters, vibration operating parameters, and acoustic operating parameters within the same time window form a corresponding relationship. Abnormal sampling points are identified for the multimodal operating parameters within each time window. By analyzing the abrupt changes in the temporal continuity and numerical changes of the multimodal operating parameters, the abrupt changes include abrupt changes in temporal continuity and abrupt changes in numerical changes. Abrupt changes in temporal continuity refer to changes between adjacent sampling times, while abrupt changes in numerical changes refer to a significant deviation of the values ​​of the multimodal operating parameters from the overall distribution characteristics within the same time window. Abnormal sampling points that deviate from the normal range of change are marked, and abnormal sampling points are removed or corrected. Noise suppression processing is performed on the high-frequency interference components and random fluctuation components in the multimodal operating parameters to reduce the impact of environmental interference and measurement disturbances on the multimodal operating parameters during operation, generating preprocessed multimodal operating parameters. For different types of operating parameters in the preprocessed multimodal operating parameters, dimensional normalization processing is performed to map operating parameters with different physical dimensions to a unified numerical range, forming a standardized operating dataset.

[0030] S2. Based on the standardized operation dataset, construct health status indicators, deterioration trend indicators and risk characteristic indicators according to equipment type to form an indicator sequence.

[0031] S2.1. Group the standardized operational dataset according to device identifier and device type, and divide it into multiple continuous time windows; for each time window, calculate the baseline deviation of each operational data relative to the historical healthy samples of the same device type, and generate health status indicators through monotonic mapping relationship and weighted fusion.

[0032] Specifically, based on the equipment identification and equipment type information recorded in the standardized operation dataset, the standardized operation dataset is classified and collected, and the standardized operation dataset in each group is divided into multiple continuous time windows according to the time order, so that each continuous time window contains voltage operation data, current operation data, temperature operation data, vibration operation data and acoustic operation data within the corresponding time range; For each continuous time window, a baseline statistical distribution formed from historical health samples of the same equipment type is extracted. The various operational data within the time window are compared with this baseline statistical distribution to calculate the degree of deviation of the operational data in terms of numerical level and fluctuation characteristics. Fluctuation characteristics refer to the dispersion and stability of operational data over time within the same continuous time window, used to characterize the fluctuations in operational status over time. When calculating fluctuation characteristics, for voltage, current, temperature, vibration, and acoustic operational data within each continuous time window, the distribution of changes in the operational data within the time window is analyzed. The fluctuation characteristics obtained within the time window are compared with the fluctuation characteristics corresponding to the baseline statistical distribution formed from historical health samples of the same equipment type. The degree of deviation of the operational data within the current time window in terms of fluctuation amplitude and stability relative to the baseline statistical distribution is calculated. Based on the degree of deviation, a comprehensive characterization of various operational data within the time window is performed to generate a health status index.

[0033] It should be noted that the expression for calculating the baseline deviation is: ; in, Indicates the baseline deviation. This represents the mean of a certain type of running data within the current time window. This represents the mean of the corresponding operational data in the historical health samples of the same device type. This represents the standard deviation of corresponding operational data in historical health samples of the same equipment type. This represents the smallest positive number that prevents the denominator from being zero.

[0034] The expression for calculating health status indicators is: ; in, Indicators representing health status Indicates the operation of data type index, Indicates the first The weights corresponding to the class's running data.

[0035] S2.2. A degradation trend index is generated by jointly characterizing the health status index and the rate of change of baseline deviation over time in adjacent time windows; a risk characteristic index is generated based on the fluctuation amplitude change, cross-modal correlation change, and frequency change of abnormal sampling points in the standardized operating dataset within the time window.

[0036] Specifically, the health status indicators of adjacent time windows are compared and analyzed in chronological order. The magnitude and direction of change of the health status indicators between adjacent time windows are calculated. At the same time, the deviation of the operating data in each time window from the baseline of historical health samples of the same equipment type is combined to calculate the rate of change of the deviation over time. The changes of the health status indicators and the rate of change of the deviation over time are comprehensively characterized to generate a degradation trend indicator. For each time window, the fluctuation amplitude of voltage, current, temperature, vibration, and acoustic operating data within the standardized operating dataset is analyzed to characterize changes in operational stability. Simultaneously, the correlation between different types of operating parameters is analyzed to characterize abnormal changes in cross-modal operational coupling. For example, within the same time window, pairwise correlation analyses are performed on voltage and current, vibration and acoustic, and temperature and vibration data to calculate the correlation between each operating parameter pair within the time window. These correlations are then compared with the corresponding correlation baseline formed from historical health samples of the same equipment type. When the correlation significantly increases, decreases, or changes from positive to negative, an abnormal change in cross-modal operational coupling is identified, and the frequency of abnormal sampling points within the time window is statistically analyzed. The changes in fluctuation amplitude, cross-modal correlation, and the frequency of abnormal sampling points are comprehensively characterized to generate risk characteristic indicators.

[0037] It should be noted that the expression for calculating the rate of change of the deviation over time is: ; in, This indicates the rate of change of the deviation over time. Indicates the index of the current time window. This represents the time interval between the center moments of adjacent time windows.

[0038] The expression for calculating the degradation trend index is: ; in, Indicators indicating a deterioration trend This indicates the weight of the rate of change of health status indicators in the deterioration trend indicators. This indicates the weight of the sum of deviation changes in the deterioration trend indicator. This represents a summary of the rate of change of deviations of various operational data over time.

[0039] S2.3. Associate the health status indicators, deterioration trend indicators and risk characteristic indicators within the same time window to form a time window level indicator combination, and arrange them in chronological order to form an indicator sequence.

[0040] Specifically, health status indicators, degradation trend indicators, and risk characteristic indicators are collected based on equipment identification, ensuring that all these indicators originate from the same urban rail transit equipment. The health status indicators, degradation trend indicators, and risk characteristic indicators are arranged according to the chronological order of time windows. Based on this chronological arrangement, the health status indicators, degradation trend indicators, and risk characteristic indicators within each time window are associated and combined to form an indicator combination reflecting the equipment's operating status, status change trends, and risk characteristics within that time window. Multiple consecutive time windows are then sequentially connected to form an indicator sequence.

[0041] S3. Based on the indicator sequence, perform fault prediction calculations in a big data analysis environment to obtain the probability of fault occurrence, health status score and remaining life assessment results.

[0042] S3.1. Based on the indicator sequence, organize it according to the device identifier, and use a sliding time window to divide the continuous time series. Then, splice multiple time window level indicator combinations in each sliding time window in chronological order to form a prediction input sequence.

[0043] Specifically, based on the equipment identification information recorded in the indicator sequence, health status indicators, degradation trend indicators, and risk characteristic indicators are collected and organized, and sorted according to the order of time windows to ensure that the health status indicators, degradation trend indicators, and risk characteristic indicators remain continuous and consistent in the time dimension, forming a predictive indicator sequence. The predictive indicator sequence is continuously divided using a sliding time window method, so that each sliding time window covers the indicator content corresponding to multiple adjacent time windows. For each sliding time window, the health status indicators, degradation trend indicators, and risk characteristic indicators within the sliding time window are combined in chronological order, so that each sliding time window forms an indicator combination that can completely reflect the operating status, status change trend, and risk characteristics of the equipment within the corresponding time range, generating a predictive input sequence.

[0044] S3.2. In a big data analytics environment, perform joint analysis on the predicted input sequence to extract equipment operating status features; based on the equipment operating status features, perform fault prediction calculation to obtain the probability of fault occurrence, and perform status score calculation to obtain a health status score.

[0045] Specifically, the predicted input sequence is loaded into a big data analytics environment. Within this environment, health status indicators, degradation trend indicators, and risk characteristic indicators contained in the predicted input sequence are jointly analyzed. This joint analysis includes two parts: time-dimensional analysis and indicator correlation analysis. In the time-dimensional analysis, the health status indicators, degradation trend indicators, and risk characteristic indicators arranged chronologically within the same sliding time window are statistically analyzed for their current time point values, average changes within the time window, fluctuation amplitudes within the time window, and directions of change over time. This is used to characterize the overall level, stability, and evolution trend of the equipment's operating status. In the indicator correlation analysis, pairwise comparisons are made between the health status indicators, degradation trend indicators, and risk characteristic indicators within the same sliding time window to analyze the consistency and synchronicity of indicator changes over time. This is used to characterize changes in equipment health and degradation. The correlation between development and risk characteristics is analyzed. The level, fluctuation, and trend characteristics of health status indicators within a time window, the intensity and direction of change of deterioration trend indicators within a time window, and the concentration and change characteristics of risk characteristic indicators within a time window are summarized with the correlation analysis results of each indicator to form equipment operating status characteristics. Equipment operating status characteristics reflecting the overall operating status of urban rail transit equipment are extracted. Based on these characteristics, the future operating status of the equipment is assessed and calculated. By comprehensively analyzing the degree of abnormality, deterioration, and risk level reflected by the equipment operating status characteristics, the probability of failure is obtained. Simultaneously, based on the same equipment operating status characteristics, the current operating status of the equipment is quantitatively evaluated to form a health status score reflecting the overall health level of the equipment.

[0046] S3.3. Based on historical fault logs, statistical analysis of fault events is performed to form a fault occurrence time distribution. The equipment operating status characteristics are matched with the equipment operating status characteristics corresponding to the historical degradation stages. Based on the fault occurrence time distribution, lifetime estimation is performed to obtain the remaining lifetime assessment results.

[0047] Specifically, after extracting the equipment operating status features and obtaining the failure probability and health status score, the failure occurrence time distribution corresponding to the equipment identifier is obtained. The failure occurrence time distribution is then grouped according to equipment type and operating condition. The equipment operating status features are matched with the failure occurrence time distribution. By comparing the similarity of the equipment operating status features in historical samples, the historical degradation stages that match the equipment operating status features are determined, such as the initial health stage, early degradation stage, mid-term degradation stage, late degradation stage, and near-failure stage. The failure occurrence time distribution features corresponding to the historical degradation stages are then extracted. Based on the failure occurrence time distribution features, the expected time range from the current moment to the failure occurrence moment is determined and converted into the remaining life assessment result.

[0048] S3.4. Big data analytics environment refers to the data processing and operation environment formed on the basis of continuous collection and unified organization of multi-source operation data, historical fault logs and maintenance records of urban rail transit equipment. The big data analytics environment can perform parallel statistical analysis, predictive calculation and historical distribution update on indicator sequences organized by time window.

[0049] Specifically, multi-source operational data, historical fault logs, and maintenance records of urban rail transit equipment are collected and organized according to equipment identification, equipment type, and collection time. A traceable correspondence is established for these data. The indicator sequences formed based on the standardized operational dataset are organized according to time windows, and parallel statistical analysis is performed on the indicator sequences in a big data analytics environment to simultaneously complete the statistical summary of health status indicators, deterioration trend indicators, and risk characteristic indicators for multiple devices and multiple time windows. Predictive calculations are performed on the indicator sequences to support obtaining fault occurrence probability, health status score, and remaining life assessment results within future time windows. The fault occurrence probability, health status score, and remaining life assessment results are continuously collected, and historical statistical distribution information and fault occurrence time distribution are updated in conjunction with historical fault logs and maintenance records.

[0050] S4. Perform uncertainty assessment on the probability of failure, health status score and historical statistical distribution information to generate risk quantification value, and map the risk quantification value to risk level.

[0051] S4.1. Based on the probability of failure and health status score, the data is organized according to the equipment identification and prediction time to form prediction result data.

[0052] Specifically, based on the failure probability and health status score, and after verifying the consistency between the failure probability and health status score to confirm that both correspond to the same equipment identifier and the same future time window, the equipment identifier information and prediction time information corresponding to the failure probability and health status score are extracted, and then associated with the failure probability and health status score. The failure probability and health status score are then collected and organized according to the equipment identifier, so that the failure probability and health status score corresponding to different equipment identifiers form independent records. The failure probability and health status score corresponding to each equipment identifier are sorted according to the prediction time order, so that the failure probability and health status score form a continuous sequence in the prediction time dimension. Finally, the failure probability and health status scores collected by equipment identifier and arranged according to the prediction time order are summarized to form the prediction result data.

[0053] S4.2. Collect historical statistical distribution information and establish historical distributions of failure probability and health status to form historical distribution data; compare and analyze the prediction results data with the historical distribution data, and calculate the degree of deviation of failure probability and health status score from the historical distribution to generate deviation amount.

[0054] Specifically, historical statistical distribution information is collected, and a set of fault occurrence probability samples and a set of health status score samples corresponding to equipment type, operating conditions and time window are extracted. The set of fault occurrence probability samples and the set of health status score samples are then aggregated and organized. Based on the set of fault occurrence probability samples, a historical distribution of fault probability is statistically generated, and based on the set of health status score samples, a historical distribution of health status is statistically generated, forming historical distribution data that includes the historical distribution of fault probability and the historical distribution of health status. Based on the corresponding equipment identifier and the probability of failure and health status score in the prediction result data, the prediction result data is matched with the historical distribution data according to equipment type and operating condition. The probability of failure is compared and analyzed with the historical distribution of failure probability and the degree of deviation of the probability of failure in the historical distribution of failure probability is calculated. At the same time, the health status score is compared and analyzed with the historical distribution of health status and the degree of deviation of the health status score in the historical distribution of health status is calculated. The deviation is then summarized to form the deviation amount.

[0055] S4.3. Based on the deviation, calculate the uncertainty characterization of the prediction results data, and jointly calculate the uncertainty characterization with the failure probability and health status score to generate a risk quantification value.

[0056] Specifically, based on the deviation, the deviation corresponding to the failure probability and the deviation corresponding to the health status score are normalized to obtain comparable deviation representations. These deviation representations are then combined to form the uncertainty representation of the prediction result data. Based on the failure probability and health status score in the prediction result data, a consistent correlation is established between the failure probability and health status score according to the equipment identifier and future time window. The process of jointly calculating the uncertainty representation and the failure probability includes: scaling both the uncertainty representation and the failure probability to ensure they are within the same numerical range; and then scaling the uncertainty representation... The uncertainty representation and the probability of failure are each assigned a weight. Based on these weights, the uncertainty representation and the probability of failure are weighted and synthesized to obtain a first joint result representing the combined impact of the probability of failure and the predicted volatility. The process of jointly calculating the uncertainty representation and the health status score includes scaling the uncertainty representation and the health status score separately, assigning weights to the uncertainty representation and the health status score separately, and weighting and synthesizing the uncertainty representation and the health status score based on these weights to obtain a second joint result representing the combined impact of the health level and the predicted volatility. The first joint result and the second joint result are scaled and then weighted and fused to obtain the risk quantification value.

[0057] S4.4. Based on historical statistical distribution information, the relative position of the risk quantification value in the overall distribution is characterized to obtain the distribution position characterization.

[0058] Specifically, based on the equipment identifier and future time window corresponding to the risk quantification value, the equipment identifier, future time window and risk quantification value are associated and recorded; the risk quantification values ​​are collected and organized according to the equipment identifier, so that the risk quantification values ​​corresponding to different equipment identifiers form independent records, and the risk quantification values ​​are arranged in the order of future time windows within each equipment identifier to form risk quantification data; Extract a set of historical risk quantification values ​​corresponding to equipment type, operating conditions and time windows from historical statistical distribution information, and establish an overall distribution of risk quantification values ​​based on the set of historical risk quantification values; compare and analyze the risk quantification values ​​in the risk quantification data with the overall distribution of risk quantification values, calculate the relative position of the risk quantification values ​​in the overall distribution of risk quantification values, and represent the relative position of the risk quantification values ​​as a distribution position representation, thus generating a distribution position representation.

[0059] S4.5. Based on the distribution location representation, determine the relative position range of the risk quantification value in the historical statistical distribution information, and assign the corresponding risk level according to the position range.

[0060] Specifically, the distribution location representation is associated with the risk quantification value to form a distribution location record of the risk quantification value; based on the overall distribution of the risk quantification value obtained from historical statistical distribution information, the first and second quantile boundaries of the overall distribution of the risk quantification value are calculated. The first and second quantile boundaries are used to divide the overall distribution of the risk quantification value into three adjacent and non-overlapping value intervals; when the risk quantification value is less than the first quantile boundary, the risk quantification value is classified as a low-risk interval and the risk level is determined to be low-risk; when the risk quantification value is between the first and second quantile boundaries, the risk quantification value is classified as a medium-risk interval and the risk level is determined to be medium-risk; when the risk quantification value is greater than the second quantile boundary, the risk quantification value is classified as a high-risk interval and the risk level is determined to be high-risk; the equipment identifier, future time window, risk quantification value, and risk level are summarized to complete the mapping of risk quantification value to risk level.

[0061] S5. Based on the risk level and risk change trend, determine the priority and timing of equipment operation and maintenance, and generate operation and maintenance management decision information.

[0062] S5.1. Organize the risk levels and their corresponding risk quantification values ​​to form a risk level time series.

[0063] Specifically, based on the risk level and the corresponding risk quantification value, the equipment identification information and time window information corresponding to the risk level and risk quantification value are extracted; the equipment identification information and time window information are associated with the risk level and risk quantification value to form a risk level record containing equipment identification, time window, risk level, and risk quantification value; the risk level records are collected and organized according to equipment identification, so that the risk level records corresponding to different equipment identifications form independent sets; after the equipment identification collection and organization is completed, the risk level records corresponding to each equipment identification are sorted in chronological order, so that the risk level and risk quantification value form a continuous arrangement relationship in the time dimension; the risk levels and risk quantification values ​​collected by equipment identification and arranged in chronological order are summarized to form a risk level time series.

[0064] S5.2. Based on the risk level time series analysis, analyze the direction and magnitude of risk level changes over time and output the risk change trend; based on the risk level within the current time window and combined with the risk change trend, comprehensively assess the urgency of equipment operation and maintenance and form the equipment operation and maintenance priority.

[0065] Specifically, the risk level time series is compared with adjacent time windows in chronological order. The rise and fall of the risk level between adjacent time windows is statistically analyzed to determine the direction of risk level change over time. The range of risk level change within consecutive time windows is also statistically analyzed to determine the magnitude of risk level change over time. The direction and magnitude of change are combined to represent the risk change trend. The system reads the risk level within the current time window and performs correlation analysis between the risk level and risk change trend to generate the risk level and risk evolution status of the equipment within the current time window. Based on the risk level and risk change trend within the current time window, the system comprehensively assesses the urgency of equipment maintenance. Higher urgency is assigned to situations with high risk levels and rising risk change trends, while lower urgency is assigned to situations with low risk levels and falling risk change trends, thus forming a priority for equipment maintenance.

[0066] S5.3. Based on the operation and maintenance handling priority and the rate of change corresponding to the risk change trend, sort the handling time sequence of the equipment to form the equipment handling sequence; based on the operation and maintenance handling priority and the equipment handling sequence, generate operation and maintenance management decision information.

[0067] Specifically, based on the maintenance and handling priority of each device and the risk change trend corresponding to the maintenance and handling priority, the rate of change of risk level in the risk change trend is further obtained; the maintenance and handling priority and the rate of change are jointly organized, and the devices are initially sorted based on the maintenance and handling priority, so that the devices with higher maintenance and handling priorities are placed in the first position. When the maintenance and handling priorities are the same, the devices are further sorted according to the rate of change of risk change trend, so that the devices with faster risk change rate are prioritized over the devices with slower risk change rate, forming the time order of device handling and generating the device handling sequence; the device identification, maintenance and handling priority, and device handling sequence are summarized and organized to form maintenance management decision information.

[0068] S6. Summarize the failure probability, health status score, remaining life assessment results and risk level to generate failure prediction information and risk warning information for operation and maintenance management.

[0069] S6.1. Based on the probability of failure, health status score, remaining life assessment results and risk level, the data is compiled according to equipment identification and future time window to form a summary data.

[0070] Specifically, based on the equipment identifier and future time window corresponding to the failure probability, health status score, remaining life assessment result, and risk level, the process involves: verifying the consistency between the equipment identifier and the future time window to confirm that the failure probability, health status score, remaining life assessment result, and risk level correspond to the same equipment identifier and the same future time window; associating and recording the failure probability, health status score, remaining life assessment result, and risk level with the equipment identifier and future time window, and then collecting and organizing them according to the equipment identifier; finally, sorting and summarizing the failure probability, health status score, remaining life assessment result, and risk level according to the future time window order to form summary data.

[0071] S6.2. Perform field alignment and correlation processing on the summarized data to form the correspondence between failure probability, health status score, remaining life assessment results and risk level, and generate predictive element correlation data.

[0072] Specifically, based on the equipment identifier, future time window, failure probability, health status score, remaining useful life assessment result, and risk level fields contained in the aggregated data, the equipment identifier field and the future time window field are checked for integrity. Using the equipment identifier and the future time window as the association key, the failure probability, health status score, remaining useful life assessment result, and risk level in the aggregated data are aligned. This ensures that the failure probability, health status score, remaining useful life assessment result, and risk level of the same equipment identifier and corresponding to the same future time window are in a one-to-one correspondence within the same record. The records after field alignment, consistency checks, and association processing are aggregated to generate predictive element association data.

[0073] S6.3. Based on the associated data of the predictive elements, the future time window is used as the time index to organize and summarize the results of the failure probability and remaining life assessment in sequence, and to express the possibility of failure and the change of remaining life within the future time window in a centralized manner, thus forming failure prediction information.

[0074] Specifically, the system reads the equipment identifier, future time window, failure probability, health status score, remaining life assessment result, and risk level contained in the associated data of the prediction elements one by one, and verifies the completeness of the associated data of the prediction elements; it also collects and organizes the future time window, failure probability, and remaining life assessment result according to the equipment identifier, so that the future time window, failure probability, and remaining life assessment result corresponding to the same equipment form a continuous record; and it summarizes and describes the changes in failure probability and remaining life assessment results within the future time window to form failure prediction information for operation and maintenance management.

[0075] S6.4. Extract equipment identifiers, future time windows, and risk levels from the associated data of predictive elements, and generate risk warning records based on changes in risk levels. Summarize the risk warning records to form risk warning information for operation and maintenance management.

[0076] Specifically, the system extracts equipment identifiers, future time windows, and risk levels from the associated data of predictive elements. Risk levels are then grouped according to equipment identifiers and arranged in the order of future time windows to generate a risk level sequence corresponding to each equipment identifier. The system compares adjacent future time windows of the risk level sequence, recording changes in risk levels (increase, remain the same, or decrease). The system combines equipment identifiers, future time windows, risk levels, and changes in risk levels to generate risk warning records. These risk warning records are then summarized and organized according to equipment identifiers and future time windows to form risk warning information for operation and maintenance management.

[0077] In summary, this invention constructs a multi-dimensional dynamic indicator sequence that integrates health status, degradation trends, and risk characteristics, and combines historical statistical distributions to assess the uncertainty of prediction results, generating a risk quantification value that can be mapped to a risk level. Under the framework of big data analysis, this invention effectively enhances the credibility, comparability, and decision-making guidance value of fault prediction results, supports the formation of intelligent operation and maintenance decisions with clear priorities and reasonable timing, and thus significantly improves the safety, reliability, and resource utilization efficiency of urban rail transit equipment operation and maintenance.

[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting faults in urban rail transit equipment based on big data analysis, characterized in that: include, Collect voltage, current, temperature, vibration and acoustic operation data and combine them with historical fault logs and maintenance records. Perform preprocessing according to time windows to generate a standardized operation dataset. Based on standardized operational datasets, health status indicators, degradation trend indicators, and risk characteristic indicators are constructed according to equipment type to form an indicator sequence. Based on the indicator sequence, fault prediction calculations are performed in a big data analysis environment to obtain the probability of fault occurrence, health status score and remaining life assessment results. Uncertainty assessment is performed on the probability of failure, health status score and historical statistical distribution information to generate risk quantification value, and the risk quantification value is mapped to risk level; Based on the risk level and risk change trend, determine the priority and timing of equipment operation and maintenance, and generate operation and maintenance management decision information; The probability of failure, health status score, remaining life assessment results and risk level are summarized to generate failure prediction information and risk warning information for operation and maintenance management.

2. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for generating the standardized runtime dataset are as follows. Collect voltage, current, temperature, vibration and acoustic operating data generated by urban rail transit equipment during operation, and combine them with historical fault logs and maintenance records to generate multimodal operating parameters; Based on the multimodal operating parameters organized according to time windows, abnormal sampling points are identified and noise suppression is performed to generate preprocessed multimodal operating parameters; Dimensional normalization is performed on different types of operating parameters in the preprocessed multimodal operating parameters to form a standardized operating dataset.

3. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for forming the indicator sequence are as follows: The standardized operational dataset is grouped according to device identifier and device type, and divided into multiple continuous time windows. For each time window, the baseline deviation of each operational data point relative to historical healthy samples of the same device type is calculated, and a health status index is generated through monotonic mapping relationship and weighted fusion. By jointly characterizing the health status indicators and the rate of change of baseline deviation over time in adjacent time windows, a deterioration trend indicator is generated; risk characteristic indicators are generated based on the fluctuation amplitude of the standardized operational dataset within the time window, the cross-modal correlation, and the frequency of abnormal sampling points. Health status indicators, deterioration trend indicators, and risk characteristic indicators within the same time window are correlated to form a time window-level indicator combination, which is then arranged in chronological order to form an indicator sequence.

4. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for obtaining the failure probability, health status score, and remaining life assessment results are as follows. Based on the indicator sequence organized by device identifier, and using a sliding time window to segment the continuous time series, multiple time window level indicator combinations within each sliding time window are spliced ​​together in chronological order to form the prediction input sequence; In a big data analytics environment, joint analysis of the predicted input sequence is performed to extract the characteristics of the equipment's operating status. Based on the characteristics of equipment operating status, fault prediction calculation is performed to obtain the probability of fault occurrence, and a status score is calculated to obtain a health status score. Based on historical fault logs, fault events are statistically analyzed to form a fault occurrence time distribution. The equipment operating status characteristics are matched with the equipment operating status characteristics corresponding to the historical degradation stages. Based on the fault occurrence time distribution, lifetime estimation is performed to obtain the remaining lifetime assessment results.

5. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 4, characterized in that: The big data analytics environment refers to a data processing and operating environment formed on the basis of collecting and uniformly organizing multi-source operating data, historical fault logs, and maintenance records of urban rail transit equipment.

6. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for generating the risk quantification value are as follows: Based on the probability of failure and health status score, the data is organized according to equipment identification and prediction time to form prediction results data; Collect historical statistical distribution information and establish historical distributions of failure probability and health status to form historical distribution data; compare and analyze the prediction results data with the historical distribution data, and calculate the degree of deviation of failure probability and health status score from the historical distribution to generate deviation amount; The uncertainty characterization of the prediction results is calculated based on the deviation, and the uncertainty characterization is jointly calculated with the failure probability and health status score to generate a risk quantification value.

7. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for mapping risk quantification values ​​to risk levels are as follows. Based on historical statistical distribution information, the relative position of the risk quantification value in the overall distribution is characterized to obtain the distribution position characterization; Based on the distribution location representation, the relative position range of the risk quantification value in the historical statistical distribution information is determined, and the corresponding risk level is assigned according to the position range.

8. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for generating operation and maintenance management decision information are as follows: The risk levels and their corresponding quantitative values ​​are compiled to form a time series of risk levels; Based on time series analysis of risk levels, the direction and magnitude of risk level changes over time are analyzed, and the trend of risk change is output. Based on the risk level within the current time window and the trend of risk changes, a comprehensive assessment of the urgency of equipment maintenance is conducted to determine the priority of equipment maintenance. Based on the priority of operation and maintenance (O&M) handling and the rate of change corresponding to the trend of risk change, the order of equipment handling time is sorted to form the equipment handling sequence; based on the O&M handling priority and the equipment handling sequence, O&M management decision information is generated.

9. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The process of summarizing the probability of failure, health status score, remaining life assessment results, and risk level is as follows: Based on the probability of failure, health status score, remaining life assessment results and risk level, the data is compiled according to equipment identification and future time window to form a summary data; The aggregated data is aligned and correlated to form a correspondence between failure probability, health status score, remaining life assessment results and risk level, and generate predictive element correlation data.

10. The urban rail transit equipment fault prediction method based on big data analysis as described in claim 1, characterized in that: The specific steps for generating fault prediction information and risk warning information for operation and maintenance management are as follows: Based on the associated data of predictive elements, and using the future time window as the time index, the results of the failure probability and remaining life assessment are organized and summarized in sequence, and the failure prediction information is formed by centrally expressing the possibility of failure and the change of remaining life within the future time window. The device identifier, future time window, and risk level are extracted from the associated data of the predictive elements, and combined with the changes in the risk level, risk warning records are generated. The risk warning records are then summarized to form risk warning information for operation and maintenance management.

Images