A rail transit equipment health degree prediction method and system based on multi-source information fusion

Abstract

The application discloses a kind of track traffic equipment health degree prediction method and system based on multi-source information fusion, it is related to track traffic equipment health degree prediction technical field, including: the multidimensional fault index data of track traffic equipment is collected;Data dimension reduction processing is carried out;The time series analysis is carried out to the fault feature vector after dimension reduction processing in combination with time stamp data;Health degree prediction is carried out by integrating algorithm model coupling time series analysis result and multi-source data;Based on health degree prediction value, the overall operation state of equipment and potential failure risk are evaluated and early warning.The application realizes the cooperation of data level, algorithm level and optimization level, the fault feature output by PCA model is one of the inputs of Prophet and XGBoost model, and XGBoost also combines other multi-source data for comprehensive prediction, the accuracy, real-time performance and reliability of prediction are ensured through effective model cooperation, which helps to discover potential faults and problems in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rail transit equipment health degree prediction, and more particularly to a rail transit equipment health degree prediction method and system based on multi-source information fusion. BACKGROUND

[0002] As an important part of public transportation, the running stability and reliability of rail transit equipment are of great significance to ensure the smoothness of urban traffic. Currently, due to the complexity of equipment, the variability of operating environment and other reasons, various abnormal situations often occur during the operation of rail transit equipment, leading to equipment failure and even affecting the normal operation of the entire rail transit system. In order to ensure the stable operation of rail transit equipment, the existing technical means usually include regular maintenance, inspection and fault elimination of the equipment. However, these methods can only deal with the equipment after failure, and cannot timely find the potential problems of the equipment, cannot achieve preventive maintenance, and affect the operation efficiency of the rail transit equipment.

[0003] Therefore, how to propose a rail transit equipment health degree prediction method and system based on multi-source information fusion, timely find potential problems of the equipment, predict the health degree of the equipment, use principal component analysis, Prophet model and XGBoost model and other means to realize accurate diagnosis of rail transit equipment failure, and improve the running stability and reliability of rail transit equipment is a problem that needs to be solved by those skilled in the art. SUMMARY

[0004] Therefore, the present application provides a rail transit equipment health degree prediction method and system based on multi-source information fusion, which collects equipment operation data, monitors the abnormal state of the equipment in real time, uses a fault diagnosis system to locate and diagnose the abnormality, and combines fault tree and equipment account information to predict possible failures and risks of the equipment, realizes timely discovery of potential problems of the equipment, prediction of the health degree of the equipment, and early warning, effectively reduces the failure rate of rail transit equipment, improves the operation efficiency, provides a reference for health degree prediction and warning of similar equipment in other fields, and provides technical support and guarantee for sustainable development of the rail transit industry. In order to achieve the above purpose, the present application adopts the following technical scheme:

[0005] A rail transit equipment health degree prediction method based on multi-source information fusion, comprising:

[0006] Collecting multi-dimensional fault index data of rail transit equipment;

[0007] Performing dimension reduction processing on the collected multi-dimensional fault index data;

[0008] Performing time series analysis on the fault feature vector after dimension reduction processing combined with timestamp data;

[0009] The health degree prediction is performed by coupling the time series analysis result and the multi-source data through an integrated algorithm model.

[0010] The overall operation state and potential failure risk of the equipment are evaluated and warned based on the health degree prediction value.

[0011] Optionally, the multi-dimensional fault indicator data comprises rail transit equipment data, abnormal monitoring data, fault diagnosis data and equipment account data.

[0012] Optionally, the dimension reduction processing of the collected multi-dimensional fault indicator data comprises:

[0013] The PCA model receives the collected multi-dimensional fault indicator data.

[0014] The integrated data is subjected to standardization processing to eliminate the influence of different dimensions and orders of magnitude.

[0015] The covariance matrix of the data set is calculated to analyze the correlation between the fault indicators.

[0016] The most important principal components are extracted from the covariance matrix through eigenvalue decomposition or singular value decomposition.

[0017] The original data is projected onto the most important principal components to obtain the dimension-reduced fault feature vector.

[0018] Optionally, the dimension-reduced fault feature vector is used as the data input for subsequent time series analysis.

[0019] Optionally, the time series analysis of the dimension-reduced fault feature vector combined with the timestamp data comprises:

[0020] The Prophet model receives the fault feature vector output by the PCA model and performs time series analysis combined with the timestamp data.

[0021] The time series is decomposed into a trend item, a seasonal item and a holiday effect item, and is fitted through linear regression and Fourier transform method.

[0022] The parameters of the Prophet model are optimized through cross-validation method.

[0023] The optimized model is used to predict the fault feature at a future time point, and the confidence interval of the prediction value is calculated.

[0024] Optionally, the confidence interval of the prediction value is used to evaluate the future health condition of the rail transit equipment.

[0025] Optionally, the health degree prediction by coupling the time series analysis result and the multi-source data through the integrated algorithm model comprises:

[0026] receiving the time series analysis result of the Prophet model and the multi-source data by the XGBoost model;

[0027] The XGBoost model integrates multiple decision tree models and optimizes the prediction performance by using a gradient boosting algorithm.

[0028] The hyperparameters of the model are adjusted by using a grid search and a random search method.

[0029] Optionally, the XGBoost model outputs a predicted value of the health degree of the rail transit equipment, which is used to evaluate the overall operation state and potential failure risk of the rail transit equipment.

[0030] Optionally, the multi-source data includes environmental temperature, humidity and load conditions.

[0031] Optionally, a health degree prediction system of rail transit equipment based on multi-source information fusion comprises:

[0032] a collection module configured to collect multi-dimensional fault indicator data of the rail transit equipment;

[0033] a dimension reduction module configured to perform dimension reduction processing on the collected multi-dimensional fault indicator data;

[0034] a time series analysis module configured to perform time series analysis on the fault feature vector after the dimension reduction processing in combination with timestamp data;

[0035] a prediction module configured to perform health degree prediction by coupling the time series analysis result and the multi-source data by using an integrated algorithm model;

[0036] an evaluation and early warning module configured to evaluate and early warn the overall operation state and potential failure risk of the equipment based on the health degree prediction value.

[0037] Compared with the prior art, the health degree prediction method and system of rail transit equipment based on multi-source information fusion have the following beneficial effects:

[0038] The health degree prediction method of rail transit equipment based on multi-source information fusion comprises the following steps: collecting multi-dimensional fault indicator data of the rail transit equipment; performing dimension reduction processing on the collected multi-dimensional fault indicator data; performing time series analysis on the fault feature vector after the dimension reduction processing in combination with timestamp data; performing health degree prediction by coupling the time series analysis result and the multi-source data by using an integrated algorithm model; and evaluating and early warning the overall operation state and potential failure risk of the equipment based on the health degree prediction value.

[0039] The present application realizes data-level collaboration: the models collaborate through shared data. The fault features output by the PCA model serve as one of the inputs for the Prophet and XGBoost models, and the XGBoost also integrates other multi-source data for comprehensive prediction.

[0040] Algorithm-level collaboration: the PCA model reduces data redundancy and noise through dimensionality reduction, improving the efficiency and accuracy of subsequent models; the Prophet model masters the future trend of device indicators through time series prediction; the XGBoost model realizes complex relationship modeling and prediction through integrating multiple decision tree models. The multiple model fusion structure complements each other at the algorithm level, and together improves the prediction ability of the entire system.

[0041] Optimization-level collaboration: during the model training and optimization process, optimization strategies and techniques can be shared among models. For example, the dimensionality reduction effect of the PCA model can affect the input quality of the Prophet and XGBoost models; the time series prediction results of the Prophet model can be used as one of the features of the XGBoost model; the hyperparameter tuning of the XGBoost model can also refer to the model performance of PCA and Prophet.

[0042] Through effective model collaboration, the accuracy, real-time performance and reliability of the prediction are ensured, which helps to discover potential faults and problems in time and provides strong support for the maintenance and operation of rail transit equipment. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on the provided drawings.

[0044] Figure 1 A flowchart of a rail transit equipment health degree prediction method based on multi-source information fusion is provided.

[0045] Figure 2 A structural framework diagram of a rail transit equipment health degree prediction system based on multi-source information fusion is provided.

[0046] Figure 3 A flowchart of a device health degree prediction and early warning method based on multi-source information fusion is provided.

[0047] Figure 4(a) is a PCA data dimensionality reduction diagram provided by the present application.

[0048] Figure 4(b) is a fault detection diagram provided by the present application.

[0049] Figure 5 The device timing index trend prediction analysis diagram provided by the present application.

[0050] Figure 6 The device fault diagnosis identification analysis diagram provided by the present application.

[0051] Figure 7 The health degree prediction diagram of multi-source information fusion provided by the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] The present application relates to a rail transit equipment health degree prediction method and system based on multi-source information fusion, which adopts principal component analysis (PCA), Prophet model and XGBoost model and other technical means to realize accurate diagnosis of rail transit equipment faults and prediction of future health status of the equipment.

[0054] The present application adopts principal component analysis (PCA) method to construct a fault diagnosis model, which is used to extract the most important information from numerous fault indicators to more accurately diagnose the fault causes of the equipment. The PCA analysis reduces the dimension of the fault indicators, improves the efficiency and accuracy of fault diagnosis. Principal component analysis extracts the most important information from a large number of equipment fault indicators, maximizes the variance of the original data, and reduces the high-dimensional fault indicator data to low dimension while retaining the most important information. This dimension reduction technique helps to improve the efficiency and accuracy of fault diagnosis. At the same time, the Prophet model, as a prediction model suitable for time series data, takes into account factors such as nonlinear trends and seasonal changes of time series, effectively fitting the attenuation and periodic change trends of rail transit equipment, so the Prophet model is used for time series prediction of equipment indicators, better mastering the future health status of the equipment, and providing index trend change information for equipment maintenance and fault prevention. Finally, the present application combines the XGBoost model to predict the health degree of the rail transit equipment. In order to more accurately predict the health degree of the equipment, the model combines index abnormal data, fault diagnosis data and equipment account data, more comprehensively considers the historical operation status and future development trend of the equipment, and thus more accurately predicts the health degree of the equipment.

[0055] The embodiment of the application discloses a rail transit equipment health degree prediction method based on multi-source information fusion, as shown in the formula (I) : Figure 1 The embodiment of the application discloses a rail transit equipment health degree prediction method based on multi-source information fusion, as shown in the formula (I) :

[0056] Collecting multi-dimensional fault index data of the rail transit equipment;

[0057] Performing dimension reduction processing on the collected multi-dimensional fault index data;

[0058] Performing time series analysis on the fault feature vectors after the dimension reduction processing in combination with timestamp data;

[0059] Performing health degree prediction through an integrated algorithm model coupling the time series analysis result and the multi-source data;

[0060] Performing equipment overall operation state and potential fault risk assessment and early warning based on the health degree prediction value.

[0061] In the specific embodiment, a device health degree prediction and early warning method based on multi-source information fusion, as shown in the formula (II), comprises the following steps: Figure 3

[0062] S1: Construction based on a device fault diagnosis identification module

[0063] Model input: The PCA model receives multi-dimensional fault index data from a device monitoring system, including rail transit equipment data and model output abnormal monitoring, fault diagnosis data and device account data, covering device operation state, fault condition, maintenance record and other related information;

[0064] Model processing: first, standardize the integrated data to eliminate the influence of different dimensions and orders of magnitude; second, calculate the covariance matrix of the data set to analyze the correlation between the fault indicators; and third, extract the most important principal components from the covariance matrix through eigenvalue decomposition or singular value decomposition, i.e., the dimensions with the largest amount of information. Finally, project the original data onto these principal components to obtain the fault feature vectors after dimension reduction;

[0065] Model output: The PCA model outputs the fault feature vectors after dimension reduction, which are used as the input of the subsequent model;

[0066] As shown in FIG. 4(a), the PCA model finds the maximum variance direction (i.e., the principal component) of the data by calculating the covariance matrix of the data, and projects the original data onto these directions; by selecting the principal components with higher cumulative contribution rate of eigenvalues, the calculation amount is reduced while most of the information is retained, so that the strong correlation relationship contained in the process data can be extracted, the dimension of the data is reduced, and most of the characteristics in the original data are retained;

[0067] ​As shown in Figure 4(b), the PCA model further divides the space into principal component subspaces and residual subspaces, and detects the process by checking whether statistics exceed control limits. In fault diagnosis, this is achieved by monitoring statistics such as T... 2 When performing fault diagnosis with SPE, and the PCA model detects a fault, further fault isolation can be implemented to determine the main source of the fault.

[0068] S2: Construction of a Prediction and Early Warning Module Based on Equipment Time-Series Indicators

[0069] Model input: The Prophet model receives the fault feature vector output by the PCA model and performs time series analysis in conjunction with timestamp data;

[0070] Model processing: The time series is decomposed into trend, seasonal, and holiday effects, and these terms are fitted using methods such as linear regression and Fourier transform. Model parameters, such as trend smoothness and seasonal cycle length, are optimized using methods like cross-validation. The optimized model is then used to predict fault characteristics at future time points, and the confidence intervals of the predicted values ​​are calculated.

[0071] Model output: Outputs future predicted values ​​of equipment indicators and their confidence intervals, used to assess the future health status of the equipment;

[0072] like Figure 5 As shown, the Prophet model can not only effectively identify and predict mutation points, but also handle data with missing values ​​without interpolation, improving the efficiency and accuracy of data processing. Furthermore, the model has a fast fitting speed, meeting the needs of real-time prediction. In addition, the Prophet model is highly interpretable; users can intuitively understand and adjust model parameters to better interpret and predict dynamic changes in data.

[0073] S3: Construction of a Device Health Prediction Module Based on Multi-Source Device Information

[0074] Model input: The XGBoost model receives the trend analysis output of the time series model and other multi-source data (such as ambient temperature, humidity, load conditions, etc.);

[0075] Model processing: The XGBoost model integrates multiple decision tree models and continuously optimizes prediction performance using the gradient boosting algorithm, comprehensively considering the impact of multi-source data on device health. Hyperparameters (such as the number of trees, depth, and learning rate) are adjusted using methods such as grid search and random search to improve the model's prediction performance.

[0076] Model optimization: In the process of model training and optimization, in order to improve the overall prediction performance, the dimension reduction effect of PCA model is used to improve the input feature quality of XGBoost model, and the time series prediction result of Prophet model is used as the additional feature of XGBoost model. Specifically, first, evaluate the effect of PCA model in dimension reduction and the prediction accuracy of Prophet model, then adjust the hyperparameters of XGBoost model according to the data characteristics processed by PCA model, such as reducing the number of features may need to adjust the maximum depth of tree or column sampling ratio, while considering the optimization of learning rate and iteration times. In addition, the prediction results of Prophet model (such as trend, seasonal component) are used as the input features of XGBoost model, and the related hyperparameters of XGBoost model are adjusted according to the importance of these new features. Based on the continuous iteration of the model, and the model performance of PCA model and Prophet model are optimized to maximize the synergistic effect between models, so as to improve the accuracy and robustness of the overall prediction system;

[0077] Model output: output the predicted value of equipment health degree, which is used to evaluate the overall operation state and potential failure risk of the equipment;

[0078] As shown in Figure 7 , a multi-source information fusion prediction model based on XGBoost model is constructed, which deeply fuses multi-dimensional information such as abnormal data, fault diagnosis results, equipment account information and historical maintenance records. Feature engineering techniques such as feature cross and encoding conversion are used to further mine the potential association between data. Through hyperparameter tuning, Bayesian optimization, cross-validation and model integration strategy, the prediction accuracy and robustness of the model are significantly improved.

[0079] In the specific embodiment, the specific steps of S1 include:

[0080] S11: X∈R n×m is defined as n samples and m variables, and the matrix X is decomposed, and the decomposition definition formula is as follows:

[0081]

[0082] In the formula, is the load matrix, T∈R n×k is the score matrix, and E is the residual matrix.

[0083] S12: Definition of score vector, the definition formula is as follows:

[0084]

[0085] S13: Calculation of X predicted value , the expression is as follows:

[0086]

[0087] In the formula, t is the score vector. This is the transpose of the load matrix.

[0088] S14: The calculation of residuals is defined as follows:

[0089]

[0090] S15: Statistic T 2 The calculations of SPE are defined as follows:

[0091] T 2 The statistic is defined as the sum of squares of the score vectors. It reflects the changes in data during the process by the fluctuations in the magnitudes of the principal component vectors within the principal component model. Its definition is:

[0092]

[0093] In the formula, t is the score vector. It is composed of the first k eigenvalues ​​of the covariance matrix.

[0094] The SPE statistic is defined as the sum of squares of the residuals between the sampled and estimated values, reflecting the degree of deviation of the measured values ​​from the principal component model at a given time. Its definition is as follows:

[0095] SPE = e·e T ;

[0096] S16: As Figure 6 As shown, based on T 2 The SPE statistic is used to measure the fault threshold. When the threshold is exceeded, it indicates that a fault has occurred. The control thresholds are defined as follows:

[0097] T 2 Statistical control line:

[0098]

[0099] In the formula, F k,n-k,α Let F represent the F distribution with degrees of freedom k and nk under confidence interval α.

[0100] SPE statistic control line:

[0101]

[0102] In the formula, c α λ is the standard normal deviation (percentile for the upper bound 1-α); where λ is the eigenvalue of the characteristic space covariance matrix.

[0103] In the specific embodiment, the specific steps of S2 include:

[0104] S21: Construction of Prophet model, in the form of:

[0105] y(t) = g(t) + s(t) + h(t) + ε;

[0106] In the formula, g(t) is a trend function, s(t) is a periodic function, h(t) is a holiday function, and ε is an error.

[0107] S22: Construction of Prophet model function, the Logistic function model is as follows:

[0108]

[0109] In the formula, C is the maximum asymptotic value of the function, k is the growth rate, and m is the midpoint of the curve.

[0110] S23: Calculation of model trend function, the calculation formula is as follows:

[0111]

[0112] In the formula, C(t) is the set upper limit value, k is the initial growth rate, t represents the time t, and γ represents the growth rate change amount on the time stamp; Where a(t) is an indicator function, its expression on the time stamp s is a(t) = (a(t), …, a s (t)) T .

[0113] S24: The calculation of the definition of piecewise linear function is as follows:

[0114] g(t) = (k + a(t) T δ) × [t + (m + a(t) T γ) ;

[0115] In the formula, k is the model growth rate, δ is the growth amount under the growth rate, and m represents the bias amount.

[0116] In the specific embodiment, the specific steps of S3 include:

[0117] S31: XGBoost model is a strong learner model composed of decision tree as base learner, and its construction formula is as follows:

[0118]

[0119] In the formula, is the predicted value; x i is the i-th sample; k is the number of decision trees, and f kFor decision tree weight, n is the number of samples.

[0120] S32: Model training based on XGBoost, the objective function at the tth iteration is:

[0121]

[0122] wherein, is a loss function, wherein is the model prediction value at the t-1th moment; g i and h i are the first and second derivatives of the error function at f t , respectively; W(f t ) is the regularization term of the tth tree model complexity.

[0123] S33: Based on the iteration of the model, the objective function is as follows:

[0124]

[0125] wherein, j is a leaf node, I j is the current node data set, T is the current sub-tree depth, η is a parameter in the regularization replacement term, and σ is a calculation parameter of the leaf node.

[0126] In the specific embodiment, a rail transit equipment health degree prediction system based on multi-source information fusion, as shown in Figure 2 , comprises:

[0127] The acquisition module is configured to acquire multi-dimensional fault indicator data of the rail transit equipment.

[0128] The dimension reduction module is configured to perform dimension reduction processing on the acquired multi-dimensional fault indicator data.

[0129] The time series analysis module is configured to perform time series analysis on the fault feature vector after the dimension reduction processing in combination with timestamp data.

[0130] The prediction module is configured to perform health degree prediction by coupling the time series analysis result and the multi-source data through an integrated algorithm model.

[0131] The evaluation and early warning module is configured to perform evaluation and early warning on the overall operation state and potential fault risk of the equipment based on the health degree prediction value.

[0132] The present application relates to a health degree prediction and early warning method of rail transit equipment based on abnormal monitoring index alarm, fault diagnosis system, fault tree and equipment account information, which comprehensively utilizes principal component analysis, Prophet model and XGBoost algorithm model, and realizes the prediction of the health condition of rail transit equipment. By comprehensively analyzing abnormal data, fault diagnosis data and equipment account information, fully considering the past operation state and index trend of the equipment, the equipment health degree prediction and early warning model based on multi-source information fusion is constructed; by collecting equipment operation data, the abnormal state of the equipment is monitored in real time, the abnormality is positioned and diagnosed by using the fault diagnosis system, and the possible faults and risks of the equipment are predicted by combining the fault tree and the equipment account information. The potential problems of the equipment can be found in time, the health degree of the equipment can be predicted, and early prediction and early warning can be realized, so as to effectively reduce the failure rate of the rail transit equipment, thereby improving the operation efficiency; not only can the scientific basis be provided for the maintenance and repair of the rail transit equipment, and the operation efficiency and reliability of the equipment are improved, but also the application range of the health degree prediction and early warning of the equipment is far beyond the specific equipment, and has a wide application prospect.

[0133] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0134] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A rail transit equipment health degree prediction method based on multi-source information fusion, characterized in that, The method comprises the following steps: Collecting multi-dimensional fault indicator data of rail transit equipment; Performing dimensionality reduction processing on the collected multi-dimensional fault indicator data; Performing time series analysis on the fault feature vectors after dimensionality reduction processing combined with timestamp data; The time series analysis on the fault feature vectors after dimensionality reduction processing combined with timestamp data comprises: Receiving the fault feature vectors output by the PCA model through the Prophet model, and performing time series analysis combined with timestamp data; Decomposing the time series into trend items, seasonal items and holiday effect items, and fitting through linear regression and Fourier transform methods; Optimizing the parameters of the Prophet model through cross-validation method; Using the optimized model to predict the fault features at future time points, and calculating the confidence interval of the prediction value; The confidence interval of the prediction value is used to evaluate the future health status of the rail transit equipment; Performing health degree prediction by coupling the time series analysis results and multi-source data through an integrated algorithm model; The health degree prediction by coupling the time series analysis results and multi-source data through an integrated algorithm model comprises: Receiving the time series analysis results of the Prophet model and multi-source data through the XGBoost model; The XGBoost model integrates multiple decision tree models, and optimizes the prediction performance through gradient boosting algorithm; And the hyperparameters of the model are adjusted through grid search and random search methods; The XGBoost model outputs the prediction value of the health degree of the rail transit equipment, which is used to evaluate the overall operation state and potential fault risk of the rail transit equipment; The multi-source data includes environmental temperature, humidity and load condition; Based on the health degree prediction value, the overall operation state and potential fault risk of the equipment are evaluated and warned. 2.The rail transit equipment health degree prediction method based on multi-source information fusion according to claim 1, characterized in that, The multi-dimensional fault indicator data includes rail transit equipment data, abnormal monitoring data, fault diagnosis data and equipment account data. 3.The rail transit equipment health degree prediction method based on multi-source information fusion of claim 1, wherein, The dimensionality reduction processing on the collected multi-dimensional fault indicator data comprises: Receiving the collected multi-dimensional fault indicator data through the PCA model; Standardizing the integrated data to eliminate the influence of different dimensions and orders of magnitude; Calculating the covariance matrix of the data set to analyze the correlation between the fault indicators; Extracting the most important principal components from the covariance matrix through eigenvalue decomposition or singular value decomposition; Projecting the original data onto the most important principal components to obtain the fault feature vectors after dimensionality reduction.

4. The track transportation equipment health degree prediction method based on multi-source information fusion according to claim 3, characterized in that, The fault feature vectors after dimensionality reduction are used as data input for subsequent time series analysis.

5. A rail transit equipment health degree prediction system based on multi-source information fusion, characterized in that, The method comprises the following steps: A collection module for collecting multi-dimensional fault indicator data of rail transit equipment; A dimensionality reduction module for performing dimensionality reduction processing on the collected multi-dimensional fault indicator data; A time series analysis module for performing time series analysis on the fault feature vectors after dimensionality reduction processing combined with timestamp data; The time series analysis on the fault feature vectors after dimensionality reduction processing combined with timestamp data comprises: Receiving the fault feature vectors output by the PCA model through the Prophet model, and performing time series analysis combined with timestamp data; Decomposing the time series into trend items, seasonal items and holiday effect items, and fitting through linear regression and Fourier transform methods; Optimizing the parameters of the Prophet model through cross-validation method; Using the optimized model to predict the fault features at future time points, and calculating the confidence interval of the prediction value; The confidence interval of the prediction value is used to evaluate the future health status of the rail transit equipment; Performing health degree prediction by coupling the time series analysis results and multi-source data through an integrated algorithm model; The health degree prediction by coupling the time series analysis results and multi-source data through an integrated algorithm model comprises: Receiving the time series analysis results of the Prophet model and multi-source data through the XGBoost model; The XGBoost model integrates multiple decision tree models, and optimizes the prediction performance through gradient boosting algorithm; And the hyperparameters of the model are adjusted through grid search and random search methods; The XGBoost model outputs the prediction value of the health degree of the rail transit equipment, which is used to evaluate the overall operation state and potential fault risk of the rail transit equipment; The multi-source data includes environmental temperature, humidity and load condition; Based on the health degree prediction value, the overall operation state and potential fault risk of the equipment are evaluated and warned. The parameters of the Prophet model are optimized by a cross-validation method; The optimized model is used to predict the failure characteristics at future time points and calculate the confidence interval of the predicted values; The confidence interval of the predicted values is used to evaluate the future health condition of the rail transit equipment; The prediction module is used to perform health degree prediction by coupling the time series analysis results and the multi-source data through an integrated algorithm model; The health degree prediction by coupling the time series analysis results and the multi-source data through the integrated algorithm model includes: The time series analysis results of the Prophet model and the multi-source data are received by an XGBoost model; The XGBoost model integrates multiple decision tree models and optimizes the prediction performance by using a gradient boosting algorithm; The hyperparameters of the model are adjusted by a grid search and a random search method; The XGBoost model outputs the predicted values of the health degree of the rail transit equipment, which are used to evaluate the overall operation state and the potential failure risk of the rail transit equipment; The multi-source data includes environmental temperature, humidity, and load conditions; The evaluation and warning module is used to evaluate and warn the overall operation state and the potential failure risk of the equipment based on the predicted values of the health degree.

Images