Rail transit equipment fault prediction method and system

Abstract

The present invention provides a rail transit equipment fault prediction method and system. The method includes: obtaining operating data and environmental data of the rail transit equipment in the current time period, determining the time series correlation coefficient and statistical characteristic dispersion of the operating data, and the environmental impact factor of the environmental data affecting the rail transit equipment fault; determining the weight parameters of a time series prediction model based on the time series correlation coefficient, statistical characteristic dispersion, and environmental impact factor, and determining the weight parameters of a classification model based on the weight parameters of the time series prediction model; inputting the operating data of the current time period into the time series prediction model, and obtaining the operating data of the rail transit equipment in the future time period based on the weight parameters of the time series prediction model; inputting the operating data of the future time period into the classification model, obtaining the failure probability of the rail transit equipment based on the weight parameters of the classification model, and determining whether the rail transit equipment has failed based on the failure probability. The present invention improves the accuracy and reliability of fault prediction.

Description

Technical Field

[0001] The present invention relates to the field of intelligent operation and maintenance technology for rail transit, and in particular to a method and system for predicting faults of rail transit equipment. Background Art

[0002] Currently, with the rapid development of rail transportation, especially high-speed railways, the complexity and difficulty of railway equipment operation and maintenance are increasing. Traditional railway operation and maintenance methods mainly rely on manual regular maintenance and emergency response and post-fault repairs after failures. This method has many disadvantages:

[0003] 1. Inefficiency: The frequency of manual inspections is limited by human resources, making it difficult to achieve real-time monitoring and dynamic perception of the status of rail transit equipment, which can easily lead to over-maintenance or under-maintenance.

[0004] 2. High cost: Frequent manual maintenance and emergency fault handling result in high operation and maintenance costs for rail transit.

[0005] 3. Safety hazards: If a fault is not handled until it occurs, it is very likely to cause a serious safety accident and pose a huge threat to the safety of railway transportation.

[0006] In recent years, with the development and support of next-generation information technologies such as big data, the Internet of Things (IoT), digital twins, artificial intelligence (AI), and the railway-specific 5G communication protocol (5G-Railway, 5G-R), predictive maintenance (PM) has gradually become a research hotspot in the field of intelligent rail transit O&M. By monitoring the status of railway equipment in real time and screening, diagnosing, and predicting potential fault risks at an early stage, it is possible to effectively improve rail transit O&M efficiency, reduce costs, and mitigate safety hazards. However, existing models generally exhibit poor generalization capabilities when faced with the complex and changing operating environment of railway equipment. For example, under the influence of factors such as varying climate conditions, train density, and equipment aging, the model's fault prediction accuracy can significantly decrease, making it difficult to meet actual O&M requirements. Summary of the Invention

[0007] The present invention provides a rail transit equipment fault prediction method and system, which are used to solve the defect of low fault prediction accuracy of rail transit equipment in the prior art and improve the fault prediction accuracy of rail transit equipment.

[0008] The present invention provides a rail transit equipment fault prediction method, comprising:

[0009] Acquire operation data and environmental data of rail transit equipment in a current time period, determine a time series correlation coefficient and a statistical characteristic dispersion of the operation data, and an environmental impact factor of the environmental data on the rail transit equipment failure;

[0010] Determining weight parameters of a time series prediction model based on the time series correlation coefficient, the statistical feature dispersion, and the environmental influencing factor, and determining weight parameters of a classification model based on the weight parameters of the time series prediction model;

[0011] Inputting the operation data of the current time period into the time series prediction model, and obtaining the operation data of the rail transit equipment in the future time period according to the weight parameters of the time series prediction model;

[0012] The operation data of the future time period is input into the classification model, the failure probability of the rail transit equipment is obtained according to the weight parameter of the classification model, and whether the rail transit equipment is faulty is determined according to the failure probability.

[0013] According to a rail transit equipment fault prediction method provided by the present invention, the weight parameters of the time series prediction model are determined according to the time series correlation coefficient, the statistical characteristic dispersion and the environmental influencing factor by the following formula, and the weight parameters of the classification model are determined according to the weight parameters of the time series prediction model:

[0014]

[0015] Among them, W LSTM and W RF are the current weight parameters of the time series prediction model and the classification model, α (t) and β (t) are the time series correlation coefficient and statistical characteristic dispersion corresponding to the current time period including the current time t, is the environmental impact factor.

[0016] According to a rail transit equipment fault prediction method provided by the present invention, the calculation formula of the time series correlation coefficient α is:

[0017]

[0018] Among them, x i is the operating data at the i-th moment in the current time period, is the mean value of the running data in the current time period, k is the lag order, N is the duration of the current time period, w i is the time attenuation weight at the i-th moment, and the calculation formula is:

[0019]

[0020] in, is the time attenuation coefficient, which is obtained by fitting historical data.

[0021] According to a rail transit equipment fault prediction method provided by the present invention, the calculation formula of the statistical characteristic dispersion β is:

[0022]

[0023] Among them, p i is the distribution probability density of the running data at the i-th moment in the current time period, Var(x) is the variance of the running data in the current time period, Var max is the historical maximum variance, δ is the volatility weight coefficient, and n is the length of the current time period.

[0024] According to a rail transit equipment fault prediction method provided by the present invention, the environmental impact factor The calculation formula is:

[0025]

[0026] in, is the aging coefficient of the rail transit equipment, a, b and c are fitting coefficients, H is the ambient humidity, H max is the preset maximum ambient humidity, G is the ambient wind speed, G max is the preset maximum ambient wind speed, T is the ambient temperature, T max The preset maximum ambient temperature.

[0027] According to a rail transit equipment fault prediction method provided by the present invention, after determining whether the rail transit equipment has failed according to the fault probability, the method further includes:

[0028] determining a current prediction error rate according to whether the rail transit equipment is faulty;

[0029] When the current prediction error rate is greater than a preset threshold, the allocation weight of the time series prediction model in the next time period is determined by the following formula:

[0030]

[0031] in, is the allocation weight of the time series prediction model in the next time period including time t+1, is the current weight parameter of the time series prediction model, E t is the current prediction error rate, is the learning rate;

[0032] Inputting the operating data of the rail transit equipment in the next time period into the time series prediction model, and obtaining the operating data of the rail transit equipment in the future time period according to the weight parameter of the time series prediction model in the next time period;

[0033] The operating data of the future time period obtained according to the weight parameter of the time series prediction model in the next time period is multiplied by the allocated weight of the time series prediction model in the next time period and input into the classification model. The failure probability of the rail transit equipment is obtained according to the weight parameter of the classification model in the next time period, and whether the rail transit equipment will fail at the next moment is determined according to the failure probability.

[0034] According to a rail transit equipment fault prediction method provided by the present invention, determining whether the rail transit equipment is faulty according to the fault probability includes:

[0035] Determine the critical value P of the failure probability according to the environmental data e , P e The calculation formula is:

[0036]

[0037] in, is the aging coefficient of the rail transit equipment, H is the ambient humidity, G is the ambient wind speed, T is the ambient temperature, is the equipment operating time factor, 、 、 and They are the influence coefficients of ambient humidity, ambient wind speed, ambient temperature and equipment operating time factors, The calculation formula is:

[0038]

[0039] Wherein, t is the running time of the rail transit equipment, t max The design life of the rail transit equipment;

[0040] When the failure probability of the rail transit equipment is greater than a critical value of the failure probability, it is determined that the rail transit equipment has failed, and an early warning is issued for the rail transit equipment.

[0041] According to a rail transit equipment fault prediction method provided by the present invention, the warning level of the rail transit equipment is determined by the following formula:

[0042]

[0043] Among them, P fis the failure probability of the rail transit equipment, L is the remaining life of the rail transit equipment, L max is the design life of the rail transit equipment, H is the ambient humidity, and H max To preset the maximum ambient humidity, is the weight coefficient, and round is the rounding function.

[0044] According to a rail transit equipment fault prediction method provided by the present invention, the remaining life R of the rail transit equipment is predicted by Weibull distribution and equipment degradation model, and the formula is as follows:

[0045]

[0046] in, and are the characteristic life and shape parameters of the Weibull distribution parameters, and t is the current time.

[0047] The present invention also provides a rail transit equipment fault prediction system, comprising:

[0048] A calculation module is used to obtain operating data and environmental data of the rail transit equipment in the current time period, determine the time series correlation coefficient and statistical characteristic dispersion of the operating data, and the environmental impact factor of the environmental data on the rail transit equipment failure;

[0049] A determination module, configured to determine weight parameters of a time series prediction model based on the time series correlation coefficient, the statistical feature dispersion, and the environmental influencing factor, and to determine weight parameters of a classification model based on the weight parameters of the time series prediction model;

[0050] a first prediction module, configured to input the operation data of the current time period into the time series prediction model, and obtain the operation data of the rail transit equipment in a future time period according to the weight parameters of the time series prediction model;

[0051] The second prediction module is used to input the operation data of the future time period into the classification model, obtain the failure probability of the rail transit equipment according to the weight parameters of the classification model, and determine whether the rail transit equipment is faulty according to the failure probability.

[0052] The rail transit equipment fault prediction method and system provided by the present invention use a time series prediction model to predict the future trend of the equipment status, use a classification model to determine whether the equipment is likely to fail, and improve the accuracy and reliability of fault prediction through multi-model fusion; at the same time, the weight parameters of the time series prediction model and the classification model are dynamically determined according to the real-time operating data and environmental data of the rail transit equipment, which is applicable to fault prediction of different types of data, further improving the accuracy and universality of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 It is a flow chart of the rail transit equipment fault prediction method provided by the present invention;

[0055] Figure 2 This is a schematic diagram of the architecture of a digital railway predictive operation and maintenance system driven by multi-source heterogeneous data in the rail transit equipment fault prediction method provided by the present invention;

[0056] Figure 3 This is a schematic diagram of the data collection and preprocessing process in the rail transit equipment fault prediction method provided by the present invention;

[0057] Figure 4 This is a schematic diagram of the fault prediction model training and reasoning process in the rail transit equipment fault prediction method provided by the present invention;

[0058] Figure 5 This is a schematic diagram of the intelligent early warning and decision support process in the rail transit equipment fault prediction method provided by the present invention;

[0059] Figure 6 It is a structural diagram of the rail transit equipment fault prediction system provided by the present invention. DETAILED DESCRIPTION

[0060] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0061] The following combination Figure 1 A rail transit equipment fault prediction method of the present invention includes:

[0062] Step 101: Acquire operation data and environmental data of rail transit equipment in a current time period, determine a time series correlation coefficient and a statistical characteristic dispersion of the operation data, and an environmental impact factor of the environmental data on the rail transit equipment failure;

[0063] Step 102: determining weight parameters of a time series prediction model based on the time series correlation coefficient, the statistical feature dispersion, and the environmental impact factor, and determining weight parameters of a classification model based on the weight parameters of the time series prediction model;

[0064] Step 103: inputting the operation data of the current time period into the time series prediction model, and obtaining the operation data of the rail transit equipment in the future time period according to the weight parameters of the time series prediction model;

[0065] Step 104 : Input the operation data of the future time period into the classification model, obtain the failure probability of the rail transit equipment according to the weight parameter of the classification model, and determine whether the rail transit equipment is faulty according to the failure probability.

[0066] This embodiment is primarily applicable to the health management of key equipment, including railway track (including components such as turnout points and stock rails), catenary systems (including insulators and catenary cables), rolling stock (such as key components like traction motors and wheelset bearings), and signaling systems (such as switches, signaling devices, and track circuits). Its core objectives are to achieve early screening of potential faults, predict remaining service life, provide intelligent monitoring and early warning, and optimize operations and maintenance decisions. By deeply integrating big data and artificial intelligence technologies, a closed-loop management mechanism encompassing "data collection and monitoring - preprocessing and analysis - fault prediction - safety warning - and operations and maintenance decision-making" is established, thereby comprehensively improving the safe operation efficiency of rail transit and reducing safety hazards.

[0067] Sensors (such as acceleration sensors, temperature sensors, current sensors, fiber optic sensors, etc.) are installed at key parts of railway equipment (such as turnout machines, contact network insulators, locomotive traction motors and signal machines) to collect equipment operation data in real time.

[0068] like Figure 2 As shown, the digital railway predictive operation and maintenance system architecture driven by multi-source heterogeneous data includes a data acquisition layer, an edge computing layer, a cloud analysis layer, and a decision output layer. It shows that the data acquisition module transmits multi-source heterogeneous data to the edge computing layer and the cloud analysis layer in real time through the railway 5G-R dedicated protocol, and the data preprocessing module cleans and integrates the data. The fault prediction module makes predictions based on LSTM and RF models. The intelligent early warning module generates early warning information and pushes it to operation and maintenance personnel through a visual interface or mobile terminal. The decision support module provides rule-based operation and maintenance decision recommendations. The data interaction and functional collaboration between the modules are clearly presented.

[0069] like Figure 3As shown, a photoelectro-mechanical-electrical sensor array (PMSA) network, distributed optical fiber sensing (DOFS), and IoT devices can be used to collect real-time operational data (such as vibration, temperature, current, voltage, strain, displacement, etc.) and environmental data (such as temperature, humidity, and wind speed) from rail transit equipment, ensuring that the collected data fully reflects the equipment's operating status. The PMSA network includes MEMS accelerometers (accuracy requirement: ±0.1g), infrared temperature sensors (resolution requirement: 0.1°C), and Hall effect current sensors (accuracy requirement: ±0.5%). The sampling frequency can be dynamically adjusted based on actual needs, ranging from 5Hz to 1kHz.

[0070] Collected data is transmitted via an IoT gateway and synchronized to cloud servers and edge computing nodes, forming a Cloud-Edge Collaborative Architecture (CECA). This enables real-time data backhaul via the railway-specific 5G-R protocol, with edge node pre-processing latency ≤ 30ms (with a required verification interval of 10-200ms). This CECA includes edge computing nodes and cloud servers.

[0071] During data collection, transmission, and storage, the AES-256 encryption algorithm (compliant with FIPS 197 standards) is used, and digital signatures are performed using the national secret SM2 algorithm to ensure all-round data security and privacy protection, effectively preventing the risk of data leakage and tampering.

[0072] like Figure 3 As shown in the figure, the collected raw data can be cleaned to remove outliers and noise caused by sensor failures or external interference. Multi-source data is fused and normalized to generate a unified feature vector. Wavelet Packet Decomposition (WPD) is used to process non-stationary signals and extract key features that accurately reflect the device status. This significantly improves feature extraction efficiency, resulting in a high-quality dataset that provides a reliable data foundation for subsequent fault prediction. The preprocessed data is then input into the dynamic weighted multi-model prediction engine for fault prediction.

[0073] The time series prediction model in this embodiment can be a long short-term memory (LSTM) network, used to predict future trends in device status. The classification model can employ machine learning models, such as deep learning, random forests (RF), and support vector machines, to determine whether a device is likely to fail. A fault prediction model is constructed based on the time series prediction model and the classification model to achieve real-time prediction of device failures. This hybrid model integration effectively improves prediction accuracy and enhances the model's generalization and adaptability.

[0074] A dynamic weight allocation model for the time series prediction model and classification model is constructed, and the weight parameters of the time series prediction model and classification model are dynamically adjusted according to data characteristics.

[0075] Figure 4 Demonstrates the model training process, including data input, model selection, and parameter optimization, as well as the inference process, including real-time data input and fault probability output. The LSTM model is trained using historical data, effectively capturing long-term dependencies in time series data such as vibration, temperature, and current, and accurately predicting future trends in device status. Furthermore, it integrates RF models to classify and predict device fault types, enabling early diagnosis of equipment components with potential failure risks.

[0076] Existing technologies have low data utilization rates and insufficient capabilities for integrating and processing heterogeneous data from multiple sources. This makes it difficult to effectively integrate data collected by sensors from different devices, resulting in significant data silos. For example, data from disparate sources, such as track vibration data, catenary temperature data, and locomotive current data, cannot be interconnected and collaboratively analyzed.

[0077] This embodiment uses a time series prediction model to predict future trends in equipment status and a classification model to determine whether equipment failures are likely to occur, thereby improving the accuracy and reliability of fault prediction through multi-model fusion. At the same time, the weight parameters of the time series prediction model and the classification model are dynamically determined based on the real-time operating data and environmental data of rail transit equipment, making it suitable for fault prediction of different types of data, thereby further improving the accuracy and universality of fault prediction.

[0078] Based on the above embodiment, this embodiment determines the weight parameters of the time series prediction model according to the time series correlation coefficient, the statistical feature dispersion and the environmental influencing factor through the following formula, and determines the weight parameters of the classification model according to the weight parameters of the time series prediction model:

[0079]

[0080] Among them, W LSTM and W RFare the current weight parameters of the time series prediction model and the classification model, α (t) and β (t) are the time series correlation coefficient and statistical characteristic dispersion corresponding to the current time period including the current time t, is the environmental impact factor.

[0081] The weight parameters of LSTM and RF (random forest) are dynamically allocated based on α and β. Indicates environmental impact factors, mainly quantifying environmental humidity H (unit: %RH), wind speed G (unit: m / s), temperature T (unit: o C) The impact of the three on equipment failure; when environmental factors When it approaches 1, the proportion of LSTM weight parameters increases, which is suitable for scenarios with significant time series features (such as vibration signal analysis); when β is large (data dispersion is high) and When the value is smaller, the RF weight parameter is increased, making it suitable for non-stationary signals. This dynamic weight parameter model is executed in real time by the FPGA module of the edge computing node through the hardware collaboration of "sensor array-edge computing-5G-R transmission".

[0082] Based on the above embodiment, the calculation formula of the time series correlation coefficient α in this embodiment is:

[0083]

[0084] Among them, x i is the operating data at the i-th moment in the current time period, is the mean of the running data in the current time period, k is the lag order (dynamically selected by Akaike information criterion, ranging from 1 to 10), N is the duration of the current time period, w i is the time attenuation weight at the i-th moment, and the calculation formula is:

[0085]

[0086] in, is the time attenuation coefficient, which is obtained by fitting historical data.

[0087] α is the data time series correlation coefficient (the value range is 0~1, reflecting the periodicity or trend of the time series. The closer the value is to 1, the stronger the time series correlation is). The correlation of time series data can be calculated through the autocorrelation function.

[0088] Based on the above embodiment, the calculation formula of the statistical characteristic dispersion β in this embodiment is:

[0089]

[0090] Among them, p i is the distribution probability density of the running data at the i-th moment in the current time period, Var(x) is the variance of the running data in the current time period, Var max is the historical maximum variance, δ is the volatility weight coefficient, which is determined by experimental data, and n is the length of the current time period.

[0091] β is the statistical characteristic dispersion, which can be used to calculate the degree of dispersion of data distribution based on Shannon entropy.

[0092] Based on the above embodiments, in this embodiment is the environmental impact factor (related to humidity H, wind speed G, and temperature T), and the calculation formula is:

[0093]

[0094] in, is the aging coefficient of the rail transit equipment (fitted by historical operation and maintenance data, with a fitting error of ≤2%), a, b, and c are fitting coefficients determined by historical data, H is the ambient humidity, and H max is the preset maximum ambient humidity (indicates the maximum value allowed by the design), G is the ambient wind speed, G max is the preset maximum ambient wind speed (indicates the maximum value allowed by the design), T is the ambient temperature, T max The preset maximum ambient temperature (indicates the maximum value allowed by the design).

[0095] Based on the above embodiment, after determining whether the rail transit equipment is faulty according to the fault probability, this embodiment further includes:

[0096] determining a current prediction error rate according to whether the rail transit equipment is faulty;

[0097] When the current prediction error rate is greater than a preset threshold, the allocation weight of the time series prediction model in the next time period is determined by the following formula:

[0098]

[0099] in, is the allocation weight of the time series prediction model in the next time period including time t+1, is the current weight parameter of the time series prediction model, E t is the current prediction error rate, is the learning rate.

[0100] Establish a model update mechanism and trigger conditions for incremental learning: When the model prediction error rate threshold exceeds the preset tolerance (e.g., E>5%), activate the incremental learning mechanism to ensure continuous model optimization. Introduce a dynamic feedback adaptive learning mechanism to establish real-time adjustment of allocation weights based on prediction error.

[0101] Based on the above embodiments, in this embodiment, determining whether the rail transit equipment is faulty according to the fault probability includes:

[0102] Determine the critical value P of the failure probability according to the environmental data e , P e The calculation formula is:

[0103]

[0104] in, is the aging coefficient of the rail transit equipment (fitted by historical operation and maintenance data, with a fitting error of ≤2%), H is the ambient humidity, G is the ambient wind speed, and T is the ambient temperature. is the equipment operating time factor, 、 、 and They are the influence coefficients of ambient humidity, ambient wind speed, ambient temperature and equipment operating time factors, The calculation formula is:

[0105]

[0106] Wherein, t is the running time of the rail transit equipment, t max is the design life of the rail transit equipment; when the equipment operating time factor τ>1, the equipment has exceeded its service life and requires special attention.

[0107] When the failure probability of the rail transit equipment is greater than a critical value of the failure probability, it is determined that the rail transit equipment has failed, and an early warning is issued for the rail transit equipment.

[0108] Based on the above embodiments, the warning level of the rail transit equipment is determined by the following formula in this embodiment:

[0109]

[0110] Among them, P f is the failure probability of the rail transit equipment, L is the remaining life of the rail transit equipment, L max is the design life of the rail transit equipment, H is the ambient humidity, and H max To preset the maximum ambient humidity, is the weight coefficient, satisfying , round is the rounding function.

[0111] For example, the rules for dynamic adjustment of warning levels are:

[0112] Red warning: Failure probability P f >80% and the remaining life of the equipment L<30 days, or the failure probability P f >90% and the equipment operating time factor τ>90%. At this time, a red alert is triggered and a shutdown instruction is generated.

[0113] Orange warning: Failure probability P f >60% and ambient wind speed G>15m / s, or failure probability P f >70% and ambient humidity H>85%.

[0114] Yellow warning: Failure probability P f >40% and the equipment operation time exceeds 80% of the design life, or the failure probability P f >50% and Var(x)>10.

[0115] The existing early warning mechanism is imperfect. The early warning system lacks intelligent warning strategies and cannot dynamically adjust warning thresholds and levels based on equipment status. Furthermore, the decision support system is relatively simplistic, making it difficult to provide comprehensive and accurate operation and maintenance recommendations and resource optimization solutions.

[0116] like Figure 5 As shown, this embodiment generates warning messages of varying levels, such as red and orange, based on fault prediction results and critical equipment thresholds (e.g., vibration amplitude exceeding a set value, abnormal temperature rise, or excessive current fluctuation). Warning levels are dynamically adjusted based on the fault probability and critical thresholds. Warning information is promptly pushed to relevant maintenance personnel via a visual interface (e.g., the instrument panel of a rail operation and maintenance monitoring center) and smart mobile devices (e.g., their mobile phone apps), allowing them to quickly take appropriate measures to prevent the occurrence and escalation of faults and ensure safe equipment operation.

[0117] In addition, a rules-based decision support system is provided. Based on early warning information and historical turnout operation and maintenance data, it helps operators develop detailed track maintenance plans, such as scheduling regular turnout inspections, replacing worn parts, and optimizing the allocation of maintenance resources, thereby improving operation and maintenance efficiency and reducing costs. Based on turnout failure predictions and remaining service life, the system can rationally arrange maintenance personnel's working hours and the allocation of maintenance tools, ensuring the normal operation of turnouts while minimizing maintenance costs and the impact on rail transportation.

[0118] Based on the above embodiments, the remaining life R of the rail transit equipment in this embodiment is predicted by Weibull distribution and equipment degradation model, and the formula is as follows:

[0119]

[0120] in, and They are the characteristic life (63.2% of the devices fail before this time) and shape parameter (determines the failure mode) in the Weibull distribution parameters. <1 is early failure, =1 means random failure, >1 represents wear failure), obtained by fitting historical data with Weibull distribution, and t represents the current time.

[0121] Predict remaining useful life (≤5%) based on Monte Carlo simulation, combined with Weibull distribution and equipment degradation models.

[0122] Based on the maximum likelihood estimation (MLE) method, the Weibull distribution parameters ( , ), the iterative convergence condition is that the sum of squared residuals ≤ 0.01.

[0123] Derived from the fault prediction model, the Monte Carlo simulation steps are as follows: generate random samples of equipment life based on Weibull distribution (the sample size is required to be ≥1000); fit the Weibull distribution parameters (shape parameter φ and characteristic life η) through historical data; and calculate the confidence interval of the remaining life of the equipment (the confidence level is required to be 95%).

[0124] This embodiment predicts the remaining service life of the turnout based on Monte Carlo simulation, provides a scientific basis for operation and maintenance decision-making, and reasonably arranges maintenance plans and resource allocation.

[0125] The advantages of the present invention are as follows:

[0126] 1. Real-time: Through the Internet of Things and edge computing technologies, real-time monitoring and prediction of equipment status can be achieved, which can promptly detect potential failure risks and improve the efficiency of rail transit operation and maintenance.

[0127] 2. High precision: Based on big data and artificial intelligence technologies, through multi-model fusion (such as combining deep learning and traditional machine learning), the accuracy and reliability of fault prediction are improved, which can provide strong support for the intelligent operation and maintenance decision-making of digital rail transit.

[0128] 3. Intelligence: The intelligent early warning and decision support system can generate early warning levels and operation and maintenance suggestions based on fault prediction results, combined with historical equipment data and operation and maintenance experience, and push them to rail transit operation and maintenance personnel through a visual interface or mobile terminal so that they can quickly take appropriate measures to avoid further expansion of the fault, optimize the operation and maintenance process, and significantly reduce operation and maintenance costs.

[0129] 4. Scalability: The system supports multi-source data fusion and model updates, is applicable to different types of railway equipment, and has good adaptability and scalability.

[0130] The rail transit equipment fault prediction system provided by the present invention is described below. The rail transit equipment fault prediction system described below and the rail transit equipment fault prediction method described above can be referenced to each other.

[0131] like Figure 6 As shown, the system includes a calculation module 601, a determination module 602, a first prediction module 603 and a second prediction module 604, wherein:

[0132] The calculation module 601 is used to obtain the operation data and environmental data of the rail transit equipment in the current time period, determine the time series correlation coefficient and statistical characteristic dispersion of the operation data, and the environmental impact factor of the environmental data affecting the failure of the rail transit equipment;

[0133] The determination module 602 is used to determine the weight parameters of the time series prediction model according to the time series correlation coefficient, the statistical feature dispersion and the environmental influencing factor, and determine the weight parameters of the classification model according to the weight parameters of the time series prediction model;

[0134] The first prediction module 603 is used to input the operation data of the current time period into the time series prediction model, and obtain the operation data of the rail transit equipment in the future time period according to the weight parameters of the time series prediction model;

[0135] The second prediction module 604 is used to input the operation data of the future time period into the classification model, obtain the failure probability of the rail transit equipment according to the weight parameters of the classification model, and determine whether the rail transit equipment is faulty according to the failure probability.

[0136] This embodiment uses a time series prediction model to predict future trends in equipment status and a classification model to determine whether equipment failures are likely to occur, thereby improving the accuracy and reliability of fault prediction through multi-model fusion. At the same time, the weight parameters of the time series prediction model and the classification model are dynamically determined based on the real-time operating data and environmental data of rail transit equipment, making it suitable for fault prediction of different types of data, thereby further improving the accuracy and universality of fault prediction.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A rail transit equipment fault prediction method, characterized in that: include: Acquire operation data and environmental data of rail transit equipment in a current time period, determine a time series correlation coefficient and a statistical characteristic dispersion of the operation data, and an environmental impact factor of the environmental data on the rail transit equipment failure; Determining weight parameters of a time series prediction model based on the time series correlation coefficient, the statistical feature dispersion, and the environmental influencing factor, and determining weight parameters of a classification model based on the weight parameters of the time series prediction model; Inputting the operation data of the current time period into the time series prediction model, and obtaining the operation data of the rail transit equipment in the future time period according to the weight parameters of the time series prediction model; inputting the operation data of the future time period into the classification model, obtaining a failure probability of the rail transit equipment according to a weight parameter of the classification model, and determining whether the rail transit equipment is faulty according to the failure probability; The weight parameters of the time series prediction model are determined according to the time series correlation coefficient, the statistical feature dispersion and the environmental influencing factor by the following formula, and the weight parameters of the classification model are determined according to the weight parameters of the time series prediction model: ； Among them, W LSTM and W RF are the current weight parameters of the time series prediction model and the classification model, α (t) and β (t) are the time series correlation coefficient and statistical characteristic dispersion corresponding to the current time period including the current time t, is the environmental impact factor; The calculation formula of the time series correlation coefficient α is: ； Among them, x i is the operating data at the i-th moment in the current time period, is the mean value of the running data in the current time period, k is the lag order, N is the duration of the current time period, w i is the time attenuation weight at the i-th moment, and the calculation formula is: ； in, is the time attenuation coefficient, obtained by fitting historical data; The calculation formula of the statistical characteristic dispersion β is: ； Among them, p i is the distribution probability density of the running data at the i-th moment in the current time period, Var(x) is the variance of the running data in the current time period, Var max is the historical maximum variance, δ is the volatility weight coefficient, and n is the length of the current time period; The environmental impact factors The calculation formula is: ； in, is the aging coefficient of the rail transit equipment, a, b and c are fitting coefficients, H is the ambient humidity, H max is the preset maximum ambient humidity, G is the ambient wind speed, G max is the preset maximum ambient wind speed, T is the ambient temperature, T max The preset maximum ambient temperature.

2. The rail transit equipment fault prediction method according to claim 1, characterized in that: After determining whether the rail transit equipment is faulty according to the fault probability, the method further includes: determining a current prediction error rate according to whether the rail transit equipment is faulty; When the current prediction error rate is greater than a preset threshold, the allocation weight of the time series prediction model in the next time period is determined by the following formula: ； in, is the allocation weight of the time series prediction model in the next time period including time t+1, is the current weight parameter of the time series prediction model, E t is the current prediction error rate, is the learning rate; Inputting the operating data of the rail transit equipment in the next time period into the time series prediction model, and obtaining the operating data of the rail transit equipment in the future time period according to the weight parameter of the time series prediction model in the next time period; The operating data of the future time period obtained according to the weight parameter of the time series prediction model in the next time period is multiplied by the allocated weight of the time series prediction model in the next time period and input into the classification model. The failure probability of the rail transit equipment is obtained according to the weight parameter of the classification model in the next time period, and whether the rail transit equipment will fail at the next moment is determined according to the failure probability.

3. The rail transit equipment fault prediction method according to claim 1 or 2, characterized in that: Determining whether the rail transit equipment is faulty according to the fault probability includes: Determine the critical value P of the failure probability according to the environmental data e , P e The calculation formula is: ； in, is the aging coefficient of the rail transit equipment, H is the ambient humidity, G is the ambient wind speed, T is the ambient temperature, is the equipment operating time factor, 、 、 and They are the influence coefficients of ambient humidity, ambient wind speed, ambient temperature and equipment operating time factors, The calculation formula is: ； Wherein, t is the running time of the rail transit equipment, t max The design life of the rail transit equipment; When the failure probability of the rail transit equipment is greater than a critical value of the failure probability, it is determined that the rail transit equipment has failed, and an early warning is issued for the rail transit equipment.

4. The rail transit equipment fault prediction method according to claim 3, characterized in that: The warning level of the rail transit equipment is determined by the following formula: ； Among them, P f is the failure probability of the rail transit equipment, L is the remaining life of the rail transit equipment, L max is the design life of the rail transit equipment, H is the ambient humidity, and H max To preset the maximum ambient humidity, is the weight coefficient, and round is the rounding function.

5. The rail transit equipment fault prediction method according to claim 4, characterized in that: The remaining life R of the rail transit equipment is predicted by Weibull distribution and equipment degradation model, and the formula is as follows: ； in, and are the characteristic life and shape parameters of the Weibull distribution parameters, and t is the current time.

6. A rail transit equipment fault prediction system, characterized in that: The rail transit equipment fault prediction method applied to any one of claims 1 to 5 comprises: A calculation module is used to obtain operating data and environmental data of the rail transit equipment in the current time period, determine the time series correlation coefficient and statistical characteristic dispersion of the operating data, and the environmental impact factor of the environmental data on the rail transit equipment failure; A determination module, configured to determine weight parameters of a time series prediction model based on the time series correlation coefficient, the statistical feature dispersion, and the environmental influencing factor, and to determine weight parameters of a classification model based on the weight parameters of the time series prediction model; a first prediction module, configured to input the operation data of the current time period into the time series prediction model, and obtain the operation data of the rail transit equipment in a future time period according to the weight parameters of the time series prediction model; The second prediction module is used to input the operation data of the future time period into the classification model, obtain the failure probability of the rail transit equipment according to the weight parameters of the classification model, and determine whether the rail transit equipment is faulty according to the failure probability.

Images